2 * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016 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 int vlan
; /* -1=trunk port, else a 12-bit VLAN ID. */
98 unsigned long *trunks
; /* Bitmap of trunked VLANs, if 'vlan' == -1.
99 * NULL if all VLANs are trunked. */
100 struct lacp
*lacp
; /* LACP if LACP is enabled, otherwise NULL. */
101 struct bond
*bond
; /* Nonnull iff more than one port. */
102 bool use_priority_tags
; /* Use 802.1p tag for frames in VLAN 0? */
104 bool protected; /* Protected port mode */
107 bool floodable
; /* True if no port has OFPUTIL_PC_NO_FLOOD set. */
110 static void bundle_remove(struct ofport
*);
111 static void bundle_update(struct ofbundle
*);
112 static void bundle_destroy(struct ofbundle
*);
113 static void bundle_del_port(struct ofport_dpif
*);
114 static void bundle_run(struct ofbundle
*);
115 static void bundle_wait(struct ofbundle
*);
116 static void bundle_flush_macs(struct ofbundle
*, bool);
117 static void bundle_move(struct ofbundle
*, struct ofbundle
*);
119 static void stp_run(struct ofproto_dpif
*ofproto
);
120 static void stp_wait(struct ofproto_dpif
*ofproto
);
121 static int set_stp_port(struct ofport
*,
122 const struct ofproto_port_stp_settings
*);
124 static void rstp_run(struct ofproto_dpif
*ofproto
);
125 static void set_rstp_port(struct ofport
*,
126 const struct ofproto_port_rstp_settings
*);
129 struct hmap_node odp_port_node
; /* In dpif_backer's "odp_to_ofport_map". */
133 struct ofbundle
*bundle
; /* Bundle that contains this port, if any. */
134 struct ovs_list bundle_node
;/* In struct ofbundle's "ports" list. */
135 struct cfm
*cfm
; /* Connectivity Fault Management, if any. */
136 struct bfd
*bfd
; /* BFD, if any. */
137 struct lldp
*lldp
; /* lldp, if any. */
138 bool may_enable
; /* May be enabled in bonds. */
139 bool is_tunnel
; /* This port is a tunnel. */
140 bool is_layer3
; /* This is a layer 3 port. */
141 long long int carrier_seq
; /* Carrier status changes. */
142 struct ofport_dpif
*peer
; /* Peer if patch port. */
145 struct stp_port
*stp_port
; /* Spanning Tree Protocol, if any. */
146 enum stp_state stp_state
; /* Always STP_DISABLED if STP not in use. */
147 long long int stp_state_entered
;
149 /* Rapid Spanning Tree. */
150 struct rstp_port
*rstp_port
; /* Rapid Spanning Tree Protocol, if any. */
151 enum rstp_state rstp_state
; /* Always RSTP_DISABLED if RSTP not in use. */
153 /* Queue to DSCP mapping. */
154 struct ofproto_port_queue
*qdscp
;
158 static odp_port_t
ofp_port_to_odp_port(const struct ofproto_dpif
*,
161 static ofp_port_t
odp_port_to_ofp_port(const struct ofproto_dpif
*,
164 static struct ofport_dpif
*
165 ofport_dpif_cast(const struct ofport
*ofport
)
167 return ofport
? CONTAINER_OF(ofport
, struct ofport_dpif
, up
) : NULL
;
170 static void port_run(struct ofport_dpif
*);
171 static int set_bfd(struct ofport
*, const struct smap
*);
172 static int set_cfm(struct ofport
*, const struct cfm_settings
*);
173 static int set_lldp(struct ofport
*ofport_
, const struct smap
*cfg
);
174 static void ofport_update_peer(struct ofport_dpif
*);
176 COVERAGE_DEFINE(rev_reconfigure
);
177 COVERAGE_DEFINE(rev_stp
);
178 COVERAGE_DEFINE(rev_rstp
);
179 COVERAGE_DEFINE(rev_bond
);
180 COVERAGE_DEFINE(rev_port_toggled
);
181 COVERAGE_DEFINE(rev_flow_table
);
182 COVERAGE_DEFINE(rev_mac_learning
);
183 COVERAGE_DEFINE(rev_mcast_snooping
);
185 /* All existing ofproto_backer instances, indexed by ofproto->up.type. */
186 struct shash all_dpif_backers
= SHASH_INITIALIZER(&all_dpif_backers
);
188 /* All existing ofproto_dpif instances, indexed by ->up.name. */
189 struct hmap all_ofproto_dpifs
= HMAP_INITIALIZER(&all_ofproto_dpifs
);
191 static bool ofproto_use_tnl_push_pop
= true;
192 static void ofproto_unixctl_init(void);
194 static inline struct ofproto_dpif
*
195 ofproto_dpif_cast(const struct ofproto
*ofproto
)
197 ovs_assert(ofproto
->ofproto_class
== &ofproto_dpif_class
);
198 return CONTAINER_OF(ofproto
, struct ofproto_dpif
, up
);
201 /* Global variables. */
202 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
204 /* Initial mappings of port to bridge mappings. */
205 static struct shash init_ofp_ports
= SHASH_INITIALIZER(&init_ofp_ports
);
207 /* Initialize 'ofm' for a learn action. If the rule already existed, reference
208 * to that rule is taken, otherwise a new rule is created. 'ofm' keeps the
209 * rule reference in both cases. */
211 ofproto_dpif_flow_mod_init_for_learn(struct ofproto_dpif
*ofproto
,
212 const struct ofputil_flow_mod
*fm
,
213 struct ofproto_flow_mod
*ofm
)
215 /* This will not take the global 'ofproto_mutex'. */
216 return ofproto_flow_mod_init_for_learn(&ofproto
->up
, fm
, ofm
);
219 /* Appends 'am' to the queue of asynchronous messages to be sent to the
220 * controller. Takes ownership of 'am' and any data it points to. */
222 ofproto_dpif_send_async_msg(struct ofproto_dpif
*ofproto
,
223 struct ofproto_async_msg
*am
)
225 if (!guarded_list_push_back(&ofproto
->ams
, &am
->list_node
, 1024)) {
226 COVERAGE_INC(packet_in_overflow
);
227 ofproto_async_msg_free(am
);
230 /* Wakes up main thread for packet-in I/O. */
231 seq_change(ofproto
->ams_seq
);
234 /* Factory functions. */
237 init(const struct shash
*iface_hints
)
239 struct shash_node
*node
;
241 /* Make a local copy, since we don't own 'iface_hints' elements. */
242 SHASH_FOR_EACH(node
, iface_hints
) {
243 const struct iface_hint
*orig_hint
= node
->data
;
244 struct iface_hint
*new_hint
= xmalloc(sizeof *new_hint
);
246 new_hint
->br_name
= xstrdup(orig_hint
->br_name
);
247 new_hint
->br_type
= xstrdup(orig_hint
->br_type
);
248 new_hint
->ofp_port
= orig_hint
->ofp_port
;
250 shash_add(&init_ofp_ports
, node
->name
, new_hint
);
253 ofproto_unixctl_init();
254 ofproto_dpif_trace_init();
259 enumerate_types(struct sset
*types
)
261 dp_enumerate_types(types
);
265 enumerate_names(const char *type
, struct sset
*names
)
267 struct ofproto_dpif
*ofproto
;
270 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
271 if (strcmp(type
, ofproto
->up
.type
)) {
274 sset_add(names
, ofproto
->up
.name
);
281 del(const char *type
, const char *name
)
286 error
= dpif_open(name
, type
, &dpif
);
288 error
= dpif_delete(dpif
);
295 port_open_type(const char *datapath_type
, const char *port_type
)
297 return dpif_port_open_type(datapath_type
, port_type
);
300 /* Type functions. */
302 static void process_dpif_port_changes(struct dpif_backer
*);
303 static void process_dpif_all_ports_changed(struct dpif_backer
*);
304 static void process_dpif_port_change(struct dpif_backer
*,
305 const char *devname
);
306 static void process_dpif_port_error(struct dpif_backer
*, int error
);
308 static struct ofproto_dpif
*
309 lookup_ofproto_dpif_by_port_name(const char *name
)
311 struct ofproto_dpif
*ofproto
;
313 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
314 if (sset_contains(&ofproto
->ports
, name
)) {
323 type_run(const char *type
)
325 struct dpif_backer
*backer
;
327 backer
= shash_find_data(&all_dpif_backers
, type
);
329 /* This is not necessarily a problem, since backers are only
330 * created on demand. */
334 if (dpif_run(backer
->dpif
)) {
335 backer
->need_revalidate
= REV_RECONFIGURE
;
338 udpif_run(backer
->udpif
);
340 /* If vswitchd started with other_config:flow_restore_wait set as "true",
341 * and the configuration has now changed to "false", enable receiving
342 * packets from the datapath. */
343 if (!backer
->recv_set_enable
&& !ofproto_get_flow_restore_wait()) {
346 backer
->recv_set_enable
= true;
348 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
350 VLOG_ERR("Failed to enable receiving packets in dpif.");
353 dpif_flow_flush(backer
->dpif
);
354 backer
->need_revalidate
= REV_RECONFIGURE
;
357 if (backer
->recv_set_enable
) {
358 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
361 if (backer
->need_revalidate
) {
362 struct ofproto_dpif
*ofproto
;
363 struct simap_node
*node
;
364 struct simap tmp_backers
;
366 /* Handle tunnel garbage collection. */
367 simap_init(&tmp_backers
);
368 simap_swap(&backer
->tnl_backers
, &tmp_backers
);
370 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
371 struct ofport_dpif
*iter
;
373 if (backer
!= ofproto
->backer
) {
377 HMAP_FOR_EACH (iter
, up
.hmap_node
, &ofproto
->up
.ports
) {
378 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
381 if (!iter
->is_tunnel
) {
385 dp_port
= netdev_vport_get_dpif_port(iter
->up
.netdev
,
386 namebuf
, sizeof namebuf
);
387 node
= simap_find(&tmp_backers
, dp_port
);
389 simap_put(&backer
->tnl_backers
, dp_port
, node
->data
);
390 simap_delete(&tmp_backers
, node
);
391 node
= simap_find(&backer
->tnl_backers
, dp_port
);
393 node
= simap_find(&backer
->tnl_backers
, dp_port
);
395 odp_port_t odp_port
= ODPP_NONE
;
397 if (!dpif_port_add(backer
->dpif
, iter
->up
.netdev
,
399 simap_put(&backer
->tnl_backers
, dp_port
,
400 odp_to_u32(odp_port
));
401 node
= simap_find(&backer
->tnl_backers
, dp_port
);
406 iter
->odp_port
= node
? u32_to_odp(node
->data
) : ODPP_NONE
;
407 if (tnl_port_reconfigure(iter
, iter
->up
.netdev
,
409 ovs_native_tunneling_is_on(ofproto
), dp_port
)) {
410 backer
->need_revalidate
= REV_RECONFIGURE
;
415 SIMAP_FOR_EACH (node
, &tmp_backers
) {
416 dpif_port_del(backer
->dpif
, u32_to_odp(node
->data
));
418 simap_destroy(&tmp_backers
);
420 switch (backer
->need_revalidate
) {
421 case REV_RECONFIGURE
: COVERAGE_INC(rev_reconfigure
); break;
422 case REV_STP
: COVERAGE_INC(rev_stp
); break;
423 case REV_RSTP
: COVERAGE_INC(rev_rstp
); break;
424 case REV_BOND
: COVERAGE_INC(rev_bond
); break;
425 case REV_PORT_TOGGLED
: COVERAGE_INC(rev_port_toggled
); break;
426 case REV_FLOW_TABLE
: COVERAGE_INC(rev_flow_table
); break;
427 case REV_MAC_LEARNING
: COVERAGE_INC(rev_mac_learning
); break;
428 case REV_MCAST_SNOOPING
: COVERAGE_INC(rev_mcast_snooping
); break;
430 backer
->need_revalidate
= 0;
433 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
434 struct ofport_dpif
*ofport
;
435 struct ofbundle
*bundle
;
437 if (ofproto
->backer
!= backer
) {
441 xlate_ofproto_set(ofproto
, ofproto
->up
.name
,
442 ofproto
->backer
->dpif
, ofproto
->ml
,
443 ofproto
->stp
, ofproto
->rstp
, ofproto
->ms
,
444 ofproto
->mbridge
, ofproto
->sflow
, ofproto
->ipfix
,
446 ofproto
->up
.forward_bpdu
,
447 connmgr_has_in_band(ofproto
->up
.connmgr
),
448 &ofproto
->backer
->support
);
450 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
451 xlate_bundle_set(ofproto
, bundle
, bundle
->name
,
452 bundle
->vlan_mode
, bundle
->vlan
,
453 bundle
->trunks
, bundle
->use_priority_tags
,
454 bundle
->bond
, bundle
->lacp
,
455 bundle
->floodable
, bundle
->protected);
458 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
459 int stp_port
= ofport
->stp_port
460 ? stp_port_no(ofport
->stp_port
)
462 xlate_ofport_set(ofproto
, ofport
->bundle
, ofport
,
463 ofport
->up
.ofp_port
, ofport
->odp_port
,
464 ofport
->up
.netdev
, ofport
->cfm
, ofport
->bfd
,
465 ofport
->lldp
, ofport
->peer
, stp_port
,
466 ofport
->rstp_port
, ofport
->qdscp
,
467 ofport
->n_qdscp
, ofport
->up
.pp
.config
,
468 ofport
->up
.pp
.state
, ofport
->is_tunnel
,
474 udpif_revalidate(backer
->udpif
);
477 process_dpif_port_changes(backer
);
482 /* Check for and handle port changes in 'backer''s dpif. */
484 process_dpif_port_changes(struct dpif_backer
*backer
)
490 error
= dpif_port_poll(backer
->dpif
, &devname
);
496 process_dpif_all_ports_changed(backer
);
500 process_dpif_port_change(backer
, devname
);
505 process_dpif_port_error(backer
, error
);
512 process_dpif_all_ports_changed(struct dpif_backer
*backer
)
514 struct ofproto_dpif
*ofproto
;
515 struct dpif_port dpif_port
;
516 struct dpif_port_dump dump
;
517 struct sset devnames
;
520 sset_init(&devnames
);
521 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
522 if (ofproto
->backer
== backer
) {
523 struct ofport
*ofport
;
525 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
526 sset_add(&devnames
, netdev_get_name(ofport
->netdev
));
530 DPIF_PORT_FOR_EACH (&dpif_port
, &dump
, backer
->dpif
) {
531 sset_add(&devnames
, dpif_port
.name
);
534 SSET_FOR_EACH (devname
, &devnames
) {
535 process_dpif_port_change(backer
, devname
);
537 sset_destroy(&devnames
);
541 process_dpif_port_change(struct dpif_backer
*backer
, const char *devname
)
543 struct ofproto_dpif
*ofproto
;
544 struct dpif_port port
;
546 /* Don't report on the datapath's device. */
547 if (!strcmp(devname
, dpif_base_name(backer
->dpif
))) {
551 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
,
552 &all_ofproto_dpifs
) {
553 if (simap_contains(&ofproto
->backer
->tnl_backers
, devname
)) {
558 ofproto
= lookup_ofproto_dpif_by_port_name(devname
);
559 if (dpif_port_query_by_name(backer
->dpif
, devname
, &port
)) {
560 /* The port was removed. If we know the datapath,
561 * report it through poll_set(). If we don't, it may be
562 * notifying us of a removal we initiated, so ignore it.
563 * If there's a pending ENOBUFS, let it stand, since
564 * everything will be reevaluated. */
565 if (ofproto
&& ofproto
->port_poll_errno
!= ENOBUFS
) {
566 sset_add(&ofproto
->port_poll_set
, devname
);
567 ofproto
->port_poll_errno
= 0;
569 } else if (!ofproto
) {
570 /* The port was added, but we don't know with which
571 * ofproto we should associate it. Delete it. */
572 dpif_port_del(backer
->dpif
, port
.port_no
);
574 struct ofport_dpif
*ofport
;
576 ofport
= ofport_dpif_cast(shash_find_data(
577 &ofproto
->up
.port_by_name
, devname
));
579 && ofport
->odp_port
!= port
.port_no
580 && !odp_port_to_ofport(backer
, port
.port_no
))
582 /* 'ofport''s datapath port number has changed from
583 * 'ofport->odp_port' to 'port.port_no'. Update our internal data
584 * structures to match. */
585 ovs_rwlock_wrlock(&backer
->odp_to_ofport_lock
);
586 hmap_remove(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
);
587 ofport
->odp_port
= port
.port_no
;
588 hmap_insert(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
,
589 hash_odp_port(port
.port_no
));
590 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
591 backer
->need_revalidate
= REV_RECONFIGURE
;
594 dpif_port_destroy(&port
);
597 /* Propagate 'error' to all ofprotos based on 'backer'. */
599 process_dpif_port_error(struct dpif_backer
*backer
, int error
)
601 struct ofproto_dpif
*ofproto
;
603 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
604 if (ofproto
->backer
== backer
) {
605 sset_clear(&ofproto
->port_poll_set
);
606 ofproto
->port_poll_errno
= error
;
612 type_wait(const char *type
)
614 struct dpif_backer
*backer
;
616 backer
= shash_find_data(&all_dpif_backers
, type
);
618 /* This is not necessarily a problem, since backers are only
619 * created on demand. */
623 dpif_wait(backer
->dpif
);
626 /* Basic life-cycle. */
628 static int add_internal_flows(struct ofproto_dpif
*);
630 static struct ofproto
*
633 struct ofproto_dpif
*ofproto
= xzalloc(sizeof *ofproto
);
638 dealloc(struct ofproto
*ofproto_
)
640 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
645 close_dpif_backer(struct dpif_backer
*backer
)
647 ovs_assert(backer
->refcount
> 0);
649 if (--backer
->refcount
) {
653 udpif_destroy(backer
->udpif
);
655 simap_destroy(&backer
->tnl_backers
);
656 ovs_rwlock_destroy(&backer
->odp_to_ofport_lock
);
657 hmap_destroy(&backer
->odp_to_ofport_map
);
658 shash_find_and_delete(&all_dpif_backers
, backer
->type
);
660 free(backer
->dp_version_string
);
661 dpif_close(backer
->dpif
);
665 /* Datapath port slated for removal from datapath. */
667 struct ovs_list list_node
;
671 static bool check_variable_length_userdata(struct dpif_backer
*backer
);
672 static void check_support(struct dpif_backer
*backer
);
675 open_dpif_backer(const char *type
, struct dpif_backer
**backerp
)
677 struct dpif_backer
*backer
;
678 struct dpif_port_dump port_dump
;
679 struct dpif_port port
;
680 struct shash_node
*node
;
681 struct ovs_list garbage_list
;
682 struct odp_garbage
*garbage
;
689 backer
= shash_find_data(&all_dpif_backers
, type
);
696 backer_name
= xasprintf("ovs-%s", type
);
698 /* Remove any existing datapaths, since we assume we're the only
699 * userspace controlling the datapath. */
701 dp_enumerate_names(type
, &names
);
702 SSET_FOR_EACH(name
, &names
) {
703 struct dpif
*old_dpif
;
705 /* Don't remove our backer if it exists. */
706 if (!strcmp(name
, backer_name
)) {
710 if (dpif_open(name
, type
, &old_dpif
)) {
711 VLOG_WARN("couldn't open old datapath %s to remove it", name
);
713 dpif_delete(old_dpif
);
714 dpif_close(old_dpif
);
717 sset_destroy(&names
);
719 backer
= xmalloc(sizeof *backer
);
721 error
= dpif_create_and_open(backer_name
, type
, &backer
->dpif
);
724 VLOG_ERR("failed to open datapath of type %s: %s", type
,
725 ovs_strerror(error
));
729 backer
->udpif
= udpif_create(backer
, backer
->dpif
);
731 backer
->type
= xstrdup(type
);
732 backer
->refcount
= 1;
733 hmap_init(&backer
->odp_to_ofport_map
);
734 ovs_rwlock_init(&backer
->odp_to_ofport_lock
);
735 backer
->need_revalidate
= 0;
736 simap_init(&backer
->tnl_backers
);
737 backer
->recv_set_enable
= !ofproto_get_flow_restore_wait();
740 if (backer
->recv_set_enable
) {
741 dpif_flow_flush(backer
->dpif
);
744 /* Loop through the ports already on the datapath and remove any
745 * that we don't need anymore. */
746 ovs_list_init(&garbage_list
);
747 dpif_port_dump_start(&port_dump
, backer
->dpif
);
748 while (dpif_port_dump_next(&port_dump
, &port
)) {
749 node
= shash_find(&init_ofp_ports
, port
.name
);
750 if (!node
&& strcmp(port
.name
, dpif_base_name(backer
->dpif
))) {
751 garbage
= xmalloc(sizeof *garbage
);
752 garbage
->odp_port
= port
.port_no
;
753 ovs_list_push_front(&garbage_list
, &garbage
->list_node
);
756 dpif_port_dump_done(&port_dump
);
758 LIST_FOR_EACH_POP (garbage
, list_node
, &garbage_list
) {
759 dpif_port_del(backer
->dpif
, garbage
->odp_port
);
763 shash_add(&all_dpif_backers
, type
, backer
);
765 check_support(backer
);
766 atomic_count_init(&backer
->tnl_count
, 0);
768 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
770 VLOG_ERR("failed to listen on datapath of type %s: %s",
771 type
, ovs_strerror(error
));
772 close_dpif_backer(backer
);
776 if (backer
->recv_set_enable
) {
777 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
780 /* This check fails if performed before udpif threads have been set,
781 * as the kernel module checks that the 'pid' in userspace action
783 backer
->support
.variable_length_userdata
784 = check_variable_length_userdata(backer
);
785 backer
->dp_version_string
= dpif_get_dp_version(backer
->dpif
);
791 ovs_native_tunneling_is_on(struct ofproto_dpif
*ofproto
)
793 return ofproto_use_tnl_push_pop
&& ofproto
->backer
->support
.tnl_push_pop
&&
794 atomic_count_get(&ofproto
->backer
->tnl_count
);
797 /* Tests whether 'backer''s datapath supports recirculation. Only newer
798 * datapaths support OVS_KEY_ATTR_RECIRC_ID in keys. We need to disable some
799 * features on older datapaths that don't support this feature.
