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.
28 #include "dynamic-string.h"
35 #include "ofp-print.h"
37 #include "ofproto/ofproto.h"
38 #include "poll-loop.h"
41 #include "socket-util.h"
42 #include "stream-ssl.h"
44 #include "system-stats.h"
48 #include "vswitchd/vswitch-idl.h"
49 #include "xenserver.h"
51 #include "sflow_api.h"
52 #include "vlan-bitmap.h"
54 VLOG_DEFINE_THIS_MODULE(bridge
);
56 COVERAGE_DEFINE(bridge_reconfigure
);
59 /* These members are always valid. */
60 struct list port_elem
; /* Element in struct port's "ifaces" list. */
61 struct hmap_node name_node
; /* In struct bridge's "iface_by_name" hmap. */
62 struct port
*port
; /* Containing port. */
63 char *name
; /* Host network device name. */
64 tag_type tag
; /* Tag associated with this interface. */
66 /* These members are valid only after bridge_reconfigure() causes them to
68 struct hmap_node ofp_port_node
; /* In struct bridge's "ifaces" hmap. */
69 int ofp_port
; /* OpenFlow port number, -1 if unknown. */
70 struct netdev
*netdev
; /* Network device. */
71 const char *type
; /* Usually same as cfg->type. */
72 const struct ovsrec_interface
*cfg
;
76 struct uuid uuid
; /* UUID of this "mirror" record in database. */
77 struct hmap_node hmap_node
; /* In struct bridge's "mirrors" hmap. */
78 struct bridge
*bridge
;
83 struct bridge
*bridge
;
84 struct hmap_node hmap_node
; /* Element in struct bridge's "ports" hmap. */
87 const struct ovsrec_port
*cfg
;
89 /* An ordinary bridge port has 1 interface.
90 * A bridge port for bonding has at least 2 interfaces. */
91 struct list ifaces
; /* List of "struct iface"s. */
95 struct hmap_node node
; /* In 'all_bridges'. */
96 char *name
; /* User-specified arbitrary name. */
97 char *type
; /* Datapath type. */
98 uint8_t ea
[ETH_ADDR_LEN
]; /* Bridge Ethernet Address. */
99 uint8_t default_ea
[ETH_ADDR_LEN
]; /* Default MAC. */
100 const struct ovsrec_bridge
*cfg
;
102 /* OpenFlow switch processing. */
103 struct ofproto
*ofproto
; /* OpenFlow switch. */
106 struct hmap ports
; /* "struct port"s indexed by name. */
107 struct hmap ifaces
; /* "struct iface"s indexed by ofp_port. */
108 struct hmap iface_by_name
; /* "struct iface"s indexed by name. */
110 /* Port mirroring. */
111 struct hmap mirrors
; /* "struct mirror" indexed by UUID. */
113 /* Synthetic local port if necessary. */
114 struct ovsrec_port synth_local_port
;
115 struct ovsrec_interface synth_local_iface
;
116 struct ovsrec_interface
*synth_local_ifacep
;
119 /* All bridges, indexed by name. */
120 static struct hmap all_bridges
= HMAP_INITIALIZER(&all_bridges
);
122 /* OVSDB IDL used to obtain configuration. */
123 static struct ovsdb_idl
*idl
;
125 /* Each time this timer expires, the bridge fetches systems and interface
126 * statistics and pushes them into the database. */
127 #define STATS_INTERVAL (5 * 1000) /* In milliseconds. */
128 static long long int stats_timer
= LLONG_MIN
;
130 /* Stores the time after which rate limited statistics may be written to the
131 * database. Only updated when changes to the database require rate limiting.
133 #define DB_LIMIT_INTERVAL (1 * 1000) /* In milliseconds. */
134 static long long int db_limiter
= LLONG_MIN
;
136 static void add_del_bridges(const struct ovsrec_open_vswitch
*);
137 static void bridge_del_ofprotos(void);
138 static bool bridge_add_ofprotos(struct bridge
*);
139 static void bridge_create(const struct ovsrec_bridge
*);
140 static void bridge_destroy(struct bridge
*);
141 static struct bridge
*bridge_lookup(const char *name
);
142 static unixctl_cb_func bridge_unixctl_dump_flows
;
143 static unixctl_cb_func bridge_unixctl_reconnect
;
144 static size_t bridge_get_controllers(const struct bridge
*br
,
145 struct ovsrec_controller
***controllersp
);
146 static void bridge_add_del_ports(struct bridge
*);
147 static void bridge_add_ofproto_ports(struct bridge
*);
148 static void bridge_del_ofproto_ports(struct bridge
*);
149 static void bridge_refresh_ofp_port(struct bridge
*);
150 static void bridge_configure_datapath_id(struct bridge
*);
151 static void bridge_configure_flow_eviction_threshold(struct bridge
*);
152 static void bridge_configure_netflow(struct bridge
*);
153 static void bridge_configure_forward_bpdu(struct bridge
*);
154 static void bridge_configure_sflow(struct bridge
*, int *sflow_bridge_number
);
155 static void bridge_configure_remotes(struct bridge
*,
156 const struct sockaddr_in
*managers
,
158 static void bridge_pick_local_hw_addr(struct bridge
*,
159 uint8_t ea
[ETH_ADDR_LEN
],
160 struct iface
**hw_addr_iface
);
161 static uint64_t bridge_pick_datapath_id(struct bridge
*,
162 const uint8_t bridge_ea
[ETH_ADDR_LEN
],
163 struct iface
*hw_addr_iface
);
164 static uint64_t dpid_from_hash(const void *, size_t nbytes
);
165 static bool bridge_has_bond_fake_iface(const struct bridge
*,
167 static bool port_is_bond_fake_iface(const struct port
*);
169 static unixctl_cb_func qos_unixctl_show
;
171 static struct port
*port_create(struct bridge
*, const struct ovsrec_port
*);
172 static void port_add_ifaces(struct port
*);
173 static void port_del_ifaces(struct port
*);
174 static void port_destroy(struct port
*);
175 static struct port
*port_lookup(const struct bridge
*, const char *name
);
176 static void port_configure(struct port
*);
177 static struct lacp_settings
*port_configure_lacp(struct port
*,
178 struct lacp_settings
*);
179 static void port_configure_bond(struct port
*, struct bond_settings
*,
180 uint32_t *bond_stable_ids
);
182 static void bridge_configure_mirrors(struct bridge
*);
183 static struct mirror
*mirror_create(struct bridge
*,
184 const struct ovsrec_mirror
*);
185 static void mirror_destroy(struct mirror
*);
186 static bool mirror_configure(struct mirror
*, const struct ovsrec_mirror
*);
188 static void iface_configure_lacp(struct iface
*, struct lacp_slave_settings
*);
189 static struct iface
*iface_create(struct port
*port
,
190 const struct ovsrec_interface
*if_cfg
);
191 static void iface_destroy(struct iface
*);
192 static struct iface
*iface_lookup(const struct bridge
*, const char *name
);
193 static struct iface
*iface_find(const char *name
);
194 static struct iface
*iface_from_ofp_port(const struct bridge
*,
196 static void iface_set_mac(struct iface
*);
197 static void iface_set_ofport(const struct ovsrec_interface
*, int64_t ofport
);
198 static void iface_clear_db_record(const struct ovsrec_interface
*if_cfg
);
199 static void iface_configure_qos(struct iface
*, const struct ovsrec_qos
*);
200 static void iface_configure_cfm(struct iface
*);
201 static void iface_refresh_cfm_stats(struct iface
*);
202 static void iface_refresh_stats(struct iface
*);
203 static void iface_refresh_status(struct iface
*);
204 static bool iface_is_synthetic(const struct iface
*);
206 static void shash_from_ovs_idl_map(char **keys
, char **values
, size_t n
,
208 static void shash_to_ovs_idl_map(struct shash
*,
209 char ***keys
, char ***values
, size_t *n
);
211 /* Public functions. */
213 /* Initializes the bridge module, configuring it to obtain its configuration
214 * from an OVSDB server accessed over 'remote', which should be a string in a
215 * form acceptable to ovsdb_idl_create(). */
217 bridge_init(const char *remote
)
219 /* Create connection to database. */
220 idl
= ovsdb_idl_create(remote
, &ovsrec_idl_class
, true);
221 ovsdb_idl_set_lock(idl
, "ovs_vswitchd");
223 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_cur_cfg
);
224 ovsdb_idl_omit_alert(idl
, &ovsrec_open_vswitch_col_statistics
);
225 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_external_ids
);
226 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_ovs_version
);
227 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_db_version
);
228 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_type
);
229 ovsdb_idl_omit(idl
, &ovsrec_open_vswitch_col_system_version
);
231 ovsdb_idl_omit_alert(idl
, &ovsrec_bridge_col_datapath_id
);
232 ovsdb_idl_omit(idl
, &ovsrec_bridge_col_external_ids
);
234 ovsdb_idl_omit(idl
, &ovsrec_port_col_external_ids
);
235 ovsdb_idl_omit(idl
, &ovsrec_port_col_fake_bridge
);
237 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_admin_state
);
238 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_duplex
);
239 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_speed
);
240 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_state
);
241 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_link_resets
);
242 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_mtu
);
243 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_ofport
);
244 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_statistics
);
245 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_status
);
246 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_fault
);
247 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_cfm_remote_mpids
);
248 ovsdb_idl_omit_alert(idl
, &ovsrec_interface_col_lacp_current
);
249 ovsdb_idl_omit(idl
, &ovsrec_interface_col_external_ids
);
251 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_is_connected
);
252 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_role
);
253 ovsdb_idl_omit_alert(idl
, &ovsrec_controller_col_status
);
254 ovsdb_idl_omit(idl
, &ovsrec_controller_col_external_ids
);
256 ovsdb_idl_omit(idl
, &ovsrec_qos_col_external_ids
);
258 ovsdb_idl_omit(idl
, &ovsrec_queue_col_external_ids
);
260 ovsdb_idl_omit(idl
, &ovsrec_mirror_col_external_ids
);
262 ovsdb_idl_omit(idl
, &ovsrec_netflow_col_external_ids
);
264 ovsdb_idl_omit(idl
, &ovsrec_sflow_col_external_ids
);
266 ovsdb_idl_omit(idl
, &ovsrec_manager_col_external_ids
);
267 ovsdb_idl_omit(idl
, &ovsrec_manager_col_inactivity_probe
);
268 ovsdb_idl_omit(idl
, &ovsrec_manager_col_is_connected
);
269 ovsdb_idl_omit(idl
, &ovsrec_manager_col_max_backoff
);
270 ovsdb_idl_omit(idl
, &ovsrec_manager_col_status
);
272 ovsdb_idl_omit(idl
, &ovsrec_ssl_col_external_ids
);
274 /* Register unixctl commands. */
275 unixctl_command_register("qos/show", "interface", qos_unixctl_show
, NULL
);
276 unixctl_command_register("bridge/dump-flows", "bridge",
277 bridge_unixctl_dump_flows
, NULL
);
278 unixctl_command_register("bridge/reconnect", "[bridge]",
279 bridge_unixctl_reconnect
, NULL
);
288 struct bridge
*br
, *next_br
;
290 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
293 ovsdb_idl_destroy(idl
);
296 /* Looks at the list of managers in 'ovs_cfg' and extracts their remote IP
297 * addresses and ports into '*managersp' and '*n_managersp'. The caller is
298 * responsible for freeing '*managersp' (with free()).
300 * You may be asking yourself "why does ovs-vswitchd care?", because
301 * ovsdb-server is responsible for connecting to the managers, and ovs-vswitchd
302 * should not be and in fact is not directly involved in that. But
303 * ovs-vswitchd needs to make sure that ovsdb-server can reach the managers, so
304 * it has to tell in-band control where the managers are to enable that.
305 * (Thus, only managers connected in-band are collected.)
