2 * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017 Nicira, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
23 #include "byte-order.h"
24 #include "connectivity.h"
29 #include "fail-open.h"
30 #include "guarded-list.h"
34 #include "mac-learning.h"
35 #include "mcast-snooping.h"
36 #include "multipath.h"
37 #include "netdev-vport.h"
42 #include "odp-execute.h"
43 #include "ofproto/ofproto-dpif.h"
44 #include "ofproto/ofproto-provider.h"
45 #include "ofproto-dpif-ipfix.h"
46 #include "ofproto-dpif-mirror.h"
47 #include "ofproto-dpif-monitor.h"
48 #include "ofproto-dpif-rid.h"
49 #include "ofproto-dpif-sflow.h"
50 #include "ofproto-dpif-trace.h"
51 #include "ofproto-dpif-upcall.h"
52 #include "ofproto-dpif-xlate.h"
53 #include "ofproto-dpif-xlate-cache.h"
54 #include "openvswitch/ofp-actions.h"
55 #include "openvswitch/dynamic-string.h"
56 #include "openvswitch/meta-flow.h"
57 #include "openvswitch/ofp-parse.h"
58 #include "openvswitch/ofp-print.h"
59 #include "openvswitch/ofp-util.h"
60 #include "openvswitch/ofpbuf.h"
61 #include "openvswitch/vlog.h"
64 #include "ovs-router.h"
65 #include "poll-loop.h"
71 #include "unaligned.h"
74 #include "vlan-bitmap.h"
76 VLOG_DEFINE_THIS_MODULE(ofproto_dpif
);
78 COVERAGE_DEFINE(ofproto_dpif_expired
);
79 COVERAGE_DEFINE(packet_in_overflow
);
83 static void rule_get_stats(struct rule
*, uint64_t *packets
, uint64_t *bytes
,
85 static struct rule_dpif
*rule_dpif_cast(const struct rule
*);
86 static void rule_expire(struct rule_dpif
*, long long now
);
89 struct hmap_node hmap_node
; /* In struct ofproto's "bundles" hmap. */
90 struct ofproto_dpif
*ofproto
; /* Owning ofproto. */
91 void *aux
; /* Key supplied by ofproto's client. */
92 char *name
; /* Identifier for log messages. */
95 struct ovs_list ports
; /* Contains "struct ofport"s. */
96 enum port_vlan_mode vlan_mode
; /* VLAN mode */
97 uint16_t qinq_ethtype
;
98 int vlan
; /* -1=trunk port, else a 12-bit VLAN ID. */
99 unsigned long *trunks
; /* Bitmap of trunked VLANs, if 'vlan' == -1.
100 * NULL if all VLANs are trunked. */
101 unsigned long *cvlans
;
102 struct lacp
*lacp
; /* LACP if LACP is enabled, otherwise NULL. */
103 struct bond
*bond
; /* Nonnull iff more than one port. */
104 bool use_priority_tags
; /* Use 802.1p tag for frames in VLAN 0? */
106 bool protected; /* Protected port mode */
109 bool floodable
; /* True if no port has OFPUTIL_PC_NO_FLOOD set. */
112 static void bundle_remove(struct ofport
*);
113 static void bundle_update(struct ofbundle
*);
114 static void bundle_destroy(struct ofbundle
*);
115 static void bundle_del_port(struct ofport_dpif
*);
116 static void bundle_run(struct ofbundle
*);
117 static void bundle_wait(struct ofbundle
*);
118 static void bundle_flush_macs(struct ofbundle
*, bool);
119 static void bundle_move(struct ofbundle
*, struct ofbundle
*);
121 static void stp_run(struct ofproto_dpif
*ofproto
);
122 static void stp_wait(struct ofproto_dpif
*ofproto
);
123 static int set_stp_port(struct ofport
*,
124 const struct ofproto_port_stp_settings
*);
126 static void rstp_run(struct ofproto_dpif
*ofproto
);
127 static void set_rstp_port(struct ofport
*,
128 const struct ofproto_port_rstp_settings
*);
131 struct hmap_node odp_port_node
; /* In dpif_backer's "odp_to_ofport_map". */
135 struct ofbundle
*bundle
; /* Bundle that contains this port, if any. */
136 struct ovs_list bundle_node
;/* In struct ofbundle's "ports" list. */
137 struct cfm
*cfm
; /* Connectivity Fault Management, if any. */
138 struct bfd
*bfd
; /* BFD, if any. */
139 struct lldp
*lldp
; /* lldp, if any. */
140 bool may_enable
; /* May be enabled in bonds. */
141 bool is_tunnel
; /* This port is a tunnel. */
142 bool is_layer3
; /* This is a layer 3 port. */
143 long long int carrier_seq
; /* Carrier status changes. */
144 struct ofport_dpif
*peer
; /* Peer if patch port. */
147 struct stp_port
*stp_port
; /* Spanning Tree Protocol, if any. */
148 enum stp_state stp_state
; /* Always STP_DISABLED if STP not in use. */
149 long long int stp_state_entered
;
151 /* Rapid Spanning Tree. */
152 struct rstp_port
*rstp_port
; /* Rapid Spanning Tree Protocol, if any. */
153 enum rstp_state rstp_state
; /* Always RSTP_DISABLED if RSTP not in use. */
155 /* Queue to DSCP mapping. */
156 struct ofproto_port_queue
*qdscp
;
160 static odp_port_t
ofp_port_to_odp_port(const struct ofproto_dpif
*,
163 static ofp_port_t
odp_port_to_ofp_port(const struct ofproto_dpif
*,
166 static struct ofport_dpif
*
167 ofport_dpif_cast(const struct ofport
*ofport
)
169 return ofport
? CONTAINER_OF(ofport
, struct ofport_dpif
, up
) : NULL
;
172 static void port_run(struct ofport_dpif
*);
173 static int set_bfd(struct ofport
*, const struct smap
*);
174 static int set_cfm(struct ofport
*, const struct cfm_settings
*);
175 static int set_lldp(struct ofport
*ofport_
, const struct smap
*cfg
);
176 static void ofport_update_peer(struct ofport_dpif
*);
178 COVERAGE_DEFINE(rev_reconfigure
);
179 COVERAGE_DEFINE(rev_stp
);
180 COVERAGE_DEFINE(rev_rstp
);
181 COVERAGE_DEFINE(rev_bond
);
182 COVERAGE_DEFINE(rev_port_toggled
);
183 COVERAGE_DEFINE(rev_flow_table
);
184 COVERAGE_DEFINE(rev_mac_learning
);
185 COVERAGE_DEFINE(rev_mcast_snooping
);
187 /* All existing ofproto_backer instances, indexed by ofproto->up.type. */
188 struct shash all_dpif_backers
= SHASH_INITIALIZER(&all_dpif_backers
);
190 /* All existing ofproto_dpif instances, indexed by ->up.name. */
191 struct hmap all_ofproto_dpifs
= HMAP_INITIALIZER(&all_ofproto_dpifs
);
193 static bool ofproto_use_tnl_push_pop
= true;
194 static void ofproto_unixctl_init(void);
196 static inline struct ofproto_dpif
*
197 ofproto_dpif_cast(const struct ofproto
*ofproto
)
199 ovs_assert(ofproto
->ofproto_class
== &ofproto_dpif_class
);
200 return CONTAINER_OF(ofproto
, struct ofproto_dpif
, up
);
203 /* Global variables. */
204 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
206 /* Initial mappings of port to bridge mappings. */
207 static struct shash init_ofp_ports
= SHASH_INITIALIZER(&init_ofp_ports
);
209 /* Initialize 'ofm' for a learn action. If the rule already existed, reference
210 * to that rule is taken, otherwise a new rule is created. 'ofm' keeps the
211 * rule reference in both cases. */
213 ofproto_dpif_flow_mod_init_for_learn(struct ofproto_dpif
*ofproto
,
214 const struct ofputil_flow_mod
*fm
,
215 struct ofproto_flow_mod
*ofm
)
217 /* This will not take the global 'ofproto_mutex'. */
218 return ofproto_flow_mod_init_for_learn(&ofproto
->up
, fm
, ofm
);
221 /* Appends 'am' to the queue of asynchronous messages to be sent to the
222 * controller. Takes ownership of 'am' and any data it points to. */
224 ofproto_dpif_send_async_msg(struct ofproto_dpif
*ofproto
,
225 struct ofproto_async_msg
*am
)
227 if (!guarded_list_push_back(&ofproto
->ams
, &am
->list_node
, 1024)) {
228 COVERAGE_INC(packet_in_overflow
);
229 ofproto_async_msg_free(am
);
232 /* Wakes up main thread for packet-in I/O. */
233 seq_change(ofproto
->ams_seq
);
236 /* Factory functions. */
239 init(const struct shash
*iface_hints
)
241 struct shash_node
*node
;
243 /* Make a local copy, since we don't own 'iface_hints' elements. */
244 SHASH_FOR_EACH(node
, iface_hints
) {
245 const struct iface_hint
*orig_hint
= node
->data
;
246 struct iface_hint
*new_hint
= xmalloc(sizeof *new_hint
);
248 new_hint
->br_name
= xstrdup(orig_hint
->br_name
);
249 new_hint
->br_type
= xstrdup(orig_hint
->br_type
);
250 new_hint
->ofp_port
= orig_hint
->ofp_port
;
252 shash_add(&init_ofp_ports
, node
->name
, new_hint
);
255 ofproto_unixctl_init();
256 ofproto_dpif_trace_init();
261 enumerate_types(struct sset
*types
)
263 dp_enumerate_types(types
);
267 enumerate_names(const char *type
, struct sset
*names
)
269 struct ofproto_dpif
*ofproto
;
272 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
273 if (strcmp(type
, ofproto
->up
.type
)) {
276 sset_add(names
, ofproto
->up
.name
);
283 del(const char *type
, const char *name
)
288 error
= dpif_open(name
, type
, &dpif
);
290 error
= dpif_delete(dpif
);
297 port_open_type(const char *datapath_type
, const char *port_type
)
299 return dpif_port_open_type(datapath_type
, port_type
);
302 /* Type functions. */
304 static void process_dpif_port_changes(struct dpif_backer
*);
305 static void process_dpif_all_ports_changed(struct dpif_backer
*);
306 static void process_dpif_port_change(struct dpif_backer
*,
307 const char *devname
);
308 static void process_dpif_port_error(struct dpif_backer
*, int error
);
310 static struct ofproto_dpif
*
311 lookup_ofproto_dpif_by_port_name(const char *name
)
313 struct ofproto_dpif
*ofproto
;
315 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
316 if (sset_contains(&ofproto
->ports
, name
)) {
325 type_run(const char *type
)
327 struct dpif_backer
*backer
;
329 backer
= shash_find_data(&all_dpif_backers
, type
);
331 /* This is not necessarily a problem, since backers are only
332 * created on demand. */
336 if (dpif_run(backer
->dpif
)) {
337 backer
->need_revalidate
= REV_RECONFIGURE
;
340 udpif_run(backer
->udpif
);
342 /* If vswitchd started with other_config:flow_restore_wait set as "true",
343 * and the configuration has now changed to "false", enable receiving
344 * packets from the datapath. */
345 if (!backer
->recv_set_enable
&& !ofproto_get_flow_restore_wait()) {
348 backer
->recv_set_enable
= true;
350 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
352 VLOG_ERR("Failed to enable receiving packets in dpif.");
355 dpif_flow_flush(backer
->dpif
);
356 backer
->need_revalidate
= REV_RECONFIGURE
;
359 if (backer
->recv_set_enable
) {
360 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
363 if (backer
->need_revalidate
) {
364 struct ofproto_dpif
*ofproto
;
365 struct simap_node
*node
;
366 struct simap tmp_backers
;
368 /* Handle tunnel garbage collection. */
369 simap_init(&tmp_backers
);
370 simap_swap(&backer
->tnl_backers
, &tmp_backers
);
372 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
373 struct ofport_dpif
*iter
;
375 if (backer
!= ofproto
->backer
) {
379 HMAP_FOR_EACH (iter
, up
.hmap_node
, &ofproto
->up
.ports
) {
380 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
383 if (!iter
->is_tunnel
) {
387 dp_port
= netdev_vport_get_dpif_port(iter
->up
.netdev
,
388 namebuf
, sizeof namebuf
);
389 node
= simap_find(&tmp_backers
, dp_port
);
391 simap_put(&backer
->tnl_backers
, dp_port
, node
->data
);
392 simap_delete(&tmp_backers
, node
);
393 node
= simap_find(&backer
->tnl_backers
, dp_port
);
395 node
= simap_find(&backer
->tnl_backers
, dp_port
);
397 odp_port_t odp_port
= ODPP_NONE
;
399 if (!dpif_port_add(backer
->dpif
, iter
->up
.netdev
,
401 simap_put(&backer
->tnl_backers
, dp_port
,
402 odp_to_u32(odp_port
));
403 node
= simap_find(&backer
->tnl_backers
, dp_port
);
408 iter
->odp_port
= node
? u32_to_odp(node
->data
) : ODPP_NONE
;
409 if (tnl_port_reconfigure(iter
, iter
->up
.netdev
,
411 ovs_native_tunneling_is_on(ofproto
), dp_port
)) {
412 backer
->need_revalidate
= REV_RECONFIGURE
;
417 SIMAP_FOR_EACH (node
, &tmp_backers
) {
418 dpif_port_del(backer
->dpif
, u32_to_odp(node
->data
));
420 simap_destroy(&tmp_backers
);
422 switch (backer
->need_revalidate
) {
423 case REV_RECONFIGURE
: COVERAGE_INC(rev_reconfigure
); break;
424 case REV_STP
: COVERAGE_INC(rev_stp
); break;
425 case REV_RSTP
: COVERAGE_INC(rev_rstp
); break;
426 case REV_BOND
: COVERAGE_INC(rev_bond
); break;
427 case REV_PORT_TOGGLED
: COVERAGE_INC(rev_port_toggled
); break;
428 case REV_FLOW_TABLE
: COVERAGE_INC(rev_flow_table
); break;
429 case REV_MAC_LEARNING
: COVERAGE_INC(rev_mac_learning
); break;
430 case REV_MCAST_SNOOPING
: COVERAGE_INC(rev_mcast_snooping
); break;
432 backer
->need_revalidate
= 0;
435 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
436 struct ofport_dpif
*ofport
;
437 struct ofbundle
*bundle
;
439 if (ofproto
->backer
!= backer
) {
443 xlate_ofproto_set(ofproto
, ofproto
->up
.name
,
444 ofproto
->backer
->dpif
, ofproto
->ml
,
445 ofproto
->stp
, ofproto
->rstp
, ofproto
->ms
,
446 ofproto
->mbridge
, ofproto
->sflow
, ofproto
->ipfix
,
448 ofproto
->up
.forward_bpdu
,
449 connmgr_has_in_band(ofproto
->up
.connmgr
),
450 &ofproto
->backer
->support
);
452 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
453 xlate_bundle_set(ofproto
, bundle
, bundle
->name
,
454 bundle
->vlan_mode
, bundle
->qinq_ethtype
,
455 bundle
->vlan
, bundle
->trunks
, bundle
->cvlans
,
456 bundle
->use_priority_tags
,
457 bundle
->bond
, bundle
->lacp
,
458 bundle
->floodable
, bundle
->protected);
461 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
462 int stp_port
= ofport
->stp_port
463 ? stp_port_no(ofport
->stp_port
)
465 xlate_ofport_set(ofproto
, ofport
->bundle
, ofport
,
466 ofport
->up
.ofp_port
, ofport
->odp_port
,
467 ofport
->up
.netdev
, ofport
->cfm
, ofport
->bfd
,
468 ofport
->lldp
, ofport
->peer
, stp_port
,
469 ofport
->rstp_port
, ofport
->qdscp
,
470 ofport
->n_qdscp
, ofport
->up
.pp
.config
,
471 ofport
->up
.pp
.state
, ofport
->is_tunnel
,
477 udpif_revalidate(backer
->udpif
);
480 process_dpif_port_changes(backer
);
485 /* Check for and handle port changes in 'backer''s dpif. */
487 process_dpif_port_changes(struct dpif_backer
*backer
)
493 error
= dpif_port_poll(backer
->dpif
, &devname
);
499 process_dpif_all_ports_changed(backer
);
503 process_dpif_port_change(backer
, devname
);
508 process_dpif_port_error(backer
, error
);
515 process_dpif_all_ports_changed(struct dpif_backer
*backer
)
517 struct ofproto_dpif
*ofproto
;
518 struct dpif_port dpif_port
;
519 struct dpif_port_dump dump
;
520 struct sset devnames
;
523 sset_init(&devnames
);
524 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
525 if (ofproto
->backer
== backer
) {
526 struct ofport
*ofport
;
528 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
529 sset_add(&devnames
, netdev_get_name(ofport
->netdev
));
533 DPIF_PORT_FOR_EACH (&dpif_port
, &dump
, backer
->dpif
) {
534 sset_add(&devnames
, dpif_port
.name
);
537 SSET_FOR_EACH (devname
, &devnames
) {
538 process_dpif_port_change(backer
, devname
);
540 sset_destroy(&devnames
);
544 process_dpif_port_change(struct dpif_backer
*backer
, const char *devname
)
546 struct ofproto_dpif
*ofproto
;
547 struct dpif_port port
;
549 /* Don't report on the datapath's device. */
550 if (!strcmp(devname
, dpif_base_name(backer
->dpif
))) {
554 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
,
555 &all_ofproto_dpifs
) {
556 if (simap_contains(&ofproto
->backer
->tnl_backers
, devname
)) {
561 ofproto
= lookup_ofproto_dpif_by_port_name(devname
);
562 if (dpif_port_query_by_name(backer
->dpif
, devname
, &port
)) {
563 /* The port was removed. If we know the datapath,
564 * report it through poll_set(). If we don't, it may be
565 * notifying us of a removal we initiated, so ignore it.