801 * Returns false if 'backer' definitely does not support recirculation, true if
802 * it seems to support recirculation or if at least the error we get is
805 check_recirc(struct dpif_backer
*backer
)
808 struct odputil_keybuf keybuf
;
811 struct odp_flow_key_parms odp_parms
= {
818 memset(&flow
, 0, sizeof flow
);
822 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
823 odp_flow_key_from_flow(&odp_parms
, &key
);
824 enable_recirc
= dpif_probe_feature(backer
->dpif
, "recirculation", &key
,
828 VLOG_INFO("%s: Datapath supports recirculation",
829 dpif_name(backer
->dpif
));
831 VLOG_INFO("%s: Datapath does not support recirculation",
832 dpif_name(backer
->dpif
));
835 return enable_recirc
;
838 /* Tests whether 'dpif' supports unique flow ids. We can skip serializing
839 * some flow attributes for datapaths that support this feature.
841 * Returns true if 'dpif' supports UFID for flow operations.
842 * Returns false if 'dpif' does not support UFID. */
844 check_ufid(struct dpif_backer
*backer
)
847 struct odputil_keybuf keybuf
;
851 struct odp_flow_key_parms odp_parms
= {
855 memset(&flow
, 0, sizeof flow
);
856 flow
.dl_type
= htons(0x1234);
858 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
859 odp_flow_key_from_flow(&odp_parms
, &key
);
860 dpif_flow_hash(backer
->dpif
, key
.data
, key
.size
, &ufid
);
862 enable_ufid
= dpif_probe_feature(backer
->dpif
, "UFID", &key
, NULL
, &ufid
);
865 VLOG_INFO("%s: Datapath supports unique flow ids",
866 dpif_name(backer
->dpif
));
868 VLOG_INFO("%s: Datapath does not support unique flow ids",
869 dpif_name(backer
->dpif
));
874 /* Tests whether 'backer''s datapath supports variable-length
875 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. We need
876 * to disable some features on older datapaths that don't support this
879 * Returns false if 'backer' definitely does not support variable-length
880 * userdata, true if it seems to support them or if at least the error we get
883 check_variable_length_userdata(struct dpif_backer
*backer
)
885 struct eth_header
*eth
;
886 struct ofpbuf actions
;
887 struct dpif_execute execute
;
888 struct dp_packet packet
;
893 /* Compose a userspace action that will cause an ERANGE error on older
894 * datapaths that don't support variable-length userdata.
896 * We really test for using userdata longer than 8 bytes, but older
897 * datapaths accepted these, silently truncating the userdata to 8 bytes.
898 * The same older datapaths rejected userdata shorter than 8 bytes, so we
899 * test for that instead as a proxy for longer userdata support. */
900 ofpbuf_init(&actions
, 64);
901 start
= nl_msg_start_nested(&actions
, OVS_ACTION_ATTR_USERSPACE
);
902 nl_msg_put_u32(&actions
, OVS_USERSPACE_ATTR_PID
,
903 dpif_port_get_pid(backer
->dpif
, ODPP_NONE
, 0));
904 nl_msg_put_unspec_zero(&actions
, OVS_USERSPACE_ATTR_USERDATA
, 4);
905 nl_msg_end_nested(&actions
, start
);
907 /* Compose a dummy ethernet packet. */
908 dp_packet_init(&packet
, ETH_HEADER_LEN
);
909 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
910 eth
->eth_type
= htons(0x1234);
912 flow_extract(&packet
, &flow
);
914 /* Execute the actions. On older datapaths this fails with ERANGE, on
915 * newer datapaths it succeeds. */
916 execute
.actions
= actions
.data
;
917 execute
.actions_len
= actions
.size
;
918 execute
.packet
= &packet
;
919 execute
.flow
= &flow
;
920 execute
.needs_help
= false;
921 execute
.probe
= true;
924 error
= dpif_execute(backer
->dpif
, &execute
);
926 dp_packet_uninit(&packet
);
927 ofpbuf_uninit(&actions
);
934 /* Variable-length userdata is not supported. */
935 VLOG_WARN("%s: datapath does not support variable-length userdata "
936 "feature (needs Linux 3.10+ or kernel module from OVS "
937 "1..11+). The NXAST_SAMPLE action will be ignored.",
938 dpif_name(backer
->dpif
));
942 /* Something odd happened. We're not sure whether variable-length
943 * userdata is supported. Default to "yes". */
944 VLOG_WARN("%s: variable-length userdata feature probe failed (%s)",
945 dpif_name(backer
->dpif
), ovs_strerror(error
));
950 /* Tests the MPLS label stack depth supported by 'backer''s datapath.
952 * Returns the number of elements in a struct flow's mpls_lse field
953 * if the datapath supports at least that many entries in an
955 * Otherwise returns the number of MPLS push actions supported by
958 check_max_mpls_depth(struct dpif_backer
*backer
)
963 for (n
= 0; n
< FLOW_MAX_MPLS_LABELS
; n
++) {
964 struct odputil_keybuf keybuf
;
966 struct odp_flow_key_parms odp_parms
= {
970 memset(&flow
, 0, sizeof flow
);
971 flow
.dl_type
= htons(ETH_TYPE_MPLS
);
972 flow_set_mpls_bos(&flow
, n
, 1);
974 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
975 odp_flow_key_from_flow(&odp_parms
, &key
);
976 if (!dpif_probe_feature(backer
->dpif
, "MPLS", &key
, NULL
, NULL
)) {
981 VLOG_INFO("%s: MPLS label stack length probed as %d",
982 dpif_name(backer
->dpif
), n
);
987 add_sample_actions(struct ofpbuf
*actions
, int nesting
)
990 nl_msg_put_odp_port(actions
, OVS_ACTION_ATTR_OUTPUT
, u32_to_odp(1));
994 size_t start
, actions_start
;
996 start
= nl_msg_start_nested(actions
, OVS_ACTION_ATTR_SAMPLE
);
997 actions_start
= nl_msg_start_nested(actions
, OVS_SAMPLE_ATTR_ACTIONS
);
998 add_sample_actions(actions
, nesting
- 1);
999 nl_msg_end_nested(actions
, actions_start
);
1000 nl_msg_put_u32(actions
, OVS_SAMPLE_ATTR_PROBABILITY
, UINT32_MAX
);
1001 nl_msg_end_nested(actions
, start
);
1004 /* Tests the nested sample actions levels supported by 'backer''s datapath.
1006 * Returns the number of nested sample actions accepted by the datapath. */
1008 check_max_sample_nesting(struct dpif_backer
*backer
)
1010 struct odputil_keybuf keybuf
;
1015 struct odp_flow_key_parms odp_parms
= {
1019 memset(&flow
, 0, sizeof flow
);
1020 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1021 odp_flow_key_from_flow(&odp_parms
, &key
);
1023 /* OVS datapath has always supported at least 3 nested levels. */
1024 for (n
= 3; n
< FLOW_MAX_SAMPLE_NESTING
; n
++) {
1025 struct ofpbuf actions
;
1028 ofpbuf_init(&actions
, 300);
1029 add_sample_actions(&actions
, n
);
1030 ok
= dpif_probe_feature(backer
->dpif
, "Sample action nesting", &key
,
1032 ofpbuf_uninit(&actions
);
1038 VLOG_INFO("%s: Max sample nesting level probed as %d",
1039 dpif_name(backer
->dpif
), n
);
1043 /* Tests whether 'backer''s datapath supports masked data in
1044 * OVS_ACTION_ATTR_SET actions. We need to disable some features on older
1045 * datapaths that don't support this feature. */
1047 check_masked_set_action(struct dpif_backer
*backer
)
1049 struct eth_header
*eth
;
1050 struct ofpbuf actions
;
1051 struct dpif_execute execute
;
1052 struct dp_packet packet
;
1055 struct ovs_key_ethernet key
, mask
;
1057 /* Compose a set action that will cause an EINVAL error on older
1058 * datapaths that don't support masked set actions.
1059 * Avoid using a full mask, as it could be translated to a non-masked
1060 * set action instead. */
1061 ofpbuf_init(&actions
, 64);
1062 memset(&key
, 0x53, sizeof key
);
1063 memset(&mask
, 0x7f, sizeof mask
);
1064 commit_masked_set_action(&actions
, OVS_KEY_ATTR_ETHERNET
, &key
, &mask
,
1067 /* Compose a dummy ethernet packet. */
1068 dp_packet_init(&packet
, ETH_HEADER_LEN
);
1069 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
1070 eth
->eth_type
= htons(0x1234);
1072 flow_extract(&packet
, &flow
);
1074 /* Execute the actions. On older datapaths this fails with EINVAL, on
1075 * newer datapaths it succeeds. */
1076 execute
.actions
= actions
.data
;
1077 execute
.actions_len
= actions
.size
;
1078 execute
.packet
= &packet
;
1079 execute
.flow
= &flow
;
1080 execute
.needs_help
= false;
1081 execute
.probe
= true;
1084 error
= dpif_execute(backer
->dpif
, &execute
);
1086 dp_packet_uninit(&packet
);
1087 ofpbuf_uninit(&actions
);
1090 /* Masked set action is not supported. */
1091 VLOG_INFO("%s: datapath does not support masked set action feature.",
1092 dpif_name(backer
->dpif
));
1097 /* Tests whether 'backer''s datapath supports truncation of a packet in
1098 * OVS_ACTION_ATTR_TRUNC. We need to disable some features on older
1099 * datapaths that don't support this feature. */
1101 check_trunc_action(struct dpif_backer
*backer
)
1103 struct eth_header
*eth
;
1104 struct ofpbuf actions
;
1105 struct dpif_execute execute
;
1106 struct dp_packet packet
;
1107 struct ovs_action_trunc
*trunc
;
1111 /* Compose an action with output(port:1,
1112 * max_len:OVS_ACTION_OUTPUT_MIN + 1).
1113 * This translates to one truncate action and one output action. */
1114 ofpbuf_init(&actions
, 64);
1115 trunc
= nl_msg_put_unspec_uninit(&actions
,
1116 OVS_ACTION_ATTR_TRUNC
, sizeof *trunc
);
1118 trunc
->max_len
= ETH_HEADER_LEN
+ 1;
1119 nl_msg_put_odp_port(&actions
, OVS_ACTION_ATTR_OUTPUT
, u32_to_odp(1));
1121 /* Compose a dummy Ethernet packet. */
1122 dp_packet_init(&packet
, ETH_HEADER_LEN
);
1123 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
1124 eth
->eth_type
= htons(0x1234);
1126 flow_extract(&packet
, &flow
);
1128 /* Execute the actions. On older datapaths this fails with EINVAL, on
1129 * newer datapaths it succeeds. */
1130 execute
.actions
= actions
.data
;
1131 execute
.actions_len
= actions
.size
;
1132 execute
.packet
= &packet
;
1133 execute
.flow
= &flow
;
1134 execute
.needs_help
= false;
1135 execute
.probe
= true;
1138 error
= dpif_execute(backer
->dpif
, &execute
);
1140 dp_packet_uninit(&packet
);
1141 ofpbuf_uninit(&actions
);
1144 VLOG_INFO("%s: Datapath does not support truncate action",
1145 dpif_name(backer
->dpif
));
1147 VLOG_INFO("%s: Datapath supports truncate action",
1148 dpif_name(backer
->dpif
));
1154 /* Tests whether 'backer''s datapath supports the clone action
1155 * OVS_ACTION_ATTR_CLONE. */
1157 check_clone(struct dpif_backer
*backer
)
1159 struct dpif_execute execute
;
1160 struct eth_header
*eth
;
1162 struct dp_packet packet
;
1163 struct ofpbuf actions
;
1167 /* Compose clone with an empty action list.
1168 * and check if datapath can decode the message. */
1169 ofpbuf_init(&actions
, 64);
1170 clone_start
= nl_msg_start_nested(&actions
, OVS_ACTION_ATTR_CLONE
);
1171 nl_msg_end_nested(&actions
, clone_start
);
1173 /* Compose a dummy Ethernet packet. */
1174 dp_packet_init(&packet
, ETH_HEADER_LEN
);
1175 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
1176 eth
->eth_type
= htons(0x1234);
1178 flow_extract(&packet
, &flow
);
1180 /* Execute the actions. On older datapaths this fails with EINVAL, on
1181 * newer datapaths it succeeds. */
1182 execute
.actions
= actions
.data
;
1183 execute
.actions_len
= actions
.size
;
1184 execute
.packet
= &packet
;
1185 execute
.flow
= &flow
;
1186 execute
.needs_help
= false;
1187 execute
.probe
= true;
1190 error
= dpif_execute(backer
->dpif
, &execute
);
1192 dp_packet_uninit(&packet
);
1193 ofpbuf_uninit(&actions
);
1196 VLOG_INFO("%s: Datapath does not support clone action",
1197 dpif_name(backer
->dpif
));
1199 VLOG_INFO("%s: Datapath supports clone action",
1200 dpif_name(backer
->dpif
));
1206 #define CHECK_FEATURE__(NAME, SUPPORT, FIELD, VALUE) \
1208 check_##NAME(struct dpif_backer *backer) \
1211 struct odputil_keybuf keybuf; \
1212 struct ofpbuf key; \
1214 struct odp_flow_key_parms odp_parms = { \
1221 memset(&flow, 0, sizeof flow); \
1222 flow.FIELD = VALUE; \
1224 ofpbuf_use_stack(&key, &keybuf, sizeof keybuf); \
1225 odp_flow_key_from_flow(&odp_parms, &key); \
1226 enable = dpif_probe_feature(backer->dpif, #NAME, &key, NULL, NULL); \
1229 VLOG_INFO("%s: Datapath supports "#NAME, dpif_name(backer->dpif)); \
1231 VLOG_INFO("%s: Datapath does not support "#NAME, \
1232 dpif_name(backer->dpif)); \
1237 #define CHECK_FEATURE(FIELD) CHECK_FEATURE__(FIELD, FIELD, FIELD, 1)
1239 CHECK_FEATURE(ct_state
)
1240 CHECK_FEATURE(ct_zone
)
1241 CHECK_FEATURE(ct_mark
)
1242 CHECK_FEATURE__(ct_label
, ct_label
, ct_label
.u64
.lo
, 1)
1243 CHECK_FEATURE__(ct_state_nat
, ct_state
, ct_state
, CS_TRACKED
|CS_SRC_NAT
)
1244 CHECK_FEATURE__(ct_orig_tuple
, ct_orig_tuple
, ct_nw_proto
, 1)
1246 #undef CHECK_FEATURE
1247 #undef CHECK_FEATURE__
1250 check_support(struct dpif_backer
*backer
)
1252 /* This feature needs to be tested after udpif threads are set. */
1253 backer
->support
.variable_length_userdata
= false;
1256 backer
->support
.odp
.recirc
= check_recirc(backer
);
1257 backer
->support
.odp
.max_mpls_depth
= check_max_mpls_depth(backer
);
1258 backer
->support
.masked_set_action
= check_masked_set_action(backer
);
1259 backer
->support
.trunc
= check_trunc_action(backer
);
1260 backer
->support
.ufid
= check_ufid(backer
);
1261 backer
->support
.tnl_push_pop
= dpif_supports_tnl_push_pop(backer
->dpif
);
1262 backer
->support
.clone
= check_clone(backer
);
1263 backer
->support
.sample_nesting
= check_max_sample_nesting(backer
);
1266 backer
->support
.odp
.ct_state
= check_ct_state(backer
);
1267 backer
->support
.odp
.ct_zone
= check_ct_zone(backer
);
1268 backer
->support
.odp
.ct_mark
= check_ct_mark(backer
);
1269 backer
->support
.odp
.ct_label
= check_ct_label(backer
);
1271 backer
->support
.odp
.ct_state_nat
= check_ct_state_nat(backer
);
1272 backer
->support
.odp
.ct_orig_tuple
= check_ct_orig_tuple(backer
);
1276 construct(struct ofproto
*ofproto_
)
1278 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1279 struct shash_node
*node
, *next
;
1282 /* Tunnel module can get used right after the udpif threads are running. */
1283 ofproto_tunnel_init();
1285 error
= open_dpif_backer(ofproto
->up
.type
, &ofproto
->backer
);
1290 uuid_generate(&ofproto
->uuid
);
1291 atomic_init(&ofproto
->tables_version
, OVS_VERSION_MIN
);
1292 ofproto
->netflow
= NULL
;
1293 ofproto
->sflow
= NULL
;
1294 ofproto
->ipfix
= NULL
;
1295 ofproto
->stp
= NULL
;
1296 ofproto
->rstp
= NULL
;
1297 ofproto
->dump_seq
= 0;
1298 hmap_init(&ofproto
->bundles
);
1299 ofproto
->ml
= mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME
);
1301 ofproto
->mbridge
= mbridge_create();
1302 ofproto
->has_bonded_bundles
= false;
1303 ofproto
->lacp_enabled
= false;
1304 ovs_mutex_init_adaptive(&ofproto
->stats_mutex
);
1306 guarded_list_init(&ofproto
->ams
);
1308 sset_init(&ofproto
->ports
);
1309 sset_init(&ofproto
->ghost_ports
);
1310 sset_init(&ofproto
->port_poll_set
);
1311 ofproto
->port_poll_errno
= 0;
1312 ofproto
->change_seq
= 0;
1313 ofproto
->ams_seq
= seq_create();
1314 ofproto
->ams_seqno
= seq_read(ofproto
->ams_seq
);
1317 SHASH_FOR_EACH_SAFE (node
, next
, &init_ofp_ports
) {
1318 struct iface_hint
*iface_hint
= node
->data
;
1320 if (!strcmp(iface_hint
->br_name
, ofproto
->up
.name
)) {
1321 /* Check if the datapath already has this port. */
1322 if (dpif_port_exists(ofproto
->backer
->dpif
, node
->name
)) {
1323 sset_add(&ofproto
->ports
, node
->name
);
1326 free(iface_hint
->br_name
);
1327 free(iface_hint
->br_type
);
1329 shash_delete(&init_ofp_ports
, node
);
1333 hmap_insert(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
,
1334 hash_string(ofproto
->up
.name
, 0));
1335 memset(&ofproto
->stats
, 0, sizeof ofproto
->stats
);
1337 ofproto_init_tables(ofproto_
, N_TABLES
);
1338 error
= add_internal_flows(ofproto
);
1340 ofproto
->up
.tables
[TBL_INTERNAL
].flags
= OFTABLE_HIDDEN
| OFTABLE_READONLY
;
1346 add_internal_miss_flow(struct ofproto_dpif
*ofproto
, int id
,
1347 const struct ofpbuf
*ofpacts
, struct rule_dpif
**rulep
)
1353 match_init_catchall(&match
);
1354 match_set_reg(&match
, 0, id
);
1356 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 0, 0, ofpacts
,
1358 *rulep
= error
? NULL
: rule_dpif_cast(rule
);
1364 add_internal_flows(struct ofproto_dpif
*ofproto
)
1366 struct ofpact_controller
*controller
;
1367 uint64_t ofpacts_stub
[128 / 8];
1368 struct ofpbuf ofpacts
;
1369 struct rule
*unused_rulep OVS_UNUSED
;
1374 ofpbuf_use_stack(&ofpacts
, ofpacts_stub
, sizeof ofpacts_stub
);
1377 controller
= ofpact_put_CONTROLLER(&ofpacts
);
1378 controller
->max_len
= UINT16_MAX
;
1379 controller
->controller_id
= 0;
1380 controller
->reason
= OFPR_IMPLICIT_MISS
;
1381 ofpact_finish_CONTROLLER(&ofpacts
, &controller
);
1383 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1384 &ofproto
->miss_rule
);
1389 ofpbuf_clear(&ofpacts
);
1390 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1391 &ofproto
->no_packet_in_rule
);
1396 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1397 &ofproto
->drop_frags_rule
);
1402 /* Drop any run away non-recirc rule lookups. Recirc_id has to be
1403 * zero when reaching this rule.
