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-upcall.h"
51 #include "ofproto-dpif-xlate.h"
52 #include "openvswitch/ofp-actions.h"
53 #include "openvswitch/dynamic-string.h"
54 #include "openvswitch/meta-flow.h"
55 #include "openvswitch/ofp-parse.h"
56 #include "openvswitch/ofp-print.h"
57 #include "openvswitch/ofp-util.h"
58 #include "openvswitch/ofpbuf.h"
59 #include "openvswitch/vlog.h"
62 #include "ovs-router.h"
63 #include "poll-loop.h"
69 #include "unaligned.h"
71 #include "vlan-bitmap.h"
73 VLOG_DEFINE_THIS_MODULE(ofproto_dpif
);
75 COVERAGE_DEFINE(ofproto_dpif_expired
);
76 COVERAGE_DEFINE(packet_in_overflow
);
85 * - Do include packets and bytes from datapath flows which have not
86 * recently been processed by a revalidator. */
87 struct ovs_mutex stats_mutex
;
88 struct dpif_flow_stats stats OVS_GUARDED
;
90 /* In non-NULL, will point to a new rule (for which a reference is held) to
91 * which all the stats updates should be forwarded. This exists only
92 * transitionally when flows are replaced.
94 * Protected by stats_mutex. If both 'rule->stats_mutex' and
95 * 'rule->new_rule->stats_mutex' must be held together, acquire them in that
97 struct rule_dpif
*new_rule OVS_GUARDED
;
99 /* If non-zero then the recirculation id that has
100 * been allocated for use with this rule.
101 * The recirculation id and associated internal flow should
102 * be freed when the rule is freed */
106 /* RULE_CAST() depends on this. */
107 BUILD_ASSERT_DECL(offsetof(struct rule_dpif
, up
) == 0);
109 static void rule_get_stats(struct rule
*, uint64_t *packets
, uint64_t *bytes
,
110 long long int *used
);
111 static struct rule_dpif
*rule_dpif_cast(const struct rule
*);
112 static void rule_expire(struct rule_dpif
*, long long now
);
119 * - Do include packets and bytes from datapath flows which have not
120 * recently been processed by a revalidator. */
121 struct ovs_mutex stats_mutex
;
122 uint64_t packet_count OVS_GUARDED
; /* Number of packets received. */
123 uint64_t byte_count OVS_GUARDED
; /* Number of bytes received. */
127 struct hmap_node hmap_node
; /* In struct ofproto's "bundles" hmap. */
128 struct ofproto_dpif
*ofproto
; /* Owning ofproto. */
129 void *aux
; /* Key supplied by ofproto's client. */
130 char *name
; /* Identifier for log messages. */
133 struct ovs_list ports
; /* Contains "struct ofport"s. */
134 enum port_vlan_mode vlan_mode
; /* VLAN mode */
135 int vlan
; /* -1=trunk port, else a 12-bit VLAN ID. */
136 unsigned long *trunks
; /* Bitmap of trunked VLANs, if 'vlan' == -1.
137 * NULL if all VLANs are trunked. */
138 struct lacp
*lacp
; /* LACP if LACP is enabled, otherwise NULL. */
139 struct bond
*bond
; /* Nonnull iff more than one port. */
140 bool use_priority_tags
; /* Use 802.1p tag for frames in VLAN 0? */
143 bool floodable
; /* True if no port has OFPUTIL_PC_NO_FLOOD set. */
146 static void bundle_remove(struct ofport
*);
147 static void bundle_update(struct ofbundle
*);
148 static void bundle_destroy(struct ofbundle
*);
149 static void bundle_del_port(struct ofport_dpif
*);
150 static void bundle_run(struct ofbundle
*);
151 static void bundle_wait(struct ofbundle
*);
152 static void bundle_flush_macs(struct ofbundle
*, bool);
153 static void bundle_move(struct ofbundle
*, struct ofbundle
*);
155 static void stp_run(struct ofproto_dpif
*ofproto
);
156 static void stp_wait(struct ofproto_dpif
*ofproto
);
157 static int set_stp_port(struct ofport
*,
158 const struct ofproto_port_stp_settings
*);
160 static void rstp_run(struct ofproto_dpif
*ofproto
);
161 static void set_rstp_port(struct ofport
*,
162 const struct ofproto_port_rstp_settings
*);
165 struct hmap_node odp_port_node
; /* In dpif_backer's "odp_to_ofport_map". */
169 struct ofbundle
*bundle
; /* Bundle that contains this port, if any. */
170 struct ovs_list bundle_node
;/* In struct ofbundle's "ports" list. */
171 struct cfm
*cfm
; /* Connectivity Fault Management, if any. */
172 struct bfd
*bfd
; /* BFD, if any. */
173 struct lldp
*lldp
; /* lldp, if any. */
174 bool may_enable
; /* May be enabled in bonds. */
175 bool is_tunnel
; /* This port is a tunnel. */
176 bool is_layer3
; /* This is a layer 3 port. */
177 long long int carrier_seq
; /* Carrier status changes. */
178 struct ofport_dpif
*peer
; /* Peer if patch port. */
181 struct stp_port
*stp_port
; /* Spanning Tree Protocol, if any. */
182 enum stp_state stp_state
; /* Always STP_DISABLED if STP not in use. */
183 long long int stp_state_entered
;
185 /* Rapid Spanning Tree. */
186 struct rstp_port
*rstp_port
; /* Rapid Spanning Tree Protocol, if any. */
187 enum rstp_state rstp_state
; /* Always RSTP_DISABLED if RSTP not in use. */
189 /* Queue to DSCP mapping. */
190 struct ofproto_port_queue
*qdscp
;
194 static odp_port_t
ofp_port_to_odp_port(const struct ofproto_dpif
*,
197 static ofp_port_t
odp_port_to_ofp_port(const struct ofproto_dpif
*,
200 static struct ofport_dpif
*
201 ofport_dpif_cast(const struct ofport
*ofport
)
203 return ofport
? CONTAINER_OF(ofport
, struct ofport_dpif
, up
) : NULL
;
206 static void port_run(struct ofport_dpif
*);
207 static int set_bfd(struct ofport
*, const struct smap
*);
208 static int set_cfm(struct ofport
*, const struct cfm_settings
*);
209 static int set_lldp(struct ofport
*ofport_
, const struct smap
*cfg
);
210 static void ofport_update_peer(struct ofport_dpif
*);
212 /* Reasons that we might need to revalidate every datapath flow, and
213 * corresponding coverage counters.
215 * A value of 0 means that there is no need to revalidate.