566 * If there's a pending ENOBUFS, let it stand, since
567 * everything will be reevaluated. */
568 if (ofproto
&& ofproto
->port_poll_errno
!= ENOBUFS
) {
569 sset_add(&ofproto
->port_poll_set
, devname
);
570 ofproto
->port_poll_errno
= 0;
572 } else if (!ofproto
) {
573 /* The port was added, but we don't know with which
574 * ofproto we should associate it. Delete it. */
575 dpif_port_del(backer
->dpif
, port
.port_no
);
577 struct ofport_dpif
*ofport
;
579 ofport
= ofport_dpif_cast(shash_find_data(
580 &ofproto
->up
.port_by_name
, devname
));
582 && ofport
->odp_port
!= port
.port_no
583 && !odp_port_to_ofport(backer
, port
.port_no
))
585 /* 'ofport''s datapath port number has changed from
586 * 'ofport->odp_port' to 'port.port_no'. Update our internal data
587 * structures to match. */
588 ovs_rwlock_wrlock(&backer
->odp_to_ofport_lock
);
589 hmap_remove(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
);
590 ofport
->odp_port
= port
.port_no
;
591 hmap_insert(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
,
592 hash_odp_port(port
.port_no
));
593 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
594 backer
->need_revalidate
= REV_RECONFIGURE
;
597 dpif_port_destroy(&port
);
600 /* Propagate 'error' to all ofprotos based on 'backer'. */
602 process_dpif_port_error(struct dpif_backer
*backer
, int error
)
604 struct ofproto_dpif
*ofproto
;
606 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
607 if (ofproto
->backer
== backer
) {
608 sset_clear(&ofproto
->port_poll_set
);
609 ofproto
->port_poll_errno
= error
;
615 type_wait(const char *type
)
617 struct dpif_backer
*backer
;
619 backer
= shash_find_data(&all_dpif_backers
, type
);
621 /* This is not necessarily a problem, since backers are only
622 * created on demand. */
626 dpif_wait(backer
->dpif
);
629 /* Basic life-cycle. */
631 static int add_internal_flows(struct ofproto_dpif
*);
633 static struct ofproto
*
636 struct ofproto_dpif
*ofproto
= xzalloc(sizeof *ofproto
);
641 dealloc(struct ofproto
*ofproto_
)
643 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
648 close_dpif_backer(struct dpif_backer
*backer
)
650 ovs_assert(backer
->refcount
> 0);
652 if (--backer
->refcount
) {
656 udpif_destroy(backer
->udpif
);
658 simap_destroy(&backer
->tnl_backers
);
659 ovs_rwlock_destroy(&backer
->odp_to_ofport_lock
);
660 hmap_destroy(&backer
->odp_to_ofport_map
);
661 shash_find_and_delete(&all_dpif_backers
, backer
->type
);
663 free(backer
->dp_version_string
);
664 dpif_close(backer
->dpif
);
668 /* Datapath port slated for removal from datapath. */
670 struct ovs_list list_node
;
674 static bool check_variable_length_userdata(struct dpif_backer
*backer
);
675 static void check_support(struct dpif_backer
*backer
);
678 open_dpif_backer(const char *type
, struct dpif_backer
**backerp
)
680 struct dpif_backer
*backer
;
681 struct dpif_port_dump port_dump
;
682 struct dpif_port port
;
683 struct shash_node
*node
;
684 struct ovs_list garbage_list
;
685 struct odp_garbage
*garbage
;
692 backer
= shash_find_data(&all_dpif_backers
, type
);
699 backer_name
= xasprintf("ovs-%s", type
);
701 /* Remove any existing datapaths, since we assume we're the only
702 * userspace controlling the datapath. */
704 dp_enumerate_names(type
, &names
);
705 SSET_FOR_EACH(name
, &names
) {
706 struct dpif
*old_dpif
;
708 /* Don't remove our backer if it exists. */
709 if (!strcmp(name
, backer_name
)) {
713 if (dpif_open(name
, type
, &old_dpif
)) {
714 VLOG_WARN("couldn't open old datapath %s to remove it", name
);
716 dpif_delete(old_dpif
);
717 dpif_close(old_dpif
);
720 sset_destroy(&names
);
722 backer
= xmalloc(sizeof *backer
);
724 error
= dpif_create_and_open(backer_name
, type
, &backer
->dpif
);
727 VLOG_ERR("failed to open datapath of type %s: %s", type
,
728 ovs_strerror(error
));
732 backer
->udpif
= udpif_create(backer
, backer
->dpif
);
734 backer
->type
= xstrdup(type
);
735 backer
->refcount
= 1;
736 hmap_init(&backer
->odp_to_ofport_map
);
737 ovs_rwlock_init(&backer
->odp_to_ofport_lock
);
738 backer
->need_revalidate
= 0;
739 simap_init(&backer
->tnl_backers
);
740 backer
->recv_set_enable
= !ofproto_get_flow_restore_wait();
743 if (backer
->recv_set_enable
) {
744 dpif_flow_flush(backer
->dpif
);
747 /* Loop through the ports already on the datapath and remove any
748 * that we don't need anymore. */
749 ovs_list_init(&garbage_list
);
750 dpif_port_dump_start(&port_dump
, backer
->dpif
);
751 while (dpif_port_dump_next(&port_dump
, &port
)) {
752 node
= shash_find(&init_ofp_ports
, port
.name
);
753 if (!node
&& strcmp(port
.name
, dpif_base_name(backer
->dpif
))) {
754 garbage
= xmalloc(sizeof *garbage
);
755 garbage
->odp_port
= port
.port_no
;
756 ovs_list_push_front(&garbage_list
, &garbage
->list_node
);
759 dpif_port_dump_done(&port_dump
);
761 LIST_FOR_EACH_POP (garbage
, list_node
, &garbage_list
) {
762 dpif_port_del(backer
->dpif
, garbage
->odp_port
);
766 shash_add(&all_dpif_backers
, type
, backer
);
768 check_support(backer
);
769 atomic_count_init(&backer
->tnl_count
, 0);
771 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
773 VLOG_ERR("failed to listen on datapath of type %s: %s",
774 type
, ovs_strerror(error
));
775 close_dpif_backer(backer
);
779 if (backer
->recv_set_enable
) {
780 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
783 /* This check fails if performed before udpif threads have been set,
784 * as the kernel module checks that the 'pid' in userspace action
786 backer
->support
.variable_length_userdata
787 = check_variable_length_userdata(backer
);
788 backer
->dp_version_string
= dpif_get_dp_version(backer
->dpif
);
794 ovs_native_tunneling_is_on(struct ofproto_dpif
*ofproto
)
796 return ofproto_use_tnl_push_pop
&& ofproto
->backer
->support
.tnl_push_pop
&&
797 atomic_count_get(&ofproto
->backer
->tnl_count
);
800 /* Tests whether 'backer''s datapath supports recirculation. Only newer
801 * datapaths support OVS_KEY_ATTR_RECIRC_ID in keys. We need to disable some
802 * features on older datapaths that don't support this feature.
804 * Returns false if 'backer' definitely does not support recirculation, true if
805 * it seems to support recirculation or if at least the error we get is
808 check_recirc(struct dpif_backer
*backer
)
811 struct odputil_keybuf keybuf
;
814 struct odp_flow_key_parms odp_parms
= {
821 memset(&flow
, 0, sizeof flow
);
825 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
826 odp_flow_key_from_flow(&odp_parms
, &key
);
827 enable_recirc
= dpif_probe_feature(backer
->dpif
, "recirculation", &key
,
831 VLOG_INFO("%s: Datapath supports recirculation",
832 dpif_name(backer
->dpif
));
834 VLOG_INFO("%s: Datapath does not support recirculation",
835 dpif_name(backer
->dpif
));
838 return enable_recirc
;
841 /* Tests whether 'dpif' supports unique flow ids. We can skip serializing
842 * some flow attributes for datapaths that support this feature.
844 * Returns true if 'dpif' supports UFID for flow operations.
845 * Returns false if 'dpif' does not support UFID. */
847 check_ufid(struct dpif_backer
*backer
)
850 struct odputil_keybuf keybuf
;
854 struct odp_flow_key_parms odp_parms
= {
858 memset(&flow
, 0, sizeof flow
);
859 flow
.dl_type
= htons(0x1234);
861 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
862 odp_flow_key_from_flow(&odp_parms
, &key
);
863 dpif_flow_hash(backer
->dpif
, key
.data
, key
.size
, &ufid
);
865 enable_ufid
= dpif_probe_feature(backer
->dpif
, "UFID", &key
, NULL
, &ufid
);
868 VLOG_INFO("%s: Datapath supports unique flow ids",
869 dpif_name(backer
->dpif
));
871 VLOG_INFO("%s: Datapath does not support unique flow ids",
872 dpif_name(backer
->dpif
));
877 /* Tests whether 'backer''s datapath supports variable-length
878 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. We need
879 * to disable some features on older datapaths that don't support this
882 * Returns false if 'backer' definitely does not support variable-length
883 * userdata, true if it seems to support them or if at least the error we get
886 check_variable_length_userdata(struct dpif_backer
*backer
)
888 struct eth_header
*eth
;
889 struct ofpbuf actions
;
890 struct dpif_execute execute
;
891 struct dp_packet packet
;
896 /* Compose a userspace action that will cause an ERANGE error on older
897 * datapaths that don't support variable-length userdata.
899 * We really test for using userdata longer than 8 bytes, but older
900 * datapaths accepted these, silently truncating the userdata to 8 bytes.
901 * The same older datapaths rejected userdata shorter than 8 bytes, so we
902 * test for that instead as a proxy for longer userdata support. */
903 ofpbuf_init(&actions
, 64);
904 start
= nl_msg_start_nested(&actions
, OVS_ACTION_ATTR_USERSPACE
);
905 nl_msg_put_u32(&actions
, OVS_USERSPACE_ATTR_PID
,
906 dpif_port_get_pid(backer
->dpif
, ODPP_NONE
, 0));
907 nl_msg_put_unspec_zero(&actions
, OVS_USERSPACE_ATTR_USERDATA
, 4);
908 nl_msg_end_nested(&actions
, start
);
910 /* Compose a dummy ethernet packet. */
911 dp_packet_init(&packet
, ETH_HEADER_LEN
);
912 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
913 eth
->eth_type
= htons(0x1234);
915 flow_extract(&packet
, &flow
);
917 /* Execute the actions. On older datapaths this fails with ERANGE, on
918 * newer datapaths it succeeds. */
919 execute
.actions
= actions
.data
;
920 execute
.actions_len
= actions
.size
;
921 execute
.packet
= &packet
;
922 execute
.flow
= &flow
;
923 execute
.needs_help
= false;
924 execute
.probe
= true;
927 error
= dpif_execute(backer
->dpif
, &execute
);
929 dp_packet_uninit(&packet
);
930 ofpbuf_uninit(&actions
);
937 /* Variable-length userdata is not supported. */
938 VLOG_WARN("%s: datapath does not support variable-length userdata "
939 "feature (needs Linux 3.10+ or kernel module from OVS "
940 "1..11+). The NXAST_SAMPLE action will be ignored.",
941 dpif_name(backer
->dpif
));
945 /* Something odd happened. We're not sure whether variable-length
946 * userdata is supported. Default to "yes". */
947 VLOG_WARN("%s: variable-length userdata feature probe failed (%s)",
948 dpif_name(backer
->dpif
), ovs_strerror(error
));
953 /* Tests number of 802.1q VLAN headers supported by 'backer''s datapath.
955 * Returns the number of elements in a struct flow's vlan
956 * if the datapath supports at least that many VLAN headers. */
958 check_max_vlan_headers(struct dpif_backer
*backer
)
961 struct odp_flow_key_parms odp_parms
= {
967 memset(&flow
, 0, sizeof flow
);
968 flow
.dl_type
= htons(ETH_TYPE_IP
);
969 for (n
= 0; n
< FLOW_MAX_VLAN_HEADERS
; n
++) {
970 struct odputil_keybuf keybuf
;
973 flow_push_vlan_uninit(&flow
, NULL
);
974 flow
.vlans
[0].tpid
= htons(ETH_TYPE_VLAN
);
975 flow
.vlans
[0].tci
= htons(1) | htons(VLAN_CFI
);
977 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
978 odp_flow_key_from_flow(&odp_parms
, &key
);
979 if (!dpif_probe_feature(backer
->dpif
, "VLAN", &key
, NULL
, NULL
)) {
984 VLOG_INFO("%s: VLAN header stack length probed as %d",
985 dpif_name(backer
->dpif
), n
);
988 /* Tests the MPLS label stack depth supported by 'backer''s datapath.
990 * Returns the number of elements in a struct flow's mpls_lse field
991 * if the datapath supports at least that many entries in an
993 * Otherwise returns the number of MPLS push actions supported by
996 check_max_mpls_depth(struct dpif_backer
*backer
)
1001 for (n
= 0; n
< FLOW_MAX_MPLS_LABELS
; n
++) {
1002 struct odputil_keybuf keybuf
;
1004 struct odp_flow_key_parms odp_parms
= {
1008 memset(&flow
, 0, sizeof flow
);
1009 flow
.dl_type
= htons(ETH_TYPE_MPLS
);
1010 flow_set_mpls_bos(&flow
, n
, 1);
1012 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1013 odp_flow_key_from_flow(&odp_parms
, &key
);
1014 if (!dpif_probe_feature(backer
->dpif
, "MPLS", &key
, NULL
, NULL
)) {
1019 VLOG_INFO("%s: MPLS label stack length probed as %d",
1020 dpif_name(backer
->dpif
), n
);
1025 add_sample_actions(struct ofpbuf
*actions
, int nesting
)
1028 nl_msg_put_odp_port(actions
, OVS_ACTION_ATTR_OUTPUT
, u32_to_odp(1));
1032 size_t start
, actions_start
;
1034 start
= nl_msg_start_nested(actions
, OVS_ACTION_ATTR_SAMPLE
);
1035 actions_start
= nl_msg_start_nested(actions
, OVS_SAMPLE_ATTR_ACTIONS
);
1036 add_sample_actions(actions
, nesting
- 1);
1037 nl_msg_end_nested(actions
, actions_start
);
1038 nl_msg_put_u32(actions
, OVS_SAMPLE_ATTR_PROBABILITY
, UINT32_MAX
);
1039 nl_msg_end_nested(actions
, start
);
1042 /* Tests the nested sample actions levels supported by 'backer''s datapath.
1044 * Returns the number of nested sample actions accepted by the datapath. */
1046 check_max_sample_nesting(struct dpif_backer
*backer
)
1048 struct odputil_keybuf keybuf
;
1053 struct odp_flow_key_parms odp_parms
= {
1057 memset(&flow
, 0, sizeof flow
);
1058 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1059 odp_flow_key_from_flow(&odp_parms
, &key
);
1061 /* OVS datapath has always supported at least 3 nested levels. */
1062 for (n
= 3; n
< FLOW_MAX_SAMPLE_NESTING
; n
++) {
1063 struct ofpbuf actions
;
1066 ofpbuf_init(&actions
, 300);
1067 add_sample_actions(&actions
, n
);
1068 ok
= dpif_probe_feature(backer
->dpif
, "Sample action nesting", &key
,
1070 ofpbuf_uninit(&actions
);
1076 VLOG_INFO("%s: Max sample nesting level probed as %d",
1077 dpif_name(backer
->dpif
), n
);
1081 /* Tests whether 'backer''s datapath supports masked data in
1082 * OVS_ACTION_ATTR_SET actions. We need to disable some features on older
1083 * datapaths that don't support this feature. */
1085 check_masked_set_action(struct dpif_backer
*backer
)
1087 struct eth_header
*eth
;
1088 struct ofpbuf actions
;
1089 struct dpif_execute execute
;
1090 struct dp_packet packet
;
1093 struct ovs_key_ethernet key
, mask
;
1095 /* Compose a set action that will cause an EINVAL error on older
1096 * datapaths that don't support masked set actions.
1097 * Avoid using a full mask, as it could be translated to a non-masked
1098 * set action instead. */
1099 ofpbuf_init(&actions
, 64);
1100 memset(&key
, 0x53, sizeof key
);
1101 memset(&mask
, 0x7f, sizeof mask
);
1102 commit_masked_set_action(&actions
, OVS_KEY_ATTR_ETHERNET
, &key
, &mask
,
1105 /* Compose a dummy ethernet packet. */
1106 dp_packet_init(&packet
, ETH_HEADER_LEN
);
1107 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
1108 eth
->eth_type
= htons(0x1234);
1110 flow_extract(&packet
, &flow
);
1112 /* Execute the actions. On older datapaths this fails with EINVAL, on
1113 * newer datapaths it succeeds. */
1114 execute
.actions
= actions
.data
;
1115 execute
.actions_len
= actions
.size
;
1116 execute
.packet
= &packet
;
1117 execute
.flow
= &flow
;
1118 execute
.needs_help
= false;
1119 execute
.probe
= true;
1122 error
= dpif_execute(backer
->dpif
, &execute
);
1124 dp_packet_uninit(&packet
);
1125 ofpbuf_uninit(&actions
);
1128 /* Masked set action is not supported. */
1129 VLOG_INFO("%s: datapath does not support masked set action feature.",
1130 dpif_name(backer
->dpif
));
1135 /* Tests whether 'backer''s datapath supports truncation of a packet in
1136 * OVS_ACTION_ATTR_TRUNC. We need to disable some features on older
1137 * datapaths that don't support this feature. */
1139 check_trunc_action(struct dpif_backer
*backer
)
1141 struct eth_header
*eth
;
1142 struct ofpbuf actions
;
1143 struct dpif_execute execute
;
1144 struct dp_packet packet
;
1145 struct ovs_action_trunc
*trunc
;
1149 /* Compose an action with output(port:1,
1150 * max_len:OVS_ACTION_OUTPUT_MIN + 1).
1151 * This translates to one truncate action and one output action. */
1152 ofpbuf_init(&actions
, 64);
1153 trunc
= nl_msg_put_unspec_uninit(&actions
,
1154 OVS_ACTION_ATTR_TRUNC
, sizeof *trunc
);
1156 trunc
->max_len
= ETH_HEADER_LEN
+ 1;
1157 nl_msg_put_odp_port(&actions
, OVS_ACTION_ATTR_OUTPUT
, u32_to_odp(1));
1159 /* Compose a dummy Ethernet packet. */
1160 dp_packet_init(&packet
, ETH_HEADER_LEN
);
1161 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
1162 eth
->eth_type
= htons(0x1234);
1164 flow_extract(&packet
, &flow
);
1166 /* Execute the actions. On older datapaths this fails with EINVAL, on
1167 * newer datapaths it succeeds. */
1168 execute
.actions
= actions
.data
;
1169 execute
.actions_len
= actions
.size
;
1170 execute
.packet
= &packet
;
1171 execute
.flow
= &flow
;
1172 execute
.needs_help
= false;
1173 execute
.probe
= true;
1176 error
= dpif_execute(backer
->dpif
, &execute
);
1178 dp_packet_uninit(&packet
);
1179 ofpbuf_uninit(&actions
);
1182 VLOG_INFO("%s: Datapath does not support truncate action",
1183 dpif_name(backer
->dpif
));
1185 VLOG_INFO("%s: Datapath supports truncate action",
1186 dpif_name(backer
->dpif
));
1192 /* Tests whether 'backer''s datapath supports the clone action
1193 * OVS_ACTION_ATTR_CLONE. */
1195 check_clone(struct dpif_backer
*backer
)
1197 struct dpif_execute execute
;
1198 struct eth_header
*eth
;
1200 struct dp_packet packet
;
1201 struct ofpbuf actions
;
1205 /* Compose clone with an empty action list.
1206 * and check if datapath can decode the message. */
1207 ofpbuf_init(&actions
, 64);
1208 clone_start
= nl_msg_start_nested(&actions
, OVS_ACTION_ATTR_CLONE
);
1209 nl_msg_end_nested(&actions
, clone_start
);
1211 /* Compose a dummy Ethernet packet. */
1212 dp_packet_init(&packet
, ETH_HEADER_LEN
);
1213 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
1214 eth
->eth_type
= htons(0x1234);
1216 flow_extract(&packet
, &flow
);
1218 /* Execute the actions. On older datapaths this fails with EINVAL, on
1219 * newer datapaths it succeeds. */
1220 execute
.actions
= actions
.data
;
1221 execute
.actions_len
= actions
.size
;
1222 execute
.packet
= &packet
;
1223 execute
.flow
= &flow
;
1224 execute
.needs_help
= false;
1225 execute
.probe
= true;
1228 error
= dpif_execute(backer
->dpif
, &execute
);
1230 dp_packet_uninit(&packet
);
1231 ofpbuf_uninit(&actions
);
1234 VLOG_INFO("%s: Datapath does not support clone action",
1235 dpif_name(backer
->dpif
));
1237 VLOG_INFO("%s: Datapath supports clone action",
1238 dpif_name(backer
->dpif
));
1244 #define CHECK_FEATURE__(NAME, SUPPORT, FIELD, VALUE) \
1246 check_##NAME(struct dpif_backer *backer) \
1249 struct odputil_keybuf keybuf; \
1250 struct ofpbuf key; \
1252 struct odp_flow_key_parms odp_parms = { \
1259 memset(&flow, 0, sizeof flow); \
1260 flow.FIELD = VALUE; \
1262 ofpbuf_use_stack(&key, &keybuf, sizeof keybuf); \
1263 odp_flow_key_from_flow(&odp_parms, &key); \
1264 enable = dpif_probe_feature(backer->dpif, #NAME, &key, NULL, NULL); \
1267 VLOG_INFO("%s: Datapath supports "#NAME, dpif_name(backer->dpif)); \
1269 VLOG_INFO("%s: Datapath does not support "#NAME, \
1270 dpif_name(backer->dpif)); \
1275 #define CHECK_FEATURE(FIELD) CHECK_FEATURE__(FIELD, FIELD, FIELD, 1)
1277 CHECK_FEATURE(ct_state
)
1278 CHECK_FEATURE(ct_zone
)
1279 CHECK_FEATURE(ct_mark
)
1280 CHECK_FEATURE__(ct_label
, ct_label
, ct_label
.u64
.lo
, 1)
1281 CHECK_FEATURE__(ct_state_nat
, ct_state
, ct_state
, CS_TRACKED
|CS_SRC_NAT
)
1282 CHECK_FEATURE__(ct_orig_tuple
, ct_orig_tuple
, ct_nw_proto
, 1)
1284 #undef CHECK_FEATURE
1285 #undef CHECK_FEATURE__
1288 check_support(struct dpif_backer
*backer
)
1290 /* This feature needs to be tested after udpif threads are set. */
1291 backer
->support
.variable_length_userdata
= false;
1294 backer
->support
.odp
.recirc
= check_recirc(backer
);
1295 backer
->support
.odp
.max_vlan_headers
= check_max_vlan_headers(backer
);
1296 backer
->support
.odp
.max_mpls_depth
= check_max_mpls_depth(backer
);
1297 backer
->support
.masked_set_action
= check_masked_set_action(backer
);
1298 backer
->support
.trunc
= check_trunc_action(backer
);
1299 backer
->support
.ufid
= check_ufid(backer
);
1300 backer
->support
.tnl_push_pop
= dpif_supports_tnl_push_pop(backer
->dpif
);
1301 backer
->support
.clone
= check_clone(backer
);
1302 backer
->support
.sample_nesting
= check_max_sample_nesting(backer
);
1305 backer
->support
.odp
.ct_state
= check_ct_state(backer
);
1306 backer
->support
.odp
.ct_zone
= check_ct_zone(backer
);
1307 backer
->support
.odp
.ct_mark
= check_ct_mark(backer
);
1308 backer
->support
.odp
.ct_label
= check_ct_label(backer
);
1310 backer
->support
.odp
.ct_state_nat
= check_ct_state_nat(backer
);
1311 backer
->support
.odp
.ct_orig_tuple
= check_ct_orig_tuple(backer
);
1315 construct(struct ofproto
*ofproto_
)
1317 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1318 struct shash_node
*node
, *next
;
1321 /* Tunnel module can get used right after the udpif threads are running. */
1322 ofproto_tunnel_init();
1324 error
= open_dpif_backer(ofproto
->up
.type
, &ofproto
->backer
);
1329 uuid_generate(&ofproto
->uuid
);
1330 atomic_init(&ofproto
->tables_version
, OVS_VERSION_MIN
);
1331 ofproto
->netflow
= NULL
;
1332 ofproto
->sflow
= NULL
;
1333 ofproto
->ipfix
= NULL
;
1334 ofproto
->stp
= NULL
;
1335 ofproto
->rstp
= NULL
;
1336 ofproto
->dump_seq
= 0;
1337 hmap_init(&ofproto
->bundles
);
1338 ofproto
->ml
= mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME
);
1340 ofproto
->mbridge
= mbridge_create();
1341 ofproto
->has_bonded_bundles
= false;
1342 ofproto
->lacp_enabled
= false;
1343 ovs_mutex_init_adaptive(&ofproto
->stats_mutex
);
1345 guarded_list_init(&ofproto
->ams
);
1347 sset_init(&ofproto
->ports
);
1348 sset_init(&ofproto
->ghost_ports
);
1349 sset_init(&ofproto
->port_poll_set
);
1350 ofproto
->port_poll_errno
= 0;
1351 ofproto
->change_seq
= 0;
1352 ofproto
->ams_seq
= seq_create();
1353 ofproto
->ams_seqno
= seq_read(ofproto
->ams_seq
);
1356 SHASH_FOR_EACH_SAFE (node
, next
, &init_ofp_ports
) {
1357 struct iface_hint
*iface_hint
= node
->data
;
1359 if (!strcmp(iface_hint
->br_name
, ofproto
->up
.name
)) {
1360 /* Check if the datapath already has this port. */
1361 if (dpif_port_exists(ofproto
->backer
->dpif
, node
->name
)) {
1362 sset_add(&ofproto
->ports
, node
->name
);
1365 free(iface_hint
->br_name
);
1366 free(iface_hint
->br_type
);
1368 shash_delete(&init_ofp_ports
, node
);
1372 hmap_insert(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
,
1373 hash_string(ofproto
->up
.name
, 0));
1374 memset(&ofproto
->stats
, 0, sizeof ofproto
->stats
);
1376 ofproto_init_tables(ofproto_
, N_TABLES
);
1377 error
= add_internal_flows(ofproto
);
1379 ofproto
->up
.tables
[TBL_INTERNAL
].flags
= OFTABLE_HIDDEN
| OFTABLE_READONLY
;
1385 add_internal_miss_flow(struct ofproto_dpif
*ofproto
, int id
,
1386 const struct ofpbuf
*ofpacts
, struct rule_dpif
**rulep
)
1392 match_init_catchall(&match
);
1393 match_set_reg(&match
, 0, id
);
1395 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 0, 0, ofpacts
,
1397 *rulep
= error
? NULL
: rule_dpif_cast(rule
);
1403 add_internal_flows(struct ofproto_dpif
*ofproto
)
1405 struct ofpact_controller
*controller
;
1406 uint64_t ofpacts_stub
[128 / 8];
1407 struct ofpbuf ofpacts
;
1408 struct rule
*unused_rulep OVS_UNUSED
;
1413 ofpbuf_use_stack(&ofpacts
, ofpacts_stub
, sizeof ofpacts_stub
);
1416 controller
= ofpact_put_CONTROLLER(&ofpacts
);
1417 controller
->max_len
= UINT16_MAX
;
1418 controller
->controller_id
= 0;
1419 controller
->reason
= OFPR_IMPLICIT_MISS
;
1420 ofpact_finish_CONTROLLER(&ofpacts
, &controller
);
1422 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1423 &ofproto
->miss_rule
);
1428 ofpbuf_clear(&ofpacts
);
1429 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1430 &ofproto
->no_packet_in_rule
);
1435 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1436 &ofproto
->drop_frags_rule
);
1441 /* Drop any run away non-recirc rule lookups. Recirc_id has to be
1442 * zero when reaching this rule.