1405 * (priority=2), recirc_id=0, actions=drop
1407 ofpbuf_clear(&ofpacts
);
1408 match_init_catchall(&match
);
1409 match_set_recirc_id(&match
, 0);
1410 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 2, 0, &ofpacts
,
1416 destruct(struct ofproto
*ofproto_
)
1418 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1419 struct ofproto_async_msg
*am
;
1420 struct rule_dpif
*rule
;
1421 struct oftable
*table
;
1422 struct ovs_list ams
;
1424 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1426 xlate_remove_ofproto(ofproto
);
1429 /* Ensure that the upcall processing threads have no remaining references
1430 * to the ofproto or anything in it. */
1431 udpif_synchronize(ofproto
->backer
->udpif
);
1433 hmap_remove(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
);
1435 OFPROTO_FOR_EACH_TABLE (table
, &ofproto
->up
) {
1436 CLS_FOR_EACH (rule
, up
.cr
, &table
->cls
) {
1437 ofproto_rule_delete(&ofproto
->up
, &rule
->up
);
1440 ofproto_group_delete_all(&ofproto
->up
);
1442 guarded_list_pop_all(&ofproto
->ams
, &ams
);
1443 LIST_FOR_EACH_POP (am
, list_node
, &ams
) {
1444 ofproto_async_msg_free(am
);
1446 guarded_list_destroy(&ofproto
->ams
);
1448 recirc_free_ofproto(ofproto
, ofproto
->up
.name
);
1450 mbridge_unref(ofproto
->mbridge
);
1452 netflow_unref(ofproto
->netflow
);
1453 dpif_sflow_unref(ofproto
->sflow
);
1454 dpif_ipfix_unref(ofproto
->ipfix
);
1455 hmap_destroy(&ofproto
->bundles
);
1456 mac_learning_unref(ofproto
->ml
);
1457 mcast_snooping_unref(ofproto
->ms
);
1459 sset_destroy(&ofproto
->ports
);
1460 sset_destroy(&ofproto
->ghost_ports
);
1461 sset_destroy(&ofproto
->port_poll_set
);
1463 ovs_mutex_destroy(&ofproto
->stats_mutex
);
1465 seq_destroy(ofproto
->ams_seq
);
1467 close_dpif_backer(ofproto
->backer
);
1471 run(struct ofproto
*ofproto_
)
1473 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1474 uint64_t new_seq
, new_dump_seq
;
1476 if (mbridge_need_revalidate(ofproto
->mbridge
)) {
1477 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1478 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1479 mac_learning_flush(ofproto
->ml
);
1480 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1481 mcast_snooping_mdb_flush(ofproto
->ms
);
1484 /* Always updates the ofproto->ams_seqno to avoid frequent wakeup during
1485 * flow restore. Even though nothing is processed during flow restore,
1486 * all queued 'ams' will be handled immediately when flow restore
1488 ofproto
->ams_seqno
= seq_read(ofproto
->ams_seq
);
1490 /* Do not perform any periodic activity required by 'ofproto' while
1491 * waiting for flow restore to complete. */
1492 if (!ofproto_get_flow_restore_wait()) {
1493 struct ofproto_async_msg
*am
;
1494 struct ovs_list ams
;
1496 guarded_list_pop_all(&ofproto
->ams
, &ams
);
1497 LIST_FOR_EACH_POP (am
, list_node
, &ams
) {
1498 connmgr_send_async_msg(ofproto
->up
.connmgr
, am
);
1499 ofproto_async_msg_free(am
);
1503 if (ofproto
->netflow
) {
1504 netflow_run(ofproto
->netflow
);
1506 if (ofproto
->sflow
) {
1507 dpif_sflow_run(ofproto
->sflow
);
1509 if (ofproto
->ipfix
) {
1510 dpif_ipfix_run(ofproto
->ipfix
);
1513 new_seq
= seq_read(connectivity_seq_get());
1514 if (ofproto
->change_seq
!= new_seq
) {
1515 struct ofport_dpif
*ofport
;
1517 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1521 ofproto
->change_seq
= new_seq
;
1523 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1524 struct ofbundle
*bundle
;
1526 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1533 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1534 if (mac_learning_run(ofproto
->ml
)) {
1535 ofproto
->backer
->need_revalidate
= REV_MAC_LEARNING
;
1537 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1539 if (mcast_snooping_run(ofproto
->ms
)) {
1540 ofproto
->backer
->need_revalidate
= REV_MCAST_SNOOPING
;
1543 new_dump_seq
= seq_read(udpif_dump_seq(ofproto
->backer
->udpif
));
1544 if (ofproto
->dump_seq
!= new_dump_seq
) {
1545 struct rule
*rule
, *next_rule
;
1546 long long now
= time_msec();
1548 /* We know stats are relatively fresh, so now is a good time to do some
1550 ofproto
->dump_seq
= new_dump_seq
;
1552 /* Expire OpenFlow flows whose idle_timeout or hard_timeout
1554 ovs_mutex_lock(&ofproto_mutex
);
1555 LIST_FOR_EACH_SAFE (rule
, next_rule
, expirable
,
1556 &ofproto
->up
.expirable
) {
1557 rule_expire(rule_dpif_cast(rule
), now
);
1559 ovs_mutex_unlock(&ofproto_mutex
);
1561 /* All outstanding data in existing flows has been accounted, so it's a
1562 * good time to do bond rebalancing. */
1563 if (ofproto
->has_bonded_bundles
) {
1564 struct ofbundle
*bundle
;
1566 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1568 bond_rebalance(bundle
->bond
);
1577 ofproto_dpif_wait(struct ofproto
*ofproto_
)
1579 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1581 if (ofproto_get_flow_restore_wait()) {
1585 if (ofproto
->sflow
) {
1586 dpif_sflow_wait(ofproto
->sflow
);
1588 if (ofproto
->ipfix
) {
1589 dpif_ipfix_wait(ofproto
->ipfix
);
1591 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1592 struct ofbundle
*bundle
;
1594 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1595 bundle_wait(bundle
);
1598 if (ofproto
->netflow
) {
1599 netflow_wait(ofproto
->netflow
);
1601 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
1602 mac_learning_wait(ofproto
->ml
);
1603 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1604 mcast_snooping_wait(ofproto
->ms
);
1606 if (ofproto
->backer
->need_revalidate
) {
1607 poll_immediate_wake();
1610 seq_wait(udpif_dump_seq(ofproto
->backer
->udpif
), ofproto
->dump_seq
);
1611 seq_wait(ofproto
->ams_seq
, ofproto
->ams_seqno
);
1615 type_get_memory_usage(const char *type
, struct simap
*usage
)
1617 struct dpif_backer
*backer
;
1619 backer
= shash_find_data(&all_dpif_backers
, type
);
1621 udpif_get_memory_usage(backer
->udpif
, usage
);
1626 flush(struct ofproto
*ofproto_
)
1628 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1629 struct dpif_backer
*backer
= ofproto
->backer
;
1632 udpif_flush(backer
->udpif
);
1637 query_tables(struct ofproto
*ofproto
,
1638 struct ofputil_table_features
*features
,
1639 struct ofputil_table_stats
*stats
)
1641 strcpy(features
->name
, "classifier");
1646 for (i
= 0; i
< ofproto
->n_tables
; i
++) {
1647 unsigned long missed
, matched
;
1649 atomic_read_relaxed(&ofproto
->tables
[i
].n_matched
, &matched
);
1650 atomic_read_relaxed(&ofproto
->tables
[i
].n_missed
, &missed
);
1652 stats
[i
].matched_count
= matched
;
1653 stats
[i
].lookup_count
= matched
+ missed
;
1659 set_tables_version(struct ofproto
*ofproto_
, ovs_version_t version
)
1661 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1663 /* Use memory_order_release to signify that any prior memory accesses can
1664 * not be reordered to happen after this atomic store. This makes sure the
1665 * new version is properly set up when the readers can read this 'version'
1667 atomic_store_explicit(&ofproto
->tables_version
, version
,
1668 memory_order_release
);
1669 /* 'need_revalidate' can be reordered to happen before the atomic_store
1670 * above, but it does not matter as this variable is not accessed by other
1672 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
1675 static struct ofport
*
1678 struct ofport_dpif
*port
= xzalloc(sizeof *port
);
1683 port_dealloc(struct ofport
*port_
)
1685 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1690 port_construct(struct ofport
*port_
)
1692 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1693 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1694 const struct netdev
*netdev
= port
->up
.netdev
;
1695 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1696 const char *dp_port_name
;
1697 struct dpif_port dpif_port
;
1700 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1701 port
->bundle
= NULL
;
1705 port
->may_enable
= false;
1706 port
->stp_port
= NULL
;
1707 port
->stp_state
= STP_DISABLED
;
1708 port
->rstp_port
= NULL
;
1709 port
->rstp_state
= RSTP_DISABLED
;
1710 port
->is_tunnel
= false;
1714 port
->carrier_seq
= netdev_get_carrier_resets(netdev
);
1715 port
->is_layer3
= netdev_vport_is_layer3(netdev
);
1717 if (netdev_vport_is_patch(netdev
)) {
1718 /* By bailing out here, we don't submit the port to the sFlow module
1719 * to be considered for counter polling export. This is correct
1720 * because the patch port represents an interface that sFlow considers
1721 * to be "internal" to the switch as a whole, and therefore not a
1722 * candidate for counter polling. */
1723 port
->odp_port
= ODPP_NONE
;
1724 ofport_update_peer(port
);
1728 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
1729 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
, dp_port_name
,
1735 port
->odp_port
= dpif_port
.port_no
;
1737 if (netdev_get_tunnel_config(netdev
)) {
1738 atomic_count_inc(&ofproto
->backer
->tnl_count
);
1739 error
= tnl_port_add(port
, port
->up
.netdev
, port
->odp_port
,
1740 ovs_native_tunneling_is_on(ofproto
), dp_port_name
);
1742 atomic_count_dec(&ofproto
->backer
->tnl_count
);
1743 dpif_port_destroy(&dpif_port
);
1747 port
->is_tunnel
= true;
1748 if (ofproto
->ipfix
) {
1749 dpif_ipfix_add_tunnel_port(ofproto
->ipfix
, port_
, port
->odp_port
);
1752 /* Sanity-check that a mapping doesn't already exist. This
1753 * shouldn't happen for non-tunnel ports. */
1754 if (odp_port_to_ofp_port(ofproto
, port
->odp_port
) != OFPP_NONE
) {
1755 VLOG_ERR("port %s already has an OpenFlow port number",
1757 dpif_port_destroy(&dpif_port
);
1761 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1762 hmap_insert(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
,
1763 hash_odp_port(port
->odp_port
));
1764 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1766 dpif_port_destroy(&dpif_port
);
1768 if (ofproto
->sflow
) {
1769 dpif_sflow_add_port(ofproto
->sflow
, port_
, port
->odp_port
);
1776 port_destruct(struct ofport
*port_
, bool del
)
1778 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1779 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1780 const char *devname
= netdev_get_name(port
->up
.netdev
);
1781 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1782 const char *dp_port_name
;
1784 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1786 xlate_ofport_remove(port
);
1789 dp_port_name
= netdev_vport_get_dpif_port(port
->up
.netdev
, namebuf
,
1791 if (del
&& dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
1792 /* The underlying device is still there, so delete it. This
1793 * happens when the ofproto is being destroyed, since the caller
1794 * assumes that removal of attached ports will happen as part of
1796 if (!port
->is_tunnel
) {
1797 dpif_port_del(ofproto
->backer
->dpif
, port
->odp_port
);
1802 port
->peer
->peer
= NULL
;
1806 if (port
->odp_port
!= ODPP_NONE
&& !port
->is_tunnel
) {
1807 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1808 hmap_remove(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
);
1809 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1812 if (port
->is_tunnel
) {
1813 atomic_count_dec(&ofproto
->backer
->tnl_count
);
1816 if (port
->is_tunnel
&& ofproto
->ipfix
) {
1817 dpif_ipfix_del_tunnel_port(ofproto
->ipfix
, port
->odp_port
);
1821 sset_find_and_delete(&ofproto
->ports
, devname
);
1822 sset_find_and_delete(&ofproto
->ghost_ports
, devname
);
1823 bundle_remove(port_
);
1824 set_cfm(port_
, NULL
);
1825 set_bfd(port_
, NULL
);
1826 set_lldp(port_
, NULL
);
1827 if (port
->stp_port
) {
1828 stp_port_disable(port
->stp_port
);
1830 set_rstp_port(port_
, NULL
);
1831 if (ofproto
->sflow
) {
1832 dpif_sflow_del_port(ofproto
->sflow
, port
->odp_port
);
1839 port_modified(struct ofport
*port_
)
1841 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1842 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1843 const char *dp_port_name
;
1844 struct netdev
*netdev
= port
->up
.netdev
;
1846 if (port
->bundle
&& port
->bundle
->bond
) {
1847 bond_slave_set_netdev(port
->bundle
->bond
, port
, netdev
);
1851 cfm_set_netdev(port
->cfm
, netdev
);
1855 bfd_set_netdev(port
->bfd
, netdev
);
1858 ofproto_dpif_monitor_port_update(port
, port
->bfd
, port
->cfm
,
1859 port
->lldp
, &port
->up
.pp
.hw_addr
);
1861 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
1863 if (port
->is_tunnel
) {
1864 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1866 if (tnl_port_reconfigure(port
, netdev
, port
->odp_port
,
1867 ovs_native_tunneling_is_on(ofproto
),
1869 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1873 ofport_update_peer(port
);
1877 port_reconfigured(struct ofport
*port_
, enum ofputil_port_config old_config
)
1879 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1880 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1881 enum ofputil_port_config changed
= old_config
^ port
->up
.pp
.config
;
1883 if (changed
& (OFPUTIL_PC_NO_RECV
| OFPUTIL_PC_NO_RECV_STP
|
1884 OFPUTIL_PC_NO_FWD
| OFPUTIL_PC_NO_FLOOD
|
1885 OFPUTIL_PC_NO_PACKET_IN
)) {
1886 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1888 if (changed
& OFPUTIL_PC_NO_FLOOD
&& port
->bundle
) {
1889 bundle_update(port
->bundle
);
1895 set_sflow(struct ofproto
*ofproto_
,
1896 const struct ofproto_sflow_options
*sflow_options
)
1898 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1899 struct dpif_sflow
*ds
= ofproto
->sflow
;
1901 if (sflow_options
) {
1902 uint32_t old_probability
= ds
? dpif_sflow_get_probability(ds
) : 0;
1904 struct ofport_dpif
*ofport
;
1906 ds
= ofproto
->sflow
= dpif_sflow_create();
1907 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1908 dpif_sflow_add_port(ds
, &ofport
->up
, ofport
->odp_port
);
1911 dpif_sflow_set_options(ds
, sflow_options
);
1912 if (dpif_sflow_get_probability(ds
) != old_probability
) {
1913 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1917 dpif_sflow_unref(ds
);
1918 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1919 ofproto
->sflow
= NULL
;
1927 struct ofproto
*ofproto_
,
1928 const struct ofproto_ipfix_bridge_exporter_options
*bridge_exporter_options
,
1929 const struct ofproto_ipfix_flow_exporter_options
*flow_exporters_options
,
1930 size_t n_flow_exporters_options
)
1932 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1933 struct dpif_ipfix
*di
= ofproto
->ipfix
;
1934 bool has_options
= bridge_exporter_options
|| flow_exporters_options
;
1935 bool new_di
= false;
1937 if (has_options
&& !di
) {
1938 di
= ofproto
->ipfix
= dpif_ipfix_create();
1943 /* Call set_options in any case to cleanly flush the flow
1944 * caches in the last exporters that are to be destroyed. */
1945 dpif_ipfix_set_options(
1946 di
, bridge_exporter_options
, flow_exporters_options
,
1947 n_flow_exporters_options
);
1949 /* Add tunnel ports only when a new ipfix created */
1950 if (new_di
== true) {
1951 struct ofport_dpif
*ofport
;
1952 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1953 if (ofport
->is_tunnel
== true) {
1954 dpif_ipfix_add_tunnel_port(di
, &ofport
->up
, ofport
->odp_port
);
1960 dpif_ipfix_unref(di
);
1961 ofproto
->ipfix
= NULL
;
1969 get_ipfix_stats(const struct ofproto
*ofproto_
,
1971 struct ovs_list
*replies
)
1973 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1974 struct dpif_ipfix
*di
= ofproto
->ipfix
;
1977 return OFPERR_NXST_NOT_CONFIGURED
;
1980 return dpif_ipfix_get_stats(di
, bridge_ipfix
, replies
);
1984 set_cfm(struct ofport
*ofport_
, const struct cfm_settings
*s
)
1986 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1987 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
1988 struct cfm
*old
= ofport
->cfm
;
1993 ofport
->cfm
= cfm_create(ofport
->up
.netdev
);
1996 if (cfm_configure(ofport
->cfm
, s
)) {
2003 cfm_unref(ofport
->cfm
);
2006 if (ofport
->cfm
!= old
) {
2007 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2009 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
2010 ofport
->lldp
, &ofport
->up
.pp
.hw_addr
);
2015 cfm_status_changed(struct ofport
*ofport_
)
2017 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2019 return ofport
->cfm
? cfm_check_status_change(ofport
->cfm
) : true;
2023 get_cfm_status(const struct ofport
*ofport_
,
2024 struct cfm_status
*status
)
2026 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2030 cfm_get_status(ofport
->cfm
, status
);
2039 set_bfd(struct ofport