217 * It would be nice to have some cleaner way to integrate with coverage
218 * counters, but with only a few reasons I guess this is good enough for
220 enum revalidate_reason
{
221 REV_RECONFIGURE
= 1, /* Switch configuration changed. */
222 REV_STP
, /* Spanning tree protocol port status change. */
223 REV_RSTP
, /* RSTP port status change. */
224 REV_BOND
, /* Bonding changed. */
225 REV_PORT_TOGGLED
, /* Port enabled or disabled by CFM, LACP, ...*/
226 REV_FLOW_TABLE
, /* Flow table changed. */
227 REV_MAC_LEARNING
, /* Mac learning changed. */
228 REV_MCAST_SNOOPING
, /* Multicast snooping changed. */
230 COVERAGE_DEFINE(rev_reconfigure
);
231 COVERAGE_DEFINE(rev_stp
);
232 COVERAGE_DEFINE(rev_rstp
);
233 COVERAGE_DEFINE(rev_bond
);
234 COVERAGE_DEFINE(rev_port_toggled
);
235 COVERAGE_DEFINE(rev_flow_table
);
236 COVERAGE_DEFINE(rev_mac_learning
);
237 COVERAGE_DEFINE(rev_mcast_snooping
);
239 /* All datapaths of a given type share a single dpif backer instance. */
246 struct ovs_rwlock odp_to_ofport_lock
;
247 struct hmap odp_to_ofport_map OVS_GUARDED
; /* Contains "struct ofport"s. */
249 struct simap tnl_backers
; /* Set of dpif ports backing tunnels. */
251 enum revalidate_reason need_revalidate
; /* Revalidate all flows. */
253 bool recv_set_enable
; /* Enables or disables receiving packets. */
255 /* Version string of the datapath stored in OVSDB. */
256 char *dp_version_string
;
258 /* Datapath feature support. */
259 struct dpif_backer_support support
;
260 struct atomic_count tnl_count
;
263 /* All existing ofproto_backer instances, indexed by ofproto->up.type. */
264 static struct shash all_dpif_backers
= SHASH_INITIALIZER(&all_dpif_backers
);
266 struct ofproto_dpif
{
267 struct hmap_node all_ofproto_dpifs_node
; /* In 'all_ofproto_dpifs'. */
269 struct dpif_backer
*backer
;
271 /* Unique identifier for this instantiation of this bridge in this running
275 ATOMIC(cls_version_t
) tables_version
; /* For classifier lookups. */
277 uint64_t dump_seq
; /* Last read of udpif_dump_seq(). */
279 /* Special OpenFlow rules. */
280 struct rule_dpif
*miss_rule
; /* Sends flow table misses to controller. */
281 struct rule_dpif
*no_packet_in_rule
; /* Drops flow table misses. */
282 struct rule_dpif
*drop_frags_rule
; /* Used in OFPUTIL_FRAG_DROP mode. */
285 struct netflow
*netflow
;
286 struct dpif_sflow
*sflow
;
287 struct dpif_ipfix
*ipfix
;
288 struct hmap bundles
; /* Contains "struct ofbundle"s. */
289 struct mac_learning
*ml
;
290 struct mcast_snooping
*ms
;
291 bool has_bonded_bundles
;
293 struct mbridge
*mbridge
;
295 struct ovs_mutex stats_mutex
;
296 struct netdev_stats stats OVS_GUARDED
; /* To account packets generated and
297 * consumed in userspace. */
301 long long int stp_last_tick
;
303 /* Rapid Spanning Tree. */
305 long long int rstp_last_tick
;
308 struct sset ports
; /* Set of standard port names. */
309 struct sset ghost_ports
; /* Ports with no datapath port. */
310 struct sset port_poll_set
; /* Queued names for port_poll() reply. */
311 int port_poll_errno
; /* Last errno for port_poll() reply. */
312 uint64_t change_seq
; /* Connectivity status changes. */
315 struct guarded_list ams
; /* Contains "struct ofproto_async_msgs"s. */
316 struct seq
*ams_seq
; /* For notifying 'ams' reception. */
320 /* All existing ofproto_dpif instances, indexed by ->up.name. */
321 static struct hmap all_ofproto_dpifs
= HMAP_INITIALIZER(&all_ofproto_dpifs
);
323 static bool ofproto_use_tnl_push_pop
= true;
324 static void ofproto_unixctl_init(void);
326 static inline struct ofproto_dpif
*
327 ofproto_dpif_cast(const struct ofproto
*ofproto
)
329 ovs_assert(ofproto
->ofproto_class
== &ofproto_dpif_class
);
330 return CONTAINER_OF(ofproto
, struct ofproto_dpif
, up
);
334 ofproto_dpif_get_enable_ufid(const struct dpif_backer
*backer
)
336 return backer
->support
.ufid
;
339 struct dpif_backer_support
*
340 ofproto_dpif_get_support(const struct ofproto_dpif
*ofproto
)
342 return &ofproto
->backer
->support
;
345 static void ofproto_trace(struct ofproto_dpif
*, struct flow
*,
346 const struct dp_packet
*packet
,
347 const struct ofpact
[], size_t ofpacts_len
,
350 /* Global variables. */
351 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
353 /* Initial mappings of port to bridge mappings. */
354 static struct shash init_ofp_ports
= SHASH_INITIALIZER(&init_ofp_ports
);
356 /* Executes 'fm'. The caller retains ownership of 'fm' and everything in
359 ofproto_dpif_flow_mod(struct ofproto_dpif
*ofproto
,
360 const struct ofputil_flow_mod
*fm
)
362 struct ofproto_flow_mod ofm
;
364 /* Multiple threads may do this for the same 'fm' at the same time.
365 * Allocate ofproto_flow_mod with execution context from stack.
367 * Note: This copy could be avoided by making ofproto_flow_mod more
368 * complex, but that may not be desireable, and a learn action is not that
369 * fast to begin with. */
371 ofproto_flow_mod(&ofproto
->up
, &ofm
);
374 /* Appends 'am' to the queue of asynchronous messages to be sent to the
375 * controller. Takes ownership of 'am' and any data it points to. */
377 ofproto_dpif_send_async_msg(struct ofproto_dpif
*ofproto
,
378 struct ofproto_async_msg
*am
)
380 if (!guarded_list_push_back(&ofproto
->ams
, &am
->list_node
, 1024)) {
381 COVERAGE_INC(packet_in_overflow
);
382 ofproto_async_msg_free(am
);
385 /* Wakes up main thread for packet-in I/O. */
386 seq_change(ofproto
->ams_seq
);
389 /* The default "table-miss" behaviour for OpenFlow1.3+ is to drop the
390 * packet rather than to send the packet to the controller.
392 * This function returns false to indicate that a packet_in message
393 * for a "table-miss" should be sent to at least one controller.
394 * False otherwise. */
396 ofproto_dpif_wants_packet_in_on_miss(struct ofproto_dpif
*ofproto
)
398 return connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
);
401 /* Factory functions. */
404 init(const struct shash
*iface_hints
)
406 struct shash_node
*node
;
408 /* Make a local copy, since we don't own 'iface_hints' elements. */
409 SHASH_FOR_EACH(node
, iface_hints
) {
410 const struct iface_hint
*orig_hint
= node
->data
;
411 struct iface_hint
*new_hint
= xmalloc(sizeof *new_hint
);
413 new_hint
->br_name
= xstrdup(orig_hint
->br_name
);
414 new_hint
->br_type
= xstrdup(orig_hint
->br_type
);
415 new_hint
->ofp_port
= orig_hint
->ofp_port
;
417 shash_add(&init_ofp_ports
, node
->name
, new_hint
);
420 ofproto_unixctl_init();
425 enumerate_types(struct sset
*types
)
427 dp_enumerate_types(types
);
431 enumerate_names(const char *type
, struct sset
*names
)
433 struct ofproto_dpif
*ofproto
;
436 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
437 if (strcmp(type
, ofproto
->up
.type
)) {
440 sset_add(names
, ofproto
->up
.name
);
447 del(const char *type
, const char *name
)
452 error
= dpif_open(name
, type
, &dpif
);
454 error
= dpif_delete(dpif
);