1444 * (priority=2), recirc_id=0, actions=drop
1446 ofpbuf_clear(&ofpacts
);
1447 match_init_catchall(&match
);
1448 match_set_recirc_id(&match
, 0);
1449 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 2, 0, &ofpacts
,
1455 destruct(struct ofproto
*ofproto_
)
1457 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1458 struct ofproto_async_msg
*am
;
1459 struct rule_dpif
*rule
;
1460 struct oftable
*table
;
1461 struct ovs_list ams
;
1463 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1465 xlate_remove_ofproto(ofproto
);
1468 /* Ensure that the upcall processing threads have no remaining references
1469 * to the ofproto or anything in it. */
1470 udpif_synchronize(ofproto
->backer
->udpif
);
1472 hmap_remove(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
);
1474 OFPROTO_FOR_EACH_TABLE (table
, &ofproto
->up
) {
1475 CLS_FOR_EACH (rule
, up
.cr
, &table
->cls
) {
1476 ofproto_rule_delete(&ofproto
->up
, &rule
->up
);
1479 ofproto_group_delete_all(&ofproto
->up
);
1481 guarded_list_pop_all(&ofproto
->ams
, &ams
);
1482 LIST_FOR_EACH_POP (am
, list_node
, &ams
) {
1483 ofproto_async_msg_free(am
);
1485 guarded_list_destroy(&ofproto
->ams
);
1487 recirc_free_ofproto(ofproto
, ofproto
->up
.name
);
1489 mbridge_unref(ofproto
->mbridge
);
1491 netflow_unref(ofproto
->netflow
);
1492 dpif_sflow_unref(ofproto
->sflow
);
1493 dpif_ipfix_unref(ofproto
->ipfix
);
1494 hmap_destroy(&ofproto
->bundles
);
1495 mac_learning_unref(ofproto
->ml
);
1496 mcast_snooping_unref(ofproto
->ms
);
1498 sset_destroy(&ofproto
->ports
);
1499 sset_destroy(&ofproto
->ghost_ports
);
1500 sset_destroy(&ofproto
->port_poll_set
);
1502 ovs_mutex_destroy(&ofproto
->stats_mutex
);
1504 seq_destroy(ofproto
->ams_seq
);
1506 close_dpif_backer(ofproto
->backer
);
1510 run(struct ofproto
*ofproto_
)
1512 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1513 uint64_t new_seq
, new_dump_seq
;
1515 if (mbridge_need_revalidate(ofproto
->mbridge
)) {
1516 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1517 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1518 mac_learning_flush(ofproto
->ml
);
1519 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1520 mcast_snooping_mdb_flush(ofproto
->ms
);
1523 /* Always updates the ofproto->ams_seqno to avoid frequent wakeup during
1524 * flow restore. Even though nothing is processed during flow restore,
1525 * all queued 'ams' will be handled immediately when flow restore
1527 ofproto
->ams_seqno
= seq_read(ofproto
->ams_seq
);
1529 /* Do not perform any periodic activity required by 'ofproto' while
1530 * waiting for flow restore to complete. */
1531 if (!ofproto_get_flow_restore_wait()) {
1532 struct ofproto_async_msg
*am
;
1533 struct ovs_list ams
;
1535 guarded_list_pop_all(&ofproto
->ams
, &ams
);
1536 LIST_FOR_EACH_POP (am
, list_node
, &ams
) {
1537 connmgr_send_async_msg(ofproto
->up
.connmgr
, am
);
1538 ofproto_async_msg_free(am
);
1542 if (ofproto
->netflow
) {
1543 netflow_run(ofproto
->netflow
);
1545 if (ofproto
->sflow
) {
1546 dpif_sflow_run(ofproto
->sflow
);
1548 if (ofproto
->ipfix
) {
1549 dpif_ipfix_run(ofproto
->ipfix
);
1552 new_seq
= seq_read(connectivity_seq_get());
1553 if (ofproto
->change_seq
!= new_seq
) {
1554 struct ofport_dpif
*ofport
;
1556 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1560 ofproto
->change_seq
= new_seq
;
1562 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1563 struct ofbundle
*bundle
;
1565 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1572 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1573 if (mac_learning_run(ofproto
->ml
)) {
1574 ofproto
->backer
->need_revalidate
= REV_MAC_LEARNING
;
1576 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1578 if (mcast_snooping_run(ofproto
->ms
)) {
1579 ofproto
->backer
->need_revalidate
= REV_MCAST_SNOOPING
;
1582 new_dump_seq
= seq_read(udpif_dump_seq(ofproto
->backer
->udpif
));
1583 if (ofproto
->dump_seq
!= new_dump_seq
) {
1584 struct rule
*rule
, *next_rule
;
1585 long long now
= time_msec();
1587 /* We know stats are relatively fresh, so now is a good time to do some
1589 ofproto
->dump_seq
= new_dump_seq
;
1591 /* Expire OpenFlow flows whose idle_timeout or hard_timeout
1593 ovs_mutex_lock(&ofproto_mutex
);
1594 LIST_FOR_EACH_SAFE (rule
, next_rule
, expirable
,
1595 &ofproto
->up
.expirable
) {
1596 rule_expire(rule_dpif_cast(rule
), now
);
1598 ovs_mutex_unlock(&ofproto_mutex
);
1600 /* All outstanding data in existing flows has been accounted, so it's a
1601 * good time to do bond rebalancing. */
1602 if (ofproto
->has_bonded_bundles
) {
1603 struct ofbundle
*bundle
;
1605 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1607 bond_rebalance(bundle
->bond
);
1616 ofproto_dpif_wait(struct ofproto
*ofproto_
)
1618 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1620 if (ofproto_get_flow_restore_wait()) {
1624 if (ofproto
->sflow
) {
1625 dpif_sflow_wait(ofproto
->sflow
);
1627 if (ofproto
->ipfix
) {
1628 dpif_ipfix_wait(ofproto
->ipfix
);
1630 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1631 struct ofbundle
*bundle
;
1633 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1634 bundle_wait(bundle
);
1637 if (ofproto
->netflow
) {
1638 netflow_wait(ofproto
->netflow
);
1640 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
1641 mac_learning_wait(ofproto
->ml
);
1642 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1643 mcast_snooping_wait(ofproto
->ms
);
1645 if (ofproto
->backer
->need_revalidate
) {
1646 poll_immediate_wake();
1649 seq_wait(udpif_dump_seq(ofproto
->backer
->udpif
), ofproto
->dump_seq
);
1650 seq_wait(ofproto
->ams_seq
, ofproto
->ams_seqno
);
1654 type_get_memory_usage(const char *type
, struct simap
*usage
)
1656 struct dpif_backer
*backer
;
1658 backer
= shash_find_data(&all_dpif_backers
, type
);
1660 udpif_get_memory_usage(backer
->udpif
, usage
);
1665 flush(struct ofproto
*ofproto_
)
1667 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1668 struct dpif_backer
*backer
= ofproto
->backer
;
1671 udpif_flush(backer
->udpif
);
1676 query_tables(struct ofproto
*ofproto
,
1677 struct ofputil_table_features
*features
,
1678 struct ofputil_table_stats
*stats
)
1680 strcpy(features
->name
, "classifier");
1685 for (i
= 0; i
< ofproto
->n_tables
; i
++) {
1686 unsigned long missed
, matched
;
1688 atomic_read_relaxed(&ofproto
->tables
[i
].n_matched
, &matched
);
1689 atomic_read_relaxed(&ofproto
->tables
[i
].n_missed
, &missed
);
1691 stats
[i
].matched_count
= matched
;
1692 stats
[i
].lookup_count
= matched
+ missed
;
1698 set_tables_version(struct ofproto
*ofproto_
, ovs_version_t version
)
1700 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1702 /* Use memory_order_release to signify that any prior memory accesses can
1703 * not be reordered to happen after this atomic store. This makes sure the
1704 * new version is properly set up when the readers can read this 'version'
1706 atomic_store_explicit(&ofproto
->tables_version
, version
,
1707 memory_order_release
);
1708 /* 'need_revalidate' can be reordered to happen before the atomic_store
1709 * above, but it does not matter as this variable is not accessed by other
1711 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
1714 static struct ofport
*
1717 struct ofport_dpif
*port
= xzalloc(sizeof *port
);
1722 port_dealloc(struct ofport
*port_
)
1724 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1729 port_construct(struct ofport
*port_
)
1731 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1732 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1733 const struct netdev
*netdev
= port
->up
.netdev
;
1734 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1735 const char *dp_port_name
;
1736 struct dpif_port dpif_port
;
1739 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1740 port
->bundle
= NULL
;
1744 port
->may_enable
= false;
1745 port
->stp_port
= NULL
;
1746 port
->stp_state
= STP_DISABLED
;
1747 port
->rstp_port
= NULL
;
1748 port
->rstp_state
= RSTP_DISABLED
;
1749 port
->is_tunnel
= false;
1753 port
->carrier_seq
= netdev_get_carrier_resets(netdev
);
1754 port
->is_layer3
= netdev_vport_is_layer3(netdev
);
1756 if (netdev_vport_is_patch(netdev
)) {
1757 /* By bailing out here, we don't submit the port to the sFlow module
1758 * to be considered for counter polling export. This is correct
1759 * because the patch port represents an interface that sFlow considers
1760 * to be "internal" to the switch as a whole, and therefore not a
1761 * candidate for counter polling. */
1762 port
->odp_port
= ODPP_NONE
;
1763 ofport_update_peer(port
);
1767 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
1768 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
, dp_port_name
,
1774 port
->odp_port
= dpif_port
.port_no
;
1776 if (netdev_get_tunnel_config(netdev
)) {
1777 atomic_count_inc(&ofproto
->backer
->tnl_count
);
1778 error
= tnl_port_add(port
, port
->up
.netdev
, port
->odp_port
,
1779 ovs_native_tunneling_is_on(ofproto
), dp_port_name
);
1781 atomic_count_dec(&ofproto
->backer
->tnl_count
);
1782 dpif_port_destroy(&dpif_port
);
1786 port
->is_tunnel
= true;
1787 if (ofproto
->ipfix
) {
1788 dpif_ipfix_add_tunnel_port(ofproto
->ipfix
, port_
, port
->odp_port
);
1791 /* Sanity-check that a mapping doesn't already exist. This
1792 * shouldn't happen for non-tunnel ports. */
1793 if (odp_port_to_ofp_port(ofproto
, port
->odp_port
) != OFPP_NONE
) {
1794 VLOG_ERR("port %s already has an OpenFlow port number",
1796 dpif_port_destroy(&dpif_port
);
1800 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1801 hmap_insert(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
,
1802 hash_odp_port(port
->odp_port
));
1803 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1805 dpif_port_destroy(&dpif_port
);
1807 if (ofproto
->sflow
) {
1808 dpif_sflow_add_port(ofproto
->sflow
, port_
, port
->odp_port
);
1815 port_destruct(struct ofport
*port_
, bool del
)
1817 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1818 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1819 const char *devname
= netdev_get_name(port
->up
.netdev
);
1820 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1821 const char *dp_port_name
;
1823 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1825 xlate_ofport_remove(port
);
1828 dp_port_name
= netdev_vport_get_dpif_port(port
->up
.netdev
, namebuf
,
1830 if (del
&& dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
1831 /* The underlying device is still there, so delete it. This
1832 * happens when the ofproto is being destroyed, since the caller
1833 * assumes that removal of attached ports will happen as part of
1835 if (!port
->is_tunnel
) {
1836 dpif_port_del(ofproto
->backer
->dpif
, port
->odp_port
);
1841 port
->peer
->peer
= NULL
;
1845 if (port
->odp_port
!= ODPP_NONE
&& !port
->is_tunnel
) {
1846 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1847 hmap_remove(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
);
1848 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1851 if (port
->is_tunnel
) {
1852 atomic_count_dec(&ofproto
->backer
->tnl_count
);
1855 if (port
->is_tunnel
&& ofproto
->ipfix
) {
1856 dpif_ipfix_del_tunnel_port(ofproto
->ipfix
, port
->odp_port
);
1860 sset_find_and_delete(&ofproto
->ports
, devname
);
1861 sset_find_and_delete(&ofproto
->ghost_ports
, devname
);
1862 bundle_remove(port_
);
1863 set_cfm(port_
, NULL
);
1864 set_bfd(port_
, NULL
);
1865 set_lldp(port_
, NULL
);
1866 if (port
->stp_port
) {
1867 stp_port_disable(port
->stp_port
);
1869 set_rstp_port(port_
, NULL
);
1870 if (ofproto
->sflow
) {
1871 dpif_sflow_del_port(ofproto
->sflow
, port
->odp_port
);
1878 port_modified(struct ofport
*port_
)
1880 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1881 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1882 const char *dp_port_name
;
1883 struct netdev
*netdev
= port
->up
.netdev
;
1885 if (port
->bundle
&& port
->bundle
->bond
) {
1886 bond_slave_set_netdev(port
->bundle
->bond
, port
, netdev
);
1890 cfm_set_netdev(port
->cfm
, netdev
);
1894 bfd_set_netdev(port
->bfd
, netdev
);
1897 ofproto_dpif_monitor_port_update(port
, port
->bfd
, port
->cfm
,
1898 port
->lldp
, &port
->up
.pp
.hw_addr
);
1900 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
1902 if (port
->is_tunnel
) {
1903 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1905 if (tnl_port_reconfigure(port
, netdev
, port
->odp_port
,
1906 ovs_native_tunneling_is_on(ofproto
),
1908 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1912 ofport_update_peer(port
);
1916 port_reconfigured(struct ofport
*port_
, enum ofputil_port_config old_config
)
1918 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1919 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1920 enum ofputil_port_config changed
= old_config
^ port
->up
.pp
.config
;
1922 if (changed
& (OFPUTIL_PC_NO_RECV
| OFPUTIL_PC_NO_RECV_STP
|
1923 OFPUTIL_PC_NO_FWD
| OFPUTIL_PC_NO_FLOOD
|
1924 OFPUTIL_PC_NO_PACKET_IN
)) {
1925 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1927 if (changed
& OFPUTIL_PC_NO_FLOOD
&& port
->bundle
) {
1928 bundle_update(port
->bundle
);
1934 set_sflow(struct ofproto
*ofproto_
,
1935 const struct ofproto_sflow_options
*sflow_options
)
1937 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1938 struct dpif_sflow
*ds
= ofproto
->sflow
;
1940 if (sflow_options
) {
1941 uint32_t old_probability
= ds
? dpif_sflow_get_probability(ds
) : 0;
1943 struct ofport_dpif
*ofport
;
1945 ds
= ofproto
->sflow
= dpif_sflow_create();
1946 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1947 dpif_sflow_add_port(ds
, &ofport
->up
, ofport
->odp_port
);
1950 dpif_sflow_set_options(ds
, sflow_options
);
1951 if (dpif_sflow_get_probability(ds
) != old_probability
) {
1952 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1956 dpif_sflow_unref(ds
);
1957 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1958 ofproto
->sflow
= NULL
;
1966 struct ofproto
*ofproto_
,
1967 const struct ofproto_ipfix_bridge_exporter_options
*bridge_exporter_options
,
1968 const struct ofproto_ipfix_flow_exporter_options
*flow_exporters_options
,
1969 size_t n_flow_exporters_options
)
1971 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1972 struct dpif_ipfix
*di
= ofproto
->ipfix
;
1973 bool has_options
= bridge_exporter_options
|| flow_exporters_options
;
1974 bool new_di
= false;
1976 if (has_options
&& !di
) {
1977 di
= ofproto
->ipfix
= dpif_ipfix_create();
1982 /* Call set_options in any case to cleanly flush the flow
1983 * caches in the last exporters that are to be destroyed. */
1984 dpif_ipfix_set_options(
1985 di
, bridge_exporter_options
, flow_exporters_options
,
1986 n_flow_exporters_options
);
1988 /* Add tunnel ports only when a new ipfix created */
1989 if (new_di
== true) {
1990 struct ofport_dpif
*ofport
;
1991 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1992 if (ofport
->is_tunnel
== true) {
1993 dpif_ipfix_add_tunnel_port(di
, &ofport
->up
, ofport
->odp_port
);
1999 dpif_ipfix_unref(di
);
2000 ofproto
->ipfix
= NULL
;
2008 get_ipfix_stats(const struct ofproto
*ofproto_
,
2010 struct ovs_list
*replies
)
2012 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2013 struct dpif_ipfix
*di
= ofproto
->ipfix
;
2016 return OFPERR_NXST_NOT_CONFIGURED
;
2019 return dpif_ipfix_get_stats(di
, bridge_ipfix
, replies
);
2023 set_cfm(struct ofport
*ofport_
, const struct cfm_settings
*s
)
2025 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2026 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2027 struct cfm
*old
= ofport
->cfm
;
2032 ofport
->cfm
= cfm_create(ofport
->up
.netdev
);
2035 if (cfm_configure(ofport
->cfm
, s
)) {
2042 cfm_unref(ofport
->cfm
);
2045 if (ofport
->cfm
!= old
) {
2046 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2048 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
2049 ofport
->lldp
, &ofport
->up
.pp
.hw_addr
);
2054 cfm_status_changed(struct ofport
*ofport_
)
2056 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2058 return ofport
->cfm
? cfm_check_status_change(ofport
->cfm
) : true;
2062 get_cfm_status(const struct ofport
*ofport_
,
2063 struct cfm_status
*status
)
2065 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2069 cfm_get_status(ofport
->cfm
, status
);
2078 set_bfd(struct ofport