308 collect_in_band_managers(const struct ovsrec_open_vswitch
*ovs_cfg
,
309 struct sockaddr_in
**managersp
, size_t *n_managersp
)
311 struct sockaddr_in
*managers
= NULL
;
312 size_t n_managers
= 0;
316 /* Collect all of the potential targets from the "targets" columns of the
317 * rows pointed to by "manager_options", excluding any that are
320 for (i
= 0; i
< ovs_cfg
->n_manager_options
; i
++) {
321 struct ovsrec_manager
*m
= ovs_cfg
->manager_options
[i
];
323 if (m
->connection_mode
&& !strcmp(m
->connection_mode
, "out-of-band")) {
324 sset_find_and_delete(&targets
, m
->target
);
326 sset_add(&targets
, m
->target
);
330 /* Now extract the targets' IP addresses. */
331 if (!sset_is_empty(&targets
)) {
334 managers
= xmalloc(sset_count(&targets
) * sizeof *managers
);
335 SSET_FOR_EACH (target
, &targets
) {
336 struct sockaddr_in
*sin
= &managers
[n_managers
];
338 if ((!strncmp(target
, "tcp:", 4)
339 && inet_parse_active(target
+ 4, JSONRPC_TCP_PORT
, sin
)) ||
340 (!strncmp(target
, "ssl:", 4)
341 && inet_parse_active(target
+ 4, JSONRPC_SSL_PORT
, sin
))) {
346 sset_destroy(&targets
);
348 *managersp
= managers
;
349 *n_managersp
= n_managers
;
353 bridge_reconfigure(const struct ovsrec_open_vswitch
*ovs_cfg
)
355 struct sockaddr_in
*managers
;
356 struct bridge
*br
, *next
;
357 int sflow_bridge_number
;
360 COVERAGE_INC(bridge_reconfigure
);
362 /* Create and destroy "struct bridge"s, "struct port"s, and "struct
363 * iface"s according to 'ovs_cfg', with only very minimal configuration
366 * This is purely an update to bridge data structures. Nothing is pushed
367 * down to ofproto or lower layers. */
368 add_del_bridges(ovs_cfg
);
369 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
370 bridge_add_del_ports(br
);
373 /* Delete all datapaths and datapath ports that are no longer configured.
375 * The kernel will reject any attempt to add a given port to a datapath if
376 * that port already belongs to a different datapath, so we must do all
377 * port deletions before any port additions. A datapath always has a
378 * "local port" so we must delete not-configured datapaths too. */
379 bridge_del_ofprotos();
380 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
382 bridge_del_ofproto_ports(br
);
386 /* Create datapaths and datapath ports that are missing.
388 * After this is done, we have our final set of bridges, ports, and
389 * interfaces. Every "struct bridge" has an ofproto, every "struct port"
390 * has at least one iface, every "struct iface" has a valid ofp_port and
392 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
393 if (!br
->ofproto
&& !bridge_add_ofprotos(br
)) {
397 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
398 bridge_refresh_ofp_port(br
);
399 bridge_add_ofproto_ports(br
);
402 /* Complete the configuration. */
403 sflow_bridge_number
= 0;
404 collect_in_band_managers(ovs_cfg
, &managers
, &n_managers
);
405 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
408 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
411 port_configure(port
);
413 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
414 iface_configure_cfm(iface
);
415 iface_configure_qos(iface
, port
->cfg
->qos
);
416 iface_set_mac(iface
);
419 bridge_configure_mirrors(br
);
420 bridge_configure_datapath_id(br
);
421 bridge_configure_flow_eviction_threshold(br
);
422 bridge_configure_forward_bpdu(br
);
423 bridge_configure_remotes(br
, managers
, n_managers
);
424 bridge_configure_netflow(br
);
425 bridge_configure_sflow(br
, &sflow_bridge_number
);
429 /* ovs-vswitchd has completed initialization, so allow the process that
430 * forked us to exit successfully. */
431 daemonize_complete();
434 /* Iterate over all ofprotos and delete any of them that do not have a
435 * configured bridge or that are the wrong type. */
437 bridge_del_ofprotos(void)
445 ofproto_enumerate_types(&types
);
446 SSET_FOR_EACH (type
, &types
) {
449 ofproto_enumerate_names(type
, &names
);
450 SSET_FOR_EACH (name
, &names
) {
451 struct bridge
*br
= bridge_lookup(name
);
452 if (!br
|| strcmp(type
, br
->type
)) {
453 ofproto_delete(name
, type
);
457 sset_destroy(&names
);
458 sset_destroy(&types
);
462 bridge_add_ofprotos(struct bridge
*br
)
464 int error
= ofproto_create(br
->name
, br
->type
, &br
->ofproto
);
466 VLOG_ERR("failed to create bridge %s: %s", br
->name
, strerror(error
));
473 port_configure(struct port
*port
)
475 const struct ovsrec_port
*cfg
= port
->cfg
;
476 struct bond_settings bond_settings
;
477 struct lacp_settings lacp_settings
;
478 struct ofproto_bundle_settings s
;
486 s
.slaves
= xmalloc(list_size(&port
->ifaces
) * sizeof *s
.slaves
);
487 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
488 s
.slaves
[s
.n_slaves
++] = iface
->ofp_port
;
494 if (list_is_short(&port
->ifaces
)) {
495 if (*cfg
->tag
>= 0 && *cfg
->tag
<= 4095) {
499 /* It's possible that bonded, VLAN-tagged ports make sense. Maybe
500 * they even work as-is. But they have not been tested. */
501 VLOG_WARN("port %s: VLAN tags not supported on bonded ports",
506 /* Get VLAN trunks. */
509 s
.trunks
= vlan_bitmap_from_array(cfg
->trunks
, cfg
->n_trunks
);
513 if (cfg
->vlan_mode
) {
514 if (!strcmp(cfg
->vlan_mode
, "access")) {
515 s
.vlan_mode
= PORT_VLAN_ACCESS
;
516 } else if (!strcmp(cfg
->vlan_mode
, "trunk")) {
517 s
.vlan_mode
= PORT_VLAN_TRUNK
;
518 } else if (!strcmp(cfg
->vlan_mode
, "native-tagged")) {
519 s
.vlan_mode
= PORT_VLAN_NATIVE_TAGGED
;
520 } else if (!strcmp(cfg
->vlan_mode
, "native-untagged")) {
521 s
.vlan_mode
= PORT_VLAN_NATIVE_UNTAGGED
;
523 /* This "can't happen" because ovsdb-server should prevent it. */
524 VLOG_ERR("unknown VLAN mode %s", cfg
->vlan_mode
);
525 s
.vlan_mode
= PORT_VLAN_TRUNK
;
529 s
.vlan_mode
= PORT_VLAN_ACCESS
;
531 VLOG_ERR("port %s: ignoring trunks in favor of implicit vlan",
535 s
.vlan_mode
= PORT_VLAN_TRUNK
;
539 /* Get LACP settings. */
540 s
.lacp
= port_configure_lacp(port
, &lacp_settings
);
544 s
.lacp_slaves
= xmalloc(s
.n_slaves
* sizeof *s
.lacp_slaves
);
545 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
546 iface_configure_lacp(iface
, &s
.lacp_slaves
[i
++]);
549 s
.lacp_slaves
= NULL
;
552 /* Get bond settings. */
553 if (s
.n_slaves
> 1) {
554 s
.bond
= &bond_settings
;
555 s
.bond_stable_ids
= xmalloc(s
.n_slaves
* sizeof *s
.bond_stable_ids
);
556 port_configure_bond(port
, &bond_settings
, s
.bond_stable_ids
);
559 s
.bond_stable_ids
= NULL
;
561 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
562 netdev_set_miimon_interval(iface
->netdev
, 0);
567 ofproto_bundle_register(port
->bridge
->ofproto
, port
, &s
);
573 free(s
.bond_stable_ids
);
576 /* Pick local port hardware address and datapath ID for 'br'. */
578 bridge_configure_datapath_id(struct bridge
*br
)
580 uint8_t ea
[ETH_ADDR_LEN
];
582 struct iface
*local_iface
;
583 struct iface
*hw_addr_iface
;
586 bridge_pick_local_hw_addr(br
, ea
, &hw_addr_iface
);
587 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
589 int error
= netdev_set_etheraddr(local_iface
->netdev
, ea
);
591 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
592 VLOG_ERR_RL(&rl
, "bridge %s: failed to set bridge "
593 "Ethernet address: %s",
594 br
->name
, strerror(error
));
597 memcpy(br
->ea
, ea
, ETH_ADDR_LEN
);
599 dpid
= bridge_pick_datapath_id(br
, ea
, hw_addr_iface
);
600 ofproto_set_datapath_id(br
->ofproto
, dpid
);
602 dpid_string
= xasprintf("%016"PRIx64
, dpid
);
603 ovsrec_bridge_set_datapath_id(br
->cfg
, dpid_string
);
607 /* Set NetFlow configuration on 'br'. */
609 bridge_configure_netflow(struct bridge
*br
)
611 struct ovsrec_netflow
*cfg
= br
->cfg
->netflow
;
612 struct netflow_options opts
;
615 ofproto_set_netflow(br
->ofproto
, NULL
);
619 memset(&opts
, 0, sizeof opts
);
621 /* Get default NetFlow configuration from datapath.
622 * Apply overrides from 'cfg'. */
623 ofproto_get_netflow_ids(br
->ofproto
, &opts
.engine_type
, &opts
.engine_id
);
624 if (cfg
->engine_type
) {
625 opts
.engine_type
= *cfg
->engine_type
;
627 if (cfg
->engine_id
) {
628 opts
.engine_id
= *cfg
->engine_id
;
631 /* Configure active timeout interval. */
632 opts
.active_timeout
= cfg
->active_timeout
;
633 if (!opts
.active_timeout
) {
634 opts
.active_timeout
= -1;
635 } else if (opts
.active_timeout
< 0) {
636 VLOG_WARN("bridge %s: active timeout interval set to negative "
637 "value, using default instead (%d seconds)", br
->name
,
638 NF_ACTIVE_TIMEOUT_DEFAULT
);
639 opts
.active_timeout
= -1;
642 /* Add engine ID to interface number to disambiguate bridgs? */
643 opts
.add_id_to_iface
= cfg
->add_id_to_interface
;
644 if (opts
.add_id_to_iface
) {
645 if (opts
.engine_id
> 0x7f) {
646 VLOG_WARN("bridge %s: NetFlow port mangling may conflict with "
647 "another vswitch, choose an engine id less than 128",
650 if (hmap_count(&br
->ports
) > 508) {
651 VLOG_WARN("bridge %s: NetFlow port mangling will conflict with "
652 "another port when more than 508 ports are used",
658 sset_init(&opts
.collectors
);
659 sset_add_array(&opts
.collectors
, cfg
->targets
, cfg
->n_targets
);
662 if (ofproto_set_netflow(br
->ofproto
, &opts
)) {
663 VLOG_ERR("bridge %s: problem setting netflow collectors", br
->name
);
665 sset_destroy(&opts
.collectors
);
668 /* Set sFlow configuration on 'br'. */
670 bridge_configure_sflow(struct bridge
*br
, int *sflow_bridge_number
)
672 const struct ovsrec_sflow
*cfg
= br
->cfg
->sflow
;
673 struct ovsrec_controller
**controllers
;
674 struct ofproto_sflow_options oso
;
675 size_t n_controllers
;
679 ofproto_set_sflow(br
->ofproto
, NULL
);
683 memset(&oso
, 0, sizeof oso
);
685 sset_init(&oso
.targets
);
686 sset_add_array(&oso
.targets
, cfg
->targets
, cfg
->n_targets
);
688 oso
.sampling_rate
= SFL_DEFAULT_SAMPLING_RATE
;
690 oso
.sampling_rate
= *cfg
->sampling
;
693 oso
.polling_interval
= SFL_DEFAULT_POLLING_INTERVAL
;
695 oso
.polling_interval
= *cfg
->polling
;
698 oso
.header_len
= SFL_DEFAULT_HEADER_SIZE
;
700 oso
.header_len
= *cfg
->header
;
703 oso
.sub_id
= (*sflow_bridge_number
)++;
704 oso
.agent_device
= cfg
->agent
;
706 oso
.control_ip
= NULL
;
707 n_controllers
= bridge_get_controllers(br
, &controllers
);
708 for (i
= 0; i
< n_controllers
; i
++) {
709 if (controllers
[i
]->local_ip
) {
710 oso
.control_ip
= controllers
[i
]->local_ip
;
714 ofproto_set_sflow(br
->ofproto
, &oso
);
716 sset_destroy(&oso
.targets
);
720 bridge_has_bond_fake_iface(const struct bridge
*br
, const char *name
)
722 const struct port
*port
= port_lookup(br
, name
);
723 return port
&& port_is_bond_fake_iface(port
);
727 port_is_bond_fake_iface(const struct port
*port
)
729 return port
->cfg
->bond_fake_iface
&& !list_is_short(&port
->ifaces
);
733 add_del_bridges(const struct ovsrec_open_vswitch
*cfg
)
735 struct bridge
*br
, *next
;
739 /* Collect new bridges' names and types. */
741 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
742 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
743 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
745 if (strchr(br_cfg
->name
, '/')) {
746 /* Prevent remote ovsdb-server users from accessing arbitrary
747 * directories, e.g. consider a bridge named "../../../etc/". */
748 VLOG_WARN_RL(&rl
, "ignoring bridge with invalid name \"%s\"",
750 } else if (!shash_add_once(&new_br
, br_cfg
->name
, br_cfg
)) {
751 VLOG_WARN_RL(&rl
, "bridge %s specified twice", br_cfg
->name
);
755 /* Get rid of deleted bridges or those whose types have changed.