461 port_open_type(const char *datapath_type
, const char *port_type
)
463 return dpif_port_open_type(datapath_type
, port_type
);
466 /* Type functions. */
468 static void process_dpif_port_changes(struct dpif_backer
*);
469 static void process_dpif_all_ports_changed(struct dpif_backer
*);
470 static void process_dpif_port_change(struct dpif_backer
*,
471 const char *devname
);
472 static void process_dpif_port_error(struct dpif_backer
*, int error
);
474 static struct ofproto_dpif
*
475 lookup_ofproto_dpif_by_port_name(const char *name
)
477 struct ofproto_dpif
*ofproto
;
479 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
480 if (sset_contains(&ofproto
->ports
, name
)) {
489 ofproto_dpif_backer_enabled(struct dpif_backer
* backer
)
491 return backer
->recv_set_enable
;
495 type_run(const char *type
)
497 struct dpif_backer
*backer
;
499 backer
= shash_find_data(&all_dpif_backers
, type
);
501 /* This is not necessarily a problem, since backers are only
502 * created on demand. */
506 /* This must be called before dpif_run() */
507 dpif_poll_threads_set(backer
->dpif
, pmd_cpu_mask
);
509 if (dpif_run(backer
->dpif
)) {
510 backer
->need_revalidate
= REV_RECONFIGURE
;
513 udpif_run(backer
->udpif
);
515 /* If vswitchd started with other_config:flow_restore_wait set as "true",
516 * and the configuration has now changed to "false", enable receiving
517 * packets from the datapath. */
518 if (!backer
->recv_set_enable
&& !ofproto_get_flow_restore_wait()) {
521 backer
->recv_set_enable
= true;
523 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
525 VLOG_ERR("Failed to enable receiving packets in dpif.");
528 dpif_flow_flush(backer
->dpif
);
529 backer
->need_revalidate
= REV_RECONFIGURE
;
532 if (backer
->recv_set_enable
) {
533 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
536 if (backer
->need_revalidate
) {
537 struct ofproto_dpif
*ofproto
;
538 struct simap_node
*node
;
539 struct simap tmp_backers
;
541 /* Handle tunnel garbage collection. */
542 simap_init(&tmp_backers
);
543 simap_swap(&backer
->tnl_backers
, &tmp_backers
);
545 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
546 struct ofport_dpif
*iter
;
548 if (backer
!= ofproto
->backer
) {
552 HMAP_FOR_EACH (iter
, up
.hmap_node
, &ofproto
->up
.ports
) {
553 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
556 if (!iter
->is_tunnel
) {
560 dp_port
= netdev_vport_get_dpif_port(iter
->up
.netdev
,
561 namebuf
, sizeof namebuf
);
562 node
= simap_find(&tmp_backers
, dp_port
);
564 simap_put(&backer
->tnl_backers
, dp_port
, node
->data
);
565 simap_delete(&tmp_backers
, node
);
566 node
= simap_find(&backer
->tnl_backers
, dp_port
);
568 node
= simap_find(&backer
->tnl_backers
, dp_port
);
570 odp_port_t odp_port
= ODPP_NONE
;
572 if (!dpif_port_add(backer
->dpif
, iter
->up
.netdev
,
574 simap_put(&backer
->tnl_backers
, dp_port
,
575 odp_to_u32(odp_port
));
576 node
= simap_find(&backer
->tnl_backers
, dp_port
);
581 iter
->odp_port
= node
? u32_to_odp(node
->data
) : ODPP_NONE
;
582 if (tnl_port_reconfigure(iter
, iter
->up
.netdev
,
584 ovs_native_tunneling_is_on(ofproto
), dp_port
)) {
585 backer
->need_revalidate
= REV_RECONFIGURE
;
590 SIMAP_FOR_EACH (node
, &tmp_backers
) {
591 dpif_port_del(backer
->dpif
, u32_to_odp(node
->data
));
593 simap_destroy(&tmp_backers
);
595 switch (backer
->need_revalidate
) {
596 case REV_RECONFIGURE
: COVERAGE_INC(rev_reconfigure
); break;
597 case REV_STP
: COVERAGE_INC(rev_stp
); break;
598 case REV_RSTP
: COVERAGE_INC(rev_rstp
); break;
599 case REV_BOND
: COVERAGE_INC(rev_bond
); break;
600 case REV_PORT_TOGGLED
: COVERAGE_INC(rev_port_toggled
); break;
601 case REV_FLOW_TABLE
: COVERAGE_INC(rev_flow_table
); break;
602 case REV_MAC_LEARNING
: COVERAGE_INC(rev_mac_learning
); break;
603 case REV_MCAST_SNOOPING
: COVERAGE_INC(rev_mcast_snooping
); break;
605 backer
->need_revalidate
= 0;
607 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
608 struct ofport_dpif
*ofport
;
609 struct ofbundle
*bundle
;
611 if (ofproto
->backer
!= backer
) {
616 xlate_ofproto_set(ofproto
, ofproto
->up
.name
,
617 ofproto
->backer
->dpif
, ofproto
->ml
,
618 ofproto
->stp
, ofproto
->rstp
, ofproto
->ms
,
619 ofproto
->mbridge
, ofproto
->sflow
, ofproto
->ipfix
,
621 ofproto
->up
.forward_bpdu
,
622 connmgr_has_in_band(ofproto
->up
.connmgr
),
623 &ofproto
->backer
->support
);
625 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
626 xlate_bundle_set(ofproto
, bundle
, bundle
->name
,
627 bundle
->vlan_mode
, bundle
->vlan
,
628 bundle
->trunks
, bundle
->use_priority_tags
,
629 bundle
->bond
, bundle
->lacp
,
633 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
634 int stp_port
= ofport
->stp_port
635 ? stp_port_no(ofport
->stp_port
)
637 xlate_ofport_set(ofproto
, ofport
->bundle
, ofport
,
638 ofport
->up
.ofp_port
, ofport
->odp_port
,
639 ofport
->up
.netdev
, ofport
->cfm
, ofport
->bfd
,
640 ofport
->lldp
, ofport
->peer
, stp_port
,
641 ofport
->rstp_port
, ofport
->qdscp
,
642 ofport
->n_qdscp
, ofport
->up
.pp
.config
,
643 ofport
->up
.pp
.state
, ofport
->is_tunnel
,
649 udpif_revalidate(backer
->udpif
);
652 process_dpif_port_changes(backer
);
657 /* Check for and handle port changes in 'backer''s dpif. */
659 process_dpif_port_changes(struct dpif_backer
*backer
)
665 error
= dpif_port_poll(backer
->dpif
, &devname
);
671 process_dpif_all_ports_changed(backer
);
675 process_dpif_port_change(backer
, devname
);
680 process_dpif_port_error(backer
, error
);
687 process_dpif_all_ports_changed(struct dpif_backer
*backer
)
689 struct ofproto_dpif
*ofproto
;
690 struct dpif_port dpif_port
;
691 struct dpif_port_dump dump
;
692 struct sset devnames
;
695 sset_init(&devnames
);
696 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
697 if (ofproto
->backer
== backer
) {
698 struct ofport
*ofport
;
700 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
701 sset_add(&devnames
, netdev_get_name(ofport
->netdev
));
705 DPIF_PORT_FOR_EACH (&dpif_port
, &dump
, backer
->dpif
) {
706 sset_add(&devnames
, dpif_port
.name
);
709 SSET_FOR_EACH (devname
, &devnames
) {
710 process_dpif_port_change(backer
, devname
);
712 sset_destroy(&devnames
);
716 process_dpif_port_change(struct dpif_backer
*backer
, const char *devname
)
718 struct ofproto_dpif
*ofproto
;
719 struct dpif_port port
;
721 /* Don't report on the datapath's device. */
722 if (!strcmp(devname
, dpif_base_name(backer
->dpif
))) {
726 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
,
727 &all_ofproto_dpifs
) {
728 if (simap_contains(&ofproto
->backer
->tnl_backers
, devname
)) {
733 ofproto
= lookup_ofproto_dpif_by_port_name(devname
);
734 if (dpif_port_query_by_name(backer
->dpif
, devname
, &port
)) {
735 /* The port was removed. If we know the datapath,
736 * report it through poll_set(). If we don't, it may be
737 * notifying us of a removal we initiated, so ignore it.