*ofport_
, const struct smap
*cfg
)
2041 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
2042 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2046 ofport
->bfd
= bfd_configure(old
, netdev_get_name(ofport
->up
.netdev
),
2047 cfg
, ofport
->up
.netdev
);
2048 if (ofport
->bfd
!= old
) {
2049 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2051 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
2052 ofport
->lldp
, &ofport
->up
.pp
.hw_addr
);
2057 bfd_status_changed(struct ofport
*ofport_
)
2059 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2061 return ofport
->bfd
? bfd_check_status_change(ofport
->bfd
) : true;
2065 get_bfd_status(struct ofport
*ofport_
, struct smap
*smap
)
2067 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2071 bfd_get_status(ofport
->bfd
, smap
);
2080 set_lldp(struct ofport
*ofport_
,
2081 const struct smap
*cfg
)
2083 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2087 if (!ofport
->lldp
) {
2088 struct ofproto_dpif
*ofproto
;
2090 ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2091 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2092 ofport
->lldp
= lldp_create(ofport
->up
.netdev
, ofport_
->mtu
, cfg
);
2095 if (!lldp_configure(ofport
->lldp
, cfg
)) {
2100 lldp_unref(ofport
->lldp
);
2101 ofport
->lldp
= NULL
;
2104 ofproto_dpif_monitor_port_update(ofport
,
2108 &ofport
->up
.pp
.hw_addr
);
2113 get_lldp_status(const struct ofport
*ofport_
,
2114 struct lldp_status
*status OVS_UNUSED
)
2116 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2118 return ofport
->lldp
? true : false;
2122 set_aa(struct ofproto
*ofproto OVS_UNUSED
,
2123 const struct aa_settings
*s
)
2125 return aa_configure(s
);
2129 aa_mapping_set(struct ofproto
*ofproto_ OVS_UNUSED
, void *aux
,
2130 const struct aa_mapping_settings
*s
)
2132 return aa_mapping_register(aux
, s
);
2136 aa_mapping_unset(struct ofproto
*ofproto OVS_UNUSED
, void *aux
)
2138 return aa_mapping_unregister(aux
);
2142 aa_vlan_get_queued(struct ofproto
*ofproto OVS_UNUSED
, struct ovs_list
*list
)
2144 return aa_get_vlan_queued(list
);
2148 aa_vlan_get_queue_size(struct ofproto
*ofproto OVS_UNUSED
)
2150 return aa_get_vlan_queue_size();
2154 /* Spanning Tree. */
2156 /* Called while rstp_mutex is held. */
2158 rstp_send_bpdu_cb(struct dp_packet
*pkt
, void *ofport_
, void *ofproto_
)
2160 struct ofproto_dpif
*ofproto
= ofproto_
;
2161 struct ofport_dpif
*ofport
= ofport_
;
2162 struct eth_header
*eth
= dp_packet_l2(pkt
);
2164 netdev_get_etheraddr(ofport
->up
.netdev
, ð
->eth_src
);
2165 if (eth_addr_is_zero(eth
->eth_src
)) {
2166 VLOG_WARN_RL(&rl
, "%s port %d: cannot send RSTP BPDU on a port which "
2167 "does not have a configured source MAC address.",
2168 ofproto
->up
.name
, ofp_to_u16(ofport
->up
.ofp_port
));
2170 ofproto_dpif_send_packet(ofport
, false, pkt
);
2172 dp_packet_delete(pkt
);
2176 send_bpdu_cb(struct dp_packet
*pkt
, int port_num
, void *ofproto_
)
2178 struct ofproto_dpif
*ofproto
= ofproto_
;
2179 struct stp_port
*sp
= stp_get_port(ofproto
->stp
, port_num
);
2180 struct ofport_dpif
*ofport
;
2182 ofport
= stp_port_get_aux(sp
);
2184 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on unknown port %d",
2185 ofproto
->up
.name
, port_num
);
2187 struct eth_header
*eth
= dp_packet_l2(pkt
);
2189 netdev_get_etheraddr(ofport
->up
.netdev
, ð
->eth_src
);
2190 if (eth_addr_is_zero(eth
->eth_src
)) {
2191 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on port %d "
2192 "with unknown MAC", ofproto
->up
.name
, port_num
);
2194 ofproto_dpif_send_packet(ofport
, false, pkt
);
2197 dp_packet_delete(pkt
);
2200 /* Configure RSTP on 'ofproto_' using the settings defined in 's'. */
2202 set_rstp(struct ofproto
*ofproto_
, const struct ofproto_rstp_settings
*s
)
2204 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2206 /* Only revalidate flows if the configuration changed. */
2207 if (!s
!= !ofproto
->rstp
) {
2208 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2212 if (!ofproto
->rstp
) {
2213 ofproto
->rstp
= rstp_create(ofproto_
->name
, s
->address
,
2214 rstp_send_bpdu_cb
, ofproto
);
2215 ofproto
->rstp_last_tick
= time_msec();
2217 rstp_set_bridge_address(ofproto
->rstp
, s
->address
);
2218 rstp_set_bridge_priority(ofproto
->rstp
, s
->priority
);
2219 rstp_set_bridge_ageing_time(ofproto
->rstp
, s
->ageing_time
);
2220 rstp_set_bridge_force_protocol_version(ofproto
->rstp
,
2221 s
->force_protocol_version
);
2222 rstp_set_bridge_max_age(ofproto
->rstp
, s
->bridge_max_age
);
2223 rstp_set_bridge_forward_delay(ofproto
->rstp
, s
->bridge_forward_delay
);
2224 rstp_set_bridge_transmit_hold_count(ofproto
->rstp
,
2225 s
->transmit_hold_count
);
2227 struct ofport
*ofport
;
2228 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2229 set_rstp_port(ofport
, NULL
);
2231 rstp_unref(ofproto
->rstp
);
2232 ofproto
->rstp
= NULL
;
2237 get_rstp_status(struct ofproto
*ofproto_
, struct ofproto_rstp_status
*s
)
2239 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2241 if (ofproto
->rstp
) {
2243 s
->root_id
= rstp_get_root_id(ofproto
->rstp
);
2244 s
->bridge_id
= rstp_get_bridge_id(ofproto
->rstp
);
2245 s
->designated_id
= rstp_get_designated_id(ofproto
->rstp
);
2246 s
->root_path_cost
= rstp_get_root_path_cost(ofproto
->rstp
);
2247 s
->designated_port_id
= rstp_get_designated_port_id(ofproto
->rstp
);
2248 s
->bridge_port_id
= rstp_get_bridge_port_id(ofproto
->rstp
);
2255 update_rstp_port_state(struct ofport_dpif
*ofport
)
2257 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2258 enum rstp_state state
;
2260 /* Figure out new state. */
2261 state
= ofport
->rstp_port
? rstp_port_get_state(ofport
->rstp_port
)
2265 if (ofport
->rstp_state
!= state
) {
2266 enum ofputil_port_state of_state
;
2269 VLOG_DBG("port %s: RSTP state changed from %s to %s",
2270 netdev_get_name(ofport
->up
.netdev
),
2271 rstp_state_name(ofport
->rstp_state
),
2272 rstp_state_name(state
));
2274 if (rstp_learn_in_state(ofport
->rstp_state
)
2275 != rstp_learn_in_state(state
)) {
2276 /* XXX: Learning action flows should also be flushed. */
2277 if (ofport
->bundle
) {
2278 if (!rstp_shift_root_learned_address(ofproto
->rstp
)
2279 || rstp_get_old_root_aux(ofproto
->rstp
) != ofport
) {
2280 bundle_flush_macs(ofport
->bundle
, false);
2284 fwd_change
= rstp_forward_in_state(ofport
->rstp_state
)
2285 != rstp_forward_in_state(state
);
2287 ofproto
->backer
->need_revalidate
= REV_RSTP
;
2288 ofport
->rstp_state
= state
;
2290 if (fwd_change
&& ofport
->bundle
) {
2291 bundle_update(ofport
->bundle
);
2294 /* Update the RSTP state bits in the OpenFlow port description. */
2295 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2296 of_state
|= (state
== RSTP_LEARNING
? OFPUTIL_PS_STP_LEARN
2297 : state
== RSTP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2298 : state
== RSTP_DISCARDING
? OFPUTIL_PS_STP_LISTEN
2300 ofproto_port_set_state(&ofport
->up
, of_state
);
2305 rstp_run(struct ofproto_dpif
*ofproto
)
2307 if (ofproto
->rstp
) {
2308 long long int now
= time_msec();
2309 long long int elapsed
= now
- ofproto
->rstp_last_tick
;
2310 struct rstp_port
*rp
;
2311 struct ofport_dpif
*ofport
;
2313 /* Every second, decrease the values of the timers. */
2314 if (elapsed
>= 1000) {
2315 rstp_tick_timers(ofproto
->rstp
);
2316 ofproto
->rstp_last_tick
= now
;
2319 while ((ofport
= rstp_get_next_changed_port_aux(ofproto
->rstp
, &rp
))) {
2320 update_rstp_port_state(ofport
);
2324 /* FIXME: This check should be done on-event (i.e., when setting
2325 * p->fdb_flush) and not periodically.
2327 while ((ofport
= rstp_check_and_reset_fdb_flush(ofproto
->rstp
, &rp
))) {
2328 if (!rstp_shift_root_learned_address(ofproto
->rstp
)
2329 || rstp_get_old_root_aux(ofproto
->rstp
) != ofport
) {
2330 bundle_flush_macs(ofport
->bundle
, false);
2334 if (rstp_shift_root_learned_address(ofproto
->rstp
)) {
2335 struct ofport_dpif
*old_root_aux
=
2336 (struct ofport_dpif
*)rstp_get_old_root_aux(ofproto
->rstp
);
2337 struct ofport_dpif
*new_root_aux
=
2338 (struct ofport_dpif
*)rstp_get_new_root_aux(ofproto
->rstp
);
2339 if (old_root_aux
!= NULL
&& new_root_aux
!= NULL
) {
2340 bundle_move(old_root_aux
->bundle
, new_root_aux
->bundle
);
2341 rstp_reset_root_changed(ofproto
->rstp
);
2347 /* Configures STP on 'ofproto_' using the settings defined in 's'. */
2349 set_stp(struct ofproto
*ofproto_
, const struct ofproto_stp_settings
*s
)
2351 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2353 /* Only revalidate flows if the configuration changed. */
2354 if (!s
!= !ofproto
->stp
) {
2355 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2359 if (!ofproto
->stp
) {
2360 ofproto
->stp
= stp_create(ofproto_
->name
, s
->system_id
,
2361 send_bpdu_cb
, ofproto
);
2362 ofproto
->stp_last_tick
= time_msec();
2365 stp_set_bridge_id(ofproto
->stp
, s
->system_id
);
2366 stp_set_bridge_priority(ofproto
->stp
, s
->priority
);
2367 stp_set_hello_time(ofproto
->stp
, s
->hello_time
);
2368 stp_set_max_age(ofproto
->stp
, s
->max_age
);
2369 stp_set_forward_delay(ofproto
->stp
, s
->fwd_delay
);
2371 struct ofport
*ofport
;
2373 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2374 set_stp_port(ofport
, NULL
);
2377 stp_unref(ofproto
->stp
);
2378 ofproto
->stp
= NULL
;
2385 get_stp_status(struct ofproto
*ofproto_
, struct ofproto_stp_status
*s
)
2387 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2391 s
->bridge_id
= stp_get_bridge_id(ofproto
->stp
);
2392 s
->designated_root
= stp_get_designated_root(ofproto
->stp
);
2393 s
->root_path_cost
= stp_get_root_path_cost(ofproto
->stp
);
2402 update_stp_port_state(struct ofport_dpif
*ofport
)
2404 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2405 enum stp_state state
;
2407 /* Figure out new state. */
2408 state
= ofport
->stp_port
? stp_port_get_state(ofport
->stp_port
)
2412 if (ofport
->stp_state
!= state
) {
2413 enum ofputil_port_state of_state
;
2416 VLOG_DBG("port %s: STP state changed from %s to %s",
2417 netdev_get_name(ofport
->up
.netdev
),
2418 stp_state_name(ofport
->stp_state
),
2419 stp_state_name(state
));
2420 if (stp_learn_in_state(ofport
->stp_state
)
2421 != stp_learn_in_state(state
)) {
2422 /* xxx Learning action flows should also be flushed. */
2423 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2424 mac_learning_flush(ofproto
->ml
);
2425 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2426 mcast_snooping_mdb_flush(ofproto
->ms
);
2428 fwd_change
= stp_forward_in_state(ofport
->stp_state
)
2429 != stp_forward_in_state(state
);
2431 ofproto
->backer
->need_revalidate
= REV_STP
;
2432 ofport
->stp_state
= state
;
2433 ofport
->stp_state_entered
= time_msec();
2435 if (fwd_change
&& ofport
->bundle
) {
2436 bundle_update(ofport
->bundle
);
2439 /* Update the STP state bits in the OpenFlow port description. */
2440 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2441 of_state
|= (state
== STP_LISTENING
? OFPUTIL_PS_STP_LISTEN
2442 : state
== STP_LEARNING
? OFPUTIL_PS_STP_LEARN
2443 : state
== STP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2444 : state
== STP_BLOCKING
? OFPUTIL_PS_STP_BLOCK
2446 ofproto_port_set_state(&ofport
->up
, of_state
);
2450 /* Configures STP on 'ofport_' using the settings defined in 's'. The
2451 * caller is responsible for assigning STP port numbers and ensuring
2452 * there are no duplicates. */
2454 set_stp_port(struct ofport
*ofport_
,
2455 const struct ofproto_port_stp_settings
*s
)
2457 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2458 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2459 struct stp_port
*sp
= ofport
->stp_port
;
2461 if (!s
|| !s
->enable
) {
2463 ofport
->stp_port
= NULL
;
2464 stp_port_disable(sp
);
2465 update_stp_port_state(ofport
);
2468 } else if (sp
&& stp_port_no(sp
) != s
->port_num
2469 && ofport
== stp_port_get_aux(sp
)) {
2470 /* The port-id changed, so disable the old one if it's not
2471 * already in use by another port. */
2472 stp_port_disable(sp
);
2475 sp
= ofport
->stp_port
= stp_get_port(ofproto
->stp
, s
->port_num
);
2477 /* Set name before enabling the port so that debugging messages can print
2479 stp_port_set_name(sp
, netdev_get_name(ofport
->up
.netdev
));
2480 stp_port_enable(sp
);
2482 stp_port_set_aux(sp
, ofport
);
2483 stp_port_set_priority(sp
, s
->priority
);
2484 stp_port_set_path_cost(sp
, s
->path_cost
);
2486 update_stp_port_state(ofport
);
2492 get_stp_port_status(struct ofport
*ofport_
,
2493 struct ofproto_port_stp_status
*s
)
2495 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2496 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2497 struct stp_port
*sp
= ofport
->stp_port
;
2499 if (!ofproto
->stp
|| !sp
) {
2505 s
->port_id
= stp_port_get_id(sp
);
2506 s
->state
= stp_port_get_state(sp
);
2507 s
->sec_in_state
= (time_msec() - ofport
->stp_state_entered
) / 1000;
2508 s
->role
= stp_port_get_role(sp
);
2514 get_stp_port_stats(struct ofport
*ofport_
,
2515 struct ofproto_port_stp_stats
*s
)
2517 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2518 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2519 struct stp_port
*sp
= ofport
->stp_port
;
2521 if (!ofproto
->stp
|| !sp
) {
2527 stp_port_get_counts(sp
, &s
->tx_count
, &s
->rx_count
, &s
->error_count
);
2533 stp_run(struct ofproto_dpif
*ofproto
)
2536 long long int now
= time_msec();
2537 long long int elapsed
= now
- ofproto
->stp_last_tick
;
2538 struct stp_port
*sp
;
2541 stp_tick(ofproto
->stp
, MIN(INT_MAX
, elapsed
));
2542 ofproto
->stp_last_tick
= now
;
2544 while (stp_get_changed_port(ofproto
->stp
, &sp
)) {
2545 struct ofport_dpif
*ofport
= stp_port_get_aux(sp
);
2548 update_stp_port_state(ofport
);
2552 if (stp_check_and_reset_fdb_flush(ofproto
->stp
)) {
2553 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2554 mac_learning_flush(ofproto
->ml
);
2555 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2556 mcast_snooping_mdb_flush(ofproto
->ms
);
2562 stp_wait(struct ofproto_dpif
*ofproto
)
2565 poll_timer_wait(1000);
2569 /* Configures RSTP on 'ofport_' using the settings defined in 's'. The
2570 * caller is responsible for assigning RSTP port numbers and ensuring
2571 * there are no duplicates. */
2573 set_rstp_port(struct ofport
*ofport_
,
2574 const struct ofproto_port_rstp_settings
*s
)
2576 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2577 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2578 struct rstp_port
*rp
= ofport
->rstp_port
;
2580 if (!s
|| !s
->enable
) {
2582 rstp_port_set_aux(rp
, NULL
);
2583 rstp_port_set_state(rp
, RSTP_DISABLED
);
2584 rstp_port_set_mac_operational(rp
, false);
2585 ofport
->rstp_port
= NULL
;
2586 rstp_port_unref(rp
);
2587 update_rstp_port_state(ofport
);
2592 /* Check if need to add a new port. */
2594 rp
= ofport
->rstp_port
= rstp_add_port(ofproto
->rstp
);
2597 rstp_port_set(rp
, s
->port_num
, s
->priority
, s
->path_cost
,
2598 s
->admin_edge_port
, s
->auto_edge
,
2599 s
->admin_p2p_mac_state
, s
->admin_port_state
, s
->mcheck
,
2601 update_rstp_port_state(ofport
);
2602 /* Synchronize operational status. */
2603 rstp_port_set_mac_operational(rp
, ofport
->may_enable
);
2607 get_rstp_port_status(struct ofport
*ofport_
,
2608 struct ofproto_port_rstp_status
*s
)
2610 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2611 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2612 struct rstp_port
*rp
= ofport
->rstp_port
;
2614 if (!ofproto
->rstp
|| !rp
) {
2620 rstp_port_get_status(rp
, &s
->port_id
, &s
->state
, &s
->role
,
2621 &s
->designated_bridge_id
, &s
->designated_port_id
,
2622 &s
->designated_path_cost
, &s
->tx_count
,
2623 &s
->rx_count
, &s
->error_count
, &s
->uptime
);
2628 set_queues(struct ofport
*ofport_
, const struct ofproto_port_queue
*qdscp
,
2631 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2632 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2634 if (ofport
->n_qdscp
!= n_qdscp
2635 || (n_qdscp
&& memcmp(ofport
->qdscp
, qdscp
,
2636 n_qdscp
* sizeof *qdscp
))) {
2637 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2638 free(ofport
->qdscp
);
2639 ofport
->qdscp
= n_qdscp
2640 ? xmemdup(qdscp
, n_qdscp
* sizeof *qdscp
)
2642 ofport
->n_qdscp
= n_qdscp
;
2650 /* Expires all MAC learning entries associated with 'bundle' and forces its
2651 * ofproto to revalidate every flow.