756 * Update 'cfg' of bridges that still exist. */
757 HMAP_FOR_EACH_SAFE (br
, next
, node
, &all_bridges
) {
758 br
->cfg
= shash_find_data(&new_br
, br
->name
);
759 if (!br
->cfg
|| strcmp(br
->type
, ofproto_normalize_type(
760 br
->cfg
->datapath_type
))) {
765 /* Add new bridges. */
766 for (i
= 0; i
< cfg
->n_bridges
; i
++) {
767 const struct ovsrec_bridge
*br_cfg
= cfg
->bridges
[i
];
768 struct bridge
*br
= bridge_lookup(br_cfg
->name
);
770 bridge_create(br_cfg
);
774 shash_destroy(&new_br
);
777 /* Delete each ofproto port on 'br' that doesn't have a corresponding "struct
780 * The kernel will reject any attempt to add a given port to a datapath if that
781 * port already belongs to a different datapath, so we must do all port
782 * deletions before any port additions. */
784 bridge_del_ofproto_ports(struct bridge
*br
)
786 struct ofproto_port_dump dump
;
787 struct ofproto_port ofproto_port
;
789 OFPROTO_PORT_FOR_EACH (&ofproto_port
, &dump
, br
->ofproto
) {
790 const char *name
= ofproto_port
.name
;
795 /* Ignore the local port. We can't change it anyhow. */
796 if (!strcmp(name
, br
->name
)) {
800 /* Get the type that 'ofproto_port' should have (ordinarily the
801 * type of its corresponding iface) or NULL if it should be
803 iface
= iface_lookup(br
, name
);
804 type
= (iface
? iface
->type
805 : bridge_has_bond_fake_iface(br
, name
) ? "internal"
808 /* If it's the wrong type then delete the ofproto port. */
810 && !strcmp(ofproto_port
.type
, type
)
811 && (!iface
|| !iface
->netdev
812 || !strcmp(netdev_get_type(iface
->netdev
), type
))) {
815 error
= ofproto_port_del(br
->ofproto
, ofproto_port
.ofp_port
);
817 VLOG_WARN("bridge %s: failed to remove %s interface (%s)",
818 br
->name
, name
, strerror(error
));
821 netdev_close(iface
->netdev
);
822 iface
->netdev
= NULL
;
828 iface_set_ofp_port(struct iface
*iface
, int ofp_port
)
830 struct bridge
*br
= iface
->port
->bridge
;
832 assert(iface
->ofp_port
< 0 && ofp_port
>= 0);
833 iface
->ofp_port
= ofp_port
;
834 hmap_insert(&br
->ifaces
, &iface
->ofp_port_node
, hash_int(ofp_port
, 0));
835 iface_set_ofport(iface
->cfg
, ofp_port
);
839 bridge_refresh_ofp_port(struct bridge
*br
)
841 struct ofproto_port_dump dump
;
842 struct ofproto_port ofproto_port
;
845 /* Clear all the "ofp_port"es. */
846 hmap_clear(&br
->ifaces
);
847 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
850 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
851 iface
->ofp_port
= -1;
855 /* Obtain the correct "ofp_port"s from ofproto. */
856 OFPROTO_PORT_FOR_EACH (&ofproto_port
, &dump
, br
->ofproto
) {
857 struct iface
*iface
= iface_lookup(br
, ofproto_port
.name
);
859 if (iface
->ofp_port
>= 0) {
860 VLOG_WARN("bridge %s: interface %s reported twice",
861 br
->name
, ofproto_port
.name
);
862 } else if (iface_from_ofp_port(br
, ofproto_port
.ofp_port
)) {
863 VLOG_WARN("bridge %s: interface %"PRIu16
" reported twice",
864 br
->name
, ofproto_port
.ofp_port
);
866 iface_set_ofp_port(iface
, ofproto_port
.ofp_port
);
872 /* Add an ofproto port for any "struct iface" that doesn't have one.
873 * Delete any "struct iface" for which this fails.
874 * Delete any "struct port" that thereby ends up with no ifaces. */
876 bridge_add_ofproto_ports(struct bridge
*br
)
878 struct port
*port
, *next_port
;
880 HMAP_FOR_EACH_SAFE (port
, next_port
, hmap_node
, &br
->ports
) {
881 struct iface
*iface
, *next_iface
;
882 struct ofproto_port ofproto_port
;
884 LIST_FOR_EACH_SAFE (iface
, next_iface
, port_elem
, &port
->ifaces
) {
887 /* Open the netdev. */
888 if (!iface
->netdev
) {
889 error
= netdev_open(iface
->name
, iface
->type
, &iface
->netdev
);
891 VLOG_WARN("could not open network device %s (%s)",
892 iface
->name
, strerror(error
));
898 /* Configure the netdev. */
903 shash_from_ovs_idl_map(iface
->cfg
->key_options
,
904 iface
->cfg
->value_options
,
905 iface
->cfg
->n_options
, &args
);
906 error
= netdev_set_config(iface
->netdev
, &args
);
907 shash_destroy(&args
);
910 VLOG_WARN("could not configure network device %s (%s)",
911 iface
->name
, strerror(error
));
912 netdev_close(iface
->netdev
);
913 iface
->netdev
= NULL
;
917 /* Add the port, if necessary. */
918 if (iface
->netdev
&& iface
->ofp_port
< 0) {
922 error
= ofproto_port_add(br
->ofproto
, iface
->netdev
,
925 iface_set_ofp_port(iface
, ofp_port
);
927 netdev_close(iface
->netdev
);
928 iface
->netdev
= NULL
;
932 /* Populate stats columns in new Interface rows. */
933 if (iface
->netdev
&& !iface
->cfg
->mtu
) {
934 iface_refresh_stats(iface
);
935 iface_refresh_status(iface
);
938 /* Delete the iface if we failed. */
939 if (iface
->netdev
&& iface
->ofp_port
>= 0) {
940 VLOG_DBG("bridge %s: interface %s is on port %d",
941 br
->name
, iface
->name
, iface
->ofp_port
);
944 VLOG_ERR("bridge %s: missing %s interface, dropping",
945 br
->name
, iface
->name
);
947 /* We already reported a related error, don't bother
950 iface_clear_db_record(iface
->cfg
);
951 iface_destroy(iface
);
954 if (list_is_empty(&port
->ifaces
)) {
955 VLOG_WARN("%s port has no interfaces, dropping", port
->name
);
960 /* Add bond fake iface if necessary. */
961 if (port_is_bond_fake_iface(port
)) {
962 if (ofproto_port_query_by_name(br
->ofproto
, port
->name
,
964 struct netdev
*netdev
;
967 error
= netdev_open(port
->name
, "internal", &netdev
);
969 ofproto_port_add(br
->ofproto
, netdev
, NULL
);
970 netdev_close(netdev
);
972 VLOG_WARN("could not open network device %s (%s)",
973 port
->name
, strerror(error
));
976 /* Already exists, nothing to do. */
977 ofproto_port_destroy(&ofproto_port
);
984 get_ovsrec_key_value(const struct ovsdb_idl_row
*row
,
985 const struct ovsdb_idl_column
*column
,
988 const struct ovsdb_datum
*datum
;
989 union ovsdb_atom atom
;
992 datum
= ovsdb_idl_get(row
, column
, OVSDB_TYPE_STRING
, OVSDB_TYPE_STRING
);
993 atom
.string
= (char *) key
;
994 idx
= ovsdb_datum_find_key(datum
, &atom
, OVSDB_TYPE_STRING
);
995 return idx
== UINT_MAX
? NULL
: datum
->values
[idx
].string
;
999 bridge_get_other_config(const struct ovsrec_bridge
*br_cfg
, const char *key
)
1001 return get_ovsrec_key_value(&br_cfg
->header_
,
1002 &ovsrec_bridge_col_other_config
, key
);
1005 /* Set Flow eviction threshold */
1007 bridge_configure_flow_eviction_threshold(struct bridge
*br
)
1009 const char *threshold_str
;
1012 threshold_str
= bridge_get_other_config(br
->cfg
, "flow-eviction-threshold");
1013 if (threshold_str
) {
1014 threshold
= strtoul(threshold_str
, NULL
, 10);
1016 threshold
= OFPROTO_FLOW_EVICTON_THRESHOLD_DEFAULT
;
1018 ofproto_set_flow_eviction_threshold(br
->ofproto
, threshold
);
1021 /* Set forward BPDU option. */
1023 bridge_configure_forward_bpdu(struct bridge
*br
)
1025 const char *forward_bpdu_str
;
1026 bool forward_bpdu
= false;
1028 forward_bpdu_str
= bridge_get_other_config(br
->cfg
, "forward-bpdu");
1029 if (forward_bpdu_str
&& !strcmp(forward_bpdu_str
, "true")) {
1030 forward_bpdu
= true;
1032 ofproto_set_forward_bpdu(br
->ofproto
, forward_bpdu
);
1036 bridge_pick_local_hw_addr(struct bridge
*br
, uint8_t ea
[ETH_ADDR_LEN
],
1037 struct iface
**hw_addr_iface
)
1041 bool found_addr
= false;
1044 *hw_addr_iface
= NULL
;
1046 /* Did the user request a particular MAC? */
1047 hwaddr
= bridge_get_other_config(br
->cfg
, "hwaddr");
1048 if (hwaddr
&& eth_addr_from_string(hwaddr
, ea
)) {
1049 if (eth_addr_is_multicast(ea
)) {
1050 VLOG_ERR("bridge %s: cannot set MAC address to multicast "
1051 "address "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(ea
));
1052 } else if (eth_addr_is_zero(ea
)) {
1053 VLOG_ERR("bridge %s: cannot set MAC address to zero", br
->name
);
1059 /* Otherwise choose the minimum non-local MAC address among all of the
1061 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1062 uint8_t iface_ea
[ETH_ADDR_LEN
];
1063 struct iface
*candidate
;
1064 struct iface
*iface
;
1066 /* Mirror output ports don't participate. */
1067 if (ofproto_is_mirror_output_bundle(br
->ofproto
, port
)) {
1071 /* Choose the MAC address to represent the port. */
1073 if (port
->cfg
->mac
&& eth_addr_from_string(port
->cfg
->mac
, iface_ea
)) {
1074 /* Find the interface with this Ethernet address (if any) so that
1075 * we can provide the correct devname to the caller. */
1076 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1077 uint8_t candidate_ea
[ETH_ADDR_LEN
];
1078 if (!netdev_get_etheraddr(candidate
->netdev
, candidate_ea
)
1079 && eth_addr_equals(iface_ea
, candidate_ea
)) {
1084 /* Choose the interface whose MAC address will represent the port.