738 * If there's a pending ENOBUFS, let it stand, since
739 * everything will be reevaluated. */
740 if (ofproto
&& ofproto
->port_poll_errno
!= ENOBUFS
) {
741 sset_add(&ofproto
->port_poll_set
, devname
);
742 ofproto
->port_poll_errno
= 0;
744 } else if (!ofproto
) {
745 /* The port was added, but we don't know with which
746 * ofproto we should associate it. Delete it. */
747 dpif_port_del(backer
->dpif
, port
.port_no
);
749 struct ofport_dpif
*ofport
;
751 ofport
= ofport_dpif_cast(shash_find_data(
752 &ofproto
->up
.port_by_name
, devname
));
754 && ofport
->odp_port
!= port
.port_no
755 && !odp_port_to_ofport(backer
, port
.port_no
))
757 /* 'ofport''s datapath port number has changed from
758 * 'ofport->odp_port' to 'port.port_no'. Update our internal data
759 * structures to match. */
760 ovs_rwlock_wrlock(&backer
->odp_to_ofport_lock
);
761 hmap_remove(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
);
762 ofport
->odp_port
= port
.port_no
;
763 hmap_insert(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
,
764 hash_odp_port(port
.port_no
));
765 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
766 backer
->need_revalidate
= REV_RECONFIGURE
;
769 dpif_port_destroy(&port
);
772 /* Propagate 'error' to all ofprotos based on 'backer'. */
774 process_dpif_port_error(struct dpif_backer
*backer
, int error
)
776 struct ofproto_dpif
*ofproto
;
778 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
779 if (ofproto
->backer
== backer
) {
780 sset_clear(&ofproto
->port_poll_set
);
781 ofproto
->port_poll_errno
= error
;
787 type_wait(const char *type
)
789 struct dpif_backer
*backer
;
791 backer
= shash_find_data(&all_dpif_backers
, type
);
793 /* This is not necessarily a problem, since backers are only
794 * created on demand. */
798 dpif_wait(backer
->dpif
);
801 /* Basic life-cycle. */
803 static int add_internal_flows(struct ofproto_dpif
*);
805 static struct ofproto
*
808 struct ofproto_dpif
*ofproto
= xzalloc(sizeof *ofproto
);
813 dealloc(struct ofproto
*ofproto_
)
815 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
820 close_dpif_backer(struct dpif_backer
*backer
)
822 ovs_assert(backer
->refcount
> 0);
824 if (--backer
->refcount
) {
828 udpif_destroy(backer
->udpif
);
830 simap_destroy(&backer
->tnl_backers
);
831 ovs_rwlock_destroy(&backer
->odp_to_ofport_lock
);
832 hmap_destroy(&backer
->odp_to_ofport_map
);
833 shash_find_and_delete(&all_dpif_backers
, backer
->type
);
835 free(backer
->dp_version_string
);
836 dpif_close(backer
->dpif
);
840 /* Datapath port slated for removal from datapath. */
842 struct ovs_list list_node
;
846 static bool check_variable_length_userdata(struct dpif_backer
*backer
);
847 static void check_support(struct dpif_backer
*backer
);
850 open_dpif_backer(const char *type
, struct dpif_backer
**backerp
)
852 struct dpif_backer
*backer
;
853 struct dpif_port_dump port_dump
;
854 struct dpif_port port
;
855 struct shash_node
*node
;
856 struct ovs_list garbage_list
;
857 struct odp_garbage
*garbage
;
864 backer
= shash_find_data(&all_dpif_backers
, type
);
871 backer_name
= xasprintf("ovs-%s", type
);
873 /* Remove any existing datapaths, since we assume we're the only
874 * userspace controlling the datapath. */
876 dp_enumerate_names(type
, &names
);
877 SSET_FOR_EACH(name
, &names
) {
878 struct dpif
*old_dpif
;
880 /* Don't remove our backer if it exists. */
881 if (!strcmp(name
, backer_name
)) {
885 if (dpif_open(name
, type
, &old_dpif
)) {
886 VLOG_WARN("couldn't open old datapath %s to remove it", name
);
888 dpif_delete(old_dpif
);
889 dpif_close(old_dpif
);
892 sset_destroy(&names
);
894 backer
= xmalloc(sizeof *backer
);
896 error
= dpif_create_and_open(backer_name
, type
, &backer
->dpif
);
899 VLOG_ERR("failed to open datapath of type %s: %s", type
,
900 ovs_strerror(error
));
904 backer
->udpif
= udpif_create(backer
, backer
->dpif
);
906 backer
->type
= xstrdup(type
);
907 backer
->refcount
= 1;
908 hmap_init(&backer
->odp_to_ofport_map
);
909 ovs_rwlock_init(&backer
->odp_to_ofport_lock
);
910 backer
->need_revalidate
= 0;
911 simap_init(&backer
->tnl_backers
);
912 backer
->recv_set_enable
= !ofproto_get_flow_restore_wait();
915 if (backer
->recv_set_enable
) {
916 dpif_flow_flush(backer
->dpif
);
919 /* Loop through the ports already on the datapath and remove any
920 * that we don't need anymore. */
921 ovs_list_init(&garbage_list
);
922 dpif_port_dump_start(&port_dump
, backer
->dpif
);
923 while (dpif_port_dump_next(&port_dump
, &port
)) {
924 node
= shash_find(&init_ofp_ports
, port
.name
);
925 if (!node
&& strcmp(port
.name
, dpif_base_name(backer
->dpif
))) {
926 garbage
= xmalloc(sizeof *garbage
);
927 garbage
->odp_port
= port
.port_no
;
928 ovs_list_push_front(&garbage_list
, &garbage
->list_node
);
931 dpif_port_dump_done(&port_dump
);
933 LIST_FOR_EACH_POP (garbage
, list_node
, &garbage_list
) {
934 dpif_port_del(backer
->dpif
, garbage
->odp_port
);
938 shash_add(&all_dpif_backers
, type
, backer
);
940 check_support(backer
);
941 atomic_count_init(&backer
->tnl_count
, 0);
943 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
945 VLOG_ERR("failed to listen on datapath of type %s: %s",
946 type
, ovs_strerror(error
));
947 close_dpif_backer(backer
);
951 if (backer
->recv_set_enable
) {
952 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
955 /* This check fails if performed before udpif threads have been set,
956 * as the kernel module checks that the 'pid' in userspace action
958 backer
->support
.variable_length_userdata
959 = check_variable_length_userdata(backer
);
960 backer
->dp_version_string
= dpif_get_dp_version(backer
->dpif
);
966 ovs_native_tunneling_is_on(struct ofproto_dpif
*ofproto
)
968 return ofproto_use_tnl_push_pop
&& ofproto
->backer
->support
.tnl_push_pop
&&
969 atomic_count_get(&ofproto
->backer
->tnl_count
);
972 /* Tests whether 'backer''s datapath supports recirculation. Only newer
973 * datapaths support OVS_KEY_ATTR_RECIRC_ID in keys. We need to disable some
974 * features on older datapaths that don't support this feature.
976 * Returns false if 'backer' definitely does not support recirculation, true if
977 * it seems to support recirculation or if at least the error we get is
980 check_recirc(struct dpif_backer
*backer
)
983 struct odputil_keybuf keybuf
;
986 struct odp_flow_key_parms odp_parms
= {
993 memset(&flow
, 0, sizeof flow
);
997 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
998 odp_flow_key_from_flow(&odp_parms
, &key
);
999 enable_recirc
= dpif_probe_feature(backer
->dpif
, "recirculation", &key
,
1002 if (enable_recirc
) {
1003 VLOG_INFO("%s: Datapath supports recirculation",
1004 dpif_name(backer
->dpif
));
1006 VLOG_INFO("%s: Datapath does not support recirculation",
1007 dpif_name(backer
->dpif
));
1010 return enable_recirc
;
1013 /* Tests whether 'dpif' supports unique flow ids. We can skip serializing
1014 * some flow attributes for datapaths that support this feature.
1016 * Returns true if 'dpif' supports UFID for flow operations.
1017 * Returns false if 'dpif' does not support UFID. */
1019 check_ufid(struct dpif_backer
*backer
)
1022 struct odputil_keybuf keybuf
;
1026 struct odp_flow_key_parms odp_parms
= {
1030 memset(&flow
, 0, sizeof flow
);
1031 flow
.dl_type
= htons(0x1234);
1033 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1034 odp_flow_key_from_flow(&odp_parms
, &key
);
1035 dpif_flow_hash(backer
->dpif
, key
.data
, key
.size
, &ufid
);
1037 enable_ufid
= dpif_probe_feature(backer
->dpif
, "UFID", &key
, &ufid
);
1040 VLOG_INFO("%s: Datapath supports unique flow ids",
1041 dpif_name(backer
->dpif
));
1043 VLOG_INFO("%s: Datapath does not support unique flow ids",
1044 dpif_name(backer
->dpif
));
1049 /* Tests whether 'backer''s datapath supports variable-length
1050 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. We need
1051 * to disable some features on older datapaths that don't support this
1054 * Returns false if 'backer' definitely does not support variable-length
1055 * userdata, true if it seems to support them or if at least the error we get
1058 check_variable_length_userdata(struct dpif_backer
*backer
)
1060 struct eth_header
*eth
;
1061 struct ofpbuf actions
;
1062 struct dpif_execute execute
;
1063 struct dp_packet packet
;
1068 /* Compose a userspace action that will cause an ERANGE error on older
1069 * datapaths that don't support variable-length userdata.