*ofport_
, const struct smap
*cfg
)
2080 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
2081 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2085 ofport
->bfd
= bfd_configure(old
, netdev_get_name(ofport
->up
.netdev
),
2086 cfg
, ofport
->up
.netdev
);
2087 if (ofport
->bfd
!= old
) {
2088 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2090 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
2091 ofport
->lldp
, &ofport
->up
.pp
.hw_addr
);
2096 bfd_status_changed(struct ofport
*ofport_
)
2098 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2100 return ofport
->bfd
? bfd_check_status_change(ofport
->bfd
) : true;
2104 get_bfd_status(struct ofport
*ofport_
, struct smap
*smap
)
2106 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2110 bfd_get_status(ofport
->bfd
, smap
);
2119 set_lldp(struct ofport
*ofport_
,
2120 const struct smap
*cfg
)
2122 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2126 if (!ofport
->lldp
) {
2127 struct ofproto_dpif
*ofproto
;
2129 ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2130 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2131 ofport
->lldp
= lldp_create(ofport
->up
.netdev
, ofport_
->mtu
, cfg
);
2134 if (!lldp_configure(ofport
->lldp
, cfg
)) {
2139 lldp_unref(ofport
->lldp
);
2140 ofport
->lldp
= NULL
;
2143 ofproto_dpif_monitor_port_update(ofport
,
2147 &ofport
->up
.pp
.hw_addr
);
2152 get_lldp_status(const struct ofport
*ofport_
,
2153 struct lldp_status
*status OVS_UNUSED
)
2155 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2157 return ofport
->lldp
? true : false;
2161 set_aa(struct ofproto
*ofproto OVS_UNUSED
,
2162 const struct aa_settings
*s
)
2164 return aa_configure(s
);
2168 aa_mapping_set(struct ofproto
*ofproto_ OVS_UNUSED
, void *aux
,
2169 const struct aa_mapping_settings
*s
)
2171 return aa_mapping_register(aux
, s
);
2175 aa_mapping_unset(struct ofproto
*ofproto OVS_UNUSED
, void *aux
)
2177 return aa_mapping_unregister(aux
);
2181 aa_vlan_get_queued(struct ofproto
*ofproto OVS_UNUSED
, struct ovs_list
*list
)
2183 return aa_get_vlan_queued(list
);
2187 aa_vlan_get_queue_size(struct ofproto
*ofproto OVS_UNUSED
)
2189 return aa_get_vlan_queue_size();
2193 /* Spanning Tree. */
2195 /* Called while rstp_mutex is held. */
2197 rstp_send_bpdu_cb(struct dp_packet
*pkt
, void *ofport_
, void *ofproto_
)
2199 struct ofproto_dpif
*ofproto
= ofproto_
;
2200 struct ofport_dpif
*ofport
= ofport_
;
2201 struct eth_header
*eth
= dp_packet_l2(pkt
);
2203 netdev_get_etheraddr(ofport
->up
.netdev
, ð
->eth_src
);
2204 if (eth_addr_is_zero(eth
->eth_src
)) {
2205 VLOG_WARN_RL(&rl
, "%s port %d: cannot send RSTP BPDU on a port which "
2206 "does not have a configured source MAC address.",
2207 ofproto
->up
.name
, ofp_to_u16(ofport
->up
.ofp_port
));
2209 ofproto_dpif_send_packet(ofport
, false, pkt
);
2211 dp_packet_delete(pkt
);
2215 send_bpdu_cb(struct dp_packet
*pkt
, int port_num
, void *ofproto_
)
2217 struct ofproto_dpif
*ofproto
= ofproto_
;
2218 struct stp_port
*sp
= stp_get_port(ofproto
->stp
, port_num
);
2219 struct ofport_dpif
*ofport
;
2221 ofport
= stp_port_get_aux(sp
);
2223 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on unknown port %d",
2224 ofproto
->up
.name
, port_num
);
2226 struct eth_header
*eth
= dp_packet_l2(pkt
);
2228 netdev_get_etheraddr(ofport
->up
.netdev
, ð
->eth_src
);
2229 if (eth_addr_is_zero(eth
->eth_src
)) {
2230 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on port %d "
2231 "with unknown MAC", ofproto
->up
.name
, port_num
);
2233 ofproto_dpif_send_packet(ofport
, false, pkt
);
2236 dp_packet_delete(pkt
);
2239 /* Configure RSTP on 'ofproto_' using the settings defined in 's'. */
2241 set_rstp(struct ofproto
*ofproto_
, const struct ofproto_rstp_settings
*s
)
2243 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2245 /* Only revalidate flows if the configuration changed. */
2246 if (!s
!= !ofproto
->rstp
) {
2247 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2251 if (!ofproto
->rstp
) {
2252 ofproto
->rstp
= rstp_create(ofproto_
->name
, s
->address
,
2253 rstp_send_bpdu_cb
, ofproto
);
2254 ofproto
->rstp_last_tick
= time_msec();
2256 rstp_set_bridge_address(ofproto
->rstp
, s
->address
);
2257 rstp_set_bridge_priority(ofproto
->rstp
, s
->priority
);
2258 rstp_set_bridge_ageing_time(ofproto
->rstp
, s
->ageing_time
);
2259 rstp_set_bridge_force_protocol_version(ofproto
->rstp
,
2260 s
->force_protocol_version
);
2261 rstp_set_bridge_max_age(ofproto
->rstp
, s
->bridge_max_age
);
2262 rstp_set_bridge_forward_delay(ofproto
->rstp
, s
->bridge_forward_delay
);
2263 rstp_set_bridge_transmit_hold_count(ofproto
->rstp
,
2264 s
->transmit_hold_count
);
2266 struct ofport
*ofport
;
2267 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2268 set_rstp_port(ofport
, NULL
);
2270 rstp_unref(ofproto
->rstp
);
2271 ofproto
->rstp
= NULL
;
2276 get_rstp_status(struct ofproto
*ofproto_
, struct ofproto_rstp_status
*s
)
2278 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2280 if (ofproto
->rstp
) {
2282 s
->root_id
= rstp_get_root_id(ofproto
->rstp
);
2283 s
->bridge_id
= rstp_get_bridge_id(ofproto
->rstp
);
2284 s
->designated_id
= rstp_get_designated_id(ofproto
->rstp
);
2285 s
->root_path_cost
= rstp_get_root_path_cost(ofproto
->rstp
);
2286 s
->designated_port_id
= rstp_get_designated_port_id(ofproto
->rstp
);
2287 s
->bridge_port_id
= rstp_get_bridge_port_id(ofproto
->rstp
);
2294 update_rstp_port_state(struct ofport_dpif
*ofport
)
2296 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2297 enum rstp_state state
;
2299 /* Figure out new state. */
2300 state
= ofport
->rstp_port
? rstp_port_get_state(ofport
->rstp_port
)
2304 if (ofport
->rstp_state
!= state
) {
2305 enum ofputil_port_state of_state
;
2308 VLOG_DBG("port %s: RSTP state changed from %s to %s",
2309 netdev_get_name(ofport
->up
.netdev
),
2310 rstp_state_name(ofport
->rstp_state
),
2311 rstp_state_name(state
));
2313 if (rstp_learn_in_state(ofport
->rstp_state
)
2314 != rstp_learn_in_state(state
)) {
2315 /* XXX: Learning action flows should also be flushed. */
2316 if (ofport
->bundle
) {
2317 if (!rstp_shift_root_learned_address(ofproto
->rstp
)
2318 || rstp_get_old_root_aux(ofproto
->rstp
) != ofport
) {
2319 bundle_flush_macs(ofport
->bundle
, false);
2323 fwd_change
= rstp_forward_in_state(ofport
->rstp_state
)
2324 != rstp_forward_in_state(state
);
2326 ofproto
->backer
->need_revalidate
= REV_RSTP
;
2327 ofport
->rstp_state
= state
;
2329 if (fwd_change
&& ofport
->bundle
) {
2330 bundle_update(ofport
->bundle
);
2333 /* Update the RSTP state bits in the OpenFlow port description. */
2334 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2335 of_state
|= (state
== RSTP_LEARNING
? OFPUTIL_PS_STP_LEARN
2336 : state
== RSTP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2337 : state
== RSTP_DISCARDING
? OFPUTIL_PS_STP_LISTEN
2339 ofproto_port_set_state(&ofport
->up
, of_state
);
2344 rstp_run(struct ofproto_dpif
*ofproto
)
2346 if (ofproto
->rstp
) {
2347 long long int now
= time_msec();
2348 long long int elapsed
= now
- ofproto
->rstp_last_tick
;
2349 struct rstp_port
*rp
;
2350 struct ofport_dpif
*ofport
;
2352 /* Every second, decrease the values of the timers. */
2353 if (elapsed
>= 1000) {
2354 rstp_tick_timers(ofproto
->rstp
);
2355 ofproto
->rstp_last_tick
= now
;
2358 while ((ofport
= rstp_get_next_changed_port_aux(ofproto
->rstp
, &rp
))) {
2359 update_rstp_port_state(ofport
);
2363 /* FIXME: This check should be done on-event (i.e., when setting
2364 * p->fdb_flush) and not periodically.
2366 while ((ofport
= rstp_check_and_reset_fdb_flush(ofproto
->rstp
, &rp
))) {
2367 if (!rstp_shift_root_learned_address(ofproto
->rstp
)
2368 || rstp_get_old_root_aux(ofproto
->rstp
) != ofport
) {
2369 bundle_flush_macs(ofport
->bundle
, false);
2373 if (rstp_shift_root_learned_address(ofproto
->rstp
)) {
2374 struct ofport_dpif
*old_root_aux
=
2375 (struct ofport_dpif
*)rstp_get_old_root_aux(ofproto
->rstp
);
2376 struct ofport_dpif
*new_root_aux
=
2377 (struct ofport_dpif
*)rstp_get_new_root_aux(ofproto
->rstp
);
2378 if (old_root_aux
!= NULL
&& new_root_aux
!= NULL
) {
2379 bundle_move(old_root_aux
->bundle
, new_root_aux
->bundle
);
2380 rstp_reset_root_changed(ofproto
->rstp
);
2386 /* Configures STP on 'ofproto_' using the settings defined in 's'. */
2388 set_stp(struct ofproto
*ofproto_
, const struct ofproto_stp_settings
*s
)
2390 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2392 /* Only revalidate flows if the configuration changed. */
2393 if (!s
!= !ofproto
->stp
) {
2394 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2398 if (!ofproto
->stp
) {
2399 ofproto
->stp
= stp_create(ofproto_
->name
, s
->system_id
,
2400 send_bpdu_cb
, ofproto
);
2401 ofproto
->stp_last_tick
= time_msec();
2404 stp_set_bridge_id(ofproto
->stp
, s
->system_id
);
2405 stp_set_bridge_priority(ofproto
->stp
, s
->priority
);
2406 stp_set_hello_time(ofproto
->stp
, s
->hello_time
);
2407 stp_set_max_age(ofproto
->stp
, s
->max_age
);
2408 stp_set_forward_delay(ofproto
->stp
, s
->fwd_delay
);
2410 struct ofport
*ofport
;
2412 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2413 set_stp_port(ofport
, NULL
);
2416 stp_unref(ofproto
->stp
);
2417 ofproto
->stp
= NULL
;
2424 get_stp_status(struct ofproto
*ofproto_
, struct ofproto_stp_status
*s
)
2426 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2430 s
->bridge_id
= stp_get_bridge_id(ofproto
->stp
);
2431 s
->designated_root
= stp_get_designated_root(ofproto
->stp
);
2432 s
->root_path_cost
= stp_get_root_path_cost(ofproto
->stp
);
2441 update_stp_port_state(struct ofport_dpif
*ofport
)
2443 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2444 enum stp_state state
;
2446 /* Figure out new state. */
2447 state
= ofport
->stp_port
? stp_port_get_state(ofport
->stp_port
)
2451 if (ofport
->stp_state
!= state
) {
2452 enum ofputil_port_state of_state
;
2455 VLOG_DBG("port %s: STP state changed from %s to %s",
2456 netdev_get_name(ofport
->up
.netdev
),
2457 stp_state_name(ofport
->stp_state
),
2458 stp_state_name(state
));
2459 if (stp_learn_in_state(ofport
->stp_state
)
2460 != stp_learn_in_state(state
)) {
2461 /* xxx Learning action flows should also be flushed. */
2462 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2463 mac_learning_flush(ofproto
->ml
);
2464 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2465 mcast_snooping_mdb_flush(ofproto
->ms
);
2467 fwd_change
= stp_forward_in_state(ofport
->stp_state
)
2468 != stp_forward_in_state(state
);
2470 ofproto
->backer
->need_revalidate
= REV_STP
;
2471 ofport
->stp_state
= state
;
2472 ofport
->stp_state_entered
= time_msec();
2474 if (fwd_change
&& ofport
->bundle
) {
2475 bundle_update(ofport
->bundle
);
2478 /* Update the STP state bits in the OpenFlow port description. */
2479 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2480 of_state
|= (state
== STP_LISTENING
? OFPUTIL_PS_STP_LISTEN
2481 : state
== STP_LEARNING
? OFPUTIL_PS_STP_LEARN
2482 : state
== STP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2483 : state
== STP_BLOCKING
? OFPUTIL_PS_STP_BLOCK
2485 ofproto_port_set_state(&ofport
->up
, of_state
);
2489 /* Configures STP on 'ofport_' using the settings defined in 's'. The
2490 * caller is responsible for assigning STP port numbers and ensuring
2491 * there are no duplicates. */
2493 set_stp_port(struct ofport
*ofport_
,
2494 const struct ofproto_port_stp_settings
*s
)
2496 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2497 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2498 struct stp_port
*sp
= ofport
->stp_port
;
2500 if (!s
|| !s
->enable
) {
2502 ofport
->stp_port
= NULL
;
2503 stp_port_disable(sp
);
2504 update_stp_port_state(ofport
);
2507 } else if (sp
&& stp_port_no(sp
) != s
->port_num
2508 && ofport
== stp_port_get_aux(sp
)) {
2509 /* The port-id changed, so disable the old one if it's not
2510 * already in use by another port. */
2511 stp_port_disable(sp
);
2514 sp
= ofport
->stp_port
= stp_get_port(ofproto
->stp
, s
->port_num
);
2516 /* Set name before enabling the port so that debugging messages can print
2518 stp_port_set_name(sp
, netdev_get_name(ofport
->up
.netdev
));
2519 stp_port_enable(sp
);
2521 stp_port_set_aux(sp
, ofport
);
2522 stp_port_set_priority(sp
, s
->priority
);
2523 stp_port_set_path_cost(sp
, s
->path_cost
);
2525 update_stp_port_state(ofport
);
2531 get_stp_port_status(struct ofport
*ofport_
,
2532 struct ofproto_port_stp_status
*s
)
2534 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2535 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2536 struct stp_port
*sp
= ofport
->stp_port
;
2538 if (!ofproto
->stp
|| !sp
) {
2544 s
->port_id
= stp_port_get_id(sp
);
2545 s
->state
= stp_port_get_state(sp
);
2546 s
->sec_in_state
= (time_msec() - ofport
->stp_state_entered
) / 1000;
2547 s
->role
= stp_port_get_role(sp
);
2553 get_stp_port_stats(struct ofport
*ofport_
,
2554 struct ofproto_port_stp_stats
*s
)
2556 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2557 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2558 struct stp_port
*sp
= ofport
->stp_port
;
2560 if (!ofproto
->stp
|| !sp
) {
2566 stp_port_get_counts(sp
, &s
->tx_count
, &s
->rx_count
, &s
->error_count
);
2572 stp_run(struct ofproto_dpif
*ofproto
)
2575 long long int now
= time_msec();
2576 long long int elapsed
= now
- ofproto
->stp_last_tick
;
2577 struct stp_port
*sp
;
2580 stp_tick(ofproto
->stp
, MIN(INT_MAX
, elapsed
));
2581 ofproto
->stp_last_tick
= now
;
2583 while (stp_get_changed_port(ofproto
->stp
, &sp
)) {
2584 struct ofport_dpif
*ofport
= stp_port_get_aux(sp
);
2587 update_stp_port_state(ofport
);
2591 if (stp_check_and_reset_fdb_flush(ofproto
->stp
)) {
2592 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2593 mac_learning_flush(ofproto
->ml
);
2594 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2595 mcast_snooping_mdb_flush(ofproto
->ms
);
2601 stp_wait(struct ofproto_dpif
*ofproto
)
2604 poll_timer_wait(1000);
2608 /* Configures RSTP on 'ofport_' using the settings defined in 's'. The
2609 * caller is responsible for assigning RSTP port numbers and ensuring
2610 * there are no duplicates. */
2612 set_rstp_port(struct ofport
*ofport_
,
2613 const struct ofproto_port_rstp_settings
*s
)
2615 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2616 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2617 struct rstp_port
*rp
= ofport
->rstp_port
;
2619 if (!s
|| !s
->enable
) {
2621 rstp_port_set_aux(rp
, NULL
);
2622 rstp_port_set_state(rp
, RSTP_DISABLED
);
2623 rstp_port_set_mac_operational(rp
, false);
2624 ofport
->rstp_port
= NULL
;
2625 rstp_port_unref(rp
);
2626 update_rstp_port_state(ofport
);
2631 /* Check if need to add a new port. */
2633 rp
= ofport
->rstp_port
= rstp_add_port(ofproto
->rstp
);
2636 rstp_port_set(rp
, s
->port_num
, s
->priority
, s
->path_cost
,
2637 s
->admin_edge_port
, s
->auto_edge
,
2638 s
->admin_p2p_mac_state
, s
->admin_port_state
, s
->mcheck
,
2640 update_rstp_port_state(ofport
);
2641 /* Synchronize operational status. */
2642 rstp_port_set_mac_operational(rp
, ofport
->may_enable
);
2646 get_rstp_port_status(struct ofport
*ofport_
,
2647 struct ofproto_port_rstp_status
*s
)
2649 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2650 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2651 struct rstp_port
*rp
= ofport
->rstp_port
;
2653 if (!ofproto
->rstp
|| !rp
) {
2659 rstp_port_get_status(rp
, &s
->port_id
, &s
->state
, &s
->role
,
2660 &s
->designated_bridge_id
, &s
->designated_port_id
,
2661 &s
->designated_path_cost
, &s
->tx_count
,
2662 &s
->rx_count
, &s
->error_count
, &s
->uptime
);
2667 set_queues(struct ofport
*ofport_
, const struct ofproto_port_queue
*qdscp
,
2670 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2671 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2673 if (ofport
->n_qdscp
!= n_qdscp
2674 || (n_qdscp
&& memcmp(ofport
->qdscp
, qdscp
,
2675 n_qdscp
* sizeof *qdscp
))) {
2676 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2677 free(ofport
->qdscp
);
2678 ofport
->qdscp
= n_qdscp
2679 ? xmemdup(qdscp
, n_qdscp
* sizeof *qdscp
)
2681 ofport
->n_qdscp
= n_qdscp
;
2689 /* Expires all MAC learning entries associated with 'bundle' and forces its
2690 * ofproto to revalidate every flow.