1085 * The Linux kernel bonding code always chooses the MAC address of
1086 * the first slave added to a bond, and the Fedora networking
1087 * scripts always add slaves to a bond in alphabetical order, so
1088 * for compatibility we choose the interface with the name that is
1089 * first in alphabetical order. */
1090 LIST_FOR_EACH (candidate
, port_elem
, &port
->ifaces
) {
1091 if (!iface
|| strcmp(candidate
->name
, iface
->name
) < 0) {
1096 /* The local port doesn't count (since we're trying to choose its
1097 * MAC address anyway). */
1098 if (iface
->ofp_port
== OFPP_LOCAL
) {
1103 error
= netdev_get_etheraddr(iface
->netdev
, iface_ea
);
1109 /* Compare against our current choice. */
1110 if (!eth_addr_is_multicast(iface_ea
) &&
1111 !eth_addr_is_local(iface_ea
) &&
1112 !eth_addr_is_reserved(iface_ea
) &&
1113 !eth_addr_is_zero(iface_ea
) &&
1114 (!found_addr
|| eth_addr_compare_3way(iface_ea
, ea
) < 0))
1116 memcpy(ea
, iface_ea
, ETH_ADDR_LEN
);
1117 *hw_addr_iface
= iface
;
1122 VLOG_DBG("bridge %s: using bridge Ethernet address "ETH_ADDR_FMT
,
1123 br
->name
, ETH_ADDR_ARGS(ea
));
1125 memcpy(ea
, br
->default_ea
, ETH_ADDR_LEN
);
1126 *hw_addr_iface
= NULL
;
1127 VLOG_WARN("bridge %s: using default bridge Ethernet "
1128 "address "ETH_ADDR_FMT
, br
->name
, ETH_ADDR_ARGS(ea
));
1132 /* Choose and returns the datapath ID for bridge 'br' given that the bridge
1133 * Ethernet address is 'bridge_ea'. If 'bridge_ea' is the Ethernet address of
1134 * an interface on 'br', then that interface must be passed in as
1135 * 'hw_addr_iface'; if 'bridge_ea' was derived some other way, then
1136 * 'hw_addr_iface' must be passed in as a null pointer. */
1138 bridge_pick_datapath_id(struct bridge
*br
,
1139 const uint8_t bridge_ea
[ETH_ADDR_LEN
],
1140 struct iface
*hw_addr_iface
)
1143 * The procedure for choosing a bridge MAC address will, in the most
1144 * ordinary case, also choose a unique MAC that we can use as a datapath
1145 * ID. In some special cases, though, multiple bridges will end up with
1146 * the same MAC address. This is OK for the bridges, but it will confuse
1147 * the OpenFlow controller, because each datapath needs a unique datapath
1150 * Datapath IDs must be unique. It is also very desirable that they be
1151 * stable from one run to the next, so that policy set on a datapath
1154 const char *datapath_id
;
1157 datapath_id
= bridge_get_other_config(br
->cfg
, "datapath-id");
1158 if (datapath_id
&& dpid_from_string(datapath_id
, &dpid
)) {
1162 if (hw_addr_iface
) {
1164 if (!netdev_get_vlan_vid(hw_addr_iface
->netdev
, &vlan
)) {
1166 * A bridge whose MAC address is taken from a VLAN network device
1167 * (that is, a network device created with vconfig(8) or similar
1168 * tool) will have the same MAC address as a bridge on the VLAN
1169 * device's physical network device.
1171 * Handle this case by hashing the physical network device MAC
1172 * along with the VLAN identifier.
1174 uint8_t buf
[ETH_ADDR_LEN
+ 2];
1175 memcpy(buf
, bridge_ea
, ETH_ADDR_LEN
);
1176 buf
[ETH_ADDR_LEN
] = vlan
>> 8;
1177 buf
[ETH_ADDR_LEN
+ 1] = vlan
;
1178 return dpid_from_hash(buf
, sizeof buf
);
1181 * Assume that this bridge's MAC address is unique, since it
1182 * doesn't fit any of the cases we handle specially.
1187 * A purely internal bridge, that is, one that has no non-virtual
1188 * network devices on it at all, is more difficult because it has no
1189 * natural unique identifier at all.
1191 * When the host is a XenServer, we handle this case by hashing the
1192 * host's UUID with the name of the bridge. Names of bridges are
1193 * persistent across XenServer reboots, although they can be reused if
1194 * an internal network is destroyed and then a new one is later
1195 * created, so this is fairly effective.
1197 * When the host is not a XenServer, we punt by using a random MAC
1198 * address on each run.
1200 const char *host_uuid
= xenserver_get_host_uuid();
1202 char *combined
= xasprintf("%s,%s", host_uuid
, br
->name
);
1203 dpid
= dpid_from_hash(combined
, strlen(combined
));
1209 return eth_addr_to_uint64(bridge_ea
);
1213 dpid_from_hash(const void *data
, size_t n
)
1215 uint8_t hash
[SHA1_DIGEST_SIZE
];
1217 BUILD_ASSERT_DECL(sizeof hash
>= ETH_ADDR_LEN
);
1218 sha1_bytes(data
, n
, hash
);
1219 eth_addr_mark_random(hash
);
1220 return eth_addr_to_uint64(hash
);
1224 iface_refresh_status(struct iface
*iface
)
1228 enum netdev_flags flags
;
1235 if (iface_is_synthetic(iface
)) {
1241 if (!netdev_get_status(iface
->netdev
, &sh
)) {
1243 char **keys
, **values
;
1245 shash_to_ovs_idl_map(&sh
, &keys
, &values
, &n
);
1246 ovsrec_interface_set_status(iface
->cfg
, keys
, values
, n
);
1251 ovsrec_interface_set_status(iface
->cfg
, NULL
, NULL
, 0);
1254 shash_destroy_free_data(&sh
);
1256 error
= netdev_get_flags(iface
->netdev
, &flags
);
1258 ovsrec_interface_set_admin_state(iface
->cfg
, flags
& NETDEV_UP
? "up" : "down");
1261 ovsrec_interface_set_admin_state(iface
->cfg
, NULL
);
1264 error
= netdev_get_features(iface
->netdev
, ¤t
, NULL
, NULL
, NULL
);
1266 ovsrec_interface_set_duplex(iface
->cfg
,
1267 netdev_features_is_full_duplex(current
)
1269 /* warning: uint64_t -> int64_t conversion */
1270 bps
= netdev_features_to_bps(current
);
1271 ovsrec_interface_set_link_speed(iface
->cfg
, &bps
, 1);
1274 ovsrec_interface_set_duplex(iface
->cfg
, NULL
);
1275 ovsrec_interface_set_link_speed(iface
->cfg
, NULL
, 0);
1278 error
= netdev_get_mtu(iface
->netdev
, &mtu
);
1281 ovsrec_interface_set_mtu(iface
->cfg
, &mtu_64
, 1);
1284 ovsrec_interface_set_mtu(iface
->cfg
, NULL
, 0);
1288 /* Writes 'iface''s CFM statistics to the database. */
1290 iface_refresh_cfm_stats(struct iface
*iface
)
1292 const struct ovsrec_interface
*cfg
= iface
->cfg
;
1294 const uint64_t *rmps
;
1297 if (iface_is_synthetic(iface
)) {
1301 fault
= ofproto_port_get_cfm_fault(iface
->port
->bridge
->ofproto
,
1304 bool fault_bool
= fault
;
1305 ovsrec_interface_set_cfm_fault(cfg
, &fault_bool
, 1);
1307 ovsrec_interface_set_cfm_fault(cfg
, NULL
, 0);
1310 error
= ofproto_port_get_cfm_remote_mpids(iface
->port
->bridge
->ofproto
,
1311 iface
->ofp_port
, &rmps
, &n_rmps
);
1313 ovsrec_interface_set_cfm_remote_mpids(cfg
, (const int64_t *)rmps
,
1316 ovsrec_interface_set_cfm_remote_mpids(cfg
, NULL
, 0);
1321 iface_refresh_stats(struct iface
*iface
)
1323 #define IFACE_STATS \
1324 IFACE_STAT(rx_packets, "rx_packets") \
1325 IFACE_STAT(tx_packets, "tx_packets") \
1326 IFACE_STAT(rx_bytes, "rx_bytes") \
1327 IFACE_STAT(tx_bytes, "tx_bytes") \
1328 IFACE_STAT(rx_dropped, "rx_dropped") \
1329 IFACE_STAT(tx_dropped, "tx_dropped") \
1330 IFACE_STAT(rx_errors, "rx_errors") \
1331 IFACE_STAT(tx_errors, "tx_errors") \
1332 IFACE_STAT(rx_frame_errors, "rx_frame_err") \
1333 IFACE_STAT(rx_over_errors, "rx_over_err") \
1334 IFACE_STAT(rx_crc_errors, "rx_crc_err") \
1335 IFACE_STAT(collisions, "collisions")
1337 #define IFACE_STAT(MEMBER, NAME) NAME,
1338 static char *keys
[] = { IFACE_STATS
};
1340 int64_t values
[ARRAY_SIZE(keys
)];
1343 struct netdev_stats stats
;
1345 if (iface_is_synthetic(iface
)) {
1349 /* Intentionally ignore return value, since errors will set 'stats' to
1350 * all-1s, and we will deal with that correctly below. */
1351 netdev_get_stats(iface
->netdev
, &stats
);
1353 /* Copy statistics into values[] array. */
1355 #define IFACE_STAT(MEMBER, NAME) values[i++] = stats.MEMBER;
1358 assert(i
== ARRAY_SIZE(keys
));
1360 ovsrec_interface_set_statistics(iface
->cfg
, keys
, values
, ARRAY_SIZE(keys
));
1365 enable_system_stats(const struct ovsrec_open_vswitch
*cfg
)
1369 /* Use other-config:enable-system-stats by preference. */
1370 enable
= get_ovsrec_key_value(&cfg
->header_
,
1371 &ovsrec_open_vswitch_col_other_config
,
1372 "enable-statistics");
1374 return !strcmp(enable
, "true");
1377 /* Disable by default. */
1382 refresh_system_stats(const struct ovsrec_open_vswitch
*cfg
)
1384 struct ovsdb_datum datum
;
1388 if (enable_system_stats(cfg
)) {
1389 get_system_stats(&stats
);
1392 ovsdb_datum_from_shash(&datum
, &stats
);
1393 ovsdb_idl_txn_write(&cfg
->header_
, &ovsrec_open_vswitch_col_statistics
,
1397 static inline const char *
1398 nx_role_to_str(enum nx_role role
)
1403 case NX_ROLE_MASTER
:
1408 return "*** INVALID ROLE ***";
1413 refresh_controller_status(void)
1417 const struct ovsrec_controller
*cfg
;
1421 /* Accumulate status for controllers on all bridges. */
1422 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
1423 ofproto_get_ofproto_controller_info(br
->ofproto
, &info
);
1426 /* Update each controller in the database with current status. */
1427 OVSREC_CONTROLLER_FOR_EACH(cfg
, idl
) {
1428 struct ofproto_controller_info
*cinfo
=
1429 shash_find_data(&info
, cfg
->target
);
1432 ovsrec_controller_set_is_connected(cfg
, cinfo
->is_connected
);
1433 ovsrec_controller_set_role(cfg
, nx_role_to_str(cinfo
->role
));
1434 ovsrec_controller_set_status(cfg
, (char **) cinfo
->pairs
.keys
,
1435 (char **) cinfo
->pairs
.values
,
1438 ovsrec_controller_set_is_connected(cfg
, false);
1439 ovsrec_controller_set_role(cfg
, NULL
);
1440 ovsrec_controller_set_status(cfg
, NULL
, NULL
, 0);
1444 ofproto_free_ofproto_controller_info(&info
);
1448 refresh_cfm_stats(void)
1450 static struct ovsdb_idl_txn
*txn
= NULL
;
1455 txn
= ovsdb_idl_txn_create(idl
);
1457 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
1458 struct iface
*iface
;
1460 HMAP_FOR_EACH (iface
, name_node
, &br
->iface_by_name
) {
1461 iface_refresh_cfm_stats(iface
);
1466 if (ovsdb_idl_txn_commit(txn
) != TXN_INCOMPLETE
) {
1467 ovsdb_idl_txn_destroy(txn
);
1475 const struct ovsrec_open_vswitch
*cfg
;
1477 bool datapath_destroyed
;
1478 bool database_changed
;
1481 /* (Re)configure if necessary. */
1482 database_changed
= ovsdb_idl_run(idl
);
1483 if (ovsdb_idl_is_lock_contended(idl
)) {
1484 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
1485 struct bridge
*br
, *next_br
;
1487 VLOG_ERR_RL(&rl
, "another ovs-vswitchd process is running, "
1488 "disabling this process until it goes away");
1490 HMAP_FOR_EACH_SAFE (br
, next_br
, node
, &all_bridges
) {
1494 } else if (!ovsdb_idl_has_lock(idl
)) {
1497 cfg
= ovsrec_open_vswitch_first(idl
);
1499 /* Let each bridge do the work that it needs to do. */
1500 datapath_destroyed
= false;
1501 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
1502 int error
= ofproto_run(br
->ofproto
);
1504 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1505 VLOG_ERR_RL(&rl
, "bridge %s: datapath was destroyed externally, "
1506 "forcing reconfiguration", br
->name
);
1507 datapath_destroyed
= true;
1511 /* Re-configure SSL. We do this on every trip through the main loop,
1512 * instead of just when the database changes, because the contents of the
1513 * key and certificate files can change without the database changing.