1071 * We really test for using userdata longer than 8 bytes, but older
1072 * datapaths accepted these, silently truncating the userdata to 8 bytes.
1073 * The same older datapaths rejected userdata shorter than 8 bytes, so we
1074 * test for that instead as a proxy for longer userdata support. */
1075 ofpbuf_init(&actions
, 64);
1076 start
= nl_msg_start_nested(&actions
, OVS_ACTION_ATTR_USERSPACE
);
1077 nl_msg_put_u32(&actions
, OVS_USERSPACE_ATTR_PID
,
1078 dpif_port_get_pid(backer
->dpif
, ODPP_NONE
, 0));
1079 nl_msg_put_unspec_zero(&actions
, OVS_USERSPACE_ATTR_USERDATA
, 4);
1080 nl_msg_end_nested(&actions
, start
);
1082 /* Compose a dummy ethernet packet. */
1083 dp_packet_init(&packet
, ETH_HEADER_LEN
);
1084 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
1085 eth
->eth_type
= htons(0x1234);
1087 flow_extract(&packet
, &flow
);
1089 /* Execute the actions. On older datapaths this fails with ERANGE, on
1090 * newer datapaths it succeeds. */
1091 execute
.actions
= actions
.data
;
1092 execute
.actions_len
= actions
.size
;
1093 execute
.packet
= &packet
;
1094 execute
.flow
= &flow
;
1095 execute
.needs_help
= false;
1096 execute
.probe
= true;
1099 error
= dpif_execute(backer
->dpif
, &execute
);
1101 dp_packet_uninit(&packet
);
1102 ofpbuf_uninit(&actions
);
1109 /* Variable-length userdata is not supported. */
1110 VLOG_WARN("%s: datapath does not support variable-length userdata "
1111 "feature (needs Linux 3.10+ or kernel module from OVS "
1112 "1..11+). The NXAST_SAMPLE action will be ignored.",
1113 dpif_name(backer
->dpif
));
1117 /* Something odd happened. We're not sure whether variable-length
1118 * userdata is supported. Default to "yes". */
1119 VLOG_WARN("%s: variable-length userdata feature probe failed (%s)",
1120 dpif_name(backer
->dpif
), ovs_strerror(error
));
1125 /* Tests the MPLS label stack depth supported by 'backer''s datapath.
1127 * Returns the number of elements in a struct flow's mpls_lse field
1128 * if the datapath supports at least that many entries in an
1130 * Otherwise returns the number of MPLS push actions supported by
1133 check_max_mpls_depth(struct dpif_backer
*backer
)
1138 for (n
= 0; n
< FLOW_MAX_MPLS_LABELS
; n
++) {
1139 struct odputil_keybuf keybuf
;
1141 struct odp_flow_key_parms odp_parms
= {
1145 memset(&flow
, 0, sizeof flow
);
1146 flow
.dl_type
= htons(ETH_TYPE_MPLS
);
1147 flow_set_mpls_bos(&flow
, n
, 1);
1149 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1150 odp_flow_key_from_flow(&odp_parms
, &key
);
1151 if (!dpif_probe_feature(backer
->dpif
, "MPLS", &key
, NULL
)) {
1156 VLOG_INFO("%s: MPLS label stack length probed as %d",
1157 dpif_name(backer
->dpif
), n
);
1161 /* Tests whether 'backer''s datapath supports masked data in
1162 * OVS_ACTION_ATTR_SET actions. We need to disable some features on older
1163 * datapaths that don't support this feature. */
1165 check_masked_set_action(struct dpif_backer
*backer
)
1167 struct eth_header
*eth
;
1168 struct ofpbuf actions
;
1169 struct dpif_execute execute
;
1170 struct dp_packet packet
;
1173 struct ovs_key_ethernet key
, mask
;
1175 /* Compose a set action that will cause an EINVAL error on older
1176 * datapaths that don't support masked set actions.
1177 * Avoid using a full mask, as it could be translated to a non-masked
1178 * set action instead. */
1179 ofpbuf_init(&actions
, 64);
1180 memset(&key
, 0x53, sizeof key
);
1181 memset(&mask
, 0x7f, sizeof mask
);
1182 commit_masked_set_action(&actions
, OVS_KEY_ATTR_ETHERNET
, &key
, &mask
,
1185 /* Compose a dummy ethernet packet. */
1186 dp_packet_init(&packet
, ETH_HEADER_LEN
);
1187 eth
= dp_packet_put_zeros(&packet
, ETH_HEADER_LEN
);
1188 eth
->eth_type
= htons(0x1234);
1190 flow_extract(&packet
, &flow
);
1192 /* Execute the actions. On older datapaths this fails with EINVAL, on
1193 * newer datapaths it succeeds. */
1194 execute
.actions
= actions
.data
;
1195 execute
.actions_len
= actions
.size
;
1196 execute
.packet
= &packet
;
1197 execute
.flow
= &flow
;
1198 execute
.needs_help
= false;
1199 execute
.probe
= true;
1202 error
= dpif_execute(backer
->dpif
, &execute
);
1204 dp_packet_uninit(&packet
);
1205 ofpbuf_uninit(&actions
);
1208 /* Masked set action is not supported. */
1209 VLOG_INFO("%s: datapath does not support masked set action feature.",
1210 dpif_name(backer
->dpif
));
1215 #define CHECK_FEATURE__(NAME, SUPPORT, FIELD, VALUE) \
1217 check_##NAME(struct dpif_backer *backer) \
1220 struct odputil_keybuf keybuf; \
1221 struct ofpbuf key; \
1223 struct odp_flow_key_parms odp_parms = { \
1230 memset(&flow, 0, sizeof flow); \
1231 flow.FIELD = VALUE; \
1233 ofpbuf_use_stack(&key, &keybuf, sizeof keybuf); \
1234 odp_flow_key_from_flow(&odp_parms, &key); \
1235 enable = dpif_probe_feature(backer->dpif, #NAME, &key, NULL); \
1238 VLOG_INFO("%s: Datapath supports "#NAME, dpif_name(backer->dpif)); \
1240 VLOG_INFO("%s: Datapath does not support "#NAME, \
1241 dpif_name(backer->dpif)); \
1246 #define CHECK_FEATURE(FIELD) CHECK_FEATURE__(FIELD, FIELD, FIELD, 1)
1248 CHECK_FEATURE(ct_state
)
1249 CHECK_FEATURE(ct_zone
)
1250 CHECK_FEATURE(ct_mark
)
1251 CHECK_FEATURE__(ct_label
, ct_label
, ct_label
.u64
.lo
, 1)
1252 CHECK_FEATURE__(ct_state_nat
, ct_state
, ct_state
, CS_TRACKED
|CS_SRC_NAT
)
1254 #undef CHECK_FEATURE
1255 #undef CHECK_FEATURE__
1258 check_support(struct dpif_backer
*backer
)
1260 /* This feature needs to be tested after udpif threads are set. */
1261 backer
->support
.variable_length_userdata
= false;
1263 backer
->support
.odp
.recirc
= check_recirc(backer
);
1264 backer
->support
.odp
.max_mpls_depth
= check_max_mpls_depth(backer
);
1265 backer
->support
.masked_set_action
= check_masked_set_action(backer
);
1266 backer
->support
.ufid
= check_ufid(backer
);
1267 backer
->support
.tnl_push_pop
= dpif_supports_tnl_push_pop(backer
->dpif
);
1269 backer
->support
.odp
.ct_state
= check_ct_state(backer
);
1270 backer
->support
.odp
.ct_zone
= check_ct_zone(backer
);
1271 backer
->support
.odp
.ct_mark
= check_ct_mark(backer
);
1272 backer
->support
.odp
.ct_label
= check_ct_label(backer
);
1274 backer
->support
.odp
.ct_state_nat
= check_ct_state_nat(backer
);
1278 construct(struct ofproto
*ofproto_
)
1280 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1281 struct shash_node
*node
, *next
;
1284 /* Tunnel module can get used right after the udpif threads are running. */
1285 ofproto_tunnel_init();
1287 error
= open_dpif_backer(ofproto
->up
.type
, &ofproto
->backer
);
1292 uuid_generate(&ofproto
->uuid
);
1293 atomic_init(&ofproto
->tables_version
, CLS_MIN_VERSION
);
1294 ofproto
->netflow
= NULL
;
1295 ofproto
->sflow
= NULL
;
1296 ofproto
->ipfix
= NULL
;
1297 ofproto
->stp
= NULL
;
1298 ofproto
->rstp
= NULL
;
1299 ofproto
->dump_seq
= 0;
1300 hmap_init(&ofproto
->bundles
);
1301 ofproto
->ml
= mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME
);
1303 ofproto
->mbridge
= mbridge_create();
1304 ofproto
->has_bonded_bundles
= false;