2653 * Normally MAC learning entries are removed only from the ofproto associated
2654 * with 'bundle', but if 'all_ofprotos' is true, then the MAC learning entries
2655 * are removed from every ofproto. When patch ports and SLB bonds are in use
2656 * and a VM migration happens and the gratuitous ARPs are somehow lost, this
2657 * avoids a MAC_ENTRY_IDLE_TIME delay before the migrated VM can communicate
2658 * with the host from which it migrated. */
2660 bundle_flush_macs(struct ofbundle
*bundle
, bool all_ofprotos
)
2662 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2663 struct mac_learning
*ml
= ofproto
->ml
;
2664 struct mac_entry
*mac
, *next_mac
;
2666 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2667 ovs_rwlock_wrlock(&ml
->rwlock
);
2668 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2669 if (mac_entry_get_port(ml
, mac
) == bundle
) {
2671 struct ofproto_dpif
*o
;
2673 HMAP_FOR_EACH (o
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
2675 struct mac_entry
*e
;
2677 ovs_rwlock_wrlock(&o
->ml
->rwlock
);
2678 e
= mac_learning_lookup(o
->ml
, mac
->mac
, mac
->vlan
);
2680 mac_learning_expire(o
->ml
, e
);
2682 ovs_rwlock_unlock(&o
->ml
->rwlock
);
2687 mac_learning_expire(ml
, mac
);
2690 ovs_rwlock_unlock(&ml
->rwlock
);
2694 bundle_move(struct ofbundle
*old
, struct ofbundle
*new)
2696 struct ofproto_dpif
*ofproto
= old
->ofproto
;
2697 struct mac_learning
*ml
= ofproto
->ml
;
2698 struct mac_entry
*mac
, *next_mac
;
2700 ovs_assert(new->ofproto
== old
->ofproto
);
2702 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2703 ovs_rwlock_wrlock(&ml
->rwlock
);
2704 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2705 if (mac_entry_get_port(ml
, mac
) == old
) {
2706 mac_entry_set_port(ml
, mac
, new);
2709 ovs_rwlock_unlock(&ml
->rwlock
);
2712 static struct ofbundle
*
2713 bundle_lookup(const struct ofproto_dpif
*ofproto
, void *aux
)
2715 struct ofbundle
*bundle
;
2717 HMAP_FOR_EACH_IN_BUCKET (bundle
, hmap_node
, hash_pointer(aux
, 0),
2718 &ofproto
->bundles
) {
2719 if (bundle
->aux
== aux
) {
2727 bundle_update(struct ofbundle
*bundle
)
2729 struct ofport_dpif
*port
;
2731 bundle
->floodable
= true;
2732 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2733 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2735 || (bundle
->ofproto
->stp
&& !stp_forward_in_state(port
->stp_state
))
2736 || (bundle
->ofproto
->rstp
&& !rstp_forward_in_state(port
->rstp_state
))) {
2737 bundle
->floodable
= false;
2744 bundle_del_port(struct ofport_dpif
*port
)
2746 struct ofbundle
*bundle
= port
->bundle
;
2748 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2750 ovs_list_remove(&port
->bundle_node
);
2751 port
->bundle
= NULL
;
2754 lacp_slave_unregister(bundle
->lacp
, port
);
2757 bond_slave_unregister(bundle
->bond
, port
);
2760 bundle_update(bundle
);
2764 bundle_add_port(struct ofbundle
*bundle
, ofp_port_t ofp_port
,
2765 struct lacp_slave_settings
*lacp
)
2767 struct ofport_dpif
*port
;
2769 port
= ofp_port_to_ofport(bundle
->ofproto
, ofp_port
);
2774 if (port
->bundle
!= bundle
) {
2775 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2777 bundle_remove(&port
->up
);
2780 port
->bundle
= bundle
;
2781 ovs_list_push_back(&bundle
->ports
, &port
->bundle_node
);
2782 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2784 || (bundle
->ofproto
->stp
&& !stp_forward_in_state(port
->stp_state
))
2785 || (bundle
->ofproto
->rstp
&& !rstp_forward_in_state(port
->rstp_state
))) {
2786 bundle
->floodable
= false;
2790 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2791 lacp_slave_register(bundle
->lacp
, port
, lacp
);
2798 bundle_destroy(struct ofbundle
*bundle
)
2800 struct ofproto_dpif
*ofproto
;
2801 struct ofport_dpif
*port
, *next_port
;
2807 ofproto
= bundle
->ofproto
;
2808 mbridge_unregister_bundle(ofproto
->mbridge
, bundle
);
2811 xlate_bundle_remove(bundle
);
2814 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2815 bundle_del_port(port
);
2818 bundle_flush_macs(bundle
, true);
2819 mcast_snooping_flush_bundle(ofproto
->ms
, bundle
);
2820 hmap_remove(&ofproto
->bundles
, &bundle
->hmap_node
);
2822 free(bundle
->trunks
);
2823 lacp_unref(bundle
->lacp
);
2824 bond_unref(bundle
->bond
);
2829 bundle_set(struct ofproto
*ofproto_
, void *aux
,
2830 const struct ofproto_bundle_settings
*s
)
2832 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2833 bool need_flush
= false;
2834 struct ofport_dpif
*port
;
2835 struct ofbundle
*bundle
;
2836 unsigned long *trunks
;
2842 bundle_destroy(bundle_lookup(ofproto
, aux
));
2846 ovs_assert(s
->n_slaves
== 1 || s
->bond
!= NULL
);
2847 ovs_assert((s
->lacp
!= NULL
) == (s
->lacp_slaves
!= NULL
));
2849 bundle
= bundle_lookup(ofproto
, aux
);
2851 bundle
= xmalloc(sizeof *bundle
);
2853 bundle
->ofproto
= ofproto
;
2854 hmap_insert(&ofproto
->bundles
, &bundle
->hmap_node
,
2855 hash_pointer(aux
, 0));
2857 bundle
->name
= NULL
;
2859 ovs_list_init(&bundle
->ports
);
2860 bundle
->vlan_mode
= PORT_VLAN_TRUNK
;
2862 bundle
->trunks
= NULL
;
2863 bundle
->use_priority_tags
= s
->use_priority_tags
;
2864 bundle
->lacp
= NULL
;
2865 bundle
->bond
= NULL
;
2867 bundle
->floodable
= true;
2868 bundle
->protected = false;
2869 mbridge_register_bundle(ofproto
->mbridge
, bundle
);
2872 if (!bundle
->name
|| strcmp(s
->name
, bundle
->name
)) {
2874 bundle
->name
= xstrdup(s
->name
);
2879 ofproto
->lacp_enabled
= true;
2880 if (!bundle
->lacp
) {
2881 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2882 bundle
->lacp
= lacp_create();
2884 lacp_configure(bundle
->lacp
, s
->lacp
);
2886 lacp_unref(bundle
->lacp
);
2887 bundle
->lacp
= NULL
;
2890 /* Update set of ports. */
2892 for (i
= 0; i
< s
->n_slaves
; i
++) {
2893 if (!bundle_add_port(bundle
, s
->slaves
[i
],
2894 s
->lacp
? &s
->lacp_slaves
[i
] : NULL
)) {
2898 if (!ok
|| ovs_list_size(&bundle
->ports
) != s
->n_slaves
) {
2899 struct ofport_dpif
*next_port
;
2901 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2902 for (i
= 0; i
< s
->n_slaves
; i
++) {
2903 if (s
->slaves
[i
] == port
->up
.ofp_port
) {
2908 bundle_del_port(port
);
2912 ovs_assert(ovs_list_size(&bundle
->ports
) <= s
->n_slaves
);
2914 if (ovs_list_is_empty(&bundle
->ports
)) {
2915 bundle_destroy(bundle
);
2919 /* Set VLAN tagging mode */
2920 if (s
->vlan_mode
!= bundle
->vlan_mode
2921 || s
->use_priority_tags
!= bundle
->use_priority_tags
) {
2922 bundle
->vlan_mode
= s
->vlan_mode
;
2923 bundle
->use_priority_tags
= s
->use_priority_tags
;
2928 vlan
= (s
->vlan_mode
== PORT_VLAN_TRUNK
? -1
2929 : s
->vlan
>= 0 && s
->vlan
<= 4095 ? s
->vlan
2931 if (vlan
!= bundle
->vlan
) {
2932 bundle
->vlan
= vlan
;
2936 /* Get trunked VLANs. */
2937 switch (s
->vlan_mode
) {
2938 case PORT_VLAN_ACCESS
:
2942 case PORT_VLAN_TRUNK
:
2943 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2946 case PORT_VLAN_NATIVE_UNTAGGED
:
2947 case PORT_VLAN_NATIVE_TAGGED
:
2948 if (vlan
!= 0 && (!s
->trunks
2949 || !bitmap_is_set(s
->trunks
, vlan
)
2950 || bitmap_is_set(s
->trunks
, 0))) {
2951 /* Force trunking the native VLAN and prohibit trunking VLAN 0. */
2953 trunks
= bitmap_clone(s
->trunks
, 4096);
2955 trunks
= bitmap_allocate1(4096);
2957 bitmap_set1(trunks
, vlan
);
2958 bitmap_set0(trunks
, 0);
2960 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2967 if (!vlan_bitmap_equal(trunks
, bundle
->trunks
)) {
2968 free(bundle
->trunks
);
2969 if (trunks
== s
->trunks
) {
2970 bundle
->trunks
= vlan_bitmap_clone(trunks
);
2972 bundle
->trunks
= trunks
;
2977 if (trunks
!= s
->trunks
) {
2982 if (!ovs_list_is_short(&bundle
->ports
)) {
2983 bundle
->ofproto
->has_bonded_bundles
= true;
2985 if (bond_reconfigure(bundle
->bond
, s
->bond
)) {
2986 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2989 bundle
->bond
= bond_create(s
->bond
, ofproto
);
2990 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2993 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2994 bond_slave_register(bundle
->bond
, port
,
2995 port
->up
.ofp_port
, port
->up
.netdev
);
2998 bond_unref(bundle
->bond
);
2999 bundle
->bond
= NULL
;
3002 /* Set proteced port mode */
3003 if (s
->protected != bundle
->protected) {
3004 bundle
->protected = s
->protected;
3008 /* If we changed something that would affect MAC learning, un-learn
3009 * everything on this port and force flow revalidation. */
3011 bundle_flush_macs(bundle
, false);
3012 mcast_snooping_flush_bundle(ofproto
->ms
, bundle
);
3019 bundle_remove(struct ofport
*port_
)
3021 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
3022 struct ofbundle
*bundle
= port
->bundle
;
3025 bundle_del_port(port
);
3026 if (ovs_list_is_empty(&bundle
->ports
)) {
3027 bundle_destroy(bundle
);
3028 } else if (ovs_list_is_short(&bundle
->ports
)) {
3029 bond_unref(bundle
->bond
);
3030 bundle
->bond
= NULL
;
3036 send_pdu_cb(void *port_
, const void *pdu
, size_t pdu_size
)
3038 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 10);
3039 struct ofport_dpif
*port
= port_
;
3043 error
= netdev_get_etheraddr(port
->up
.netdev
, &ea
);
3045 struct dp_packet packet
;
3048 dp_packet_init(&packet
, 0);
3049 packet_pdu
= eth_compose(&packet
, eth_addr_lacp
, ea
, ETH_TYPE_LACP
,
3051 memcpy(packet_pdu
, pdu
, pdu_size
);
3053 ofproto_dpif_send_packet(port
, false, &packet
);
3054 dp_packet_uninit(&packet
);
3056 VLOG_ERR_RL(&rl
, "port %s: cannot obtain Ethernet address of iface "
3057 "%s (%s)", port
->bundle
->name
,
3058 netdev_get_name(port
->up
.netdev
), ovs_strerror(error
));
3063 bundle_send_learning_packets(struct ofbundle
*bundle
)
3065 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
3066 int error
, n_packets
, n_errors
;
3067 struct mac_entry
*e
;
3069 struct ovs_list list_node
;
3070 struct ofport_dpif
*port
;
3071 struct dp_packet
*pkt
;
3073 struct ovs_list packets
;
3075 ovs_list_init(&packets
);
3076 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
3077 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
3078 if (mac_entry_get_port(ofproto
->ml
, e
) != bundle
) {
3079 pkt_node
= xmalloc(sizeof *pkt_node
);
3080 pkt_node
->pkt
= bond_compose_learning_packet(bundle
->bond
,
3082 (void **)&pkt_node
->port
);
3083 ovs_list_push_back(&packets
, &pkt_node
->list_node
);
3086 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3088 error
= n_packets
= n_errors
= 0;
3089 LIST_FOR_EACH_POP (pkt_node
, list_node
, &packets
) {
3092 ret
= ofproto_dpif_send_packet(pkt_node
->port
, false, pkt_node
->pkt
);
3093 dp_packet_delete(pkt_node
->pkt
);
3103 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3104 VLOG_WARN_RL(&rl
, "bond %s: %d errors sending %d gratuitous learning "
3105 "packets, last error was: %s",
3106 bundle
->name
, n_errors
, n_packets
, ovs_strerror(error
));
3108 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3109 bundle
->name
, n_packets
);
3114 bundle_run(struct ofbundle
*bundle
)
3117 lacp_run(bundle
->lacp
, send_pdu_cb
);
3120 struct ofport_dpif
*port
;
3122 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
3123 bond_slave_set_may_enable(bundle
->bond
, port
, port
->may_enable
);
3126 if (bond_run(bundle
->bond
, lacp_status(bundle
->lacp
))) {
3127 bundle
->ofproto
->backer
->need_revalidate
= REV_BOND
;
3130 if (bond_should_send_learning_packets(bundle
->bond
)) {
3131 bundle_send_learning_packets(bundle
);
3137 bundle_wait(struct ofbundle
*bundle
)
3140 lacp_wait(bundle
->lacp
);
3143 bond_wait(bundle
->bond
);
3150 mirror_set__(struct ofproto
*ofproto_
, void *aux
,
3151 const struct ofproto_mirror_settings
*s
)
3153 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3154 struct ofbundle
**srcs
, **dsts
;
3159 mirror_destroy(ofproto
->mbridge
, aux
);
3163 srcs
= xmalloc(s
->n_srcs
* sizeof *srcs
);
3164 dsts
= xmalloc(s
->n_dsts
* sizeof *dsts
);
3166 for (i
= 0; i
< s
->n_srcs
; i
++) {
3167 srcs
[i
] = bundle_lookup(ofproto
, s
->srcs
[i
]);
3170 for (i
= 0; i
< s
->n_dsts
; i
++) {
3171 dsts
[i
] = bundle_lookup(ofproto
, s
->dsts
[i
]);
3174 error
= mirror_set(ofproto
->mbridge
, aux
, s
->name
, srcs
, s
->n_srcs
, dsts
,
3175 s
->n_dsts
, s
->src_vlans
,
3176 bundle_lookup(ofproto
, s
->out_bundle
),
3177 s
->snaplen
, s
->out_vlan
);
3184 mirror_get_stats__(struct ofproto
*ofproto
, void *aux
,
3185 uint64_t *packets
, uint64_t *bytes
)
3187 return mirror_get_stats(ofproto_dpif_cast(ofproto
)->mbridge
, aux
, packets
,
3192 set_flood_vlans(struct ofproto
*ofproto_
, unsigned long *flood_vlans
)
3194 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3195 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
3196 if (mac_learning_set_flood_vlans(ofproto
->ml
, flood_vlans
)) {
3197 mac_learning_flush(ofproto
->ml
);
3199 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3204 is_mirror_output_bundle(const struct ofproto
*ofproto_
, void *aux
)
3206 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3207 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3208 return bundle
&& mirror_bundle_out(ofproto
->mbridge
, bundle
) != 0;
3212 forward_bpdu_changed(struct ofproto
*ofproto_
)
3214 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3215 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3219 set_mac_table_config(struct ofproto
*ofproto_
, unsigned int idle_time
,
3222 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3223 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
3224 mac_learning_set_idle_time(ofproto
->ml
, idle_time
);
3225 mac_learning_set_max_entries(ofproto
->ml
, max_entries
);
3226 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3229 /* Configures multicast snooping on 'ofport' using the settings
3230 * defined in 's'. */
3232 set_mcast_snooping(struct ofproto
*ofproto_
,
3233 const struct ofproto_mcast_snooping_settings
*s
)
3235 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3237 /* Only revalidate flows if the configuration changed. */
3238 if (!s
!= !ofproto
->ms
) {
3239 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3244 ofproto
->ms
= mcast_snooping_create();
3247 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3248 mcast_snooping_set_idle_time(ofproto
->ms
, s
->idle_time
);
3249 mcast_snooping_set_max_entries(ofproto
->ms
, s
->max_entries
);
3250 if (mcast_snooping_set_flood_unreg(ofproto
->ms
, s
->flood_unreg
)) {
3251 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3253 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3255 mcast_snooping_unref(ofproto
->ms
);
3262 /* Configures multicast snooping port's flood settings on 'ofproto'. */
3264 set_mcast_snooping_port(struct ofproto
*ofproto_
, void *aux
,
3265 const struct ofproto_mcast_snooping_port_settings
*s
)
3267 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3268 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3270 if (ofproto
->ms
&& s
) {
3271 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3272 mcast_snooping_set_port_flood(ofproto
->ms
, bundle
, s
->flood
);
3273 mcast_snooping_set_port_flood_reports(ofproto
->ms
, bundle
,
3275 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3283 struct ofport_dpif
*
3284 ofp_port_to_ofport(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
3286 struct ofport
*ofport
= ofproto_get_port(&ofproto
->up
, ofp_port
);
3287 return ofport
? ofport_dpif_cast(ofport
) : NULL
;
3291 ofproto_port_from_dpif_port(struct ofproto_dpif
*ofproto
,
3292 struct ofproto_port
*ofproto_port
,
3293 struct dpif_port
*dpif_port
)
3295 ofproto_port
->name
= dpif_port
->name
;
3296 ofproto_port
->type
= dpif_port
->type
;
3297 ofproto_port
->ofp_port
= odp_port_to_ofp_port(ofproto
, dpif_port
->port_no
);
3301 ofport_update_peer(struct ofport_dpif
*ofport
)
3303 const struct ofproto_dpif
*ofproto
;
3304 struct dpif_backer
*backer
;
3307 if (!netdev_vport_is_patch(ofport
->up
.netdev
)) {
3311 backer
= ofproto_dpif_cast(ofport
->up
.ofproto
)->backer
;
3312 backer
->need_revalidate
= REV_RECONFIGURE
;
3315 ofport
->peer
->peer
= NULL
;
3316 ofport
->peer
= NULL
;
3319 peer_name
= netdev_vport_patch_peer(ofport
->up
.netdev
);
3324 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
3325 struct ofport
*peer_ofport
;
3326 struct ofport_dpif
*peer
;
3329 if (ofproto
->backer
!= backer
) {
3333 peer_ofport
= shash_find_data(&ofproto
->up
.port_by_name
, peer_name
);
3338 peer
= ofport_dpif_cast(peer_ofport
);
3339 peer_peer
= netdev_vport_patch_peer(peer
->up
.netdev
);
3340 if (peer_peer
&& !strcmp(netdev_get_name(ofport
->up
.netdev
),
3342 ofport
->peer
= peer
;
3343 ofport
->peer
->peer
= ofport
;
3353 port_run(struct ofport_dpif
*ofport
)
3355 long long int carrier_seq
= netdev_get_carrier_resets(ofport
->up
.netdev
);
3356 bool carrier_changed
= carrier_seq
!= ofport
->carrier_seq
;
3357 bool enable
= netdev_get_carrier(ofport
->up
.netdev
);
3358 bool cfm_enable
= false;
3359 bool bfd_enable
= false;
3361 ofport
->carrier_seq
= carrier_seq
;
3364 int cfm_opup
= cfm_get_opup(ofport
->cfm
);
3366 cfm_enable
= !cfm_get_fault(ofport
->cfm
);
3368 if (cfm_opup
>= 0) {
3369 cfm_enable
= cfm_enable
&& cfm_opup
;
3374 bfd_enable
= bfd_forwarding(ofport
->bfd
);
3377 if (ofport
->bfd
|| ofport
->cfm
) {
3378 enable
= enable
&& (cfm_enable
|| bfd_enable
);
3381 if (ofport
->bundle
) {
3382 enable
= enable
&& lacp_slave_may_enable(ofport
->bundle
->lacp
, ofport
);
3383 if (carrier_changed
) {
3384 lacp_slave_carrier_changed(ofport
->bundle
->lacp
, ofport
);
3388 if (ofport
->may_enable
!= enable
) {
3389 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3391 ofproto
->backer
->need_revalidate
= REV_PORT_TOGGLED
;
3393 if (ofport
->rstp_port
) {
3394 rstp_port_set_mac_operational(ofport
->rstp_port
, enable
);
3398 ofport
->may_enable
= enable
;
3402 port_query_by_name(const struct ofproto
*ofproto_
, const char *devname
,
3403 struct ofproto_port
*ofproto_port
)
3405 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3406 struct dpif_port dpif_port
;
3409 if (sset_contains(&ofproto
->ghost_ports
, devname
)) {
3410 const char *type
= netdev_get_type_from_name(devname
);
3412 /* We may be called before ofproto->up.port_by_name is populated with
3413 * the appropriate ofport. For this reason, we must get the name and
3414 * type from the netdev layer directly. */
3416 const struct ofport
*ofport
;
3418 ofport
= shash_find_data(&ofproto
->up
.port_by_name
, devname
);
3419 ofproto_port
->ofp_port
= ofport
? ofport
->ofp_port
: OFPP_NONE
;
3420 ofproto_port
->name
= xstrdup(devname
);
3421 ofproto_port
->type
= xstrdup(type
);
3427 if (!sset_contains(&ofproto
->ports
, devname
)) {
3430 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
3431 devname
, &dpif_port
);
3433 ofproto_port_from_dpif_port(ofproto
, ofproto_port
, &dpif_port
);
3439 port_add(struct ofproto
*ofproto_
, struct netdev
*netdev
)
3441 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3442 const char *devname
= netdev_get_name(netdev
);
3443 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
3444 const char *dp_port_name
;
3446 if (netdev_vport_is_patch(netdev
)) {
3447 sset_add(&ofproto
->ghost_ports
, netdev_get_name(netdev
));
3451 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
3452 if (!dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
3453 odp_port_t port_no
= ODPP_NONE
;
3456 error
= dpif_port_add(ofproto
->backer
->dpif
, netdev
, &port_no
);
3460 if (netdev_get_tunnel_config(netdev
)) {
3461 simap_put(&ofproto
->backer
->tnl_backers
,
3462 dp_port_name
, odp_to_u32(port_no
));
3466 if (netdev_get_tunnel_config(netdev
)) {
3467 sset_add(&ofproto
->ghost_ports
, devname
);
3469 sset_add(&ofproto
->ports
, devname
);
3475 port_del(struct ofproto
*ofproto_
, ofp_port_t ofp_port
)
3477 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3478 struct ofport_dpif
*ofport
= ofp_port_to_ofport(ofproto
, ofp_port
);
3485 sset_find_and_delete(&ofproto
->ghost_ports
,
3486 netdev_get_name(ofport
->up
.netdev
));
3487 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3488 if (!ofport
->is_tunnel
&& !netdev_vport_is_patch(ofport
->up
.netdev
)) {
3489 error
= dpif_port_del(ofproto
->backer
->dpif
, ofport
->odp_port
);
3491 /* The caller is going to close ofport->up.netdev. If this is a
3492 * bonded port, then the bond is using that netdev, so remove it
3493 * from the bond. The client will need to reconfigure everything
3494 * after deleting ports, so then the slave will get re-added. */
3495 bundle_remove(&ofport
->up
);
3502 port_set_config(const struct ofport
*ofport_
, const struct smap
*cfg
)
3504 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3505 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3507 if (sset_contains(&ofproto
->ghost_ports
,
3508 netdev_get_name(ofport
->up
.netdev
))) {
3512 return dpif_port_set_config(ofproto
->backer
->dpif
, ofport
->odp_port
, cfg
);
3516 port_get_stats(const struct ofport
*ofport_
, struct netdev_stats
*stats
)
3518 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3521 error
= netdev_get_stats(ofport
->up
.netdev
, stats
);
3523 if (!error
&& ofport_
->ofp_port
== OFPP_LOCAL
) {
3524 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3526 ovs_mutex_lock(&ofproto
->stats_mutex
);
3527 /* ofproto->stats.tx_packets represents packets that we created
3528 * internally and sent to some port (e.g. packets sent with
3529 * ofproto_dpif_send_packet()). Account for them as if they had
3530 * come from OFPP_LOCAL and got forwarded. */
3532 if (stats
->rx_packets
!= UINT64_MAX
) {
3533 stats
->rx_packets
+= ofproto
->stats
.tx_packets
;
3536 if (stats
->rx_bytes
!= UINT64_MAX
) {
3537 stats
->rx_bytes
+= ofproto
->stats
.tx_bytes
;
3540 /* ofproto->stats.rx_packets represents packets that were received on
3541 * some port and we processed internally and dropped (e.g. STP).