1515 * We do this before bridge_reconfigure() because that function might
1516 * initiate SSL connections and thus requires SSL to be configured. */
1517 if (cfg
&& cfg
->ssl
) {
1518 const struct ovsrec_ssl
*ssl
= cfg
->ssl
;
1520 stream_ssl_set_key_and_cert(ssl
->private_key
, ssl
->certificate
);
1521 stream_ssl_set_ca_cert_file(ssl
->ca_cert
, ssl
->bootstrap_ca_cert
);
1524 if (database_changed
|| datapath_destroyed
) {
1526 struct ovsdb_idl_txn
*txn
= ovsdb_idl_txn_create(idl
);
1528 bridge_reconfigure(cfg
);
1530 ovsrec_open_vswitch_set_cur_cfg(cfg
, cfg
->next_cfg
);
1531 ovsdb_idl_txn_commit(txn
);
1532 ovsdb_idl_txn_destroy(txn
); /* XXX */
1534 /* We still need to reconfigure to avoid dangling pointers to
1535 * now-destroyed ovsrec structures inside bridge data. */
1536 static const struct ovsrec_open_vswitch null_cfg
;
1538 bridge_reconfigure(&null_cfg
);
1542 /* Refresh system and interface stats if necessary. */
1543 if (time_msec() >= stats_timer
) {
1545 struct ovsdb_idl_txn
*txn
;
1547 txn
= ovsdb_idl_txn_create(idl
);
1548 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
1551 HMAP_FOR_EACH (port
, hmap_node
, &br
->ports
) {
1552 struct iface
*iface
;
1554 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
1555 iface_refresh_stats(iface
);
1556 iface_refresh_status(iface
);
1560 refresh_system_stats(cfg
);
1561 refresh_controller_status();
1562 ovsdb_idl_txn_commit(txn
);
1563 ovsdb_idl_txn_destroy(txn
); /* XXX */
1566 stats_timer
= time_msec() + STATS_INTERVAL
;
1569 if (time_msec() >= db_limiter
) {
1570 struct ovsdb_idl_txn
*txn
;
1572 txn
= ovsdb_idl_txn_create(idl
);
1573 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
1574 struct iface
*iface
;
1576 HMAP_FOR_EACH (iface
, name_node
, &br
->iface_by_name
) {
1577 const char *link_state
;
1578 int64_t link_resets
;
1581 if (iface_is_synthetic(iface
)) {
1585 current
= ofproto_port_is_lacp_current(br
->ofproto
,
1589 ovsrec_interface_set_lacp_current(iface
->cfg
, &bl
, 1);
1591 ovsrec_interface_set_lacp_current(iface
->cfg
, NULL
, 0);
1594 link_state
= netdev_get_carrier(iface
->netdev
) ? "up" : "down";
1595 ovsrec_interface_set_link_state(iface
->cfg
, link_state
);
1597 link_resets
= netdev_get_carrier_resets(iface
->netdev
);
1598 ovsrec_interface_set_link_resets(iface
->cfg
, &link_resets
, 1);
1602 if (ovsdb_idl_txn_commit(txn
) != TXN_UNCHANGED
) {
1603 db_limiter
= time_msec() + DB_LIMIT_INTERVAL
;
1605 ovsdb_idl_txn_destroy(txn
);
1608 refresh_cfm_stats();
1614 ovsdb_idl_wait(idl
);
1615 if (!hmap_is_empty(&all_bridges
)) {
1618 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
1619 ofproto_wait(br
->ofproto
);
1621 poll_timer_wait_until(stats_timer
);
1623 if (db_limiter
> time_msec()) {
1624 poll_timer_wait_until(db_limiter
);
1629 /* QoS unixctl user interface functions. */
1631 struct qos_unixctl_show_cbdata
{
1633 struct iface
*iface
;
1637 qos_unixctl_show_cb(unsigned int queue_id
,
1638 const struct shash
*details
,
1641 struct qos_unixctl_show_cbdata
*data
= aux
;
1642 struct ds
*ds
= data
->ds
;
1643 struct iface
*iface
= data
->iface
;
1644 struct netdev_queue_stats stats
;
1645 struct shash_node
*node
;
1648 ds_put_cstr(ds
, "\n");
1650 ds_put_format(ds
, "Queue %u:\n", queue_id
);
1652 ds_put_cstr(ds
, "Default:\n");
1655 SHASH_FOR_EACH (node
, details
) {
1656 ds_put_format(ds
, "\t%s: %s\n", node
->name
, (char *)node
->data
);
1659 error
= netdev_get_queue_stats(iface
->netdev
, queue_id
, &stats
);
1661 if (stats
.tx_packets
!= UINT64_MAX
) {
1662 ds_put_format(ds
, "\ttx_packets: %"PRIu64
"\n", stats
.tx_packets
);
1665 if (stats
.tx_bytes
!= UINT64_MAX
) {
1666 ds_put_format(ds
, "\ttx_bytes: %"PRIu64
"\n", stats
.tx_bytes
);
1669 if (stats
.tx_errors
!= UINT64_MAX
) {
1670 ds_put_format(ds
, "\ttx_errors: %"PRIu64
"\n", stats
.tx_errors
);
1673 ds_put_format(ds
, "\tFailed to get statistics for queue %u: %s",
1674 queue_id
, strerror(error
));
1679 qos_unixctl_show(struct unixctl_conn
*conn
,
1680 const char *args
, void *aux OVS_UNUSED
)
1682 struct ds ds
= DS_EMPTY_INITIALIZER
;
1683 struct shash sh
= SHASH_INITIALIZER(&sh
);
1684 struct iface
*iface
;
1686 struct shash_node
*node
;
1687 struct qos_unixctl_show_cbdata data
;
1690 iface
= iface_find(args
);
1692 unixctl_command_reply(conn
, 501, "no such interface");
1696 netdev_get_qos(iface
->netdev
, &type
, &sh
);
1698 if (*type
!= '\0') {
1699 ds_put_format(&ds
, "QoS: %s %s\n", iface
->name
, type
);
1701 SHASH_FOR_EACH (node
, &sh
) {
1702 ds_put_format(&ds
, "%s: %s\n", node
->name
, (char *)node
->data
);
1707 error
= netdev_dump_queues(iface
->netdev
, qos_unixctl_show_cb
, &data
);
1710 ds_put_format(&ds
, "failed to dump queues: %s", strerror(error
));
1712 unixctl_command_reply(conn
, 200, ds_cstr(&ds
));
1714 ds_put_format(&ds
, "QoS not configured on %s\n", iface
->name
);
1715 unixctl_command_reply(conn
, 501, ds_cstr(&ds
));
1718 shash_destroy_free_data(&sh
);
1722 /* Bridge reconfiguration functions. */
1724 bridge_create(const struct ovsrec_bridge
*br_cfg
)
1728 assert(!bridge_lookup(br_cfg
->name
));
1729 br
= xzalloc(sizeof *br
);
1731 br
->name
= xstrdup(br_cfg
->name
);
1732 br
->type
= xstrdup(ofproto_normalize_type(br_cfg
->datapath_type
));
1735 /* Derive the default Ethernet address from the bridge's UUID. This should
1736 * be unique and it will be stable between ovs-vswitchd runs. */
1737 memcpy(br
->default_ea
, &br_cfg
->header_
.uuid
, ETH_ADDR_LEN
);
1738 eth_addr_mark_random(br
->default_ea
);
1740 hmap_init(&br
->ports
);
1741 hmap_init(&br
->ifaces
);
1742 hmap_init(&br
->iface_by_name
);
1743 hmap_init(&br
->mirrors
);
1745 hmap_insert(&all_bridges
, &br
->node
, hash_string(br
->name
, 0));
1749 bridge_destroy(struct bridge
*br
)
1752 struct mirror
*mirror
, *next_mirror
;
1753 struct port
*port
, *next_port
;
1755 HMAP_FOR_EACH_SAFE (port
, next_port
, hmap_node
, &br
->ports
) {
1758 HMAP_FOR_EACH_SAFE (mirror
, next_mirror
, hmap_node
, &br
->mirrors
) {
1759 mirror_destroy(mirror
);
1761 hmap_remove(&all_bridges
, &br
->node
);
1762 ofproto_destroy(br
->ofproto
);
1763 hmap_destroy(&br
->ifaces
);
1764 hmap_destroy(&br
->ports
);
1765 hmap_destroy(&br
->iface_by_name
);
1766 hmap_destroy(&br
->mirrors
);
1773 static struct bridge
*
1774 bridge_lookup(const char *name
)
1778 HMAP_FOR_EACH_WITH_HASH (br
, node
, hash_string(name
, 0), &all_bridges
) {
1779 if (!strcmp(br
->name
, name
)) {
1786 /* Handle requests for a listing of all flows known by the OpenFlow
1787 * stack, including those normally hidden. */
1789 bridge_unixctl_dump_flows(struct unixctl_conn
*conn
,
1790 const char *args
, void *aux OVS_UNUSED
)
1795 br
= bridge_lookup(args
);
1797 unixctl_command_reply(conn
, 501, "Unknown bridge");
1802 ofproto_get_all_flows(br
->ofproto
, &results
);
1804 unixctl_command_reply(conn
, 200, ds_cstr(&results
));
1805 ds_destroy(&results
);
1808 /* "bridge/reconnect [BRIDGE]": makes BRIDGE drop all of its controller
1809 * connections and reconnect. If BRIDGE is not specified, then all bridges
1810 * drop their controller connections and reconnect. */
1812 bridge_unixctl_reconnect(struct unixctl_conn
*conn
,
1813 const char *args
, void *aux OVS_UNUSED
)
1816 if (args
[0] != '\0') {
1817 br
= bridge_lookup(args
);
1819 unixctl_command_reply(conn
, 501, "Unknown bridge");
1822 ofproto_reconnect_controllers(br
->ofproto
);
1824 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
1825 ofproto_reconnect_controllers(br
->ofproto
);
1828 unixctl_command_reply(conn
, 200, NULL
);
1832 bridge_get_controllers(const struct bridge
*br
,
1833 struct ovsrec_controller
***controllersp
)
1835 struct ovsrec_controller
**controllers
;
1836 size_t n_controllers
;
1838 controllers
= br
->cfg
->controller
;
1839 n_controllers
= br
->cfg
->n_controller
;
1841 if (n_controllers
== 1 && !strcmp(controllers
[0]->target
, "none")) {
1847 *controllersp
= controllers
;
1849 return n_controllers
;
1852 /* Adds and deletes "struct port"s and "struct iface"s under 'br' to match
1853 * those configured in 'br->cfg'. */
1855 bridge_add_del_ports(struct bridge
*br
)
1857 struct port
*port
, *next
;
1858 struct shash_node
*node
;
1859 struct shash new_ports
;
1862 /* Collect new ports. */
1863 shash_init(&new_ports
);
1864 for (i
= 0; i
< br
->cfg
->n_ports
; i
++) {
1865 const char *name
= br
->cfg
->ports
[i
]->name
;
1866 if (!shash_add_once(&new_ports
, name
, br
->cfg
->ports
[i
])) {
1867 VLOG_WARN("bridge %s: %s specified twice as bridge port",
1871 if (bridge_get_controllers(br
, NULL
)
1872 && !shash_find(&new_ports
, br
->name
)) {
1873 VLOG_WARN("bridge %s: no port named %s, synthesizing one",
1874 br
->name
, br
->name
);
1876 br
->synth_local_port
.interfaces
= &br
->synth_local_ifacep
;
1877 br
->synth_local_port
.n_interfaces
= 1;
1878 br
->synth_local_port
.name
= br
->name
;
1880 br
->synth_local_iface
.name
= br
->name
;
1881 br
->synth_local_iface
.type
= "internal";
1883 br
->synth_local_ifacep
= &br
->synth_local_iface
;
1885 shash_add(&new_ports
, br
->name
, &br
->synth_local_port
);
1888 /* Get rid of deleted ports.