1305 ofproto
->lacp_enabled
= false;
1306 ovs_mutex_init_adaptive(&ofproto
->stats_mutex
);
1308 guarded_list_init(&ofproto
->ams
);
1310 sset_init(&ofproto
->ports
);
1311 sset_init(&ofproto
->ghost_ports
);
1312 sset_init(&ofproto
->port_poll_set
);
1313 ofproto
->port_poll_errno
= 0;
1314 ofproto
->change_seq
= 0;
1315 ofproto
->ams_seq
= seq_create();
1316 ofproto
->ams_seqno
= seq_read(ofproto
->ams_seq
);
1319 SHASH_FOR_EACH_SAFE (node
, next
, &init_ofp_ports
) {
1320 struct iface_hint
*iface_hint
= node
->data
;
1322 if (!strcmp(iface_hint
->br_name
, ofproto
->up
.name
)) {
1323 /* Check if the datapath already has this port. */
1324 if (dpif_port_exists(ofproto
->backer
->dpif
, node
->name
)) {
1325 sset_add(&ofproto
->ports
, node
->name
);
1328 free(iface_hint
->br_name
);
1329 free(iface_hint
->br_type
);
1331 shash_delete(&init_ofp_ports
, node
);
1335 hmap_insert(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
,
1336 hash_string(ofproto
->up
.name
, 0));
1337 memset(&ofproto
->stats
, 0, sizeof ofproto
->stats
);
1339 ofproto_init_tables(ofproto_
, N_TABLES
);
1340 error
= add_internal_flows(ofproto
);
1342 ofproto
->up
.tables
[TBL_INTERNAL
].flags
= OFTABLE_HIDDEN
| OFTABLE_READONLY
;
1348 add_internal_miss_flow(struct ofproto_dpif
*ofproto
, int id
,
1349 const struct ofpbuf
*ofpacts
, struct rule_dpif
**rulep
)
1355 match_init_catchall(&match
);
1356 match_set_reg(&match
, 0, id
);
1358 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 0, 0, ofpacts
,
1360 *rulep
= error
? NULL
: rule_dpif_cast(rule
);
1366 add_internal_flows(struct ofproto_dpif
*ofproto
)
1368 struct ofpact_controller
*controller
;
1369 uint64_t ofpacts_stub
[128 / 8];
1370 struct ofpbuf ofpacts
;
1371 struct rule
*unused_rulep OVS_UNUSED
;
1376 ofpbuf_use_stack(&ofpacts
, ofpacts_stub
, sizeof ofpacts_stub
);
1379 controller
= ofpact_put_CONTROLLER(&ofpacts
);
1380 controller
->max_len
= UINT16_MAX
;
1381 controller
->controller_id
= 0;
1382 controller
->reason
= OFPR_IMPLICIT_MISS
;
1383 ofpact_finish_CONTROLLER(&ofpacts
, &controller
);
1385 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1386 &ofproto
->miss_rule
);
1391 ofpbuf_clear(&ofpacts
);
1392 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1393 &ofproto
->no_packet_in_rule
);
1398 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1399 &ofproto
->drop_frags_rule
);
1404 /* Drop any run away non-recirc rule lookups. Recirc_id has to be
1405 * zero when reaching this rule.
1407 * (priority=2), recirc_id=0, actions=drop
1409 ofpbuf_clear(&ofpacts
);
1410 match_init_catchall(&match
);
1411 match_set_recirc_id(&match
, 0);
1412 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 2, 0, &ofpacts
,
1418 destruct(struct ofproto
*ofproto_
)
1420 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1421 struct ofproto_async_msg
*am
;
1422 struct rule_dpif
*rule
;
1423 struct oftable
*table
;
1424 struct ovs_list ams
;
1426 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1428 xlate_remove_ofproto(ofproto
);
1431 /* Ensure that the upcall processing threads have no remaining references
1432 * to the ofproto or anything in it. */
1433 udpif_synchronize(ofproto
->backer
->udpif
);
1435 hmap_remove(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
);
1437 OFPROTO_FOR_EACH_TABLE (table
, &ofproto
->up
) {
1438 CLS_FOR_EACH (rule
, up
.cr
, &table
->cls
) {
1439 ofproto_rule_delete(&ofproto
->up
, &rule
->up
);
1442 ofproto_group_delete_all(&ofproto
->up
);
1444 guarded_list_pop_all(&ofproto
->ams
, &ams
);
1445 LIST_FOR_EACH_POP (am
, list_node
, &ams
) {
1446 ofproto_async_msg_free(am
);
1448 guarded_list_destroy(&ofproto
->ams
);
1450 recirc_free_ofproto(ofproto
, ofproto
->up
.name
);
1452 mbridge_unref(ofproto
->mbridge
);
1454 netflow_unref(ofproto
->netflow
);
1455 dpif_sflow_unref(ofproto
->sflow
);
1456 dpif_ipfix_unref(ofproto
->ipfix
);
1457 hmap_destroy(&ofproto
->bundles
);
1458 mac_learning_unref(ofproto
->ml
);
1459 mcast_snooping_unref(ofproto
->ms
);
1461 sset_destroy(&ofproto
->ports
);
1462 sset_destroy(&ofproto
->ghost_ports
);
1463 sset_destroy(&ofproto
->port_poll_set
);
1465 ovs_mutex_destroy(&ofproto
->stats_mutex
);
1467 seq_destroy(ofproto
->ams_seq
);
1469 close_dpif_backer(ofproto
->backer
);
1473 run(struct ofproto
*ofproto_
)
1475 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1476 uint64_t new_seq
, new_dump_seq
;
1478 if (mbridge_need_revalidate(ofproto
->mbridge
)) {
1479 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1480 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1481 mac_learning_flush(ofproto
->ml
);
1482 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1483 mcast_snooping_mdb_flush(ofproto
->ms
);
1486 /* Always updates the ofproto->ams_seqno to avoid frequent wakeup during
1487 * flow restore. Even though nothing is processed during flow restore,
1488 * all queued 'ams' will be handled immediately when flow restore
1490 ofproto
->ams_seqno
= seq_read(ofproto
->ams_seq
);
1492 /* Do not perform any periodic activity required by 'ofproto' while
1493 * waiting for flow restore to complete. */
1494 if (!ofproto_get_flow_restore_wait()) {
1495 struct ofproto_async_msg
*am
;
1496 struct ovs_list ams
;
1498 guarded_list_pop_all(&ofproto
->ams
, &ams
);
1499 LIST_FOR_EACH_POP (am
, list_node
, &ams
) {
1500 connmgr_send_async_msg(ofproto
->up
.connmgr
, am
);
1501 ofproto_async_msg_free(am
);
1505 if (ofproto
->netflow
) {
1506 netflow_run(ofproto
->netflow
);
1508 if (ofproto
->sflow
) {
1509 dpif_sflow_run(ofproto
->sflow
);
1511 if (ofproto
->ipfix
) {
1512 dpif_ipfix_run(ofproto
->ipfix
);
1515 new_seq
= seq_read(connectivity_seq_get());
1516 if (ofproto
->change_seq
!= new_seq
) {
1517 struct ofport_dpif
*ofport
;
1519 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1523 ofproto
->change_seq
= new_seq
;
1525 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1526 struct ofbundle
*bundle
;
1528 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1535 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1536 if (mac_learning_run(ofproto
->ml
)) {
1537 ofproto
->backer
->need_revalidate
= REV_MAC_LEARNING
;
1539 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1541 if (mcast_snooping_run(ofproto
->ms
)) {
1542 ofproto
->backer
->need_revalidate
= REV_MCAST_SNOOPING
;
1545 new_dump_seq
= seq_read(udpif_dump_seq(ofproto
->backer
->udpif
));
1546 if (ofproto
->dump_seq
!= new_dump_seq
) {
1547 struct rule
*rule
, *next_rule
;
1548 long long now
= time_msec();
1550 /* We know stats are relatively fresh, so now is a good time to do some
1552 ofproto
->dump_seq
= new_dump_seq
;
1554 /* Expire OpenFlow flows whose idle_timeout or hard_timeout
1556 ovs_mutex_lock(&ofproto_mutex
);
1557 LIST_FOR_EACH_SAFE (rule
, next_rule
, expirable
,
1558 &ofproto
->up
.expirable
) {
1559 rule_expire(rule_dpif_cast(rule
), now
);
1561 ovs_mutex_unlock(&ofproto_mutex
);
1563 /* All outstanding data in existing flows has been accounted, so it's a