2692 * Normally MAC learning entries are removed only from the ofproto associated
2693 * with 'bundle', but if 'all_ofprotos' is true, then the MAC learning entries
2694 * are removed from every ofproto. When patch ports and SLB bonds are in use
2695 * and a VM migration happens and the gratuitous ARPs are somehow lost, this
2696 * avoids a MAC_ENTRY_IDLE_TIME delay before the migrated VM can communicate
2697 * with the host from which it migrated. */
2699 bundle_flush_macs(struct ofbundle
*bundle
, bool all_ofprotos
)
2701 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2702 struct mac_learning
*ml
= ofproto
->ml
;
2703 struct mac_entry
*mac
, *next_mac
;
2705 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2706 ovs_rwlock_wrlock(&ml
->rwlock
);
2707 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2708 if (mac_entry_get_port(ml
, mac
) == bundle
) {
2710 struct ofproto_dpif
*o
;
2712 HMAP_FOR_EACH (o
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
2714 struct mac_entry
*e
;
2716 ovs_rwlock_wrlock(&o
->ml
->rwlock
);
2717 e
= mac_learning_lookup(o
->ml
, mac
->mac
, mac
->vlan
);
2719 mac_learning_expire(o
->ml
, e
);
2721 ovs_rwlock_unlock(&o
->ml
->rwlock
);
2726 mac_learning_expire(ml
, mac
);
2729 ovs_rwlock_unlock(&ml
->rwlock
);
2733 bundle_move(struct ofbundle
*old
, struct ofbundle
*new)
2735 struct ofproto_dpif
*ofproto
= old
->ofproto
;
2736 struct mac_learning
*ml
= ofproto
->ml
;
2737 struct mac_entry
*mac
, *next_mac
;
2739 ovs_assert(new->ofproto
== old
->ofproto
);
2741 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2742 ovs_rwlock_wrlock(&ml
->rwlock
);
2743 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2744 if (mac_entry_get_port(ml
, mac
) == old
) {
2745 mac_entry_set_port(ml
, mac
, new);
2748 ovs_rwlock_unlock(&ml
->rwlock
);
2751 static struct ofbundle
*
2752 bundle_lookup(const struct ofproto_dpif
*ofproto
, void *aux
)
2754 struct ofbundle
*bundle
;
2756 HMAP_FOR_EACH_IN_BUCKET (bundle
, hmap_node
, hash_pointer(aux
, 0),
2757 &ofproto
->bundles
) {
2758 if (bundle
->aux
== aux
) {
2766 bundle_update(struct ofbundle
*bundle
)
2768 struct ofport_dpif
*port
;
2770 bundle
->floodable
= true;
2771 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2772 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2774 || (bundle
->ofproto
->stp
&& !stp_forward_in_state(port
->stp_state
))
2775 || (bundle
->ofproto
->rstp
&& !rstp_forward_in_state(port
->rstp_state
))) {
2776 bundle
->floodable
= false;
2783 bundle_del_port(struct ofport_dpif
*port
)
2785 struct ofbundle
*bundle
= port
->bundle
;
2787 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2789 ovs_list_remove(&port
->bundle_node
);
2790 port
->bundle
= NULL
;
2793 lacp_slave_unregister(bundle
->lacp
, port
);
2796 bond_slave_unregister(bundle
->bond
, port
);
2799 bundle_update(bundle
);
2803 bundle_add_port(struct ofbundle
*bundle
, ofp_port_t ofp_port
,
2804 struct lacp_slave_settings
*lacp
)
2806 struct ofport_dpif
*port
;
2808 port
= ofp_port_to_ofport(bundle
->ofproto
, ofp_port
);
2813 if (port
->bundle
!= bundle
) {
2814 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2816 bundle_remove(&port
->up
);
2819 port
->bundle
= bundle
;
2820 ovs_list_push_back(&bundle
->ports
, &port
->bundle_node
);
2821 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2823 || (bundle
->ofproto
->stp
&& !stp_forward_in_state(port
->stp_state
))
2824 || (bundle
->ofproto
->rstp
&& !rstp_forward_in_state(port
->rstp_state
))) {
2825 bundle
->floodable
= false;
2829 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2830 lacp_slave_register(bundle
->lacp
, port
, lacp
);
2837 bundle_destroy(struct ofbundle
*bundle
)
2839 struct ofproto_dpif
*ofproto
;
2840 struct ofport_dpif
*port
, *next_port
;
2846 ofproto
= bundle
->ofproto
;
2847 mbridge_unregister_bundle(ofproto
->mbridge
, bundle
);
2850 xlate_bundle_remove(bundle
);
2853 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2854 bundle_del_port(port
);
2857 bundle_flush_macs(bundle
, true);
2858 mcast_snooping_flush_bundle(ofproto
->ms
, bundle
);
2859 hmap_remove(&ofproto
->bundles
, &bundle
->hmap_node
);
2861 free(bundle
->trunks
);
2862 free(bundle
->cvlans
);
2863 lacp_unref(bundle
->lacp
);
2864 bond_unref(bundle
->bond
);
2869 bundle_set(struct ofproto
*ofproto_
, void *aux
,
2870 const struct ofproto_bundle_settings
*s
)
2872 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2873 bool need_flush
= false;
2874 struct ofport_dpif
*port
;
2875 struct ofbundle
*bundle
;
2876 unsigned long *trunks
= NULL
;
2877 unsigned long *cvlans
= NULL
;
2883 bundle_destroy(bundle_lookup(ofproto
, aux
));
2887 ovs_assert(s
->n_slaves
== 1 || s
->bond
!= NULL
);
2888 ovs_assert((s
->lacp
!= NULL
) == (s
->lacp_slaves
!= NULL
));
2890 bundle
= bundle_lookup(ofproto
, aux
);
2892 bundle
= xmalloc(sizeof *bundle
);
2894 bundle
->ofproto
= ofproto
;
2895 hmap_insert(&ofproto
->bundles
, &bundle
->hmap_node
,
2896 hash_pointer(aux
, 0));
2898 bundle
->name
= NULL
;
2900 ovs_list_init(&bundle
->ports
);
2901 bundle
->vlan_mode
= PORT_VLAN_TRUNK
;
2902 bundle
->qinq_ethtype
= ETH_TYPE_VLAN_8021AD
;
2904 bundle
->trunks
= NULL
;
2905 bundle
->cvlans
= NULL
;
2906 bundle
->use_priority_tags
= s
->use_priority_tags
;
2907 bundle
->lacp
= NULL
;
2908 bundle
->bond
= NULL
;
2910 bundle
->floodable
= true;
2911 bundle
->protected = false;
2912 mbridge_register_bundle(ofproto
->mbridge
, bundle
);
2915 if (!bundle
->name
|| strcmp(s
->name
, bundle
->name
)) {
2917 bundle
->name
= xstrdup(s
->name
);
2922 ofproto
->lacp_enabled
= true;
2923 if (!bundle
->lacp
) {
2924 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2925 bundle
->lacp
= lacp_create();
2927 lacp_configure(bundle
->lacp
, s
->lacp
);
2929 lacp_unref(bundle
->lacp
);
2930 bundle
->lacp
= NULL
;
2933 /* Update set of ports. */
2935 for (i
= 0; i
< s
->n_slaves
; i
++) {
2936 if (!bundle_add_port(bundle
, s
->slaves
[i
],
2937 s
->lacp
? &s
->lacp_slaves
[i
] : NULL
)) {
2941 if (!ok
|| ovs_list_size(&bundle
->ports
) != s
->n_slaves
) {
2942 struct ofport_dpif
*next_port
;
2944 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2945 for (i
= 0; i
< s
->n_slaves
; i
++) {
2946 if (s
->slaves
[i
] == port
->up
.ofp_port
) {
2951 bundle_del_port(port
);
2955 ovs_assert(ovs_list_size(&bundle
->ports
) <= s
->n_slaves
);
2957 if (ovs_list_is_empty(&bundle
->ports
)) {
2958 bundle_destroy(bundle
);
2962 /* Set VLAN tagging mode */
2963 if (s
->vlan_mode
!= bundle
->vlan_mode
2964 || s
->use_priority_tags
!= bundle
->use_priority_tags
) {
2965 bundle
->vlan_mode
= s
->vlan_mode
;
2966 bundle
->use_priority_tags
= s
->use_priority_tags
;
2970 if (s
->qinq_ethtype
!= bundle
->qinq_ethtype
) {
2971 bundle
->qinq_ethtype
= s
->qinq_ethtype
;
2976 vlan
= (s
->vlan_mode
== PORT_VLAN_TRUNK
? -1
2977 : s
->vlan
>= 0 && s
->vlan
<= 4095 ? s
->vlan
2979 if (vlan
!= bundle
->vlan
) {
2980 bundle
->vlan
= vlan
;
2984 /* Get trunked VLANs. */
2985 switch (s
->vlan_mode
) {
2986 case PORT_VLAN_ACCESS
:
2990 case PORT_VLAN_TRUNK
:
2991 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2994 case PORT_VLAN_NATIVE_UNTAGGED
:
2995 case PORT_VLAN_NATIVE_TAGGED
:
2996 if (vlan
!= 0 && (!s
->trunks
2997 || !bitmap_is_set(s
->trunks
, vlan
)
2998 || bitmap_is_set(s
->trunks
, 0))) {
2999 /* Force trunking the native VLAN and prohibit trunking VLAN 0. */
3001 trunks
= bitmap_clone(s
->trunks
, 4096);
3003 trunks
= bitmap_allocate1(4096);
3005 bitmap_set1(trunks
, vlan
);
3006 bitmap_set0(trunks
, 0);
3008 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
3012 case PORT_VLAN_DOT1Q_TUNNEL
:
3013 cvlans
= CONST_CAST(unsigned long *, s
->cvlans
);
3019 if (!vlan_bitmap_equal(trunks
, bundle
->trunks
)) {
3020 free(bundle
->trunks
);
3021 if (trunks
== s
->trunks
) {
3022 bundle
->trunks
= vlan_bitmap_clone(trunks
);
3024 bundle
->trunks
= trunks
;
3029 if (trunks
!= s
->trunks
) {
3033 if (!vlan_bitmap_equal(cvlans
, bundle
->cvlans
)) {
3034 free(bundle
->cvlans
);
3035 if (cvlans
== s
->cvlans
) {
3036 bundle
->cvlans
= vlan_bitmap_clone(cvlans
);
3038 bundle
->cvlans
= cvlans
;
3043 if (cvlans
!= s
->cvlans
) {
3048 if (!ovs_list_is_short(&bundle
->ports
)) {
3049 bundle
->ofproto
->has_bonded_bundles
= true;
3051 if (bond_reconfigure(bundle
->bond
, s
->bond
)) {
3052 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3055 bundle
->bond
= bond_create(s
->bond
, ofproto
);
3056 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3059 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
3060 bond_slave_register(bundle
->bond
, port
,
3061 port
->up
.ofp_port
, port
->up
.netdev
);
3064 bond_unref(bundle
->bond
);
3065 bundle
->bond
= NULL
;
3068 /* Set proteced port mode */
3069 if (s
->protected != bundle
->protected) {
3070 bundle
->protected = s
->protected;
3074 /* If we changed something that would affect MAC learning, un-learn
3075 * everything on this port and force flow revalidation. */
3077 bundle_flush_macs(bundle
, false);
3078 mcast_snooping_flush_bundle(ofproto
->ms
, bundle
);
3085 bundle_remove(struct ofport
*port_
)
3087 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
3088 struct ofbundle
*bundle
= port
->bundle
;
3091 bundle_del_port(port
);
3092 if (ovs_list_is_empty(&bundle
->ports
)) {
3093 bundle_destroy(bundle
);
3094 } else if (ovs_list_is_short(&bundle
->ports
)) {
3095 bond_unref(bundle
->bond
);
3096 bundle
->bond
= NULL
;
3102 send_pdu_cb(void *port_
, const void *pdu
, size_t pdu_size
)
3104 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 10);
3105 struct ofport_dpif
*port
= port_
;
3109 error
= netdev_get_etheraddr(port
->up
.netdev
, &ea
);
3111 struct dp_packet packet
;
3114 dp_packet_init(&packet
, 0);
3115 packet_pdu
= eth_compose(&packet
, eth_addr_lacp
, ea
, ETH_TYPE_LACP
,
3117 memcpy(packet_pdu
, pdu
, pdu_size
);
3119 ofproto_dpif_send_packet(port
, false, &packet
);
3120 dp_packet_uninit(&packet
);
3122 VLOG_ERR_RL(&rl
, "port %s: cannot obtain Ethernet address of iface "
3123 "%s (%s)", port
->bundle
->name
,
3124 netdev_get_name(port
->up
.netdev
), ovs_strerror(error
));
3129 bundle_send_learning_packets(struct ofbundle
*bundle
)
3131 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
3132 int error
, n_packets
, n_errors
;
3133 struct mac_entry
*e
;
3135 struct ovs_list list_node
;
3136 struct ofport_dpif
*port
;
3137 struct dp_packet
*pkt
;
3139 struct ovs_list packets
;
3141 ovs_list_init(&packets
);
3142 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
3143 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
3144 if (mac_entry_get_port(ofproto
->ml
, e
) != bundle
) {
3145 pkt_node
= xmalloc(sizeof *pkt_node
);
3146 pkt_node
->pkt
= bond_compose_learning_packet(bundle
->bond
,
3148 (void **)&pkt_node
->port
);
3149 ovs_list_push_back(&packets
, &pkt_node
->list_node
);
3152 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3154 error
= n_packets
= n_errors
= 0;
3155 LIST_FOR_EACH_POP (pkt_node
, list_node
, &packets
) {
3158 ret
= ofproto_dpif_send_packet(pkt_node
->port
, false, pkt_node
->pkt
);
3159 dp_packet_delete(pkt_node
->pkt
);
3169 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3170 VLOG_WARN_RL(&rl
, "bond %s: %d errors sending %d gratuitous learning "
3171 "packets, last error was: %s",
3172 bundle
->name
, n_errors
, n_packets
, ovs_strerror(error
));
3174 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3175 bundle
->name
, n_packets
);
3180 bundle_run(struct ofbundle
*bundle
)
3183 lacp_run(bundle
->lacp
, send_pdu_cb
);
3186 struct ofport_dpif
*port
;
3188 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
3189 bond_slave_set_may_enable(bundle
->bond
, port
, port
->may_enable
);
3192 if (bond_run(bundle
->bond
, lacp_status(bundle
->lacp
))) {
3193 bundle
->ofproto
->backer
->need_revalidate
= REV_BOND
;
3196 if (bond_should_send_learning_packets(bundle
->bond
)) {
3197 bundle_send_learning_packets(bundle
);
3203 bundle_wait(struct ofbundle
*bundle
)
3206 lacp_wait(bundle
->lacp
);
3209 bond_wait(bundle
->bond
);
3216 mirror_set__(struct ofproto
*ofproto_
, void *aux
,
3217 const struct ofproto_mirror_settings
*s
)
3219 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3220 struct ofbundle
**srcs
, **dsts
;
3225 mirror_destroy(ofproto
->mbridge
, aux
);
3229 srcs
= xmalloc(s
->n_srcs
* sizeof *srcs
);
3230 dsts
= xmalloc(s
->n_dsts
* sizeof *dsts
);
3232 for (i
= 0; i
< s
->n_srcs
; i
++) {
3233 srcs
[i
] = bundle_lookup(ofproto
, s
->srcs
[i
]);
3236 for (i
= 0; i
< s
->n_dsts
; i
++) {
3237 dsts
[i
] = bundle_lookup(ofproto
, s
->dsts
[i
]);
3240 error
= mirror_set(ofproto
->mbridge
, aux
, s
->name
, srcs
, s
->n_srcs
, dsts
,
3241 s
->n_dsts
, s
->src_vlans
,
3242 bundle_lookup(ofproto
, s
->out_bundle
),
3243 s
->snaplen
, s
->out_vlan
);
3250 mirror_get_stats__(struct ofproto
*ofproto
, void *aux
,
3251 uint64_t *packets
, uint64_t *bytes
)
3253 return mirror_get_stats(ofproto_dpif_cast(ofproto
)->mbridge
, aux
, packets
,
3258 set_flood_vlans(struct ofproto
*ofproto_
, unsigned long *flood_vlans
)
3260 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3261 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
3262 if (mac_learning_set_flood_vlans(ofproto
->ml
, flood_vlans
)) {
3263 mac_learning_flush(ofproto
->ml
);
3265 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3270 is_mirror_output_bundle(const struct ofproto
*ofproto_
, void *aux
)
3272 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3273 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3274 return bundle
&& mirror_bundle_out(ofproto
->mbridge
, bundle
) != 0;
3278 forward_bpdu_changed(struct ofproto
*ofproto_
)
3280 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3281 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3285 set_mac_table_config(struct ofproto
*ofproto_
, unsigned int idle_time
,
3288 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3289 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
3290 mac_learning_set_idle_time(ofproto
->ml
, idle_time
);
3291 mac_learning_set_max_entries(ofproto
->ml
, max_entries
);
3292 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3295 /* Configures multicast snooping on 'ofport' using the settings
3296 * defined in 's'. */
3298 set_mcast_snooping(struct ofproto
*ofproto_
,
3299 const struct ofproto_mcast_snooping_settings
*s
)
3301 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3303 /* Only revalidate flows if the configuration changed. */
3304 if (!s
!= !ofproto
->ms
) {
3305 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3310 ofproto
->ms
= mcast_snooping_create();
3313 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3314 mcast_snooping_set_idle_time(ofproto
->ms
, s
->idle_time
);
3315 mcast_snooping_set_max_entries(ofproto
->ms
, s
->max_entries
);
3316 if (mcast_snooping_set_flood_unreg(ofproto
->ms
, s
->flood_unreg
)) {
3317 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3319 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3321 mcast_snooping_unref(ofproto
->ms
);
3328 /* Configures multicast snooping port's flood settings on 'ofproto'. */
3330 set_mcast_snooping_port(struct ofproto
*ofproto_
, void *aux
,
3331 const struct ofproto_mcast_snooping_port_settings
*s
)
3333 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3334 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3336 if (ofproto
->ms
&& s
) {
3337 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3338 mcast_snooping_set_port_flood(ofproto
->ms
, bundle
, s
->flood
);
3339 mcast_snooping_set_port_flood_reports(ofproto
->ms
, bundle
,
3341 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3349 struct ofport_dpif
*
3350 ofp_port_to_ofport(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
3352 struct ofport
*ofport
= ofproto_get_port(&ofproto
->up
, ofp_port
);
3353 return ofport
? ofport_dpif_cast(ofport
) : NULL
;
3357 ofproto_port_from_dpif_port(struct ofproto_dpif
*ofproto
,
3358 struct ofproto_port
*ofproto_port
,
3359 struct dpif_port
*dpif_port
)
3361 ofproto_port
->name
= dpif_port
->name
;
3362 ofproto_port
->type
= dpif_port
->type
;
3363 ofproto_port
->ofp_port
= odp_port_to_ofp_port(ofproto
, dpif_port
->port_no
);
3367 ofport_update_peer(struct ofport_dpif
*ofport
)
3369 const struct ofproto_dpif
*ofproto
;
3370 struct dpif_backer
*backer
;
3373 if (!netdev_vport_is_patch(ofport
->up
.netdev
)) {
3377 backer
= ofproto_dpif_cast(ofport
->up
.ofproto
)->backer
;
3378 backer
->need_revalidate
= REV_RECONFIGURE
;
3381 ofport
->peer
->peer
= NULL
;
3382 ofport
->peer
= NULL
;
3385 peer_name
= netdev_vport_patch_peer(ofport
->up
.netdev
);
3390 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
3391 struct ofport
*peer_ofport
;
3392 struct ofport_dpif
*peer
;
3395 if (ofproto
->backer
!= backer
) {
3399 peer_ofport
= shash_find_data(&ofproto
->up
.port_by_name
, peer_name
);
3404 peer
= ofport_dpif_cast(peer_ofport
);
3405 peer_peer
= netdev_vport_patch_peer(peer
->up
.netdev
);
3406 if (peer_peer
&& !strcmp(netdev_get_name(ofport
->up
.netdev
),
3408 ofport
->peer
= peer
;
3409 ofport
->peer
->peer
= ofport
;
3419 port_run(struct ofport_dpif
*ofport
)
3421 long long int carrier_seq
= netdev_get_carrier_resets(ofport
->up
.netdev
);
3422 bool carrier_changed
= carrier_seq
!= ofport
->carrier_seq
;
3423 bool enable
= netdev_get_carrier(ofport
->up
.netdev
);
3424 bool cfm_enable
= false;
3425 bool bfd_enable
= false;
3427 ofport
->carrier_seq
= carrier_seq
;
3430 int cfm_opup
= cfm_get_opup(ofport
->cfm
);
3432 cfm_enable
= !cfm_get_fault(ofport
->cfm
);
3434 if (cfm_opup
>= 0) {
3435 cfm_enable
= cfm_enable
&& cfm_opup
;
3440 bfd_enable
= bfd_forwarding(ofport
->bfd
);
3443 if (ofport
->bfd
|| ofport
->cfm
) {
3444 enable
= enable
&& (cfm_enable
|| bfd_enable
);
3447 if (ofport
->bundle
) {
3448 enable
= enable
&& lacp_slave_may_enable(ofport
->bundle
->lacp
, ofport
);
3449 if (carrier_changed
) {
3450 lacp_slave_carrier_changed(ofport
->bundle
->lacp
, ofport
);
3454 if (ofport
->may_enable
!= enable
) {
3455 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3457 ofproto
->backer
->need_revalidate
= REV_PORT_TOGGLED
;
3459 if (ofport
->rstp_port
) {
3460 rstp_port_set_mac_operational(ofport
->rstp_port
, enable
);
3464 ofport
->may_enable
= enable
;
3468 port_query_by_name(const struct ofproto
*ofproto_
, const char *devname
,
3469 struct ofproto_port
*ofproto_port
)
3471 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3472 struct dpif_port dpif_port
;
3475 if (sset_contains(&ofproto
->ghost_ports
, devname
)) {
3476 const char *type
= netdev_get_type_from_name(devname
);
3478 /* We may be called before ofproto->up.port_by_name is populated with
3479 * the appropriate ofport. For this reason, we must get the name and
3480 * type from the netdev layer directly. */
3482 const struct ofport
*ofport
;
3484 ofport
= shash_find_data(&ofproto
->up
.port_by_name
, devname
);
3485 ofproto_port
->ofp_port
= ofport
? ofport
->ofp_port
: OFPP_NONE
;
3486 ofproto_port
->name
= xstrdup(devname
);
3487 ofproto_port
->type
= xstrdup(type
);
3493 if (!sset_contains(&ofproto
->ports
, devname
)) {
3496 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
3497 devname
, &dpif_port
);
3499 ofproto_port_from_dpif_port(ofproto
, ofproto_port
, &dpif_port
);
3505 port_add(struct ofproto
*ofproto_
, struct netdev
*netdev
)
3507 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3508 const char *devname
= netdev_get_name(netdev
);
3509 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
3510 const char *dp_port_name
;
3512 if (netdev_vport_is_patch(netdev
)) {
3513 sset_add(&ofproto
->ghost_ports
, netdev_get_name(netdev
));
3517 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
3518 if (!dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
3519 odp_port_t port_no
= ODPP_NONE
;
3522 error
= dpif_port_add(ofproto
->backer
->dpif
, netdev
, &port_no
);
3526 if (netdev_get_tunnel_config(netdev
)) {
3527 simap_put(&ofproto
->backer
->tnl_backers
,
3528 dp_port_name
, odp_to_u32(port_no
));
3532 if (netdev_get_tunnel_config(netdev
)) {
3533 sset_add(&ofproto
->ghost_ports
, devname
);
3535 sset_add(&ofproto
->ports
, devname
);
3541 port_del(struct ofproto
*ofproto_
, ofp_port_t ofp_port
)
3543 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3544 struct ofport_dpif
*ofport
= ofp_port_to_ofport(ofproto
, ofp_port
);
3551 sset_find_and_delete(&ofproto
->ghost_ports
,
3552 netdev_get_name(ofport
->up
.netdev
));
3553 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3554 if (!ofport
->is_tunnel
&& !netdev_vport_is_patch(ofport
->up
.netdev
)) {
3555 error
= dpif_port_del(ofproto
->backer
->dpif
, ofport
->odp_port
);
3557 /* The caller is going to close ofport->up.netdev. If this is a
3558 * bonded port, then the bond is using that netdev, so remove it
3559 * from the bond. The client will need to reconfigure everything
3560 * after deleting ports, so then the slave will get re-added. */
3561 bundle_remove(&ofport
->up
);
3568 port_set_config(const struct ofport
*ofport_
, const struct smap
*cfg
)
3570 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3571 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3573 if (sset_contains(&ofproto
->ghost_ports
,
3574 netdev_get_name(ofport
->up
.netdev
))) {
3578 return dpif_port_set_config(ofproto
->backer
->dpif
, ofport
->odp_port
, cfg
);
3582 port_get_stats(const struct ofport
*ofport_
, struct netdev_stats
*stats
)
3584 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3587 error
= netdev_get_stats(ofport
->up
.netdev
, stats
);
3589 if (!error
&& ofport_
->ofp_port
== OFPP_LOCAL
) {
3590 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3592 ovs_mutex_lock(&ofproto
->stats_mutex
);
3593 /* ofproto->stats.tx_packets represents packets that we created
3594 * internally and sent to some port (e.g. packets sent with
3595 * ofproto_dpif_send_packet()). Account for them as if they had
3596 * come from OFPP_LOCAL and got forwarded. */
3598 if (stats
->rx_packets
!= UINT64_MAX
) {
3599 stats
->rx_packets
+= ofproto
->stats
.tx_packets
;
3602 if (stats
->rx_bytes
!= UINT64_MAX
) {
3603 stats
->rx_bytes
+= ofproto
->stats
.tx_bytes
;
3606 /* ofproto->stats.rx_packets represents packets that were received on
3607 * some port and we processed internally and dropped (e.g. STP).