1889 * Get rid of deleted interfaces on ports that still exist.
1890 * Update 'cfg' of ports that still exist. */
1891 HMAP_FOR_EACH_SAFE (port
, next
, hmap_node
, &br
->ports
) {
1892 port
->cfg
= shash_find_data(&new_ports
, port
->name
);
1896 port_del_ifaces(port
);
1900 /* Create new ports.
1901 * Add new interfaces to existing ports. */
1902 SHASH_FOR_EACH (node
, &new_ports
) {
1903 struct port
*port
= port_lookup(br
, node
->name
);
1905 struct ovsrec_port
*cfg
= node
->data
;
1906 port
= port_create(br
, cfg
);
1908 port_add_ifaces(port
);
1909 if (list_is_empty(&port
->ifaces
)) {
1910 VLOG_WARN("bridge %s: port %s has no interfaces, dropping",
1911 br
->name
, port
->name
);
1915 shash_destroy(&new_ports
);
1918 /* Initializes 'oc' appropriately as a management service controller for
1921 * The caller must free oc->target when it is no longer needed. */
1923 bridge_ofproto_controller_for_mgmt(const struct bridge
*br
,
1924 struct ofproto_controller
*oc
)
1926 oc
->target
= xasprintf("punix:%s/%s.mgmt", ovs_rundir(), br
->name
);
1927 oc
->max_backoff
= 0;
1928 oc
->probe_interval
= 60;
1929 oc
->band
= OFPROTO_OUT_OF_BAND
;
1931 oc
->burst_limit
= 0;
1934 /* Converts ovsrec_controller 'c' into an ofproto_controller in 'oc'. */
1936 bridge_ofproto_controller_from_ovsrec(const struct ovsrec_controller
*c
,
1937 struct ofproto_controller
*oc
)
1939 oc
->target
= c
->target
;
1940 oc
->max_backoff
= c
->max_backoff
? *c
->max_backoff
/ 1000 : 8;
1941 oc
->probe_interval
= c
->inactivity_probe
? *c
->inactivity_probe
/ 1000 : 5;
1942 oc
->band
= (!c
->connection_mode
|| !strcmp(c
->connection_mode
, "in-band")
1943 ? OFPROTO_IN_BAND
: OFPROTO_OUT_OF_BAND
);
1944 oc
->rate_limit
= c
->controller_rate_limit
? *c
->controller_rate_limit
: 0;
1945 oc
->burst_limit
= (c
->controller_burst_limit
1946 ? *c
->controller_burst_limit
: 0);
1949 /* Configures the IP stack for 'br''s local interface properly according to the
1950 * configuration in 'c'. */
1952 bridge_configure_local_iface_netdev(struct bridge
*br
,
1953 struct ovsrec_controller
*c
)
1955 struct netdev
*netdev
;
1956 struct in_addr mask
, gateway
;
1958 struct iface
*local_iface
;
1961 /* If there's no local interface or no IP address, give up. */
1962 local_iface
= iface_from_ofp_port(br
, OFPP_LOCAL
);
1963 if (!local_iface
|| !c
->local_ip
|| !inet_aton(c
->local_ip
, &ip
)) {
1967 /* Bring up the local interface. */
1968 netdev
= local_iface
->netdev
;
1969 netdev_turn_flags_on(netdev
, NETDEV_UP
, true);
1971 /* Configure the IP address and netmask. */
1972 if (!c
->local_netmask
1973 || !inet_aton(c
->local_netmask
, &mask
)
1975 mask
.s_addr
= guess_netmask(ip
.s_addr
);
1977 if (!netdev_set_in4(netdev
, ip
, mask
)) {
1978 VLOG_INFO("bridge %s: configured IP address "IP_FMT
", netmask "IP_FMT
,
1979 br
->name
, IP_ARGS(&ip
.s_addr
), IP_ARGS(&mask
.s_addr
));
1982 /* Configure the default gateway. */
1983 if (c
->local_gateway
1984 && inet_aton(c
->local_gateway
, &gateway
)
1985 && gateway
.s_addr
) {
1986 if (!netdev_add_router(netdev
, gateway
)) {
1987 VLOG_INFO("bridge %s: configured gateway "IP_FMT
,
1988 br
->name
, IP_ARGS(&gateway
.s_addr
));
1993 /* Returns true if 'a' and 'b' are the same except that any number of slashes
1994 * in either string are treated as equal to any number of slashes in the other,
1995 * e.g. "x///y" is equal to "x/y". */
1997 equal_pathnames(const char *a
, const char *b
)
2001 a
+= strspn(a
, "/");
2002 b
+= strspn(b
, "/");
2003 } else if (*a
== '\0') {
2014 bridge_configure_remotes(struct bridge
*br
,
2015 const struct sockaddr_in
*managers
, size_t n_managers
)
2017 const char *disable_ib_str
, *queue_id_str
;
2018 bool disable_in_band
= false;
2021 struct ovsrec_controller
**controllers
;
2022 size_t n_controllers
;
2024 enum ofproto_fail_mode fail_mode
;
2026 struct ofproto_controller
*ocs
;
2030 /* Check if we should disable in-band control on this bridge. */
2031 disable_ib_str
= bridge_get_other_config(br
->cfg
, "disable-in-band");
2032 if (disable_ib_str
&& !strcmp(disable_ib_str
, "true")) {
2033 disable_in_band
= true;
2036 /* Set OpenFlow queue ID for in-band control. */
2037 queue_id_str
= bridge_get_other_config(br
->cfg
, "in-band-queue");
2038 queue_id
= queue_id_str
? strtol(queue_id_str
, NULL
, 10) : -1;
2039 ofproto_set_in_band_queue(br
->ofproto
, queue_id
);
2041 if (disable_in_band
) {
2042 ofproto_set_extra_in_band_remotes(br
->ofproto
, NULL
, 0);
2044 ofproto_set_extra_in_band_remotes(br
->ofproto
, managers
, n_managers
);
2047 n_controllers
= bridge_get_controllers(br
, &controllers
);
2049 ocs
= xmalloc((n_controllers
+ 1) * sizeof *ocs
);
2052 bridge_ofproto_controller_for_mgmt(br
, &ocs
[n_ocs
++]);
2053 for (i
= 0; i
< n_controllers
; i
++) {
2054 struct ovsrec_controller
*c
= controllers
[i
];
2056 if (!strncmp(c
->target
, "punix:", 6)
2057 || !strncmp(c
->target
, "unix:", 5)) {
2058 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2061 whitelist
= xasprintf("unix:%s/%s.controller",
2062 ovs_rundir(), br
->name
);
2063 if (!equal_pathnames(c
->target
, whitelist
)) {
2064 /* Prevent remote ovsdb-server users from accessing arbitrary
2065 * Unix domain sockets and overwriting arbitrary local
2067 VLOG_ERR_RL(&rl
, "bridge %s: Not adding Unix domain socket "
2068 "controller \"%s\" due to possibility for remote "
2069 "exploit. Instead, specify whitelisted \"%s\" or "
2070 "connect to \"unix:%s/%s.mgmt\" (which is always "
2071 "available without special configuration).",
2072 br
->name
, c
->target
, whitelist
,
2073 ovs_rundir(), br
->name
);
2081 bridge_configure_local_iface_netdev(br
, c
);
2082 bridge_ofproto_controller_from_ovsrec(c
, &ocs
[n_ocs
]);
2083 if (disable_in_band
) {
2084 ocs
[n_ocs
].band
= OFPROTO_OUT_OF_BAND
;
2089 ofproto_set_controllers(br
->ofproto
, ocs
, n_ocs
);
2090 free(ocs
[0].target
); /* From bridge_ofproto_controller_for_mgmt(). */
2093 /* Set the fail-mode. */
2094 fail_mode
= !br
->cfg
->fail_mode
2095 || !strcmp(br
->cfg
->fail_mode
, "standalone")
2096 ? OFPROTO_FAIL_STANDALONE
2097 : OFPROTO_FAIL_SECURE
;
2098 ofproto_set_fail_mode(br
->ofproto
, fail_mode
);
2100 /* Configure OpenFlow controller connection snooping. */
2101 if (!ofproto_has_snoops(br
->ofproto
)) {
2105 sset_add_and_free(&snoops
, xasprintf("punix:%s/%s.snoop",
2106 ovs_rundir(), br
->name
));
2107 ofproto_set_snoops(br
->ofproto
, &snoops
);
2108 sset_destroy(&snoops
);
2112 /* Port functions. */
2114 static struct port
*
2115 port_create(struct bridge
*br
, const struct ovsrec_port
*cfg
)
2119 port
= xzalloc(sizeof *port
);
2121 port
->name
= xstrdup(cfg
->name
);
2123 list_init(&port
->ifaces
);
2125 hmap_insert(&br
->ports
, &port
->hmap_node
, hash_string(port
->name
, 0));
2127 VLOG_INFO("created port %s on bridge %s", port
->name
, br
->name
);
2133 get_port_other_config(const struct ovsrec_port
*port
, const char *key
,
2134 const char *default_value
)
2138 value
= get_ovsrec_key_value(&port
->header_
, &ovsrec_port_col_other_config
,
2140 return value
? value
: default_value
;
2144 get_interface_other_config(const struct ovsrec_interface
*iface
,
2145 const char *key
, const char *default_value
)
2149 value
= get_ovsrec_key_value(&iface
->header_
,
2150 &ovsrec_interface_col_other_config
, key
);
2151 return value
? value
: default_value
;
2154 /* Deletes interfaces from 'port' that are no longer configured for it. */
2156 port_del_ifaces(struct port
*port
)
2158 struct iface
*iface
, *next
;
2159 struct sset new_ifaces
;
2162 /* Collect list of new interfaces. */
2163 sset_init(&new_ifaces
);
2164 for (i
= 0; i
< port
->cfg
->n_interfaces
; i
++) {
2165 const char *name
= port
->cfg
->interfaces
[i
]->name
;
2166 const char *type
= port
->cfg
->interfaces
[i
]->name
;
2167 if (strcmp(type
, "null")) {
2168 sset_add(&new_ifaces
, name
);
2172 /* Get rid of deleted interfaces. */
2173 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
2174 if (!sset_contains(&new_ifaces
, iface
->name
)) {
2175 iface_destroy(iface
);
2179 sset_destroy(&new_ifaces
);
2182 /* Adds new interfaces to 'port' and updates 'type' and 'cfg' members of
2185 port_add_ifaces(struct port
*port
)
2187 struct shash new_ifaces
;
2188 struct shash_node
*node
;
2191 /* Collect new ifaces. */
2192 shash_init(&new_ifaces
);
2193 for (i
= 0; i
< port
->cfg
->n_interfaces
; i
++) {
2194 const struct ovsrec_interface
*cfg
= port
->cfg
->interfaces
[i
];
2195 if (strcmp(cfg
->type
, "null")
2196 && !shash_add_once(&new_ifaces
, cfg
->name
, cfg
)) {
2197 VLOG_WARN("port %s: %s specified twice as port interface",
2198 port
->name
, cfg
->name
);
2199 iface_clear_db_record(cfg
);
2203 /* Create new interfaces.