1564 * good time to do bond rebalancing. */
1565 if (ofproto
->has_bonded_bundles
) {
1566 struct ofbundle
*bundle
;
1568 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1570 bond_rebalance(bundle
->bond
);
1579 ofproto_dpif_wait(struct ofproto
*ofproto_
)
1581 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1583 if (ofproto_get_flow_restore_wait()) {
1587 if (ofproto
->sflow
) {
1588 dpif_sflow_wait(ofproto
->sflow
);
1590 if (ofproto
->ipfix
) {
1591 dpif_ipfix_wait(ofproto
->ipfix
);
1593 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1594 struct ofbundle
*bundle
;
1596 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1597 bundle_wait(bundle
);
1600 if (ofproto
->netflow
) {
1601 netflow_wait(ofproto
->netflow
);
1603 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
1604 mac_learning_wait(ofproto
->ml
);
1605 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1606 mcast_snooping_wait(ofproto
->ms
);
1608 if (ofproto
->backer
->need_revalidate
) {
1609 poll_immediate_wake();
1612 seq_wait(udpif_dump_seq(ofproto
->backer
->udpif
), ofproto
->dump_seq
);
1613 seq_wait(ofproto
->ams_seq
, ofproto
->ams_seqno
);
1617 type_get_memory_usage(const char *type
, struct simap
*usage
)
1619 struct dpif_backer
*backer
;
1621 backer
= shash_find_data(&all_dpif_backers
, type
);
1623 udpif_get_memory_usage(backer
->udpif
, usage
);
1628 flush(struct ofproto
*ofproto_
)
1630 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1631 struct dpif_backer
*backer
= ofproto
->backer
;
1634 udpif_flush(backer
->udpif
);
1639 query_tables(struct ofproto
*ofproto
,
1640 struct ofputil_table_features
*features
,
1641 struct ofputil_table_stats
*stats
)
1643 strcpy(features
->name
, "classifier");
1648 for (i
= 0; i
< ofproto
->n_tables
; i
++) {
1649 unsigned long missed
, matched
;
1651 atomic_read_relaxed(&ofproto
->tables
[i
].n_matched
, &matched
);
1652 atomic_read_relaxed(&ofproto
->tables
[i
].n_missed
, &missed
);
1654 stats
[i
].matched_count
= matched
;
1655 stats
[i
].lookup_count
= matched
+ missed
;
1661 set_tables_version(struct ofproto
*ofproto_
, cls_version_t version
)
1663 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1665 atomic_store_relaxed(&ofproto
->tables_version
, version
);
1666 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
1669 static struct ofport
*
1672 struct ofport_dpif
*port
= xzalloc(sizeof *port
);
1677 port_dealloc(struct ofport
*port_
)
1679 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1684 port_construct(struct ofport
*port_
)
1686 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1687 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1688 const struct netdev
*netdev
= port
->up
.netdev
;
1689 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1690 const char *dp_port_name
;
1691 struct dpif_port dpif_port
;
1694 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1695 port
->bundle
= NULL
;
1699 port
->may_enable
= false;
1700 port
->stp_port
= NULL
;
1701 port
->stp_state
= STP_DISABLED
;
1702 port
->rstp_port
= NULL
;
1703 port
->rstp_state
= RSTP_DISABLED
;
1704 port
->is_tunnel
= false;
1708 port
->carrier_seq
= netdev_get_carrier_resets(netdev
);
1709 port
->is_layer3
= netdev_vport_is_layer3(netdev
);
1711 if (netdev_vport_is_patch(netdev
)) {
1712 /* By bailing out here, we don't submit the port to the sFlow module
1713 * to be considered for counter polling export. This is correct
1714 * because the patch port represents an interface that sFlow considers
1715 * to be "internal" to the switch as a whole, and therefore not a
1716 * candidate for counter polling. */
1717 port
->odp_port
= ODPP_NONE
;
1718 ofport_update_peer(port
);
1722 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
1723 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
, dp_port_name
,
1729 port
->odp_port
= dpif_port
.port_no
;
1731 if (netdev_get_tunnel_config(netdev
)) {
1732 atomic_count_inc(&ofproto
->backer
->tnl_count
);
1733 error
= tnl_port_add(port
, port
->up
.netdev
, port
->odp_port
,
1734 ovs_native_tunneling_is_on(ofproto
), dp_port_name
);
1736 atomic_count_dec(&ofproto
->backer
->tnl_count
);
1737 dpif_port_destroy(&dpif_port
);
1741 port
->is_tunnel
= true;
1742 if (ofproto
->ipfix
) {
1743 dpif_ipfix_add_tunnel_port(ofproto
->ipfix
, port_
, port
->odp_port
);
1746 /* Sanity-check that a mapping doesn't already exist. This
1747 * shouldn't happen for non-tunnel ports. */
1748 if (odp_port_to_ofp_port(ofproto
, port
->odp_port
) != OFPP_NONE
) {
1749 VLOG_ERR("port %s already has an OpenFlow port number",
1751 dpif_port_destroy(&dpif_port
);
1755 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1756 hmap_insert(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
,
1757 hash_odp_port(port
->odp_port
));
1758 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1760 dpif_port_destroy(&dpif_port
);
1762 if (ofproto
->sflow
) {
1763 dpif_sflow_add_port(ofproto
->sflow
, port_
, port
->odp_port
);
1770 port_destruct(struct ofport
*port_
, bool del
)
1772 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1773 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1774 const char *devname
= netdev_get_name(port
->up
.netdev
);
1775 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1776 const char *dp_port_name
;
1778 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1780 xlate_ofport_remove(port
);
1783 dp_port_name
= netdev_vport_get_dpif_port(port
->up
.netdev
, namebuf
,
1785 if (del
&& dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
1786 /* The underlying device is still there, so delete it. This
1787 * happens when the ofproto is being destroyed, since the caller
1788 * assumes that removal of attached ports will happen as part of
1790 if (!port
->is_tunnel
) {
1791 dpif_port_del(ofproto
->backer
->dpif
, port
->odp_port
);
1796 port
->peer
->peer
= NULL
;
1800 if (port
->odp_port
!= ODPP_NONE
&& !port
->is_tunnel
) {
1801 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1802 hmap_remove(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
);
1803 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1806 if (port
->is_tunnel
) {
1807 atomic_count_dec(&ofproto
->backer
->tnl_count
);
1810 if (port
->is_tunnel
&& ofproto
->ipfix
) {
1811 dpif_ipfix_del_tunnel_port(ofproto
->ipfix
, port
->odp_port
);
1815 sset_find_and_delete(&ofproto
->ports
, devname
);
1816 sset_find_and_delete(&ofproto
->ghost_ports
, devname
);
1817 bundle_remove(port_
);
1818 set_cfm(port_
, NULL
);
1819 set_bfd(port_
, NULL
);
1820 set_lldp(port_
, NULL
);
1821 if (port
->stp_port
) {
1822 stp_port_disable(port
->stp_port
);
1824 set_rstp_port(port_
, NULL
);
1825 if (ofproto
->sflow
) {
1826 dpif_sflow_del_port(ofproto
->sflow
, port
->odp_port
);
1833 port_modified(struct ofport
*port_
)
1835 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1836 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1837 const char *dp_port_name
;
1838 struct netdev
*netdev
= port
->up
.netdev
;
1840 if (port