3542 * Account for them as if they had been forwarded to OFPP_LOCAL. */
3544 if (stats
->tx_packets
!= UINT64_MAX
) {
3545 stats
->tx_packets
+= ofproto
->stats
.rx_packets
;
3548 if (stats
->tx_bytes
!= UINT64_MAX
) {
3549 stats
->tx_bytes
+= ofproto
->stats
.rx_bytes
;
3551 ovs_mutex_unlock(&ofproto
->stats_mutex
);
3558 port_get_lacp_stats(const struct ofport
*ofport_
, struct lacp_slave_stats
*stats
)
3560 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3561 if (ofport
->bundle
&& ofport
->bundle
->lacp
) {
3562 if (lacp_get_slave_stats(ofport
->bundle
->lacp
, ofport
, stats
)) {
3569 struct port_dump_state
{
3570 struct sset_position pos
;
3573 struct ofproto_port port
;
3578 port_dump_start(const struct ofproto
*ofproto_ OVS_UNUSED
, void **statep
)
3580 *statep
= xzalloc(sizeof(struct port_dump_state
));
3585 port_dump_next(const struct ofproto
*ofproto_
, void *state_
,
3586 struct ofproto_port
*port
)
3588 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3589 struct port_dump_state
*state
= state_
;
3590 const struct sset
*sset
;
3591 struct sset_node
*node
;
3593 if (state
->has_port
) {
3594 ofproto_port_destroy(&state
->port
);
3595 state
->has_port
= false;
3597 sset
= state
->ghost
? &ofproto
->ghost_ports
: &ofproto
->ports
;
3598 while ((node
= sset_at_position(sset
, &state
->pos
))) {
3601 error
= port_query_by_name(ofproto_
, node
->name
, &state
->port
);
3603 *port
= state
->port
;
3604 state
->has_port
= true;
3606 } else if (error
!= ENODEV
) {
3611 if (!state
->ghost
) {
3612 state
->ghost
= true;
3613 memset(&state
->pos
, 0, sizeof state
->pos
);
3614 return port_dump_next(ofproto_
, state_
, port
);
3621 port_dump_done(const struct ofproto
*ofproto_ OVS_UNUSED
, void *state_
)
3623 struct port_dump_state
*state
= state_
;
3625 if (state
->has_port
) {
3626 ofproto_port_destroy(&state
->port
);
3633 port_poll(const struct ofproto
*ofproto_
, char **devnamep
)
3635 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3637 if (ofproto
->port_poll_errno
) {
3638 int error
= ofproto
->port_poll_errno
;
3639 ofproto
->port_poll_errno
= 0;
3643 if (sset_is_empty(&ofproto
->port_poll_set
)) {
3647 *devnamep
= sset_pop(&ofproto
->port_poll_set
);
3652 port_poll_wait(const struct ofproto
*ofproto_
)
3654 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3655 dpif_port_poll_wait(ofproto
->backer
->dpif
);
3659 port_is_lacp_current(const struct ofport
*ofport_
)
3661 const struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3662 return (ofport
->bundle
&& ofport
->bundle
->lacp
3663 ? lacp_slave_is_current(ofport
->bundle
->lacp
, ofport
)
3667 /* If 'rule' is an OpenFlow rule, that has expired according to OpenFlow rules,
3668 * then delete it entirely. */
3670 rule_expire(struct rule_dpif
*rule
, long long now
)
3671 OVS_REQUIRES(ofproto_mutex
)
3673 uint16_t hard_timeout
, idle_timeout
;
3676 hard_timeout
= rule
->up
.hard_timeout
;
3677 idle_timeout
= rule
->up
.idle_timeout
;
3679 /* Has 'rule' expired? */
3681 long long int modified
;
3683 ovs_mutex_lock(&rule
->up
.mutex
);
3684 modified
= rule
->up
.modified
;
3685 ovs_mutex_unlock(&rule
->up
.mutex
);
3687 if (now
> modified
+ hard_timeout
* 1000) {
3688 reason
= OFPRR_HARD_TIMEOUT
;
3692 if (reason
< 0 && idle_timeout
) {
3695 ovs_mutex_lock(&rule
->stats_mutex
);
3696 used
= rule
->stats
.used
;
3697 ovs_mutex_unlock(&rule
->stats_mutex
);
3699 if (now
> used
+ idle_timeout
* 1000) {
3700 reason
= OFPRR_IDLE_TIMEOUT
;
3705 COVERAGE_INC(ofproto_dpif_expired
);
3706 ofproto_rule_expire(&rule
->up
, reason
);
3711 ofproto_dpif_set_packet_odp_port(const struct ofproto_dpif
*ofproto
,
3712 ofp_port_t in_port
, struct dp_packet
*packet
)
3714 if (in_port
== OFPP_NONE
) {
3715 in_port
= OFPP_LOCAL
;
3717 packet
->md
.in_port
.odp_port
= ofp_port_to_odp_port(ofproto
, in_port
);
3721 ofproto_dpif_execute_actions__(struct ofproto_dpif
*ofproto
,
3722 ovs_version_t version
, const struct flow
*flow
,
3723 struct rule_dpif
*rule
,
3724 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3725 int depth
, int resubmits
,
3726 struct dp_packet
*packet
)
3728 struct dpif_flow_stats stats
;
3729 struct xlate_out xout
;
3730 struct xlate_in xin
;
3731 struct dpif_execute execute
;
3734 ovs_assert((rule
!= NULL
) != (ofpacts
!= NULL
));
3736 dpif_flow_stats_extract(flow
, packet
, time_msec(), &stats
);
3739 rule_dpif_credit_stats(rule
, &stats
);
3742 uint64_t odp_actions_stub
[1024 / 8];
3743 struct ofpbuf odp_actions
= OFPBUF_STUB_INITIALIZER(odp_actions_stub
);
3744 xlate_in_init(&xin
, ofproto
, version
, flow
, flow
->in_port
.ofp_port
, rule
,
3745 stats
.tcp_flags
, packet
, NULL
, &odp_actions
);
3746 xin
.ofpacts
= ofpacts
;
3747 xin
.ofpacts_len
= ofpacts_len
;
3748 xin
.resubmit_stats
= &stats
;
3750 xin
.resubmits
= resubmits
;
3751 if (xlate_actions(&xin
, &xout
) != XLATE_OK
) {
3756 execute
.actions
= odp_actions
.data
;
3757 execute
.actions_len
= odp_actions
.size
;
3759 pkt_metadata_from_flow(&packet
->md
, flow
);
3760 execute
.packet
= packet
;
3761 execute
.flow
= flow
;
3762 execute
.needs_help
= (xout
.slow
& SLOW_ACTION
) != 0;
3763 execute
.probe
= false;
3766 /* Fix up in_port. */
3767 ofproto_dpif_set_packet_odp_port(ofproto
, flow
->in_port
.ofp_port
, packet
);
3769 error
= dpif_execute(ofproto
->backer
->dpif
, &execute
);
3771 xlate_out_uninit(&xout
);
3772 ofpbuf_uninit(&odp_actions
);
3777 /* Executes, within 'ofproto', the actions in 'rule' or 'ofpacts' on 'packet'.
3778 * 'flow' must reflect the data in 'packet'. */
3780 ofproto_dpif_execute_actions(struct ofproto_dpif
*ofproto
,
3781 ovs_version_t version
, const struct flow
*flow
,
3782 struct rule_dpif
*rule
,
3783 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3784 struct dp_packet
*packet
)
3786 return ofproto_dpif_execute_actions__(ofproto
, version
, flow
, rule
,
3787 ofpacts
, ofpacts_len
, 0, 0, packet
);
3791 rule_dpif_credit_stats__(struct rule_dpif
*rule
,
3792 const struct dpif_flow_stats
*stats
,
3794 OVS_REQUIRES(rule
->stats_mutex
)
3796 if (credit_counts
) {
3797 rule
->stats
.n_packets
+= stats
->n_packets
;
3798 rule
->stats
.n_bytes
+= stats
->n_bytes
;
3800 rule
->stats
.used
= MAX(rule
->stats
.used
, stats
->used
);
3804 rule_dpif_credit_stats(struct rule_dpif
*rule
,
3805 const struct dpif_flow_stats
*stats
)
3807 ovs_mutex_lock(&rule
->stats_mutex
);
3808 if (OVS_UNLIKELY(rule
->new_rule
)) {
3809 ovs_mutex_lock(&rule
->new_rule
->stats_mutex
);
3810 rule_dpif_credit_stats__(rule
->new_rule
, stats
, rule
->forward_counts
);
3811 ovs_mutex_unlock(&rule
->new_rule
->stats_mutex
);
3813 rule_dpif_credit_stats__(rule
, stats
, true);
3815 ovs_mutex_unlock(&rule
->stats_mutex
);
3818 /* Sets 'rule''s recirculation id. */
3820 rule_dpif_set_recirc_id(struct rule_dpif
*rule
, uint32_t id
)
3821 OVS_REQUIRES(rule
->up
.mutex
)
3823 ovs_assert(!rule
->recirc_id
|| rule
->recirc_id
== id
);
3824 if (rule
->recirc_id
== id
) {
3825 /* Release the new reference to the same id. */
3828 rule
->recirc_id
= id
;
3832 /* Sets 'rule''s recirculation id. */
3834 rule_set_recirc_id(struct rule
*rule_
, uint32_t id
)
3836 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3838 ovs_mutex_lock(&rule
->up
.mutex
);
3839 rule_dpif_set_recirc_id(rule
, id
);
3840 ovs_mutex_unlock(&rule
->up
.mutex
);
3844 ofproto_dpif_get_tables_version(struct ofproto_dpif
*ofproto
)
3846 ovs_version_t version
;
3848 /* Use memory_order_acquire to signify that any following memory accesses
3849 * can not be reordered to happen before this atomic read. This makes sure
3850 * all following reads relate to this or a newer version, but never to an
3852 atomic_read_explicit(&ofproto
->tables_version
, &version
,
3853 memory_order_acquire
);
3857 /* The returned rule (if any) is valid at least until the next RCU quiescent
3858 * period. If the rule needs to stay around longer, the caller should take
3861 * 'flow' is non-const to allow for temporary modifications during the lookup.
3862 * Any changes are restored before returning. */
3863 static struct rule_dpif
*
3864 rule_dpif_lookup_in_table(struct ofproto_dpif
*ofproto
, ovs_version_t version
,
3865 uint8_t table_id
, struct flow
*flow
,
3866 struct flow_wildcards
*wc
)
3868 struct classifier
*cls
= &ofproto
->up
.tables
[table_id
].cls
;
3869 return rule_dpif_cast(rule_from_cls_rule(classifier_lookup(cls
, version
,
3874 ofproto_dpif_credit_table_stats(struct ofproto_dpif
*ofproto
, uint8_t table_id
,
3875 uint64_t n_matches
, uint64_t n_misses
)
3877 struct oftable
*tbl
= &ofproto
->up
.tables
[table_id
];
3881 atomic_add_relaxed(&tbl
->n_matched
, n_matches
, &orig
);
3884 atomic_add_relaxed(&tbl
->n_missed
, n_misses
, &orig
);
3888 /* Look up 'flow' in 'ofproto''s classifier version 'version', starting from
3889 * table '*table_id'. Returns the rule that was found, which may be one of the
3890 * special rules according to packet miss hadling. If 'may_packet_in' is
3891 * false, returning of the miss_rule (which issues packet ins for the
3892 * controller) is avoided. Updates 'wc', if nonnull, to reflect the fields
3893 * that were used during the lookup.
3895 * If 'honor_table_miss' is true, the first lookup occurs in '*table_id', but
3896 * if none is found then the table miss configuration for that table is
3897 * honored, which can result in additional lookups in other OpenFlow tables.
3898 * In this case the function updates '*table_id' to reflect the final OpenFlow
3899 * table that was searched.
3901 * If 'honor_table_miss' is false, then only one table lookup occurs, in
3904 * The rule is returned in '*rule', which is valid at least until the next
3905 * RCU quiescent period. If the '*rule' needs to stay around longer, the
3906 * caller must take a reference.
3908 * 'in_port' allows the lookup to take place as if the in port had the value
3909 * 'in_port'. This is needed for resubmit action support.
3911 * 'flow' is non-const to allow for temporary modifications during the lookup.
3912 * Any changes are restored before returning. */
3914 rule_dpif_lookup_from_table(struct ofproto_dpif
*ofproto
,
3915 ovs_version_t version
, struct flow
*flow
,
3916 struct flow_wildcards
*wc
,
3917 const struct dpif_flow_stats
*stats
,
3918 uint8_t *table_id
, ofp_port_t in_port
,
3919 bool may_packet_in
, bool honor_table_miss
,
3920 struct xlate_cache
*xcache
)
3922 ovs_be16 old_tp_src
= flow
->tp_src
, old_tp_dst
= flow
->tp_dst
;
3923 ofp_port_t old_in_port
= flow
->in_port
.ofp_port
;
3924 enum ofputil_table_miss miss_config
;
3925 struct rule_dpif
*rule
;
3928 /* We always unwildcard nw_frag (for IP), so they
3929 * need not be unwildcarded here. */
3930 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
3931 && ofproto
->up
.frag_handling
!= OFPUTIL_FRAG_NX_MATCH
) {
3932 if (ofproto
->up
.frag_handling
== OFPUTIL_FRAG_NORMAL
) {
3933 /* We must pretend that transport ports are unavailable. */
3934 flow
->tp_src
= htons(0);
3935 flow
->tp_dst
= htons(0);
3937 /* Must be OFPUTIL_FRAG_DROP (we don't have OFPUTIL_FRAG_REASM).
3938 * Use the drop_frags_rule (which cannot disappear). */
3939 rule
= ofproto
->drop_frags_rule
;
3941 struct oftable
*tbl
= &ofproto
->up
.tables
[*table_id
];
3944 atomic_add_relaxed(&tbl
->n_matched
, stats
->n_packets
, &orig
);
3947 struct xc_entry
*entry
;
3949 entry
= xlate_cache_add_entry(xcache
, XC_TABLE
);
3950 entry
->table
.ofproto
= ofproto
;
3951 entry
->table
.id
= *table_id
;
3952 entry
->table
.match
= true;
3958 /* Look up a flow with 'in_port' as the input port. Then restore the
3959 * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
3960 * have surprising behavior). */
3961 flow
->in_port
.ofp_port
= in_port
;
3963 /* Our current implementation depends on n_tables == N_TABLES, and
3964 * TBL_INTERNAL being the last table. */
3965 BUILD_ASSERT_DECL(N_TABLES
== TBL_INTERNAL
+ 1);
3967 miss_config
= OFPUTIL_TABLE_MISS_CONTINUE
;
3969 for (next_id
= *table_id
;
3970 next_id
< ofproto
->up
.n_tables
;
3971 next_id
++, next_id
+= (next_id
== TBL_INTERNAL
))
3973 *table_id
= next_id
;
3974 rule
= rule_dpif_lookup_in_table(ofproto
, version
, next_id
, flow
, wc
);
3976 struct oftable
*tbl
= &ofproto
->up
.tables
[next_id
];
3979 atomic_add_relaxed(rule
? &tbl
->n_matched
: &tbl
->n_missed
,
3980 stats
->n_packets
, &orig
);
3983 struct xc_entry
*entry
;
3985 entry
= xlate_cache_add_entry(xcache
, XC_TABLE
);
3986 entry
->table
.ofproto
= ofproto
;
3987 entry
->table
.id
= next_id
;
3988 entry
->table
.match
= (rule
!= NULL
);
3991 goto out
; /* Match. */
3993 if (honor_table_miss
) {
3994 miss_config
= ofproto_table_get_miss_config(&ofproto
->up
,
3996 if (miss_config
== OFPUTIL_TABLE_MISS_CONTINUE
) {
4003 rule
= ofproto
->no_packet_in_rule
;
4004 if (may_packet_in
) {
4005 if (miss_config
== OFPUTIL_TABLE_MISS_CONTINUE
4006 || miss_config
== OFPUTIL_TABLE_MISS_CONTROLLER
) {
4007 struct ofport_dpif
*port
;
4009 port
= ofp_port_to_ofport(ofproto
, old_in_port
);
4011 VLOG_WARN_RL(&rl
, "packet-in on unknown OpenFlow port %"PRIu32
,
4013 } else if (!(port
->up
.pp
.config
& OFPUTIL_PC_NO_PACKET_IN
)) {
4014 rule
= ofproto
->miss_rule
;
4016 } else if (miss_config
== OFPUTIL_TABLE_MISS_DEFAULT
&&
4017 connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
)) {
4018 rule
= ofproto
->miss_rule
;
4022 /* Restore port numbers, as they may have been modified above. */
4023 flow
->tp_src
= old_tp_src
;
4024 flow
->tp_dst
= old_tp_dst
;
4025 /* Restore the old in port. */
4026 flow
->in_port
.ofp_port
= old_in_port
;
4031 static struct rule_dpif
*rule_dpif_cast(const struct rule
*rule
)
4033 return rule
? CONTAINER_OF(rule
, struct rule_dpif
, up
) : NULL
;
4036 static struct rule
*
4039 struct rule_dpif
*rule
= xzalloc(sizeof *rule
);
4044 rule_dealloc(struct rule
*rule_
)
4046 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4051 check_mask(struct ofproto_dpif
*ofproto
, const struct miniflow
*flow
)
4053 const struct odp_support
*support
;
4054 uint16_t ct_state
, ct_zone
;
4058 support
= &ofproto
->backer
->support
.odp
;
4059 ct_state
= MINIFLOW_GET_U8(flow
, ct_state
);
4060 if (support
->ct_state
&& support
->ct_zone
&& support
->ct_mark
4061 && support
->ct_label
&& support
->ct_state_nat
) {
4062 return ct_state
& CS_UNSUPPORTED_MASK
? OFPERR_OFPBMC_BAD_MASK
: 0;
4065 ct_zone
= MINIFLOW_GET_U16(flow
, ct_zone
);
4066 ct_mark
= MINIFLOW_GET_U32(flow
, ct_mark
);
4067 ct_label
= MINIFLOW_GET_U128(flow
, ct_label
);
4069 if ((ct_state
&& !support
->ct_state
)
4070 || (ct_state
& CS_UNSUPPORTED_MASK
)
4071 || ((ct_state
& (CS_SRC_NAT
| CS_DST_NAT
)) && !support
->ct_state_nat
)
4072 || (ct_zone
&& !support
->ct_zone
)
4073 || (ct_mark
&& !support
->ct_mark
)
4074 || (!ovs_u128_is_zero(ct_label
) && !support
->ct_label
)) {
4075 return OFPERR_OFPBMC_BAD_MASK
;
4082 report_unsupported_ct(const char *detail
)
4084 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
4085 VLOG_WARN_RL(&rl
, "Rejecting ct action because datapath does not support "
4086 "ct action%s%s (your kernel module may be out of date)",
4088 detail
? detail
: "");
4092 check_actions(const struct ofproto_dpif
*ofproto
,
4093 const struct rule_actions
*const actions
)
4095 const struct ofpact
*ofpact
;
4097 OFPACT_FOR_EACH (ofpact
, actions
->ofpacts
, actions
->ofpacts_len
) {
4098 const struct odp_support
*support
;
4099 const struct ofpact_conntrack
*ct
;
4100 const struct ofpact
*a
;
4102 if (ofpact
->type
!= OFPACT_CT
) {
4106 ct
= CONTAINER_OF(ofpact
, struct ofpact_conntrack
, ofpact
);
4107 support
= &ofproto
->backer
->support
.odp
;
4109 if (!support
->ct_state
) {
4110 report_unsupported_ct(NULL
);
4111 return OFPERR_OFPBAC_BAD_TYPE
;
4113 if ((ct
->zone_imm
|| ct
->zone_src
.field
) && !support
->ct_zone
) {
4114 report_unsupported_ct("zone");
4115 return OFPERR_OFPBAC_BAD_ARGUMENT
;
4118 OFPACT_FOR_EACH(a
, ct
->actions
, ofpact_ct_get_action_len(ct
)) {
4119 const struct mf_field
*dst
= ofpact_get_mf_dst(a
);
4121 if (a
->type
== OFPACT_NAT
&& !support
->ct_state_nat
) {
4122 /* The backer doesn't seem to support the NAT bits in
4123 * 'ct_state': assume that it doesn't support the NAT
4125 report_unsupported_ct("nat");
4126 return OFPERR_OFPBAC_BAD_TYPE
;
4128 if (dst
&& ((dst
->id
== MFF_CT_MARK
&& !support
->ct_mark
)
4129 || (dst
->id
== MFF_CT_LABEL
&& !support
->ct_label
))) {
4130 report_unsupported_ct("setting mark and/or label");
4131 return OFPERR_OFPBAC_BAD_SET_ARGUMENT
;
4140 rule_check(struct rule
*rule
)