2204 * Update interface types and 'cfg' members. */
2205 SHASH_FOR_EACH (node
, &new_ifaces
) {
2206 const struct ovsrec_interface
*cfg
= node
->data
;
2207 const char *iface_name
= node
->name
;
2208 struct iface
*iface
;
2210 iface
= iface_lookup(port
->bridge
, iface_name
);
2212 iface
= iface_create(port
, cfg
);
2217 /* Determine interface type. The local port always has type
2218 * "internal". Other ports take their type from the database and
2219 * default to "system" if none is specified. */
2220 iface
->type
= (!strcmp(iface_name
, port
->bridge
->name
) ? "internal"
2221 : cfg
->type
[0] ? cfg
->type
2224 shash_destroy(&new_ifaces
);
2228 port_destroy(struct port
*port
)
2231 struct bridge
*br
= port
->bridge
;
2232 struct iface
*iface
, *next
;
2235 ofproto_bundle_unregister(br
->ofproto
, port
);
2238 LIST_FOR_EACH_SAFE (iface
, next
, port_elem
, &port
->ifaces
) {
2239 iface_destroy(iface
);
2242 hmap_remove(&br
->ports
, &port
->hmap_node
);
2244 VLOG_INFO("destroyed port %s on bridge %s", port
->name
, br
->name
);
2251 static struct port
*
2252 port_lookup(const struct bridge
*br
, const char *name
)
2256 HMAP_FOR_EACH_WITH_HASH (port
, hmap_node
, hash_string(name
, 0),
2258 if (!strcmp(port
->name
, name
)) {
2266 enable_lacp(struct port
*port
, bool *activep
)
2268 if (!port
->cfg
->lacp
) {
2269 /* XXX when LACP implementation has been sufficiently tested, enable by
2270 * default and make active on bonded ports. */
2272 } else if (!strcmp(port
->cfg
->lacp
, "off")) {
2274 } else if (!strcmp(port
->cfg
->lacp
, "active")) {
2277 } else if (!strcmp(port
->cfg
->lacp
, "passive")) {
2281 VLOG_WARN("port %s: unknown LACP mode %s",
2282 port
->name
, port
->cfg
->lacp
);
2287 static struct lacp_settings
*
2288 port_configure_lacp(struct port
*port
, struct lacp_settings
*s
)
2290 const char *lacp_time
;
2291 long long int custom_time
;
2294 if (!enable_lacp(port
, &s
->active
)) {
2298 s
->name
= port
->name
;
2299 memcpy(s
->id
, port
->bridge
->ea
, ETH_ADDR_LEN
);
2301 /* Prefer bondable links if unspecified. */
2302 priority
= atoi(get_port_other_config(port
->cfg
, "lacp-system-priority",
2304 s
->priority
= (priority
> 0 && priority
<= UINT16_MAX
2306 : UINT16_MAX
- !list_is_short(&port
->ifaces
));
2308 s
->heartbeat
= !strcmp(get_port_other_config(port
->cfg
,
2313 lacp_time
= get_port_other_config(port
->cfg
, "lacp-time", "slow");
2314 custom_time
= atoi(lacp_time
);
2315 if (!strcmp(lacp_time
, "fast")) {
2316 s
->lacp_time
= LACP_TIME_FAST
;
2317 } else if (!strcmp(lacp_time
, "slow")) {
2318 s
->lacp_time
= LACP_TIME_SLOW
;
2319 } else if (custom_time
> 0) {
2320 s
->lacp_time
= LACP_TIME_CUSTOM
;
2321 s
->custom_time
= custom_time
;
2323 s
->lacp_time
= LACP_TIME_SLOW
;
2330 iface_configure_lacp(struct iface
*iface
, struct lacp_slave_settings
*s
)
2332 int priority
, portid
, key
;
2334 portid
= atoi(get_interface_other_config(iface
->cfg
, "lacp-port-id", "0"));
2335 priority
= atoi(get_interface_other_config(iface
->cfg
,
2336 "lacp-port-priority", "0"));
2337 key
= atoi(get_interface_other_config(iface
->cfg
, "lacp-aggregation-key",
2340 if (portid
<= 0 || portid
> UINT16_MAX
) {
2341 portid
= iface
->ofp_port
;
2344 if (priority
<= 0 || priority
> UINT16_MAX
) {
2345 priority
= UINT16_MAX
;
2348 if (key
< 0 || key
> UINT16_MAX
) {
2352 s
->name
= iface
->name
;
2354 s
->priority
= priority
;
2359 port_configure_bond(struct port
*port
, struct bond_settings
*s
,
2360 uint32_t *bond_stable_ids
)
2362 const char *detect_s
;
2363 struct iface
*iface
;
2364 int miimon_interval
;
2367 s
->name
= port
->name
;
2368 s
->balance
= BM_SLB
;
2369 if (port
->cfg
->bond_mode
2370 && !bond_mode_from_string(&s
->balance
, port
->cfg
->bond_mode
)) {
2371 VLOG_WARN("port %s: unknown bond_mode %s, defaulting to %s",
2372 port
->name
, port
->cfg
->bond_mode
,
2373 bond_mode_to_string(s
->balance
));
2375 if (s
->balance
== BM_SLB
&& port
->bridge
->cfg
->n_flood_vlans
) {
2376 VLOG_WARN("port %s: SLB bonds are incompatible with flood_vlans, "
2377 "please use another bond type or disable flood_vlans",
2381 miimon_interval
= atoi(get_port_other_config(port
->cfg
,
2382 "bond-miimon-interval", "0"));
2383 if (miimon_interval
<= 0) {
2384 miimon_interval
= 200;
2387 detect_s
= get_port_other_config(port
->cfg
, "bond-detect-mode", "carrier");
2388 if (!strcmp(detect_s
, "carrier")) {
2389 miimon_interval
= 0;
2390 } else if (strcmp(detect_s
, "miimon")) {
2391 VLOG_WARN("port %s: unsupported bond-detect-mode %s, "
2392 "defaulting to carrier", port
->name
, detect_s
);
2393 miimon_interval
= 0;
2396 s
->up_delay
= MAX(0, port
->cfg
->bond_updelay
);
2397 s
->down_delay
= MAX(0, port
->cfg
->bond_downdelay
);
2398 s
->basis
= atoi(get_port_other_config(port
->cfg
, "bond-hash-basis", "0"));
2399 s
->rebalance_interval
= atoi(
2400 get_port_other_config(port
->cfg
, "bond-rebalance-interval", "10000"));
2401 if (s
->rebalance_interval
< 1000) {
2402 s
->rebalance_interval
= 1000;
2405 s
->fake_iface
= port
->cfg
->bond_fake_iface
;
2408 LIST_FOR_EACH (iface
, port_elem
, &port
->ifaces
) {
2409 long long stable_id
;
2411 stable_id
= atoll(get_interface_other_config(iface
->cfg
,
2412 "bond-stable-id", "0"));
2413 if (stable_id
<= 0 || stable_id
>= UINT32_MAX
) {
2414 stable_id
= iface
->ofp_port
;
2416 bond_stable_ids
[i
++] = stable_id
;
2418 netdev_set_miimon_interval(iface
->netdev
, miimon_interval
);
2422 /* Interface functions. */
2424 static struct iface
*
2425 iface_create(struct port
*port
, const struct ovsrec_interface
*if_cfg
)
2427 struct bridge
*br
= port
->bridge
;
2428 struct iface
*iface
;
2429 char *name
= if_cfg
->name
;
2431 iface
= xzalloc(sizeof *iface
);
2433 iface
->name
= xstrdup(name
);
2434 iface
->ofp_port
= -1;
2435 iface
->tag
= tag_create_random();
2436 iface
->netdev
= NULL
;
2437 iface
->cfg
= if_cfg
;
2439 hmap_insert(&br
->iface_by_name
, &iface
->name_node
, hash_string(name
, 0));
2441 list_push_back(&port
->ifaces
, &iface
->port_elem
);
2443 VLOG_DBG("attached network device %s to port %s", iface
->name
, port
->name
);
2449 iface_destroy(struct iface
*iface
)
2452 struct port
*port
= iface
->port
;
2453 struct bridge
*br
= port
->bridge
;
2455 if (br
->ofproto
&& iface
->ofp_port
>= 0) {
2456 ofproto_port_unregister(br
->ofproto
, iface
->ofp_port
);
2459 if (iface
->ofp_port
>= 0) {
2460 hmap_remove(&br
->ifaces
, &iface
->ofp_port_node
);
2463 list_remove(&iface
->port_elem
);
2464 hmap_remove(&br
->iface_by_name
, &iface
->name_node
);
2466 netdev_close(iface
->netdev
);
2473 static struct iface
*
2474 iface_lookup(const struct bridge
*br
, const char *name
)
2476 struct iface
*iface
;
2478 HMAP_FOR_EACH_WITH_HASH (iface
, name_node
, hash_string(name
, 0),
2479 &br
->iface_by_name
) {
2480 if (!strcmp(iface
->name
, name
)) {
2488 static struct iface
*
2489 iface_find(const char *name
)
2491 const struct bridge
*br
;
2493 HMAP_FOR_EACH (br
, node
, &all_bridges
) {
2494 struct iface
*iface
= iface_lookup(br
, name
);
2503 static struct iface
*
2504 iface_from_ofp_port(const struct bridge
*br
, uint16_t ofp_port
)
2506 struct iface
*iface
;
2508 HMAP_FOR_EACH_IN_BUCKET (iface
, ofp_port_node
,
2509 hash_int(ofp_port
, 0), &br
->ifaces
) {
2510 if (iface
->ofp_port
== ofp_port
) {
2517 /* Set Ethernet address of 'iface', if one is specified in the configuration
2520 iface_set_mac(struct iface
*iface
)
2522 uint8_t ea
[ETH_ADDR_LEN
];
2524 if (!strcmp(iface
->type
, "internal")
2525 && iface
->cfg
->mac
&& eth_addr_from_string(iface
->cfg
->mac
, ea
)) {
2526 if (iface
->ofp_port
== OFPP_LOCAL
) {
2527 VLOG_ERR("interface %s: ignoring mac in Interface record "
2528 "(use Bridge record to set local port's mac)",
2530 } else if (eth_addr_is_multicast(ea
)) {
2531 VLOG_ERR("interface %s: cannot set MAC to multicast address",
2534 int error
= netdev_set_etheraddr(iface
->netdev
, ea
);
2536 VLOG_ERR("interface %s: setting MAC failed (%s)",
2537 iface
->name
, strerror(error
));
2543 /* Sets the ofport column of 'if_cfg' to 'ofport'. */
2545 iface_set_ofport(const struct ovsrec_interface
*if_cfg
, int64_t ofport
)
2547 if (if_cfg
&& !ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
2548 ovsrec_interface_set_ofport(if_cfg
, &ofport
, 1);
2552 /* Clears all of the fields in 'if_cfg' that indicate interface status, and
2553 * sets the "ofport" field to -1.