3608 * Account for them as if they had been forwarded to OFPP_LOCAL. */
3610 if (stats
->tx_packets
!= UINT64_MAX
) {
3611 stats
->tx_packets
+= ofproto
->stats
.rx_packets
;
3614 if (stats
->tx_bytes
!= UINT64_MAX
) {
3615 stats
->tx_bytes
+= ofproto
->stats
.rx_bytes
;
3617 ovs_mutex_unlock(&ofproto
->stats_mutex
);
3624 port_get_lacp_stats(const struct ofport
*ofport_
, struct lacp_slave_stats
*stats
)
3626 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3627 if (ofport
->bundle
&& ofport
->bundle
->lacp
) {
3628 if (lacp_get_slave_stats(ofport
->bundle
->lacp
, ofport
, stats
)) {
3635 struct port_dump_state
{
3636 struct sset_position pos
;
3639 struct ofproto_port port
;
3644 port_dump_start(const struct ofproto
*ofproto_ OVS_UNUSED
, void **statep
)
3646 *statep
= xzalloc(sizeof(struct port_dump_state
));
3651 port_dump_next(const struct ofproto
*ofproto_
, void *state_
,
3652 struct ofproto_port
*port
)
3654 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3655 struct port_dump_state
*state
= state_
;
3656 const struct sset
*sset
;
3657 struct sset_node
*node
;
3659 if (state
->has_port
) {
3660 ofproto_port_destroy(&state
->port
);
3661 state
->has_port
= false;
3663 sset
= state
->ghost
? &ofproto
->ghost_ports
: &ofproto
->ports
;
3664 while ((node
= sset_at_position(sset
, &state
->pos
))) {
3667 error
= port_query_by_name(ofproto_
, node
->name
, &state
->port
);
3669 *port
= state
->port
;
3670 state
->has_port
= true;
3672 } else if (error
!= ENODEV
) {
3677 if (!state
->ghost
) {
3678 state
->ghost
= true;
3679 memset(&state
->pos
, 0, sizeof state
->pos
);
3680 return port_dump_next(ofproto_
, state_
, port
);
3687 port_dump_done(const struct ofproto
*ofproto_ OVS_UNUSED
, void *state_
)
3689 struct port_dump_state
*state
= state_
;
3691 if (state
->has_port
) {
3692 ofproto_port_destroy(&state
->port
);
3699 port_poll(const struct ofproto
*ofproto_
, char **devnamep
)
3701 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3703 if (ofproto
->port_poll_errno
) {
3704 int error
= ofproto
->port_poll_errno
;
3705 ofproto
->port_poll_errno
= 0;
3709 if (sset_is_empty(&ofproto
->port_poll_set
)) {
3713 *devnamep
= sset_pop(&ofproto
->port_poll_set
);
3718 port_poll_wait(const struct ofproto
*ofproto_
)
3720 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3721 dpif_port_poll_wait(ofproto
->backer
->dpif
);
3725 port_is_lacp_current(const struct ofport
*ofport_
)
3727 const struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3728 return (ofport
->bundle
&& ofport
->bundle
->lacp
3729 ? lacp_slave_is_current(ofport
->bundle
->lacp
, ofport
)
3733 /* If 'rule' is an OpenFlow rule, that has expired according to OpenFlow rules,
3734 * then delete it entirely. */
3736 rule_expire(struct rule_dpif
*rule
, long long now
)
3737 OVS_REQUIRES(ofproto_mutex
)
3739 uint16_t hard_timeout
, idle_timeout
;
3742 hard_timeout
= rule
->up
.hard_timeout
;
3743 idle_timeout
= rule
->up
.idle_timeout
;
3745 /* Has 'rule' expired? */
3747 long long int modified
;
3749 ovs_mutex_lock(&rule
->up
.mutex
);
3750 modified
= rule
->up
.modified
;
3751 ovs_mutex_unlock(&rule
->up
.mutex
);
3753 if (now
> modified
+ hard_timeout
* 1000) {
3754 reason
= OFPRR_HARD_TIMEOUT
;
3758 if (reason
< 0 && idle_timeout
) {
3761 ovs_mutex_lock(&rule
->stats_mutex
);
3762 used
= rule
->stats
.used
;
3763 ovs_mutex_unlock(&rule
->stats_mutex
);
3765 if (now
> used
+ idle_timeout
* 1000) {
3766 reason
= OFPRR_IDLE_TIMEOUT
;
3771 COVERAGE_INC(ofproto_dpif_expired
);
3772 ofproto_rule_expire(&rule
->up
, reason
);
3777 ofproto_dpif_set_packet_odp_port(const struct ofproto_dpif
*ofproto
,
3778 ofp_port_t in_port
, struct dp_packet
*packet
)
3780 if (in_port
== OFPP_NONE
) {
3781 in_port
= OFPP_LOCAL
;
3783 packet
->md
.in_port
.odp_port
= ofp_port_to_odp_port(ofproto
, in_port
);
3787 ofproto_dpif_execute_actions__(struct ofproto_dpif
*ofproto
,
3788 ovs_version_t version
, const struct flow
*flow
,
3789 struct rule_dpif
*rule
,
3790 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3791 int depth
, int resubmits
,
3792 struct dp_packet
*packet
)
3794 struct dpif_flow_stats stats
;
3795 struct xlate_out xout
;
3796 struct xlate_in xin
;
3797 struct dpif_execute execute
;
3800 ovs_assert((rule
!= NULL
) != (ofpacts
!= NULL
));
3802 dpif_flow_stats_extract(flow
, packet
, time_msec(), &stats
);
3805 rule_dpif_credit_stats(rule
, &stats
);
3808 uint64_t odp_actions_stub
[1024 / 8];
3809 struct ofpbuf odp_actions
= OFPBUF_STUB_INITIALIZER(odp_actions_stub
);
3810 xlate_in_init(&xin
, ofproto
, version
, flow
, flow
->in_port
.ofp_port
, rule
,
3811 stats
.tcp_flags
, packet
, NULL
, &odp_actions
);
3812 xin
.ofpacts
= ofpacts
;
3813 xin
.ofpacts_len
= ofpacts_len
;
3814 xin
.resubmit_stats
= &stats
;
3816 xin
.resubmits
= resubmits
;
3817 if (xlate_actions(&xin
, &xout
) != XLATE_OK
) {
3822 execute
.actions
= odp_actions
.data
;
3823 execute
.actions_len
= odp_actions
.size
;
3825 pkt_metadata_from_flow(&packet
->md
, flow
);
3826 execute
.packet
= packet
;
3827 execute
.flow
= flow
;
3828 execute
.needs_help
= (xout
.slow
& SLOW_ACTION
) != 0;
3829 execute
.probe
= false;
3832 /* Fix up in_port. */
3833 ofproto_dpif_set_packet_odp_port(ofproto
, flow
->in_port
.ofp_port
, packet
);
3835 error
= dpif_execute(ofproto
->backer
->dpif
, &execute
);
3837 xlate_out_uninit(&xout
);
3838 ofpbuf_uninit(&odp_actions
);
3843 /* Executes, within 'ofproto', the actions in 'rule' or 'ofpacts' on 'packet'.
3844 * 'flow' must reflect the data in 'packet'. */
3846 ofproto_dpif_execute_actions(struct ofproto_dpif
*ofproto
,
3847 ovs_version_t version
, const struct flow
*flow
,
3848 struct rule_dpif
*rule
,
3849 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3850 struct dp_packet
*packet
)
3852 return ofproto_dpif_execute_actions__(ofproto
, version
, flow
, rule
,
3853 ofpacts
, ofpacts_len
, 0, 0, packet
);
3857 rule_dpif_credit_stats__(struct rule_dpif
*rule
,
3858 const struct dpif_flow_stats
*stats
,
3860 OVS_REQUIRES(rule
->stats_mutex
)
3862 if (credit_counts
) {
3863 rule
->stats
.n_packets
+= stats
->n_packets
;
3864 rule
->stats
.n_bytes
+= stats
->n_bytes
;
3866 rule
->stats
.used
= MAX(rule
->stats
.used
, stats
->used
);
3870 rule_dpif_credit_stats(struct rule_dpif
*rule
,
3871 const struct dpif_flow_stats
*stats
)
3873 ovs_mutex_lock(&rule
->stats_mutex
);
3874 if (OVS_UNLIKELY(rule
->new_rule
)) {
3875 ovs_mutex_lock(&rule
->new_rule
->stats_mutex
);
3876 rule_dpif_credit_stats__(rule
->new_rule
, stats
, rule
->forward_counts
);
3877 ovs_mutex_unlock(&rule
->new_rule
->stats_mutex
);
3879 rule_dpif_credit_stats__(rule
, stats
, true);
3881 ovs_mutex_unlock(&rule
->stats_mutex
);
3884 /* Sets 'rule''s recirculation id. */
3886 rule_dpif_set_recirc_id(struct rule_dpif
*rule
, uint32_t id
)
3887 OVS_REQUIRES(rule
->up
.mutex
)
3889 ovs_assert(!rule
->recirc_id
|| rule
->recirc_id
== id
);
3890 if (rule
->recirc_id
== id
) {
3891 /* Release the new reference to the same id. */
3894 rule
->recirc_id
= id
;
3898 /* Sets 'rule''s recirculation id. */
3900 rule_set_recirc_id(struct rule
*rule_
, uint32_t id
)
3902 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3904 ovs_mutex_lock(&rule
->up
.mutex
);
3905 rule_dpif_set_recirc_id(rule
, id
);
3906 ovs_mutex_unlock(&rule
->up
.mutex
);
3910 ofproto_dpif_get_tables_version(struct ofproto_dpif
*ofproto
)
3912 ovs_version_t version
;
3914 /* Use memory_order_acquire to signify that any following memory accesses
3915 * can not be reordered to happen before this atomic read. This makes sure
3916 * all following reads relate to this or a newer version, but never to an
3918 atomic_read_explicit(&ofproto
->tables_version
, &version
,
3919 memory_order_acquire
);
3923 /* The returned rule (if any) is valid at least until the next RCU quiescent
3924 * period. If the rule needs to stay around longer, the caller should take
3927 * 'flow' is non-const to allow for temporary modifications during the lookup.
3928 * Any changes are restored before returning. */
3929 static struct rule_dpif
*
3930 rule_dpif_lookup_in_table(struct ofproto_dpif
*ofproto
, ovs_version_t version
,
3931 uint8_t table_id
, struct flow
*flow
,
3932 struct flow_wildcards
*wc
)
3934 struct classifier
*cls
= &ofproto
->up
.tables
[table_id
].cls
;
3935 return rule_dpif_cast(rule_from_cls_rule(classifier_lookup(cls
, version
,
3940 ofproto_dpif_credit_table_stats(struct ofproto_dpif
*ofproto
, uint8_t table_id
,
3941 uint64_t n_matches
, uint64_t n_misses
)
3943 struct oftable
*tbl
= &ofproto
->up
.tables
[table_id
];
3947 atomic_add_relaxed(&tbl
->n_matched
, n_matches
, &orig
);
3950 atomic_add_relaxed(&tbl
->n_missed
, n_misses
, &orig
);
3954 /* Look up 'flow' in 'ofproto''s classifier version 'version', starting from
3955 * table '*table_id'. Returns the rule that was found, which may be one of the
3956 * special rules according to packet miss hadling. If 'may_packet_in' is
3957 * false, returning of the miss_rule (which issues packet ins for the
3958 * controller) is avoided. Updates 'wc', if nonnull, to reflect the fields
3959 * that were used during the lookup.
3961 * If 'honor_table_miss' is true, the first lookup occurs in '*table_id', but
3962 * if none is found then the table miss configuration for that table is
3963 * honored, which can result in additional lookups in other OpenFlow tables.
3964 * In this case the function updates '*table_id' to reflect the final OpenFlow
3965 * table that was searched.
3967 * If 'honor_table_miss' is false, then only one table lookup occurs, in
3970 * The rule is returned in '*rule', which is valid at least until the next
3971 * RCU quiescent period. If the '*rule' needs to stay around longer, the
3972 * caller must take a reference.
3974 * 'in_port' allows the lookup to take place as if the in port had the value
3975 * 'in_port'. This is needed for resubmit action support.
3977 * 'flow' is non-const to allow for temporary modifications during the lookup.
3978 * Any changes are restored before returning. */
3980 rule_dpif_lookup_from_table(struct ofproto_dpif
*ofproto
,
3981 ovs_version_t version
, struct flow
*flow
,
3982 struct flow_wildcards
*wc
,
3983 const struct dpif_flow_stats
*stats
,
3984 uint8_t *table_id
, ofp_port_t in_port
,
3985 bool may_packet_in
, bool honor_table_miss
,
3986 struct xlate_cache
*xcache
)
3988 ovs_be16 old_tp_src
= flow
->tp_src
, old_tp_dst
= flow
->tp_dst
;
3989 ofp_port_t old_in_port
= flow
->in_port
.ofp_port
;
3990 enum ofputil_table_miss miss_config
;
3991 struct rule_dpif
*rule
;
3994 /* We always unwildcard nw_frag (for IP), so they
3995 * need not be unwildcarded here. */
3996 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
3997 && ofproto
->up
.frag_handling
!= OFPUTIL_FRAG_NX_MATCH
) {
3998 if (ofproto
->up
.frag_handling
== OFPUTIL_FRAG_NORMAL
) {
3999 /* We must pretend that transport ports are unavailable. */
4000 flow
->tp_src
= htons(0);
4001 flow
->tp_dst
= htons(0);
4003 /* Must be OFPUTIL_FRAG_DROP (we don't have OFPUTIL_FRAG_REASM).