4142 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->ofproto
);
4145 err
= check_mask(ofproto
, &rule
->cr
.match
.mask
->masks
);
4149 return check_actions(ofproto
, rule
->actions
);
4153 rule_construct(struct rule
*rule_
)
4154 OVS_NO_THREAD_SAFETY_ANALYSIS
4156 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4159 error
= rule_check(rule_
);
4164 ovs_mutex_init_adaptive(&rule
->stats_mutex
);
4165 rule
->stats
.n_packets
= 0;
4166 rule
->stats
.n_bytes
= 0;
4167 rule
->stats
.used
= rule
->up
.modified
;
4168 rule
->recirc_id
= 0;
4169 rule
->new_rule
= NULL
;
4170 rule
->forward_counts
= false;
4176 rule_insert(struct rule
*rule_
, struct rule
*old_rule_
, bool forward_counts
)
4177 OVS_REQUIRES(ofproto_mutex
)
4179 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4182 struct rule_dpif
*old_rule
= rule_dpif_cast(old_rule_
);
4184 ovs_assert(!old_rule
->new_rule
);
4186 /* Take a reference to the new rule, and refer all stats updates from
4187 * the old rule to the new rule. */
4188 ofproto_rule_ref(&rule
->up
);
4190 ovs_mutex_lock(&old_rule
->stats_mutex
);
4191 ovs_mutex_lock(&rule
->stats_mutex
);
4192 old_rule
->new_rule
= rule
; /* Forward future stats. */
4193 old_rule
->forward_counts
= forward_counts
;
4195 if (forward_counts
) {
4196 rule
->stats
= old_rule
->stats
; /* Transfer stats to the new
4199 /* Used timestamp must be forwarded whenever a rule is modified. */
4200 rule
->stats
.used
= old_rule
->stats
.used
;
4202 ovs_mutex_unlock(&rule
->stats_mutex
);
4203 ovs_mutex_unlock(&old_rule
->stats_mutex
);
4208 rule_destruct(struct rule
*rule_
)
4209 OVS_NO_THREAD_SAFETY_ANALYSIS
4211 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4213 ovs_mutex_destroy(&rule
->stats_mutex
);
4214 /* Release reference to the new rule, if any. */
4215 if (rule
->new_rule
) {
4216 ofproto_rule_unref(&rule
->new_rule
->up
);
4218 if (rule
->recirc_id
) {
4219 recirc_free_id(rule
->recirc_id
);
4224 rule_get_stats(struct rule
*rule_
, uint64_t *packets
, uint64_t *bytes
,
4225 long long int *used
)
4227 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4229 ovs_mutex_lock(&rule
->stats_mutex
);
4230 if (OVS_UNLIKELY(rule
->new_rule
)) {
4231 rule_get_stats(&rule
->new_rule
->up
, packets
, bytes
, used
);
4233 *packets
= rule
->stats
.n_packets
;
4234 *bytes
= rule
->stats
.n_bytes
;
4235 *used
= rule
->stats
.used
;
4237 ovs_mutex_unlock(&rule
->stats_mutex
);
4240 struct ofproto_dpif_packet_out
{
4241 struct xlate_cache xcache
;
4242 struct ofpbuf odp_actions
;
4243 struct recirc_refs rr
;
4248 static struct ofproto_dpif_packet_out
*
4249 ofproto_dpif_packet_out_new(void)
4251 struct ofproto_dpif_packet_out
*aux
= xmalloc(sizeof *aux
);
4252 xlate_cache_init(&aux
->xcache
);
4253 ofpbuf_init(&aux
->odp_actions
, 64);
4254 aux
->rr
= RECIRC_REFS_EMPTY_INITIALIZER
;
4255 aux
->needs_help
= false;
4261 ofproto_dpif_packet_out_delete(struct ofproto_dpif_packet_out
*aux
)
4264 xlate_cache_uninit(&aux
->xcache
);
4265 ofpbuf_uninit(&aux
->odp_actions
);
4266 recirc_refs_unref(&aux
->rr
);
4272 packet_xlate(struct ofproto
*ofproto_
, struct ofproto_packet_out
*opo
)
4273 OVS_REQUIRES(ofproto_mutex
)
4275 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4276 struct xlate_out xout
;
4277 struct xlate_in xin
;
4278 enum ofperr error
= 0;
4280 struct ofproto_dpif_packet_out
*aux
= ofproto_dpif_packet_out_new();
4282 xlate_in_init(&xin
, ofproto
, opo
->version
, opo
->flow
,
4283 opo
->flow
->in_port
.ofp_port
, NULL
, 0, opo
->packet
, NULL
,
4285 xin
.ofpacts
= opo
->ofpacts
;
4286 xin
.ofpacts_len
= opo
->ofpacts_len
;
4287 /* No learning or stats, but collect side effects to xcache. */
4288 xin
.allow_side_effects
= false;
4289 xin
.resubmit_stats
= NULL
;
4290 xin
.xcache
= &aux
->xcache
;
4292 if (xlate_actions(&xin
, &xout
) != XLATE_OK
) {
4293 error
= OFPERR_OFPFMFC_UNKNOWN
; /* Error processing actions. */
4296 /* Prepare learn actions. */
4297 struct xc_entry
*entry
;
4298 struct ofpbuf entries
= aux
->xcache
.entries
;
4300 XC_ENTRY_FOR_EACH (entry
, &entries
) {
4301 if (entry
->type
== XC_LEARN
) {
4302 struct ofproto_flow_mod
*ofm
= entry
->learn
.ofm
;
4304 error
= ofproto_flow_mod_learn_refresh(ofm
);
4308 struct rule
*rule
= ofm
->temp_rule
;
4309 ofm
->learn_adds_rule
= (rule
->state
== RULE_INITIALIZED
);
4310 if (ofm
->learn_adds_rule
) {
4311 /* If learning on a different bridge, must use its next
4312 * version number. */
4313 ofm
->version
= (rule
->ofproto
== ofproto_
)
4314 ? opo
->version
: rule
->ofproto
->tables_version
+ 1;
4315 error
= ofproto_flow_mod_learn_start(ofm
);
4324 aux
->needs_help
= (xout
.slow
& SLOW_ACTION
) != 0;
4325 recirc_refs_swap(&aux
->rr
, &xout
.recircs
); /* Hold recirc refs. */
4327 xlate_out_uninit(&xout
);
4332 xlate_out_uninit(&xout
);
4333 ofproto_dpif_packet_out_delete(aux
);
4339 packet_xlate_revert(struct ofproto
*ofproto OVS_UNUSED
,
4340 struct ofproto_packet_out
*opo
)
4341 OVS_REQUIRES(ofproto_mutex
)
4343 struct ofproto_dpif_packet_out
*aux
= opo
->aux
;
4346 /* Revert the learned flows. */
4347 struct xc_entry
*entry
;
4348 struct ofpbuf entries
= aux
->xcache
.entries
;
4350 XC_ENTRY_FOR_EACH (entry
, &entries
) {
4351 if (entry
->type
== XC_LEARN
&& entry
->learn
.ofm
->learn_adds_rule
) {
4352 ofproto_flow_mod_learn_revert(entry
->learn
.ofm
);
4356 ofproto_dpif_packet_out_delete(aux
);
4360 /* Push stats and perform side effects of flow translation. */
4362 ofproto_dpif_xcache_execute(struct ofproto_dpif
*ofproto
,
4363 struct xlate_cache
*xcache
,
4364 const struct dpif_flow_stats
*stats
)
4365 OVS_REQUIRES(ofproto_mutex
)
4367 struct xc_entry
*entry
;
4368 struct ofpbuf entries
= xcache
->entries
;
4370 XC_ENTRY_FOR_EACH (entry
, &entries
) {
4371 switch (entry
->type
) {
4373 /* Finish the learned flows. */
4374 if (entry
->learn
.ofm
->learn_adds_rule
) {
4375 ofproto_flow_mod_learn_finish(entry
->learn
.ofm
, &ofproto
->up
);
4378 case XC_FIN_TIMEOUT
:
4379 if (stats
->tcp_flags
& (TCP_FIN
| TCP_RST
)) {
4380 /* 'ofproto_mutex' already held */
4381 ofproto_rule_reduce_timeouts__(&entry
->fin
.rule
->up
,
4386 /* All the rest can be dealt with by the xlate layer. */
4397 xlate_push_stats_entry(entry
, stats
);
4406 packet_execute(struct ofproto
*ofproto_
, struct ofproto_packet_out
*opo
)
4407 OVS_REQUIRES(ofproto_mutex
)
4409 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4410 struct dpif_flow_stats stats
;
4411 struct dpif_execute execute
;
4413 struct ofproto_dpif_packet_out
*aux
= opo
->aux
;
4416 /* Run the side effects from the xcache. */
4417 dpif_flow_stats_extract(opo
->flow
, opo
->packet
, time_msec(), &stats
);
4418 ofproto_dpif_xcache_execute(ofproto
, &aux
->xcache
, &stats
);
4420 execute
.actions
= aux
->odp_actions
.data
;
4421 execute
.actions_len
= aux
->odp_actions
.size
;
4423 pkt_metadata_from_flow(&opo
->packet
->md
, opo
->flow
);
4424 execute
.packet
= opo
->packet
;
4425 execute
.flow
= opo
->flow
;
4426 execute
.needs_help
= aux
->needs_help
;
4427 execute
.probe
= false;
4430 /* Fix up in_port. */
4431 ofproto_dpif_set_packet_odp_port(ofproto
, opo
->flow
->in_port
.ofp_port
,
4434 dpif_execute(ofproto
->backer
->dpif
, &execute
);
4435 ofproto_dpif_packet_out_delete(aux
);
4439 static struct group_dpif
*group_dpif_cast(const struct ofgroup
*group
)
4441 return group
? CONTAINER_OF(group
, struct group_dpif
, up
) : NULL
;
4444 static struct ofgroup
*
4447 struct group_dpif
*group
= xzalloc(sizeof *group
);
4452 group_dealloc(struct ofgroup
*group_
)
4454 struct group_dpif
*group
= group_dpif_cast(group_
);
4459 group_construct_stats(struct group_dpif
*group
)
4460 OVS_REQUIRES(group
->stats_mutex
)
4462 group
->packet_count
= 0;
4463 group
->byte_count
= 0;
4465 struct ofputil_bucket
*bucket
;
4466 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
4467 bucket
->stats
.packet_count
= 0;
4468 bucket
->stats
.byte_count
= 0;
4473 group_dpif_credit_stats(struct group_dpif
*group
,
4474 struct ofputil_bucket
*bucket
,
4475 const struct dpif_flow_stats
*stats
)
4477 ovs_mutex_lock(&group
->stats_mutex
);
4478 group
->packet_count
+= stats
->n_packets
;
4479 group
->byte_count
+= stats
->n_bytes
;
4481 bucket
->stats
.packet_count
+= stats
->n_packets
;
4482 bucket
->stats
.byte_count
+= stats
->n_bytes
;
4483 } else { /* Credit to all buckets */
4484 struct ofputil_bucket
*bucket
;
4485 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
4486 bucket
->stats
.packet_count
+= stats
->n_packets
;
4487 bucket
->stats
.byte_count
+= stats
->n_bytes
;
4490 ovs_mutex_unlock(&group
->stats_mutex
);
4494 group_construct(struct ofgroup
*group_
)
4496 struct group_dpif
*group
= group_dpif_cast(group_
);
4498 ovs_mutex_init_adaptive(&group
->stats_mutex
);
4499 ovs_mutex_lock(&group
->stats_mutex
);
4500 group_construct_stats(group
);
4501 ovs_mutex_unlock(&group
->stats_mutex
);
4506 group_destruct(struct ofgroup
*group_
)
4508 struct group_dpif
*group
= group_dpif_cast(group_
);
4509 ovs_mutex_destroy(&group
->stats_mutex
);
4513 group_get_stats(const struct ofgroup
*group_
, struct ofputil_group_stats
*ogs
)
4515 struct group_dpif
*group
= group_dpif_cast(group_
);
4517 ovs_mutex_lock(&group
->stats_mutex
);
4518 ogs
->packet_count
= group
->packet_count
;
4519 ogs
->byte_count
= group
->byte_count
;
4521 struct bucket_counter
*bucket_stats
= ogs
->bucket_stats
;
4522 struct ofputil_bucket
*bucket
;
4523 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
4524 bucket_stats
->packet_count
= bucket
->stats
.packet_count
;
4525 bucket_stats
->byte_count
= bucket
->stats
.byte_count
;
4528 ovs_mutex_unlock(&group
->stats_mutex
);
4533 /* If the group exists, this function increments the groups's reference count.
4535 * Make sure to call ofproto_group_unref() after no longer needing to maintain
4536 * a reference to the group. */
4538 group_dpif_lookup(struct ofproto_dpif
*ofproto
, uint32_t group_id
,
4539 ovs_version_t version
, bool take_ref
)
4541 struct ofgroup
*ofgroup
= ofproto_group_lookup(&ofproto
->up
, group_id
,
4543 return ofgroup
? group_dpif_cast(ofgroup
) : NULL
;
4546 /* Sends 'packet' out 'ofport'. If 'port' is a tunnel and that tunnel type
4547 * supports a notion of an OAM flag, sets it if 'oam' is true.
4548 * May modify 'packet'.
4549 * Returns 0 if successful, otherwise a positive errno value. */
4551 ofproto_dpif_send_packet(const struct ofport_dpif
*ofport
, bool oam
,
4552 struct dp_packet
*packet
)
4554 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
4557 error
= xlate_send_packet(ofport
, oam
, packet
);
4559 ovs_mutex_lock(&ofproto
->stats_mutex
);
4560 ofproto
->stats
.tx_packets
++;
4561 ofproto
->stats
.tx_bytes
+= dp_packet_size(packet
);
4562 ovs_mutex_unlock(&ofproto
->stats_mutex
);
4566 /* Return the version string of the datapath that backs up
4570 get_datapath_version(const struct ofproto
*ofproto_
)
4572 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4574 return ofproto
->backer
->dp_version_string
;
4578 type_set_config(const char *type
, const struct smap
*other_config
)
4580 struct dpif_backer
*backer
;
4582 backer
= shash_find_data(&all_dpif_backers
, type
);
4584 /* This is not necessarily a problem, since backers are only
4585 * created on demand. */
4589 dpif_set_config(backer
->dpif
, other_config
);
4593 ct_flush(const struct ofproto
*ofproto_
, const uint16_t *zone
)
4595 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4597 ct_dpif_flush(ofproto
->backer
->dpif
, zone
);
4601 set_frag_handling(struct ofproto
*ofproto_
,
4602 enum ofputil_frag_handling frag_handling
)
4604 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4605 if (frag_handling
!= OFPUTIL_FRAG_REASM
) {
4606 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4614 nxt_resume(struct ofproto
*ofproto_
,
4615 const struct ofputil_packet_in_private
*pin
)
4617 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4619 /* Translate pin into datapath actions. */
4620 uint64_t odp_actions_stub
[1024 / 8];
4621 struct ofpbuf odp_actions
= OFPBUF_STUB_INITIALIZER(odp_actions_stub
);
4622 enum slow_path_reason slow
;
4623 enum ofperr error
= xlate_resume(ofproto
, pin
, &odp_actions
, &slow
);
4625 /* Steal 'pin->packet' and put it into a dp_packet. */
4626 struct dp_packet packet
;
4627 dp_packet_init(&packet
, pin
->public.packet_len
);
4628 dp_packet_put(&packet
, pin
->public.packet
, pin
->public.packet_len
);
4630 pkt_metadata_from_flow(&packet
.md
, &pin
->public.flow_metadata
.flow
);
4632 /* Fix up in_port. */
4633 ofproto_dpif_set_packet_odp_port(ofproto
,
4634 pin
->public.flow_metadata
.flow
.in_port
.ofp_port
,
4637 struct flow headers
;
4638 flow_extract(&packet
, &headers
);
4640 /* Execute the datapath actions on the packet. */
4641 struct dpif_execute execute
= {
4642 .actions
= odp_actions
.data
,
4643 .actions_len
= odp_actions
.size
,
4644 .needs_help
= (slow
& SLOW_ACTION
) != 0,
4648 dpif_execute(ofproto
->backer
->dpif
, &execute
);
4651 ofpbuf_uninit(&odp_actions
);
4652 dp_packet_uninit(&packet
);
4660 set_netflow(struct ofproto
*ofproto_
,
4661 const struct netflow_options
*netflow_options
)
4663 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4665 if (netflow_options
) {
4666 if (!ofproto
->netflow
) {
4667 ofproto
->netflow
= netflow_create();
4668 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4670 return netflow_set_options(ofproto
->netflow
, netflow_options
);
4671 } else if (ofproto
->netflow
) {
4672 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4673 netflow_unref(ofproto
->netflow
);
4674 ofproto
->netflow
= NULL
;
4681 get_netflow_ids(const struct ofproto
*ofproto_
,
4682 uint8_t *engine_type
, uint8_t *engine_id
)
4684 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4686 dpif_get_netflow_ids(ofproto
->backer
->dpif
, engine_type
, engine_id
);
4689 struct ofproto_dpif
*
4690 ofproto_dpif_lookup(const char *name
)
4692 struct ofproto_dpif
*ofproto
;
4694 HMAP_FOR_EACH_WITH_HASH (ofproto
, all_ofproto_dpifs_node
,
4695 hash_string(name
, 0), &all_ofproto_dpifs
) {
4696 if (!strcmp(ofproto
->up
.name
, name
)) {
4704 ofproto_unixctl_fdb_flush(struct unixctl_conn
*conn
, int argc
,
4705 const char *argv
[], void *aux OVS_UNUSED
)
4707 struct ofproto_dpif
*ofproto
;
4710 ofproto
= ofproto_dpif_lookup(argv
[1]);
4712 unixctl_command_reply_error(conn
, "no such bridge");
4715 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4716 mac_learning_flush(ofproto
->ml
);
4717 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4719 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4720 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4721 mac_learning_flush(ofproto
->ml
);
4722 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4726 unixctl_command_reply(conn
, "table successfully flushed");
4730 ofproto_unixctl_mcast_snooping_flush(struct unixctl_conn
*conn
, int argc
,
4731 const char *argv
[], void *aux OVS_UNUSED
)
4733 struct ofproto_dpif
*ofproto
;
4736 ofproto
= ofproto_dpif_lookup(argv
[1]);
4738 unixctl_command_reply_error(conn
, "no such bridge");
4742 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4743 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4746 mcast_snooping_mdb_flush(ofproto
->ms
);
4748 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4749 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4752 mcast_snooping_mdb_flush(ofproto
->ms
);
4756 unixctl_command_reply(conn
, "table successfully flushed");
4759 static struct ofport_dpif
*
4760 ofbundle_get_a_port(const struct ofbundle
*bundle
)
4762 return CONTAINER_OF(ovs_list_front(&bundle
->ports
), struct ofport_dpif
,
4767 ofproto_unixctl_fdb_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4768 const char *argv
[], void *aux OVS_UNUSED
)
4770 struct ds ds
= DS_EMPTY_INITIALIZER
;
4771 const struct ofproto_dpif
*ofproto
;
4772 const struct mac_entry
*e
;
4774 ofproto
= ofproto_dpif_lookup(argv
[1]);
4776 unixctl_command_reply_error(conn
, "no such bridge");
4780 ds_put_cstr(&ds
, " port VLAN MAC Age\n");
4781 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
4782 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
4783 struct ofbundle
*bundle
= mac_entry_get_port(ofproto
->ml
, e
);
4784 char name
[OFP_MAX_PORT_NAME_LEN
];
4786 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4788 ds_put_format(&ds
, "%5s %4d "ETH_ADDR_FMT
" %3d\n",
4789 name
, e
->vlan
, ETH_ADDR_ARGS(e
->mac
),
4790 mac_entry_age(ofproto
->ml
, e
));
4792 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4793 unixctl_command_reply(conn