2555 * This is appropriate when 'if_cfg''s interface cannot be created or is
2556 * otherwise invalid. */
2558 iface_clear_db_record(const struct ovsrec_interface
*if_cfg
)
2560 if (!ovsdb_idl_row_is_synthetic(&if_cfg
->header_
)) {
2561 iface_set_ofport(if_cfg
, -1);
2562 ovsrec_interface_set_status(if_cfg
, NULL
, NULL
, 0);
2563 ovsrec_interface_set_admin_state(if_cfg
, NULL
);
2564 ovsrec_interface_set_duplex(if_cfg
, NULL
);
2565 ovsrec_interface_set_link_speed(if_cfg
, NULL
, 0);
2566 ovsrec_interface_set_link_state(if_cfg
, NULL
);
2567 ovsrec_interface_set_mtu(if_cfg
, NULL
, 0);
2568 ovsrec_interface_set_cfm_fault(if_cfg
, NULL
, 0);
2569 ovsrec_interface_set_cfm_remote_mpids(if_cfg
, NULL
, 0);
2570 ovsrec_interface_set_lacp_current(if_cfg
, NULL
, 0);
2571 ovsrec_interface_set_statistics(if_cfg
, NULL
, NULL
, 0);
2575 /* Adds the 'n' key-value pairs in 'keys' in 'values' to 'shash'.
2577 * The value strings in '*shash' are taken directly from values[], not copied,
2578 * so the caller should not modify or free them. */
2580 shash_from_ovs_idl_map(char **keys
, char **values
, size_t n
,
2581 struct shash
*shash
)
2586 for (i
= 0; i
< n
; i
++) {
2587 shash_add(shash
, keys
[i
], values
[i
]);
2591 /* Creates 'keys' and 'values' arrays from 'shash'.
2593 * Sets 'keys' and 'values' to heap allocated arrays representing the key-value
2594 * pairs in 'shash'. The caller takes ownership of 'keys' and 'values'. They
2595 * are populated with with strings taken directly from 'shash' and thus have
2596 * the same ownership of the key-value pairs in shash.
2599 shash_to_ovs_idl_map(struct shash
*shash
,
2600 char ***keys
, char ***values
, size_t *n
)
2604 struct shash_node
*sn
;
2606 count
= shash_count(shash
);
2608 k
= xmalloc(count
* sizeof *k
);
2609 v
= xmalloc(count
* sizeof *v
);
2612 SHASH_FOR_EACH(sn
, shash
) {
2623 struct iface_delete_queues_cbdata
{
2624 struct netdev
*netdev
;
2625 const struct ovsdb_datum
*queues
;
2629 queue_ids_include(const struct ovsdb_datum
*queues
, int64_t target
)
2631 union ovsdb_atom atom
;
2633 atom
.integer
= target
;
2634 return ovsdb_datum_find_key(queues
, &atom
, OVSDB_TYPE_INTEGER
) != UINT_MAX
;
2638 iface_delete_queues(unsigned int queue_id
,
2639 const struct shash
*details OVS_UNUSED
, void *cbdata_
)
2641 struct iface_delete_queues_cbdata
*cbdata
= cbdata_
;
2643 if (!queue_ids_include(cbdata
->queues
, queue_id
)) {
2644 netdev_delete_queue(cbdata
->netdev
, queue_id
);
2649 iface_configure_qos(struct iface
*iface
, const struct ovsrec_qos
*qos
)
2651 if (!qos
|| qos
->type
[0] == '\0' || qos
->n_queues
< 1) {
2652 netdev_set_qos(iface
->netdev
, NULL
, NULL
);
2654 struct iface_delete_queues_cbdata cbdata
;
2655 struct shash details
;
2659 /* Configure top-level Qos for 'iface'. */
2660 shash_from_ovs_idl_map(qos
->key_other_config
, qos
->value_other_config
,
2661 qos
->n_other_config
, &details
);
2662 netdev_set_qos(iface
->netdev
, qos
->type
, &details
);
2663 shash_destroy(&details
);
2665 /* Deconfigure queues that were deleted. */
2666 cbdata
.netdev
= iface
->netdev
;
2667 cbdata
.queues
= ovsrec_qos_get_queues(qos
, OVSDB_TYPE_INTEGER
,
2669 netdev_dump_queues(iface
->netdev
, iface_delete_queues
, &cbdata
);
2671 /* Configure queues for 'iface'. */
2673 for (i
= 0; i
< qos
->n_queues
; i
++) {
2674 const struct ovsrec_queue
*queue
= qos
->value_queues
[i
];
2675 unsigned int queue_id
= qos
->key_queues
[i
];
2677 if (queue_id
== 0) {
2681 shash_from_ovs_idl_map(queue
->key_other_config
,
2682 queue
->value_other_config
,
2683 queue
->n_other_config
, &details
);
2684 netdev_set_queue(iface
->netdev
, queue_id
, &details
);
2685 shash_destroy(&details
);
2688 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2689 VLOG_WARN_RL(&rl
, "interface %s: QoS configured without a default "
2690 "queue (queue 0). Packets not directed to a "
2691 "correctly configured queue may be dropped.",
2696 netdev_set_policing(iface
->netdev
,
2697 iface
->cfg
->ingress_policing_rate
,
2698 iface
->cfg
->ingress_policing_burst
);
2702 iface_configure_cfm(struct iface
*iface
)
2704 const struct ovsrec_interface
*cfg
= iface
->cfg
;
2705 const char *extended_str
, *opstate_str
;
2706 struct cfm_settings s
;
2708 if (!cfg
->n_cfm_mpid
) {
2709 ofproto_port_clear_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
);
2713 s
.mpid
= *cfg
->cfm_mpid
;
2714 s
.interval
= atoi(get_interface_other_config(iface
->cfg
, "cfm_interval",
2716 if (s
.interval
<= 0) {
2720 extended_str
= get_interface_other_config(iface
->cfg
, "cfm_extended",
2722 s
.extended
= !strcasecmp("true", extended_str
);
2724 opstate_str
= get_interface_other_config(iface
->cfg
, "cfm_opstate", "up");
2725 s
.opup
= !strcasecmp("up", opstate_str
);
2727 ofproto_port_set_cfm(iface
->port
->bridge
->ofproto
, iface
->ofp_port
, &s
);
2730 /* Returns true if 'iface' is synthetic, that is, if we constructed it locally
2731 * instead of obtaining it from the database. */
2733 iface_is_synthetic(const struct iface
*iface
)
2735 return ovsdb_idl_row_is_synthetic(&iface
->cfg
->header_
);
2738 /* Port mirroring. */
2740 static struct mirror
*
2741 mirror_find_by_uuid(struct bridge
*br
, const struct uuid
*uuid
)
2745 HMAP_FOR_EACH_IN_BUCKET (m
, hmap_node
, uuid_hash(uuid
), &br
->mirrors
) {
2746 if (uuid_equals(uuid
, &m
->uuid
)) {
2754 bridge_configure_mirrors(struct bridge
*br
)
2756 const struct ovsdb_datum
*mc
;
2757 unsigned long *flood_vlans
;
2758 struct mirror
*m
, *next
;
2761 /* Get rid of deleted mirrors. */
2762 mc
= ovsrec_bridge_get_mirrors(br
->cfg
, OVSDB_TYPE_UUID
);
2763 HMAP_FOR_EACH_SAFE (m
, next
, hmap_node
, &br
->mirrors
) {
2764 union ovsdb_atom atom
;
2766 atom
.uuid
= m
->uuid
;
2767 if (ovsdb_datum_find_key(mc
, &atom
, OVSDB_TYPE_UUID
) == UINT_MAX
) {
2772 /* Add new mirrors and reconfigure existing ones. */
2773 for (i
= 0; i
< br
->cfg
->n_mirrors
; i
++) {
2774 const struct ovsrec_mirror
*cfg
= br
->cfg
->mirrors
[i
];
2775 struct mirror
*m
= mirror_find_by_uuid(br
, &cfg
->header_
.uuid
);
2777 m
= mirror_create(br
, cfg
);
2779 if (!mirror_configure(m
, cfg
)) {
2784 /* Update flooded vlans (for RSPAN). */
2785 flood_vlans
= vlan_bitmap_from_array(br
->cfg
->flood_vlans
,
2786 br
->cfg
->n_flood_vlans
);
2787 ofproto_set_flood_vlans(br
->ofproto
, flood_vlans
);
2788 bitmap_free(flood_vlans
);
2791 static struct mirror
*
2792 mirror_create(struct bridge
*br
, const struct ovsrec_mirror
*cfg
)
2796 m
= xzalloc(sizeof *m
);
2797 m
->uuid
= cfg
->header_
.uuid
;
2798 hmap_insert(&br
->mirrors
, &m
->hmap_node
, uuid_hash(&m
->uuid
));
2800 m
->name
= xstrdup(cfg
->name
);
2806 mirror_destroy(struct mirror
*m
)
2809 struct bridge
*br
= m
->bridge
;
2812 ofproto_mirror_unregister(br
->ofproto
, m
);
2815 hmap_remove(&br
->mirrors
, &m
->hmap_node
);
2822 mirror_collect_ports(struct mirror
*m
,
2823 struct ovsrec_port
**in_ports
, int n_in_ports
,
2824 void ***out_portsp
, size_t *n_out_portsp
)
2826 void **out_ports
= xmalloc(n_in_ports
* sizeof *out_ports
);
2827 size_t n_out_ports
= 0;
2830 for (i
= 0; i
< n_in_ports
; i
++) {
2831 const char *name
= in_ports
[i
]->name
;
2832 struct port
*port
= port_lookup(m
->bridge
, name
);
2834 out_ports
[n_out_ports
++] = port
;
2836 VLOG_WARN("bridge %s: mirror %s cannot match on nonexistent "
2837 "port %s", m
->bridge
->name
, m
->name
, name
);
2840 *out_portsp
= out_ports
;
2841 *n_out_portsp
= n_out_ports
;
2845 mirror_configure(struct mirror
*m
, const struct ovsrec_mirror
*cfg
)
2847 struct ofproto_mirror_settings s
;
2850 if (strcmp(cfg
->name
, m
->name
)) {
2852 m
->name
= xstrdup(cfg
->name
);
2856 /* Get output port or VLAN. */
2857 if (cfg
->output_port
) {
2858 s
.out_bundle
= port_lookup(m
->bridge
, cfg
->output_port
->name
);
2859 if (!s
.out_bundle
) {
2860 VLOG_ERR("bridge %s: mirror %s outputs to port not on bridge",
2861 m
->bridge
->name
, m
->name
);
2864 s
.out_vlan
= UINT16_MAX
;
2866 if (cfg
->output_vlan
) {
2867 VLOG_ERR("bridge %s: mirror %s specifies both output port and "
2868 "output vlan; ignoring output vlan",
2869 m
->bridge
->name
, m
->name
);
2871 } else if (cfg
->output_vlan
) {
2872 /* The database should prevent invalid VLAN values. */
2873 s
.out_bundle
= NULL
;
2874 s
.out_vlan
= *cfg
->output_vlan
;
2876 VLOG_ERR("bridge %s: mirror %s does not specify output; ignoring",
2877 m
->bridge
->name
, m
->name
);
2881 /* Get port selection. */
2882 if (cfg
->select_all
) {
2883 size_t n_ports
= hmap_count(&m
->bridge
->ports
);
2884 void **ports
= xmalloc(n_ports
* sizeof *ports
);
2889 HMAP_FOR_EACH (port
, hmap_node
, &m
->bridge
->ports
) {
2899 /* Get ports, dropping ports that don't exist.
2900 * The IDL ensures that there are no duplicates. */
2901 mirror_collect_ports(m
, cfg
->select_src_port
, cfg
->n_select_src_port
,
2902 &s
.srcs
, &s
.n_srcs
);
2903 mirror_collect_ports(m
, cfg
->select_dst_port
, cfg
->n_select_dst_port
,
2904 &s
.dsts
, &s
.n_dsts
);
2907 /* Get VLAN selection. */
2908 s
.src_vlans
= vlan_bitmap_from_array(cfg
->select_vlan
, cfg
->n_select_vlan
);
2911 ofproto_mirror_register(m
->bridge
->ofproto
, m
, &s
);
2914 if (s
.srcs
!= s
.dsts
) {