4004 * Use the drop_frags_rule (which cannot disappear). */
4005 rule
= ofproto
->drop_frags_rule
;
4007 struct oftable
*tbl
= &ofproto
->up
.tables
[*table_id
];
4010 atomic_add_relaxed(&tbl
->n_matched
, stats
->n_packets
, &orig
);
4013 struct xc_entry
*entry
;
4015 entry
= xlate_cache_add_entry(xcache
, XC_TABLE
);
4016 entry
->table
.ofproto
= ofproto
;
4017 entry
->table
.id
= *table_id
;
4018 entry
->table
.match
= true;
4024 /* Look up a flow with 'in_port' as the input port. Then restore the
4025 * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
4026 * have surprising behavior). */
4027 flow
->in_port
.ofp_port
= in_port
;
4029 /* Our current implementation depends on n_tables == N_TABLES, and
4030 * TBL_INTERNAL being the last table. */
4031 BUILD_ASSERT_DECL(N_TABLES
== TBL_INTERNAL
+ 1);
4033 miss_config
= OFPUTIL_TABLE_MISS_CONTINUE
;
4035 for (next_id
= *table_id
;
4036 next_id
< ofproto
->up
.n_tables
;
4037 next_id
++, next_id
+= (next_id
== TBL_INTERNAL
))
4039 *table_id
= next_id
;
4040 rule
= rule_dpif_lookup_in_table(ofproto
, version
, next_id
, flow
, wc
);
4042 struct oftable
*tbl
= &ofproto
->up
.tables
[next_id
];
4045 atomic_add_relaxed(rule
? &tbl
->n_matched
: &tbl
->n_missed
,
4046 stats
->n_packets
, &orig
);
4049 struct xc_entry
*entry
;
4051 entry
= xlate_cache_add_entry(xcache
, XC_TABLE
);
4052 entry
->table
.ofproto
= ofproto
;
4053 entry
->table
.id
= next_id
;
4054 entry
->table
.match
= (rule
!= NULL
);
4057 goto out
; /* Match. */
4059 if (honor_table_miss
) {
4060 miss_config
= ofproto_table_get_miss_config(&ofproto
->up
,
4062 if (miss_config
== OFPUTIL_TABLE_MISS_CONTINUE
) {
4069 rule
= ofproto
->no_packet_in_rule
;
4070 if (may_packet_in
) {
4071 if (miss_config
== OFPUTIL_TABLE_MISS_CONTINUE
4072 || miss_config
== OFPUTIL_TABLE_MISS_CONTROLLER
) {
4073 struct ofport_dpif
*port
;
4075 port
= ofp_port_to_ofport(ofproto
, old_in_port
);
4077 VLOG_WARN_RL(&rl
, "packet-in on unknown OpenFlow port %"PRIu32
,
4079 } else if (!(port
->up
.pp
.config
& OFPUTIL_PC_NO_PACKET_IN
)) {
4080 rule
= ofproto
->miss_rule
;
4082 } else if (miss_config
== OFPUTIL_TABLE_MISS_DEFAULT
&&
4083 connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
)) {
4084 rule
= ofproto
->miss_rule
;
4088 /* Restore port numbers, as they may have been modified above. */
4089 flow
->tp_src
= old_tp_src
;
4090 flow
->tp_dst
= old_tp_dst
;
4091 /* Restore the old in port. */
4092 flow
->in_port
.ofp_port
= old_in_port
;
4097 static struct rule_dpif
*rule_dpif_cast(const struct rule
*rule
)
4099 return rule
? CONTAINER_OF(rule
, struct rule_dpif
, up
) : NULL
;
4102 static struct rule
*
4105 struct rule_dpif
*rule
= xzalloc(sizeof *rule
);
4110 rule_dealloc(struct rule
*rule_
)
4112 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4117 check_mask(struct ofproto_dpif
*ofproto
, const struct miniflow
*flow
)
4119 const struct odp_support
*support
;
4120 uint16_t ct_state
, ct_zone
;
4124 support
= &ofproto
->backer
->support
.odp
;
4125 ct_state
= MINIFLOW_GET_U8(flow
, ct_state
);
4126 if (support
->ct_state
&& support
->ct_zone
&& support
->ct_mark
4127 && support
->ct_label
&& support
->ct_state_nat
) {
4128 return ct_state
& CS_UNSUPPORTED_MASK
? OFPERR_OFPBMC_BAD_MASK
: 0;
4131 ct_zone
= MINIFLOW_GET_U16(flow
, ct_zone
);
4132 ct_mark
= MINIFLOW_GET_U32(flow
, ct_mark
);
4133 ct_label
= MINIFLOW_GET_U128(flow
, ct_label
);
4135 if ((ct_state
&& !support
->ct_state
)
4136 || (ct_state
& CS_UNSUPPORTED_MASK
)
4137 || ((ct_state
& (CS_SRC_NAT
| CS_DST_NAT
)) && !support
->ct_state_nat
)
4138 || (ct_zone
&& !support
->ct_zone
)
4139 || (ct_mark
&& !support
->ct_mark
)
4140 || (!ovs_u128_is_zero(ct_label
) && !support
->ct_label
)) {
4141 return OFPERR_OFPBMC_BAD_MASK
;
4148 report_unsupported_ct(const char *detail
)
4150 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
4151 VLOG_WARN_RL(&rl
, "Rejecting ct action because datapath does not support "
4152 "ct action%s%s (your kernel module may be out of date)",
4154 detail
? detail
: "");
4158 check_actions(const struct ofproto_dpif
*ofproto
,
4159 const struct rule_actions
*const actions
)
4161 const struct ofpact
*ofpact
;
4163 OFPACT_FOR_EACH (ofpact
, actions
->ofpacts
, actions
->ofpacts_len
) {
4164 const struct odp_support
*support
;
4165 const struct ofpact_conntrack
*ct
;
4166 const struct ofpact
*a
;
4168 if (ofpact
->type
!= OFPACT_CT
) {
4172 ct
= CONTAINER_OF(ofpact
, struct ofpact_conntrack
, ofpact
);
4173 support
= &ofproto
->backer
->support
.odp
;
4175 if (!support
->ct_state
) {
4176 report_unsupported_ct(NULL
);
4177 return OFPERR_OFPBAC_BAD_TYPE
;
4179 if ((ct
->zone_imm
|| ct
->zone_src
.field
) && !support
->ct_zone
) {
4180 report_unsupported_ct("zone");
4181 return OFPERR_OFPBAC_BAD_ARGUMENT
;
4184 OFPACT_FOR_EACH(a
, ct
->actions
, ofpact_ct_get_action_len(ct
)) {
4185 const struct mf_field
*dst
= ofpact_get_mf_dst(a
);
4187 if (a
->type
== OFPACT_NAT
&& !support
->ct_state_nat
) {
4188 /* The backer doesn't seem to support the NAT bits in
4189 * 'ct_state': assume that it doesn't support the NAT
4191 report_unsupported_ct("nat");
4192 return OFPERR_OFPBAC_BAD_TYPE
;
4194 if (dst
&& ((dst
->id
== MFF_CT_MARK
&& !support
->ct_mark
)
4195 || (dst
->id
== MFF_CT_LABEL
&& !support
->ct_label
))) {
4196 report_unsupported_ct("setting mark and/or label");
4197 return OFPERR_OFPBAC_BAD_SET_ARGUMENT
;
4206 rule_check(struct rule
*rule
)
4208 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->ofproto
);
4211 err
= check_mask(ofproto
, &rule
->cr
.match
.mask
->masks
);
4215 return check_actions(ofproto
, rule
->actions
);
4219 rule_construct(struct rule
*rule_
)
4220 OVS_NO_THREAD_SAFETY_ANALYSIS
4222 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4225 error
= rule_check(rule_
);
4230 ovs_mutex_init_adaptive(&rule
->stats_mutex
);
4231 rule
->stats
.n_packets
= 0;
4232 rule
->stats
.n_bytes
= 0;
4233 rule
->stats
.used
= rule
->up
.modified
;
4234 rule
->recirc_id
= 0;
4235 rule
->new_rule
= NULL
;
4236 rule
->forward_counts
= false;
4242 rule_insert(struct rule
*rule_
, struct rule
*old_rule_
, bool forward_counts
)
4243 OVS_REQUIRES(ofproto_mutex
)
4245 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4248 struct rule_dpif
*old_rule
= rule_dpif_cast(old_rule_
);
4250 ovs_assert(!old_rule
->new_rule
);
4252 /* Take a reference to the new rule, and refer all stats updates from
4253 * the old rule to the new rule. */
4254 ofproto_rule_ref(&rule
->up
);
4256 ovs_mutex_lock(&old_rule
->stats_mutex
);
4257 ovs_mutex_lock(&rule
->stats_mutex
);
4258 old_rule
->new_rule
= rule
; /* Forward future stats. */
4259 old_rule
->forward_counts
= forward_counts
;
4261 if (forward_counts
) {
4262 rule
->stats
= old_rule
->stats
; /* Transfer stats to the new
4265 /* Used timestamp must be forwarded whenever a rule is modified. */
4266 rule
->stats
.used
= old_rule
->stats
.used
;
4268 ovs_mutex_unlock(&rule
->stats_mutex
);
4269 ovs_mutex_unlock(&old_rule
->stats_mutex
);
4274 rule_destruct(struct rule
*rule_
)
4275 OVS_NO_THREAD_SAFETY_ANALYSIS
4277 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4279 ovs_mutex_destroy(&rule
->stats_mutex
);
4280 /* Release reference to the new rule, if any. */
4281 if (rule
->new_rule
) {
4282 ofproto_rule_unref(&rule
->new_rule
->up
);
4284 if (rule
->recirc_id
) {
4285 recirc_free_id(rule
->recirc_id
);
4290 rule_get_stats(struct rule
*rule_
, uint64_t *packets
, uint64_t *bytes
,
4291 long long int *used
)
4293 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4295 ovs_mutex_lock(&rule
->stats_mutex
);
4296 if (OVS_UNLIKELY(rule
->new_rule
)) {
4297 rule_get_stats(&rule
->new_rule
->up
, packets
, bytes
, used
);
4299 *packets
= rule
->stats
.n_packets
;
4300 *bytes
= rule
->stats
.n_bytes
;
4301 *used
= rule
->stats
.used
;
4303 ovs_mutex_unlock(&rule
->stats_mutex
);
4306 struct ofproto_dpif_packet_out
{
4307 struct xlate_cache xcache
;
4308 struct ofpbuf odp_actions
;
4309 struct recirc_refs rr
;
4314 static struct ofproto_dpif_packet_out
*
4315 ofproto_dpif_packet_out_new(void)
4317 struct ofproto_dpif_packet_out
*aux
= xmalloc(sizeof *aux
);
4318 xlate_cache_init(&aux
->xcache
);
4319 ofpbuf_init(&aux
->odp_actions
, 64);
4320 aux
->rr
= RECIRC_REFS_EMPTY_INITIALIZER
;
4321 aux
->needs_help
= false;
4327 ofproto_dpif_packet_out_delete(struct ofproto_dpif_packet_out
*aux
)
4330 xlate_cache_uninit(&aux
->xcache
);
4331 ofpbuf_uninit(&aux
->odp_actions
);
4332 recirc_refs_unref(&aux
->rr
);
4338 packet_xlate(struct ofproto
*ofproto_
, struct ofproto_packet_out
*opo
)
4339 OVS_REQUIRES(ofproto_mutex
)
4341 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4342 struct xlate_out xout
;
4343 struct xlate_in xin
;
4344 enum ofperr error
= 0;
4346 struct ofproto_dpif_packet_out
*aux
= ofproto_dpif_packet_out_new();
4348 xlate_in_init(&xin
, ofproto
, opo
->version
, opo
->flow
,
4349 opo
->flow
->in_port
.ofp_port
, NULL
, 0, opo
->packet
, NULL
,
4351 xin
.ofpacts
= opo
->ofpacts
;
4352 xin
.ofpacts_len
= opo
->ofpacts_len
;
4353 /* No learning or stats, but collect side effects to xcache. */
4354 xin
.allow_side_effects
= false;
4355 xin
.resubmit_stats
= NULL
;
4356 xin
.xcache
= &aux
->xcache
;
4358 if (xlate_actions(&xin
, &xout
) != XLATE_OK
) {
4359 error
= OFPERR_OFPFMFC_UNKNOWN
; /* Error processing actions. */
4362 /* Prepare learn actions. */
4363 struct xc_entry
*entry
;
4364 struct ofpbuf entries
= aux
->xcache
.entries
;
4366 XC_ENTRY_FOR_EACH (entry
, &entries
) {
4367 if (entry
->type
== XC_LEARN
) {
4368 struct ofproto_flow_mod
*ofm
= entry
->learn
.ofm
;
4370 error
= ofproto_flow_mod_learn_refresh(ofm
);
4374 struct rule
*rule
= ofm
->temp_rule
;
4375 ofm
->learn_adds_rule
= (rule
->state
== RULE_INITIALIZED
);
4376 if (ofm
->learn_adds_rule
) {
4377 /* If learning on a different bridge, must use its next
4378 * version number. */
4379 ofm
->version
= (rule
->ofproto
== ofproto_
)
4380 ? opo
->version
: rule
->ofproto
->tables_version
+ 1;
4381 error
= ofproto_flow_mod_learn_start(ofm
);
4390 aux
->needs_help
= (xout
.slow
& SLOW_ACTION
) != 0;
4391 recirc_refs_swap(&aux
->rr
, &xout
.recircs
); /* Hold recirc refs. */
4393 xlate_out_uninit(&xout
);
4398 xlate_out_uninit(&xout
);
4399 ofproto_dpif_packet_out_delete(aux
);
4405 packet_xlate_revert(struct ofproto
*ofproto OVS_UNUSED
,
4406 struct ofproto_packet_out
*opo
)
4407 OVS_REQUIRES(ofproto_mutex
)
4409 struct ofproto_dpif_packet_out
*aux
= opo
->aux
;
4412 /* Revert the learned flows. */
4413 struct xc_entry
*entry
;
4414 struct ofpbuf entries
= aux
->xcache
.entries
;
4416 XC_ENTRY_FOR_EACH (entry
, &entries
) {
4417 if (entry
->type
== XC_LEARN
&& entry
->learn
.ofm
->learn_adds_rule
) {
4418 ofproto_flow_mod_learn_revert(entry
->learn
.ofm
);
4422 ofproto_dpif_packet_out_delete(aux
);
4426 /* Push stats and perform side effects of flow translation. */
4428 ofproto_dpif_xcache_execute(struct ofproto_dpif
*ofproto
,
4429 struct xlate_cache
*xcache
,
4430 const struct dpif_flow_stats
*stats
)
4431 OVS_REQUIRES(ofproto_mutex
)
4433 struct xc_entry
*entry
;
4434 struct ofpbuf entries
= xcache
->entries
;
4436 XC_ENTRY_FOR_EACH (entry
, &entries
) {
4437 switch (entry
->type
) {
4439 /* Finish the learned flows. */
4440 if (entry
->learn
.ofm
->learn_adds_rule
) {
4441 ofproto_flow_mod_learn_finish(entry
->learn
.ofm
, &ofproto
->up
);
4444 case XC_FIN_TIMEOUT
:
4445 if (stats
->tcp_flags
& (TCP_FIN
| TCP_RST
)) {
4446 /* 'ofproto_mutex' already held */
4447 ofproto_rule_reduce_timeouts__(&entry
->fin
.rule
->up
,
4452 /* All the rest can be dealt with by the xlate layer. */
4463 xlate_push_stats_entry(entry
, stats
);
4472 packet_execute(struct ofproto
*ofproto_
, struct ofproto_packet_out
*opo
)
4473 OVS_REQUIRES(ofproto_mutex
)
4475 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4476 struct dpif_flow_stats stats
;
4477 struct dpif_execute execute
;
4479 struct ofproto_dpif_packet_out
*aux
= opo
->aux
;
4482 /* Run the side effects from the xcache. */
4483 dpif_flow_stats_extract(opo
->flow
, opo
->packet
, time_msec(), &stats
);
4484 ofproto_dpif_xcache_execute(ofproto
, &aux
->xcache
, &stats
);
4486 execute
.actions
= aux
->odp_actions
.data
;
4487 execute
.actions_len
= aux
->odp_actions
.size
;
4489 pkt_metadata_from_flow(&opo
->packet
->md
, opo
->flow
);
4490 execute
.packet
= opo
->packet
;
4491 execute
.flow
= opo
->flow
;
4492 execute
.needs_help
= aux
->needs_help
;
4493 execute
.probe
= false;
4496 /* Fix up in_port. */
4497 ofproto_dpif_set_packet_odp_port(ofproto
, opo
->flow
->in_port
.ofp_port
,
4500 dpif_execute(ofproto
->backer
->dpif
, &execute
);
4501 ofproto_dpif_packet_out_delete(aux
);
4505 static struct group_dpif
*group_dpif_cast(const struct ofgroup
*group
)
4507 return group
? CONTAINER_OF(group
, struct group_dpif
, up
) : NULL
;
4510 static struct ofgroup
*
4513 struct group_dpif
*group
= xzalloc(sizeof *group
);
4518 group_dealloc(struct ofgroup
*group_
)
4520 struct group_dpif
*group
= group_dpif_cast(group_
);
4525 group_construct_stats(struct group_dpif
*group
)
4526 OVS_REQUIRES(group
->stats_mutex
)
4528 group
->packet_count
= 0;
4529 group
->byte_count
= 0;
4531 struct ofputil_bucket
*bucket
;
4532 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
4533 bucket
->stats
.packet_count
= 0;
4534 bucket
->stats
.byte_count
= 0;
4539 group_dpif_credit_stats(struct group_dpif
*group
,
4540 struct ofputil_bucket
*bucket
,
4541 const struct dpif_flow_stats
*stats
)
4543 ovs_mutex_lock(&group
->stats_mutex
);
4544 group
->packet_count
+= stats
->n_packets
;
4545 group
->byte_count
+= stats
->n_bytes
;
4547 bucket
->stats
.packet_count
+= stats
->n_packets
;
4548 bucket
->stats
.byte_count
+= stats
->n_bytes
;
4549 } else { /* Credit to all buckets */
4550 struct ofputil_bucket
*bucket
;
4551 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
4552 bucket
->stats
.packet_count
+= stats
->n_packets
;
4553 bucket
->stats
.byte_count
+= stats
->n_bytes
;
4556 ovs_mutex_unlock(&group
->stats_mutex
);
4560 group_construct(struct ofgroup
*group_
)
4562 struct group_dpif
*group
= group_dpif_cast(group_
);
4564 ovs_mutex_init_adaptive(&group
->stats_mutex
);
4565 ovs_mutex_lock(&group
->stats_mutex
);
4566 group_construct_stats(group
);
4567 ovs_mutex_unlock(&group
->stats_mutex
);
4572 group_destruct(struct ofgroup
*group_
)
4574 struct group_dpif
*group
= group_dpif_cast(group_
);
4575 ovs_mutex_destroy(&group
->stats_mutex
);
4579 group_get_stats(const struct ofgroup
*group_
, struct ofputil_group_stats
*ogs
)
4581 struct group_dpif
*group
= group_dpif_cast(group_
);
4583 ovs_mutex_lock(&group
->stats_mutex
);
4584 ogs
->packet_count
= group
->packet_count
;
4585 ogs
->byte_count
= group
->byte_count
;
4587 struct bucket_counter
*bucket_stats
= ogs
->bucket_stats
;
4588 struct ofputil_bucket
*bucket
;
4589 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
4590 bucket_stats
->packet_count
= bucket
->stats
.packet_count
;
4591 bucket_stats
->byte_count
= bucket
->stats
.byte_count
;
4594 ovs_mutex_unlock(&group
->stats_mutex
);
4599 /* If the group exists, this function increments the groups's reference count.
4601 * Make sure to call ofproto_group_unref() after no longer needing to maintain
4602 * a reference to the group. */
4604 group_dpif_lookup(struct ofproto_dpif
*ofproto
, uint32_t group_id
,
4605 ovs_version_t version
, bool take_ref
)
4607 struct ofgroup
*ofgroup
= ofproto_group_lookup(&ofproto
->up
, group_id
,
4609 return ofgroup
? group_dpif_cast(ofgroup
) : NULL
;
4612 /* Sends 'packet' out 'ofport'. If 'port' is a tunnel and that tunnel type
4613 * supports a notion of an OAM flag, sets it if 'oam' is true.
4614 * May modify 'packet'.