, ds_cstr(&ds
));
4798 ofproto_unixctl_mcast_snooping_show(struct unixctl_conn
*conn
,
4799 int argc OVS_UNUSED
,
4801 void *aux OVS_UNUSED
)
4803 struct ds ds
= DS_EMPTY_INITIALIZER
;
4804 const struct ofproto_dpif
*ofproto
;
4805 const struct ofbundle
*bundle
;
4806 const struct mcast_group
*grp
;
4807 struct mcast_group_bundle
*b
;
4808 struct mcast_mrouter_bundle
*mrouter
;
4810 ofproto
= ofproto_dpif_lookup(argv
[1]);
4812 unixctl_command_reply_error(conn
, "no such bridge");
4816 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4817 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4821 ds_put_cstr(&ds
, " port VLAN GROUP Age\n");
4822 ovs_rwlock_rdlock(&ofproto
->ms
->rwlock
);
4823 LIST_FOR_EACH (grp
, group_node
, &ofproto
->ms
->group_lru
) {
4824 LIST_FOR_EACH(b
, bundle_node
, &grp
->bundle_lru
) {
4825 char name
[OFP_MAX_PORT_NAME_LEN
];
4828 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4830 ds_put_format(&ds
, "%5s %4d ", name
, grp
->vlan
);
4831 ipv6_format_mapped(&grp
->addr
, &ds
);
4832 ds_put_format(&ds
, " %3d\n",
4833 mcast_bundle_age(ofproto
->ms
, b
));
4837 /* ports connected to multicast routers */
4838 LIST_FOR_EACH(mrouter
, mrouter_node
, &ofproto
->ms
->mrouter_lru
) {
4839 char name
[OFP_MAX_PORT_NAME_LEN
];
4841 bundle
= mrouter
->port
;
4842 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4844 ds_put_format(&ds
, "%5s %4d querier %3d\n",
4845 name
, mrouter
->vlan
,
4846 mcast_mrouter_age(ofproto
->ms
, mrouter
));
4848 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
4849 unixctl_command_reply(conn
, ds_cstr(&ds
));
4853 /* Store the current ofprotos in 'ofproto_shash'. Returns a sorted list
4854 * of the 'ofproto_shash' nodes. It is the responsibility of the caller
4855 * to destroy 'ofproto_shash' and free the returned value. */
4856 static const struct shash_node
**
4857 get_ofprotos(struct shash
*ofproto_shash
)
4859 const struct ofproto_dpif
*ofproto
;
4861 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4862 char *name
= xasprintf("%s@%s", ofproto
->up
.type
, ofproto
->up
.name
);
4863 shash_add_nocopy(ofproto_shash
, name
, ofproto
);
4866 return shash_sort(ofproto_shash
);
4870 ofproto_unixctl_dpif_dump_dps(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4871 const char *argv
[] OVS_UNUSED
,
4872 void *aux OVS_UNUSED
)
4874 struct ds ds
= DS_EMPTY_INITIALIZER
;
4875 struct shash ofproto_shash
;
4876 const struct shash_node
**sorted_ofprotos
;
4879 shash_init(&ofproto_shash
);
4880 sorted_ofprotos
= get_ofprotos(&ofproto_shash
);
4881 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4882 const struct shash_node
*node
= sorted_ofprotos
[i
];
4883 ds_put_format(&ds
, "%s\n", node
->name
);
4886 shash_destroy(&ofproto_shash
);
4887 free(sorted_ofprotos
);
4889 unixctl_command_reply(conn
, ds_cstr(&ds
));
4894 show_dp_feature_b(struct ds
*ds
, const char *feature
, bool b
)
4896 ds_put_format(ds
, "%s: %s\n", feature
, b
? "Yes" : "No");
4900 show_dp_feature_s(struct ds
*ds
, const char *feature
, size_t s
)
4902 ds_put_format(ds
, "%s: %"PRIuSIZE
"\n", feature
, s
);
4906 dpif_show_support(const struct dpif_backer_support
*support
, struct ds
*ds
)
4908 show_dp_feature_b(ds
, "Variable length userdata",
4909 support
->variable_length_userdata
);
4910 show_dp_feature_b(ds
, "Masked set action", support
->masked_set_action
);
4911 show_dp_feature_b(ds
, "Tunnel push pop", support
->tnl_push_pop
);
4912 show_dp_feature_b(ds
, "Ufid", support
->ufid
);
4913 show_dp_feature_b(ds
, "Trunc action", support
->trunc
);
4914 show_dp_feature_b(ds
, "Clone action", support
->clone
);
4915 show_dp_feature_s(ds
, "Max MPLS depth", support
->odp
.max_mpls_depth
);
4916 show_dp_feature_b(ds
, "Recirc", support
->odp
.recirc
);
4917 show_dp_feature_b(ds
, "CT state", support
->odp
.ct_state
);
4918 show_dp_feature_b(ds
, "CT zone", support
->odp
.ct_zone
);
4919 show_dp_feature_b(ds
, "CT mark", support
->odp
.ct_mark
);
4920 show_dp_feature_b(ds
, "CT label", support
->odp
.ct_label
);
4921 show_dp_feature_b(ds
, "CT State NAT", support
->odp
.ct_state_nat
);
4922 show_dp_feature_s(ds
, "Max sample nesting", support
->sample_nesting
);
4926 dpif_show_backer(const struct dpif_backer
*backer
, struct ds
*ds
)
4928 const struct shash_node
**ofprotos
;
4929 struct dpif_dp_stats dp_stats
;
4930 struct shash ofproto_shash
;
4933 dpif_get_dp_stats(backer
->dpif
, &dp_stats
);
4934 ds_put_format(ds
, "%s: hit:%"PRIu64
" missed:%"PRIu64
"\n",
4935 dpif_name(backer
->dpif
), dp_stats
.n_hit
, dp_stats
.n_missed
);
4937 shash_init(&ofproto_shash
);
4938 ofprotos
= get_ofprotos(&ofproto_shash
);
4939 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4940 struct ofproto_dpif
*ofproto
= ofprotos
[i
]->data
;
4941 const struct shash_node
**ports
;
4944 if (ofproto
->backer
!= backer
) {
4948 ds_put_format(ds
, "\t%s:\n", ofproto
->up
.name
);
4950 ports
= shash_sort(&ofproto
->up
.port_by_name
);
4951 for (j
= 0; j
< shash_count(&ofproto
->up
.port_by_name
); j
++) {
4952 const struct shash_node
*node
= ports
[j
];
4953 struct ofport
*ofport
= node
->data
;
4955 odp_port_t odp_port
;
4957 ds_put_format(ds
, "\t\t%s %u/", netdev_get_name(ofport
->netdev
),
4960 odp_port
= ofp_port_to_odp_port(ofproto
, ofport
->ofp_port
);
4961 if (odp_port
!= ODPP_NONE
) {
4962 ds_put_format(ds
, "%"PRIu32
":", odp_port
);
4964 ds_put_cstr(ds
, "none:");
4967 ds_put_format(ds
, " (%s", netdev_get_type(ofport
->netdev
));
4970 if (!netdev_get_config(ofport
->netdev
, &config
)) {
4971 const struct smap_node
**nodes
;
4974 nodes
= smap_sort(&config
);
4975 for (i
= 0; i
< smap_count(&config
); i
++) {
4976 const struct smap_node
*node
= nodes
[i
];
4977 ds_put_format(ds
, "%c %s=%s", i
? ',' : ':',
4978 node
->key
, node
->value
);
4982 smap_destroy(&config
);
4984 ds_put_char(ds
, ')');
4985 ds_put_char(ds
, '\n');
4989 shash_destroy(&ofproto_shash
);
4994 ofproto_unixctl_dpif_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4995 const char *argv
[] OVS_UNUSED
, void *aux OVS_UNUSED
)
4997 struct ds ds
= DS_EMPTY_INITIALIZER
;
4998 const struct shash_node
**backers
;
5001 backers
= shash_sort(&all_dpif_backers
);
5002 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
5003 dpif_show_backer(backers
[i
]->data
, &ds
);
5007 unixctl_command_reply(conn
, ds_cstr(&ds
));
5012 ofproto_unixctl_dpif_dump_flows(struct unixctl_conn
*conn
,
5013 int argc OVS_UNUSED
, const char *argv
[],
5014 void *aux OVS_UNUSED
)
5016 const struct ofproto_dpif
*ofproto
;
5018 struct ds ds
= DS_EMPTY_INITIALIZER
;
5019 bool verbosity
= false;
5021 struct dpif_port dpif_port
;
5022 struct dpif_port_dump port_dump
;
5023 struct hmap portno_names
;
5025 struct dpif_flow_dump
*flow_dump
;
5026 struct dpif_flow_dump_thread
*flow_dump_thread
;
5030 ofproto
= ofproto_dpif_lookup(argv
[argc
- 1]);
5032 unixctl_command_reply_error(conn
, "no such bridge");
5036 if (argc
> 2 && !strcmp(argv
[1], "-m")) {
5040 hmap_init(&portno_names
);
5041 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, ofproto
->backer
->dpif
) {
5042 odp_portno_names_set(&portno_names
, dpif_port
.port_no
, dpif_port
.name
);
5046 flow_dump
= dpif_flow_dump_create(ofproto
->backer
->dpif
, false);
5047 flow_dump_thread
= dpif_flow_dump_thread_create(flow_dump
);
5048 while (dpif_flow_dump_next(flow_dump_thread
, &f
, 1)) {
5051 if (odp_flow_key_to_flow(f
.key
, f
.key_len
, &flow
) == ODP_FIT_ERROR
5052 || xlate_lookup_ofproto(ofproto
->backer
, &flow
, NULL
) != ofproto
) {
5057 odp_format_ufid(&f
.ufid
, &ds
);
5058 ds_put_cstr(&ds
, " ");
5060 odp_flow_format(f
.key
, f
.key_len
, f
.mask
, f
.mask_len
,
5061 &portno_names
, &ds
, verbosity
);
5062 ds_put_cstr(&ds
, ", ");
5063 dpif_flow_stats_format(&f
.stats
, &ds
);
5064 ds_put_cstr(&ds
, ", actions:");
5065 format_odp_actions(&ds
, f
.actions
, f
.actions_len
);
5066 ds_put_char(&ds
, '\n');
5068 dpif_flow_dump_thread_destroy(flow_dump_thread
);
5069 error
= dpif_flow_dump_destroy(flow_dump
);
5073 ds_put_format(&ds
, "dpif/dump_flows failed: %s", ovs_strerror(errno
));
5074 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
5076 unixctl_command_reply(conn
, ds_cstr(&ds
));
5078 odp_portno_names_destroy(&portno_names
);
5079 hmap_destroy(&portno_names
);
5084 ofproto_revalidate_all_backers(void)
5086 const struct shash_node
**backers
;
5089 backers
= shash_sort(&all_dpif_backers
);
5090 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
5091 struct dpif_backer
*backer
= backers
[i
]->data
;
5092 backer
->need_revalidate
= REV_RECONFIGURE
;
5098 disable_tnl_push_pop(struct unixctl_conn
*conn OVS_UNUSED
, int argc OVS_UNUSED
,
5099 const char *argv
[], void *aux OVS_UNUSED
)
5101 if (!strcasecmp(argv
[1], "off")) {
5102 ofproto_use_tnl_push_pop
= false;
5103 unixctl_command_reply(conn
, "Tunnel push-pop off");
5104 ofproto_revalidate_all_backers();
5105 } else if (!strcasecmp(argv
[1], "on")) {
5106 ofproto_use_tnl_push_pop
= true;
5107 unixctl_command_reply(conn
, "Tunnel push-pop on");
5108 ofproto_revalidate_all_backers();
5110 unixctl_command_reply_error(conn
, "Invalid argument");
5115 disable_datapath_truncate(struct unixctl_conn
*conn OVS_UNUSED
,
5116 int argc OVS_UNUSED
,
5117 const char *argv
[] OVS_UNUSED
,
5118 void *aux OVS_UNUSED
)
5120 const struct shash_node
**backers
;
5123 backers
= shash_sort(&all_dpif_backers
);
5124 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
5125 struct dpif_backer
*backer
= backers
[i
]->data
;
5126 backer
->support
.trunc
= false;
5129 unixctl_command_reply(conn
, "Datapath truncate action diabled");
5133 disable_datapath_clone(struct unixctl_conn
*conn OVS_UNUSED
,
5134 int argc
, const char *argv
[],
5135 void *aux OVS_UNUSED
)
5137 struct ds ds
= DS_EMPTY_INITIALIZER
;
5138 const char *br
= argv
[argc
-1];
5139 struct ofproto_dpif
*ofproto
;
5141 ofproto
= ofproto_dpif_lookup(br
);
5143 unixctl_command_reply_error(conn
, "no such bridge");
5146 xlate_disable_dp_clone(ofproto
);
5147 udpif_flush(ofproto
->backer
->udpif
);
5148 ds_put_format(&ds
, "Datapath clone action disabled for bridge %s", br
);
5149 unixctl_command_reply(conn
, ds_cstr(&ds
));
5153 ofproto_unixctl_dpif_show_dp_features(struct unixctl_conn
*conn
,
5154 int argc
, const char *argv
[],
5155 void *aux OVS_UNUSED
)
5157 struct ds ds
= DS_EMPTY_INITIALIZER
;
5158 const char *br
= argv
[argc
-1];
5159 struct ofproto_dpif
*ofproto
= ofproto_dpif_lookup(br
);
5162 unixctl_command_reply_error(conn
, "no such bridge");
5166 dpif_show_support(&ofproto
->backer
->support
, &ds
);
5167 unixctl_command_reply(conn
, ds_cstr(&ds
));
5171 ofproto_unixctl_init(void)
5173 static bool registered
;
5179 unixctl_command_register("fdb/flush", "[bridge]", 0, 1,
5180 ofproto_unixctl_fdb_flush
, NULL
);
5181 unixctl_command_register("fdb/show", "bridge", 1, 1,
5182 ofproto_unixctl_fdb_show
, NULL
);
5183 unixctl_command_register("mdb/flush", "[bridge]", 0, 1,
5184 ofproto_unixctl_mcast_snooping_flush
, NULL
);
5185 unixctl_command_register("mdb/show", "bridge", 1, 1,
5186 ofproto_unixctl_mcast_snooping_show
, NULL
);
5187 unixctl_command_register("dpif/dump-dps", "", 0, 0,
5188 ofproto_unixctl_dpif_dump_dps
, NULL
);
5189 unixctl_command_register("dpif/show", "", 0, 0, ofproto_unixctl_dpif_show
,
5191 unixctl_command_register("dpif/show-dp-features", "bridge", 1, 1,
5192 ofproto_unixctl_dpif_show_dp_features
, NULL
);
5193 unixctl_command_register("dpif/dump-flows", "[-m] bridge", 1, 2,
5194 ofproto_unixctl_dpif_dump_flows
, NULL
);
5196 unixctl_command_register("ofproto/tnl-push-pop", "[on]|[off]", 1, 1,
5197 disable_tnl_push_pop
, NULL
);
5199 unixctl_command_register("dpif/disable-truncate", "", 0, 0,
5200 disable_datapath_truncate
, NULL
);
5202 unixctl_command_register("dpif/disable-dp-clone", "bridge", 1, 1,
5203 disable_datapath_clone
, NULL
);
5207 ofp_port_to_odp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
5209 const struct ofport_dpif
*ofport
= ofp_port_to_ofport(ofproto
, ofp_port
);
5210 return ofport
? ofport
->odp_port
: ODPP_NONE
;
5213 struct ofport_dpif
*
5214 odp_port_to_ofport(const struct dpif_backer
*backer
, odp_port_t odp_port
)
5216 struct ofport_dpif
*port
;
5218 ovs_rwlock_rdlock(&backer
->odp_to_ofport_lock
);
5219 HMAP_FOR_EACH_IN_BUCKET (port
, odp_port_node
, hash_odp_port(odp_port
),
5220 &backer
->odp_to_ofport_map
) {
5221 if (port
->odp_port
== odp_port
) {
5222 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5227 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5232 odp_port_to_ofp_port(const struct ofproto_dpif
*ofproto
, odp_port_t odp_port
)
5234 struct ofport_dpif
*port
;
5236 port
= odp_port_to_ofport(ofproto
->backer
, odp_port
);
5237 if (port
&& &ofproto
->up
== port
->up
.ofproto
) {
5238 return port
->up
.ofp_port
;
5245 ofproto_dpif_add_internal_flow(struct ofproto_dpif
*ofproto
,
5246 const struct match
*match
, int priority
,
5247 uint16_t idle_timeout
,
5248 const struct ofpbuf
*ofpacts
,
5249 struct rule
**rulep
)
5251 struct ofputil_flow_mod fm
;
5252 struct rule_dpif
*rule
;
5255 fm
= (struct ofputil_flow_mod
) {
5256 .buffer_id
= UINT32_MAX
,
5258 .priority
= priority
,
5259 .table_id
= TBL_INTERNAL
,
5260 .command
= OFPFC_ADD
,
5261 .idle_timeout
= idle_timeout
,
5262 .flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
,
5263 .ofpacts
= ofpacts
->data
,
5264 .ofpacts_len
= ofpacts
->size
,
5267 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5269 VLOG_ERR_RL(&rl
, "failed to add internal flow (%s)",
5270 ofperr_to_string(error
));
5275 rule
= rule_dpif_lookup_in_table(ofproto
,
5276 ofproto_dpif_get_tables_version(ofproto
),
5277 TBL_INTERNAL
, &fm
.match
.flow
,
5288 ofproto_dpif_delete_internal_flow(struct ofproto_dpif
*ofproto
,
5289 struct match
*match
, int priority
)
5291 struct ofputil_flow_mod fm
;
5294 fm
= (struct ofputil_flow_mod
) {
5295 .buffer_id
= UINT32_MAX
,
5297 .priority
= priority
,
5298 .table_id
= TBL_INTERNAL
,
5299 .out_port
= OFPP_ANY
,
5300 .out_group
= OFPG_ANY
,
5301 .flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
,
5302 .command
= OFPFC_DELETE_STRICT
,
5305 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5307 VLOG_ERR_RL(&rl
, "failed to delete internal flow (%s)",
5308 ofperr_to_string(error
));
5316 meter_get_features(const struct ofproto
*ofproto_
,
5317 struct ofputil_meter_features
*features
)
5319 const struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
5321 dpif_meter_get_features(ofproto
->backer
->dpif
, features
);
5325 meter_set(struct ofproto
*ofproto_
, ofproto_meter_id
*meter_id
,
5326 struct ofputil_meter_config
*config
)
5328 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
5330 switch (dpif_meter_set(ofproto
->backer
->dpif
, meter_id
, config
)) {
5333 case EFBIG
: /* meter_id out of range */
5334 case ENOMEM
: /* Cannot allocate meter */
5335 return OFPERR_OFPMMFC_OUT_OF_METERS
;
5336 case EBADF
: /* Unsupported flags */
5337 return OFPERR_OFPMMFC_BAD_FLAGS
;
5338 case EINVAL
: /* Too many bands */
5339 return OFPERR_OFPMMFC_OUT_OF_BANDS
;
5340 case ENODEV
: /* Unsupported band type */
5341 return OFPERR_OFPMMFC_BAD_BAND
;
5343 return OFPERR_OFPMMFC_UNKNOWN
;
5348 meter_get(const struct ofproto
*ofproto_
, ofproto_meter_id meter_id
,
5349 struct ofputil_meter_stats
*stats
, uint16_t n_bands
)
5351 const struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
5353 if (!dpif_meter_get(ofproto
->backer
->dpif
, meter_id
, stats
, n_bands
)) {
5356 return OFPERR_OFPMMFC_UNKNOWN_METER
;
5360 meter_del(struct ofproto
*ofproto_
, ofproto_meter_id meter_id
)
5362 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
5364 dpif_meter_del(ofproto
->backer
->dpif
, meter_id
, NULL
, 0);
5367 const struct ofproto_class ofproto_dpif_class
= {
5381 NULL
, /* get_memory_usage. */
5382 type_get_memory_usage
,
5402 port_is_lacp_current
,
5403 port_get_lacp_stats
,
5404 NULL
, /* rule_choose_table */
5408 NULL
, /* rule_delete */
5413 packet_xlate_revert
,
5431 aa_vlan_get_queue_size
,
5438 get_stp_port_status
,
5443 get_rstp_port_status
,
5450 is_mirror_output_bundle
,
5451 forward_bpdu_changed
,
5452 set_mac_table_config
,
5454 set_mcast_snooping_port
,
5459 group_alloc
, /* group_alloc */
5460 group_construct
, /* group_construct */
5461 group_destruct
, /* group_destruct */
5462 group_dealloc
, /* group_dealloc */
5463 NULL
, /* group_modify */
5464 group_get_stats
, /* group_get_stats */
5465 get_datapath_version
, /* get_datapath_version */
5467 ct_flush
, /* ct_flush */