4615 * Returns 0 if successful, otherwise a positive errno value. */
4617 ofproto_dpif_send_packet(const struct ofport_dpif
*ofport
, bool oam
,
4618 struct dp_packet
*packet
)
4620 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
4623 error
= xlate_send_packet(ofport
, oam
, packet
);
4625 ovs_mutex_lock(&ofproto
->stats_mutex
);
4626 ofproto
->stats
.tx_packets
++;
4627 ofproto
->stats
.tx_bytes
+= dp_packet_size(packet
);
4628 ovs_mutex_unlock(&ofproto
->stats_mutex
);
4632 /* Return the version string of the datapath that backs up
4636 get_datapath_version(const struct ofproto
*ofproto_
)
4638 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4640 return ofproto
->backer
->dp_version_string
;
4644 type_set_config(const char *type
, const struct smap
*other_config
)
4646 struct dpif_backer
*backer
;
4648 backer
= shash_find_data(&all_dpif_backers
, type
);
4650 /* This is not necessarily a problem, since backers are only
4651 * created on demand. */
4655 dpif_set_config(backer
->dpif
, other_config
);
4659 ct_flush(const struct ofproto
*ofproto_
, const uint16_t *zone
)
4661 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4663 ct_dpif_flush(ofproto
->backer
->dpif
, zone
);
4667 set_frag_handling(struct ofproto
*ofproto_
,
4668 enum ofputil_frag_handling frag_handling
)
4670 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4671 if (frag_handling
!= OFPUTIL_FRAG_REASM
) {
4672 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4680 nxt_resume(struct ofproto
*ofproto_
,
4681 const struct ofputil_packet_in_private
*pin
)
4683 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4685 /* Translate pin into datapath actions. */
4686 uint64_t odp_actions_stub
[1024 / 8];
4687 struct ofpbuf odp_actions
= OFPBUF_STUB_INITIALIZER(odp_actions_stub
);
4688 enum slow_path_reason slow
;
4689 enum ofperr error
= xlate_resume(ofproto
, pin
, &odp_actions
, &slow
);
4691 /* Steal 'pin->packet' and put it into a dp_packet. */
4692 struct dp_packet packet
;
4693 dp_packet_init(&packet
, pin
->public.packet_len
);
4694 dp_packet_put(&packet
, pin
->public.packet
, pin
->public.packet_len
);
4696 pkt_metadata_from_flow(&packet
.md
, &pin
->public.flow_metadata
.flow
);
4698 /* Fix up in_port. */
4699 ofproto_dpif_set_packet_odp_port(ofproto
,
4700 pin
->public.flow_metadata
.flow
.in_port
.ofp_port
,
4703 struct flow headers
;
4704 flow_extract(&packet
, &headers
);
4706 /* Execute the datapath actions on the packet. */
4707 struct dpif_execute execute
= {
4708 .actions
= odp_actions
.data
,
4709 .actions_len
= odp_actions
.size
,
4710 .needs_help
= (slow
& SLOW_ACTION
) != 0,
4714 dpif_execute(ofproto
->backer
->dpif
, &execute
);
4717 ofpbuf_uninit(&odp_actions
);
4718 dp_packet_uninit(&packet
);
4726 set_netflow(struct ofproto
*ofproto_
,
4727 const struct netflow_options
*netflow_options
)
4729 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4731 if (netflow_options
) {
4732 if (!ofproto
->netflow
) {
4733 ofproto
->netflow
= netflow_create();
4734 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4736 return netflow_set_options(ofproto
->netflow
, netflow_options
);
4737 } else if (ofproto
->netflow
) {
4738 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4739 netflow_unref(ofproto
->netflow
);
4740 ofproto
->netflow
= NULL
;
4747 get_netflow_ids(const struct ofproto
*ofproto_
,
4748 uint8_t *engine_type
, uint8_t *engine_id
)
4750 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4752 dpif_get_netflow_ids(ofproto
->backer
->dpif
, engine_type
, engine_id
);
4755 struct ofproto_dpif
*
4756 ofproto_dpif_lookup(const char *name
)
4758 struct ofproto_dpif
*ofproto
;
4760 HMAP_FOR_EACH_WITH_HASH (ofproto
, all_ofproto_dpifs_node
,
4761 hash_string(name
, 0), &all_ofproto_dpifs
) {
4762 if (!strcmp(ofproto
->up
.name
, name
)) {
4770 ofproto_unixctl_fdb_flush(struct unixctl_conn
*conn
, int argc
,
4771 const char *argv
[], void *aux OVS_UNUSED
)
4773 struct ofproto_dpif
*ofproto
;
4776 ofproto
= ofproto_dpif_lookup(argv
[1]);
4778 unixctl_command_reply_error(conn
, "no such bridge");
4781 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4782 mac_learning_flush(ofproto
->ml
);
4783 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4785 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4786 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4787 mac_learning_flush(ofproto
->ml
);
4788 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4792 unixctl_command_reply(conn
, "table successfully flushed");
4796 ofproto_unixctl_mcast_snooping_flush(struct unixctl_conn
*conn
, int argc
,
4797 const char *argv
[], void *aux OVS_UNUSED
)
4799 struct ofproto_dpif
*ofproto
;
4802 ofproto
= ofproto_dpif_lookup(argv
[1]);
4804 unixctl_command_reply_error(conn
, "no such bridge");
4808 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4809 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4812 mcast_snooping_mdb_flush(ofproto
->ms
);
4814 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4815 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4818 mcast_snooping_mdb_flush(ofproto
->ms
);
4822 unixctl_command_reply(conn
, "table successfully flushed");
4825 static struct ofport_dpif
*
4826 ofbundle_get_a_port(const struct ofbundle
*bundle
)
4828 return CONTAINER_OF(ovs_list_front(&bundle
->ports
), struct ofport_dpif
,
4833 ofproto_unixctl_fdb_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4834 const char *argv
[], void *aux OVS_UNUSED
)
4836 struct ds ds
= DS_EMPTY_INITIALIZER
;
4837 const struct ofproto_dpif
*ofproto
;
4838 const struct mac_entry
*e
;
4840 ofproto
= ofproto_dpif_lookup(argv
[1]);
4842 unixctl_command_reply_error(conn
, "no such bridge");
4846 ds_put_cstr(&ds
, " port VLAN MAC Age\n");
4847 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
4848 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
4849 struct ofbundle
*bundle
= mac_entry_get_port(ofproto
->ml
, e
);
4850 char name
[OFP_MAX_PORT_NAME_LEN
];
4852 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4854 ds_put_format(&ds
, "%5s %4d "ETH_ADDR_FMT
" %3d\n",
4855 name
, e
->vlan
, ETH_ADDR_ARGS(e
->mac
),
4856 mac_entry_age(ofproto
->ml
, e
));
4858 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4859 unixctl_command_reply(conn
, ds_cstr(&ds
));
4864 ofproto_unixctl_mcast_snooping_show(struct unixctl_conn
*conn
,
4865 int argc OVS_UNUSED
,
4867 void *aux OVS_UNUSED
)
4869 struct ds ds
= DS_EMPTY_INITIALIZER
;
4870 const struct ofproto_dpif
*ofproto
;
4871 const struct ofbundle
*bundle
;
4872 const struct mcast_group
*grp
;
4873 struct mcast_group_bundle
*b
;
4874 struct mcast_mrouter_bundle
*mrouter
;
4876 ofproto
= ofproto_dpif_lookup(argv
[1]);
4878 unixctl_command_reply_error(conn
, "no such bridge");
4882 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4883 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4887 ds_put_cstr(&ds
, " port VLAN GROUP Age\n");
4888 ovs_rwlock_rdlock(&ofproto
->ms
->rwlock
);
4889 LIST_FOR_EACH (grp
, group_node
, &ofproto
->ms
->group_lru
) {
4890 LIST_FOR_EACH(b
, bundle_node
, &grp
->bundle_lru
) {
4891 char name
[OFP_MAX_PORT_NAME_LEN
];
4894 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4896 ds_put_format(&ds
, "%5s %4d ", name
, grp
->vlan
);
4897 ipv6_format_mapped(&grp
->addr
, &ds
);
4898 ds_put_format(&ds
, " %3d\n",
4899 mcast_bundle_age(ofproto
->ms
, b
));
4903 /* ports connected to multicast routers */
4904 LIST_FOR_EACH(mrouter
, mrouter_node
, &ofproto
->ms
->mrouter_lru
) {
4905 char name
[OFP_MAX_PORT_NAME_LEN
];
4907 bundle
= mrouter
->port
;
4908 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4910 ds_put_format(&ds
, "%5s %4d querier %3d\n",
4911 name
, mrouter
->vlan
,
4912 mcast_mrouter_age(ofproto
->ms
, mrouter
));
4914 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
4915 unixctl_command_reply(conn
, ds_cstr(&ds
));
4919 /* Store the current ofprotos in 'ofproto_shash'. Returns a sorted list
4920 * of the 'ofproto_shash' nodes. It is the responsibility of the caller
4921 * to destroy 'ofproto_shash' and free the returned value. */
4922 static const struct shash_node
**
4923 get_ofprotos(struct shash
*ofproto_shash
)
4925 const struct ofproto_dpif
*ofproto
;
4927 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4928 char *name
= xasprintf("%s@%s", ofproto
->up
.type
, ofproto
->up
.name
);
4929 shash_add_nocopy(ofproto_shash
, name
, ofproto
);
4932 return shash_sort(ofproto_shash
);
4936 ofproto_unixctl_dpif_dump_dps(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4937 const char *argv
[] OVS_UNUSED
,
4938 void *aux OVS_UNUSED
)
4940 struct ds ds
= DS_EMPTY_INITIALIZER
;
4941 struct shash ofproto_shash
;
4942 const struct shash_node
**sorted_ofprotos
;
4945 shash_init(&ofproto_shash
);
4946 sorted_ofprotos
= get_ofprotos(&ofproto_shash
);
4947 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4948 const struct shash_node
*node
= sorted_ofprotos
[i
];
4949 ds_put_format(&ds
, "%s\n", node
->name
);
4952 shash_destroy(&ofproto_shash
);
4953 free(sorted_ofprotos
);
4955 unixctl_command_reply(conn
, ds_cstr(&ds
));
4960 show_dp_feature_b(struct ds
*ds
, const char *feature
, bool b
)
4962 ds_put_format(ds
, "%s: %s\n", feature
, b
? "Yes" : "No");
4966 show_dp_feature_s(struct ds
*ds
, const char *feature
, size_t s
)
4968 ds_put_format(ds
, "%s: %"PRIuSIZE
"\n", feature
, s
);
4972 dpif_show_support(const struct dpif_backer_support
*support
, struct ds
*ds
)
4974 show_dp_feature_b(ds
, "Variable length userdata",
4975 support
->variable_length_userdata
);
4976 show_dp_feature_b(ds
, "Masked set action", support
->masked_set_action
);
4977 show_dp_feature_b(ds
, "Tunnel push pop", support
->tnl_push_pop
);
4978 show_dp_feature_b(ds
, "Ufid", support
->ufid
);
4979 show_dp_feature_b(ds
, "Trunc action", support
->trunc
);
4980 show_dp_feature_b(ds
, "Clone action", support
->clone
);
4981 show_dp_feature_s(ds
, "Max MPLS depth", support
->odp
.max_mpls_depth
);
4982 show_dp_feature_b(ds
, "Recirc", support
->odp
.recirc
);
4983 show_dp_feature_b(ds
, "CT state", support
->odp
.ct_state
);
4984 show_dp_feature_b(ds
, "CT zone", support
->odp
.ct_zone
);
4985 show_dp_feature_b(ds
, "CT mark", support
->odp
.ct_mark
);
4986 show_dp_feature_b(ds
, "CT label", support
->odp
.ct_label
);
4987 show_dp_feature_b(ds
, "CT State NAT", support
->odp
.ct_state_nat
);
4988 show_dp_feature_s(ds
, "Max sample nesting", support
->sample_nesting
);
4992 dpif_show_backer(const struct dpif_backer
*backer
, struct ds
*ds
)
4994 const struct shash_node
**ofprotos
;
4995 struct dpif_dp_stats dp_stats
;
4996 struct shash ofproto_shash
;
4999 dpif_get_dp_stats(backer
->dpif
, &dp_stats
);
5000 ds_put_format(ds
, "%s: hit:%"PRIu64
" missed:%"PRIu64
"\n",
5001 dpif_name(backer
->dpif
), dp_stats
.n_hit
, dp_stats
.n_missed
);
5003 shash_init(&ofproto_shash
);
5004 ofprotos
= get_ofprotos(&ofproto_shash
);
5005 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
5006 struct ofproto_dpif
*ofproto
= ofprotos
[i
]->data
;
5007 const struct shash_node
**ports
;
5010 if (ofproto
->backer
!= backer
) {
5014 ds_put_format(ds
, "\t%s:\n", ofproto
->up
.name
);
5016 ports
= shash_sort(&ofproto
->up
.port_by_name
);
5017 for (j
= 0; j
< shash_count(&ofproto
->up
.port_by_name
); j
++) {
5018 const struct shash_node
*node
= ports
[j
];
5019 struct ofport
*ofport
= node
->data
;
5021 odp_port_t odp_port
;
5023 ds_put_format(ds
, "\t\t%s %u/", netdev_get_name(ofport
->netdev
),
5026 odp_port
= ofp_port_to_odp_port(ofproto
, ofport
->ofp_port
);
5027 if (odp_port
!= ODPP_NONE
) {
5028 ds_put_format(ds
, "%"PRIu32
":", odp_port
);
5030 ds_put_cstr(ds
, "none:");
5033 ds_put_format(ds
, " (%s", netdev_get_type(ofport
->netdev
));
5036 if (!netdev_get_config(ofport
->netdev
, &config
)) {
5037 const struct smap_node
**nodes
;
5040 nodes
= smap_sort(&config
);
5041 for (i
= 0; i
< smap_count(&config
); i
++) {
5042 const struct smap_node
*node
= nodes
[i
];
5043 ds_put_format(ds
, "%c %s=%s", i
? ',' : ':',
5044 node
->key
, node
->value
);
5048 smap_destroy(&config
);
5050 ds_put_char(ds
, ')');
5051 ds_put_char(ds
, '\n');
5055 shash_destroy(&ofproto_shash
);
5060 ofproto_unixctl_dpif_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
5061 const char *argv
[] OVS_UNUSED
, void *aux OVS_UNUSED
)
5063 struct ds ds
= DS_EMPTY_INITIALIZER
;
5064 const struct shash_node
**backers
;
5067 backers
= shash_sort(&all_dpif_backers
);
5068 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
5069 dpif_show_backer(backers
[i
]->data
, &ds
);
5073 unixctl_command_reply(conn
, ds_cstr(&ds
));
5078 ofproto_unixctl_dpif_dump_flows(struct unixctl_conn
*conn
,
5079 int argc OVS_UNUSED
, const char *argv
[],
5080 void *aux OVS_UNUSED
)
5082 const struct ofproto_dpif
*ofproto
;
5084 struct ds ds
= DS_EMPTY_INITIALIZER
;
5085 bool verbosity
= false;
5087 struct dpif_port dpif_port
;
5088 struct dpif_port_dump port_dump
;
5089 struct hmap portno_names
;
5091 struct dpif_flow_dump
*flow_dump
;
5092 struct dpif_flow_dump_thread
*flow_dump_thread
;
5096 ofproto
= ofproto_dpif_lookup(argv
[argc
- 1]);
5098 unixctl_command_reply_error(conn
, "no such bridge");
5102 if (argc
> 2 && !strcmp(argv
[1], "-m")) {
5106 hmap_init(&portno_names
);
5107 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, ofproto
->backer
->dpif
) {
5108 odp_portno_names_set(&portno_names
, dpif_port
.port_no
, dpif_port
.name
);
5112 flow_dump
= dpif_flow_dump_create(ofproto
->backer
->dpif
, false);
5113 flow_dump_thread
= dpif_flow_dump_thread_create(flow_dump
);
5114 while (dpif_flow_dump_next(flow_dump_thread
, &f
, 1)) {
5117 if (odp_flow_key_to_flow(f
.key
, f
.key_len
, &flow
) == ODP_FIT_ERROR
5118 || xlate_lookup_ofproto(ofproto
->backer
, &flow
, NULL
) != ofproto
) {
5123 odp_format_ufid(&f
.ufid
, &ds
);
5124 ds_put_cstr(&ds
, " ");
5126 odp_flow_format(f
.key
, f
.key_len
, f
.mask
, f
.mask_len
,
5127 &portno_names
, &ds
, verbosity
);
5128 ds_put_cstr(&ds
, ", ");
5129 dpif_flow_stats_format(&f
.stats
, &ds
);
5130 ds_put_cstr(&ds
, ", actions:");
5131 format_odp_actions(&ds
, f
.actions
, f
.actions_len
);
5132 ds_put_char(&ds
, '\n');
5134 dpif_flow_dump_thread_destroy(flow_dump_thread
);
5135 error
= dpif_flow_dump_destroy(flow_dump
);
5139 ds_put_format(&ds
, "dpif/dump_flows failed: %s", ovs_strerror(errno
));
5140 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
5142 unixctl_command_reply(conn
, ds_cstr(&ds
));
5144 odp_portno_names_destroy(&portno_names
);
5145 hmap_destroy(&portno_names
);
5150 ofproto_revalidate_all_backers(void)
5152 const struct shash_node
**backers
;
5155 backers
= shash_sort(&all_dpif_backers
);
5156 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
5157 struct dpif_backer
*backer
= backers
[i
]->data
;
5158 backer
->need_revalidate
= REV_RECONFIGURE
;
5164 disable_tnl_push_pop(struct unixctl_conn
*conn OVS_UNUSED
, int argc OVS_UNUSED
,
5165 const char *argv
[], void *aux OVS_UNUSED
)
5167 if (!strcasecmp(argv
[1], "off")) {
5168 ofproto_use_tnl_push_pop
= false;
5169 unixctl_command_reply(conn
, "Tunnel push-pop off");
5170 ofproto_revalidate_all_backers();
5171 } else if (!strcasecmp(argv
[1], "on")) {
5172 ofproto_use_tnl_push_pop
= true;
5173 unixctl_command_reply(conn
, "Tunnel push-pop on");
5174 ofproto_revalidate_all_backers();
5176 unixctl_command_reply_error(conn
, "Invalid argument");
5181 disable_datapath_truncate(struct unixctl_conn
*conn OVS_UNUSED
,
5182 int argc OVS_UNUSED
,
5183 const char *argv
[] OVS_UNUSED
,
5184 void *aux OVS_UNUSED
)
5186 const struct shash_node
**backers
;
5189 backers
= shash_sort(&all_dpif_backers
);
5190 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
5191 struct dpif_backer
*backer
= backers
[i
]->data
;
5192 backer
->support
.trunc
= false;
5195 unixctl_command_reply(conn
, "Datapath truncate action diabled");
5199 disable_datapath_clone(struct unixctl_conn
*conn OVS_UNUSED
,
5200 int argc
, const char *argv
[],
5201 void *aux OVS_UNUSED
)
5203 struct ds ds
= DS_EMPTY_INITIALIZER
;
5204 const char *br
= argv
[argc
-1];
5205 struct ofproto_dpif
*ofproto
;
5207 ofproto
= ofproto_dpif_lookup(br
);
5209 unixctl_command_reply_error(conn
, "no such bridge");
5212 xlate_disable_dp_clone(ofproto
);
5213 udpif_flush(ofproto
->backer
->udpif
);
5214 ds_put_format(&ds
, "Datapath clone action disabled for bridge %s", br
);
5215 unixctl_command_reply(conn
, ds_cstr(&ds
));
5219 ofproto_unixctl_dpif_show_dp_features(struct unixctl_conn
*conn
,
5220 int argc
, const char *argv
[],
5221 void *aux OVS_UNUSED
)
5223 struct ds ds
= DS_EMPTY_INITIALIZER
;
5224 const char *br
= argv
[argc
-1];
5225 struct ofproto_dpif
*ofproto
= ofproto_dpif_lookup(br
);
5228 unixctl_command_reply_error(conn
, "no such bridge");
5232 dpif_show_support(&ofproto
->backer
->support
, &ds
);
5233 unixctl_command_reply(conn
, ds_cstr(&ds
));
5237 ofproto_unixctl_init(void)
5239 static bool registered
;
5245 unixctl_command_register("fdb/flush", "[bridge]", 0, 1,
5246 ofproto_unixctl_fdb_flush
, NULL
);
5247 unixctl_command_register("fdb/show", "bridge", 1, 1,
5248 ofproto_unixctl_fdb_show
, NULL
);
5249 unixctl_command_register("mdb/flush", "[bridge]", 0, 1,
5250 ofproto_unixctl_mcast_snooping_flush
, NULL
);
5251 unixctl_command_register("mdb/show", "bridge", 1, 1,
5252 ofproto_unixctl_mcast_snooping_show
, NULL
);
5253 unixctl_command_register("dpif/dump-dps", "", 0, 0,
5254 ofproto_unixctl_dpif_dump_dps
, NULL
);
5255 unixctl_command_register("dpif/show", "", 0, 0, ofproto_unixctl_dpif_show
,
5257 unixctl_command_register("dpif/show-dp-features", "bridge", 1, 1,
5258 ofproto_unixctl_dpif_show_dp_features
, NULL
);
5259 unixctl_command_register("dpif/dump-flows", "[-m] bridge", 1, 2,
5260 ofproto_unixctl_dpif_dump_flows
, NULL
);
5262 unixctl_command_register("ofproto/tnl-push-pop", "[on]|[off]", 1, 1,
5263 disable_tnl_push_pop
, NULL
);
5265 unixctl_command_register("dpif/disable-truncate", "", 0, 0,
5266 disable_datapath_truncate
, NULL
);
5268 unixctl_command_register("dpif/disable-dp-clone", "bridge", 1, 1,
5269 disable_datapath_clone
, NULL
);
5273 ofp_port_to_odp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
5275 const struct ofport_dpif
*ofport
= ofp_port_to_ofport(ofproto
, ofp_port
);
5276 return ofport
? ofport
->odp_port
: ODPP_NONE
;
5279 struct ofport_dpif
*
5280 odp_port_to_ofport(const struct dpif_backer
*backer
, odp_port_t odp_port
)
5282 struct ofport_dpif
*port
;
5284 ovs_rwlock_rdlock(&backer
->odp_to_ofport_lock
);
5285 HMAP_FOR_EACH_IN_BUCKET (port
, odp_port_node
, hash_odp_port(odp_port
),
5286 &backer
->odp_to_ofport_map
) {
5287 if (port
->odp_port
== odp_port
) {
5288 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5293 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5298 odp_port_to_ofp_port(const struct ofproto_dpif
*ofproto
, odp_port_t odp_port
)
5300 struct ofport_dpif
*port
;
5302 port
= odp_port_to_ofport(ofproto
->backer
, odp_port
);
5303 if (port
&& &ofproto
->up
== port
->up
.ofproto
) {
5304 return port
->up
.ofp_port
;
5311 ofproto_dpif_add_internal_flow(struct ofproto_dpif
*ofproto
,
5312 const struct match
*match
, int priority
,
5313 uint16_t idle_timeout
,
5314 const struct ofpbuf
*ofpacts
,
5315 struct rule
**rulep
)
5317 struct ofputil_flow_mod fm
;
5318 struct rule_dpif
*rule
;
5321 fm
= (struct ofputil_flow_mod
) {
5322 .buffer_id
= UINT32_MAX
,
5324 .priority
= priority
,
5325 .table_id
= TBL_INTERNAL
,
5326 .command
= OFPFC_ADD
,
5327 .idle_timeout
= idle_timeout
,
5328 .flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
,
5329 .ofpacts
= ofpacts
->data
,
5330 .ofpacts_len
= ofpacts
->size
,
5333 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5335 VLOG_ERR_RL(&rl
, "failed to add internal flow (%s)",
5336 ofperr_to_string(error
));
5341 rule
= rule_dpif_lookup_in_table(ofproto
,
5342 ofproto_dpif_get_tables_version(ofproto
),
5343 TBL_INTERNAL
, &fm
.match
.flow
,
5354 ofproto_dpif_delete_internal_flow(struct ofproto_dpif
*ofproto
,
5355 struct match
*match
, int priority
)
5357 struct ofputil_flow_mod fm
;
5360 fm
= (struct ofputil_flow_mod
) {
5361 .buffer_id
= UINT32_MAX
,
5363 .priority
= priority
,
5364 .table_id
= TBL_INTERNAL
,
5365 .out_port
= OFPP_ANY
,
5366 .out_group
= OFPG_ANY
,
5367 .flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
,
5368 .command
= OFPFC_DELETE_STRICT
,
5371 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5373 VLOG_ERR_RL(&rl
, "failed to delete internal flow (%s)",
5374 ofperr_to_string(error
));
5382 meter_get_features(const struct ofproto
*ofproto_
,
5383 struct ofputil_meter_features
*features
)
5385 const struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
5387 dpif_meter_get_features(ofproto
->backer
->dpif
, features
);
5391 meter_set(struct ofproto
*ofproto_
, ofproto_meter_id
*meter_id
,
5392 struct ofputil_meter_config
*config
)
5394 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
5396 switch (dpif_meter_set(ofproto
->backer
->dpif
, meter_id
, config
)) {
5399 case EFBIG
: /* meter_id out of range */
5400 case ENOMEM
: /* Cannot allocate meter */
5401 return OFPERR_OFPMMFC_OUT_OF_METERS
;
5402 case EBADF
: /* Unsupported flags */
5403 return OFPERR_OFPMMFC_BAD_FLAGS
;
5404 case EINVAL
: /* Too many bands */
5405 return OFPERR_OFPMMFC_OUT_OF_BANDS
;
5406 case ENODEV
: /* Unsupported band type */
5407 return OFPERR_OFPMMFC_BAD_BAND
;
5409 return OFPERR_OFPMMFC_UNKNOWN
;
5414 meter_get(const struct ofproto
*ofproto_
, ofproto_meter_id meter_id
,
5415 struct ofputil_meter_stats
*stats
, uint16_t n_bands
)
5417 const struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
5419 if (!dpif_meter_get(ofproto
->backer
->dpif
, meter_id
, stats
, n_bands
)) {
5422 return OFPERR_OFPMMFC_UNKNOWN_METER
;
5426 meter_del(struct ofproto
*ofproto_
, ofproto_meter_id meter_id
)
5428 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
5430 dpif_meter_del(ofproto
->backer
->dpif
, meter_id
, NULL
, 0);
5433 const struct ofproto_class ofproto_dpif_class
= {
5447 NULL
, /* get_memory_usage. */
5448 type_get_memory_usage
,
5468 port_is_lacp_current
,
5469 port_get_lacp_stats
,
5470 NULL
, /* rule_choose_table */
5474 NULL
, /* rule_delete */
5479 packet_xlate_revert
,
5497 aa_vlan_get_queue_size
,
5504 get_stp_port_status
,
5509 get_rstp_port_status
,
5516 is_mirror_output_bundle
,
5517 forward_bpdu_changed
,
5518 set_mac_table_config
,
5520 set_mcast_snooping_port
,
5525 group_alloc
, /* group_alloc */
5526 group_construct
, /* group_construct */
5527 group_destruct
, /* group_destruct */
5528 group_dealloc
, /* group_dealloc */
5529 NULL
, /* group_modify */
5530 group_get_stats
, /* group_get_stats */
5531 get_datapath_version
, /* get_datapath_version */
5533 ct_flush
, /* ct_flush */