->bundle
&& port
->bundle
->bond
) {
1841 bond_slave_set_netdev(port
->bundle
->bond
, port
, netdev
);
1845 cfm_set_netdev(port
->cfm
, netdev
);
1849 bfd_set_netdev(port
->bfd
, netdev
);
1852 ofproto_dpif_monitor_port_update(port
, port
->bfd
, port
->cfm
,
1853 port
->lldp
, &port
->up
.pp
.hw_addr
);
1855 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
1857 if (port
->is_tunnel
) {
1858 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1860 if (tnl_port_reconfigure(port
, netdev
, port
->odp_port
,
1861 ovs_native_tunneling_is_on(ofproto
),
1863 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1867 ofport_update_peer(port
);
1871 port_reconfigured(struct ofport
*port_
, enum ofputil_port_config old_config
)
1873 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1874 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1875 enum ofputil_port_config changed
= old_config
^ port
->up
.pp
.config
;
1877 if (changed
& (OFPUTIL_PC_NO_RECV
| OFPUTIL_PC_NO_RECV_STP
|
1878 OFPUTIL_PC_NO_FWD
| OFPUTIL_PC_NO_FLOOD
|
1879 OFPUTIL_PC_NO_PACKET_IN
)) {
1880 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1882 if (changed
& OFPUTIL_PC_NO_FLOOD
&& port
->bundle
) {
1883 bundle_update(port
->bundle
);
1889 set_sflow(struct ofproto
*ofproto_
,
1890 const struct ofproto_sflow_options
*sflow_options
)
1892 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1893 struct dpif_sflow
*ds
= ofproto
->sflow
;
1895 if (sflow_options
) {
1896 uint32_t old_probability
= ds
? dpif_sflow_get_probability(ds
) : 0;
1898 struct ofport_dpif
*ofport
;
1900 ds
= ofproto
->sflow
= dpif_sflow_create();
1901 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1902 dpif_sflow_add_port(ds
, &ofport
->up
, ofport
->odp_port
);
1905 dpif_sflow_set_options(ds
, sflow_options
);
1906 if (dpif_sflow_get_probability(ds
) != old_probability
) {
1907 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1911 dpif_sflow_unref(ds
);
1912 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1913 ofproto
->sflow
= NULL
;
1921 struct ofproto
*ofproto_
,
1922 const struct ofproto_ipfix_bridge_exporter_options
*bridge_exporter_options
,
1923 const struct ofproto_ipfix_flow_exporter_options
*flow_exporters_options
,
1924 size_t n_flow_exporters_options
)
1926 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1927 struct dpif_ipfix
*di
= ofproto
->ipfix
;
1928 bool has_options
= bridge_exporter_options
|| flow_exporters_options
;
1929 bool new_di
= false;
1931 if (has_options
&& !di
) {
1932 di
= ofproto
->ipfix
= dpif_ipfix_create();
1937 /* Call set_options in any case to cleanly flush the flow
1938 * caches in the last exporters that are to be destroyed. */
1939 dpif_ipfix_set_options(
1940 di
, bridge_exporter_options
, flow_exporters_options
,
1941 n_flow_exporters_options
);
1943 /* Add tunnel ports only when a new ipfix created */
1944 if (new_di
== true) {
1945 struct ofport_dpif
*ofport
;
1946 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1947 if (ofport
->is_tunnel
== true) {
1948 dpif_ipfix_add_tunnel_port(di
, &ofport
->up
, ofport
->odp_port
);
1954 dpif_ipfix_unref(di
);
1955 ofproto
->ipfix
= NULL
;
1963 get_ipfix_stats(const struct ofproto
*ofproto_
,
1965 struct ovs_list
*replies
)
1967 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1968 struct dpif_ipfix
*di
= ofproto
->ipfix
;
1971 return OFPERR_NXST_NOT_CONFIGURED
;
1974 return dpif_ipfix_get_stats(di
, bridge_ipfix
, replies
);
1978 set_cfm(struct ofport
*ofport_
, const struct cfm_settings
*s
)
1980 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1981 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
1982 struct cfm
*old
= ofport
->cfm
;
1987 ofport
->cfm
= cfm_create(ofport
->up
.netdev
);
1990 if (cfm_configure(ofport
->cfm
, s
)) {
1997 cfm_unref(ofport
->cfm
);
2000 if (ofport
->cfm
!= old
) {
2001 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2003 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
2004 ofport
->lldp
, &ofport
->up
.pp
.hw_addr
);
2009 cfm_status_changed(struct ofport
*ofport_
)
2011 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2013 return ofport
->cfm
? cfm_check_status_change(ofport
->cfm
) : true;
2017 get_cfm_status(const struct ofport
*ofport_
,
2018 struct cfm_status
*status
)
2020 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2024 cfm_get_status(ofport
->cfm
, status
);
2033 set_bfd(struct ofport
*ofport_
, const struct smap
*cfg
)
2035 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
2036 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2040 ofport
->bfd
= bfd_configure(old
, netdev_get_name(ofport
->up
.netdev
),
2041 cfg
, ofport
->up
.netdev
);
2042 if (ofport
->bfd
!= old
) {
2043 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2045 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
2046 ofport
->lldp
, &ofport
->up
.pp
.hw_addr
);
2051 bfd_status_changed(struct ofport
*ofport_
)
2053 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2055 return ofport
->bfd
? bfd_check_status_change(ofport
->bfd
) : true;
2059 get_bfd_status(struct ofport
*ofport_
, struct smap
*smap
)
2061 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2065 bfd_get_status(ofport
->bfd
, smap
);
2074 set_lldp(struct ofport
*ofport_
,
2075 const struct smap
*cfg
)
2077 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2081 if (!ofport
->lldp
) {
2082 struct ofproto_dpif
*ofproto
;
2084 ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2085 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2086 ofport
->lldp
= lldp_create(ofport
->up
.netdev
, ofport_
->mtu
, cfg
);
2089 if (!lldp_configure(ofport
->lldp
, cfg
)) {
2094 lldp_unref(ofport
->lldp
);
2095 ofport
->lldp
= NULL
;
2098 ofproto_dpif_monitor_port_update(ofport
,
2102 &ofport
->up
.pp
.hw_addr
);
2107 get_lldp_status(const struct ofport
*ofport_
,
2108 struct lldp_status
*status OVS_UNUSED
)
2110 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2112 return ofport
->lldp
? true : false;
2116 set_aa(struct ofproto
*ofproto OVS_UNUSED
,
2117 const struct aa_settings
*s
)
2119 return aa_configure(s
);
2123 aa_mapping_set(struct ofproto
*ofproto_ OVS_UNUSED
, void *aux
,
2124 const struct aa_mapping_settings
*s
)
2126 return aa_mapping_register(aux
, s
);
2130 aa_mapping_unset(struct ofproto
*ofproto OVS_UNUSED
, void *aux
)
2132 return aa_mapping_unregister(aux
);
2136 aa_vlan_get_queued(struct ofproto
*ofproto OVS_UNUSED
, struct ovs_list
*list
)
2138 return aa_get_vlan_queued(list
);
2142 aa_vlan_get_queue_size(struct ofproto
*ofproto OVS_UNUSED
)
2144 return aa_get_vlan_queue_size();
2148 /* Spanning Tree. */
2150 /* Called while rstp_mutex is held. */
2152 rstp_send_bpdu_cb(struct dp_packet
*pkt
, void *ofport_
, void *ofproto_
)
2154 struct ofproto_dpif
*ofproto
= ofproto_
;
2155 struct ofport_dpif
*ofport
= ofport_
;
2156 struct eth_header
*eth
= dp_packet_l2(pkt
);
2158 netdev_get_etheraddr(ofport
->up
.netdev
, ð
->eth_src
);
2159 if (eth_addr_is_zero(eth
->eth_src
)) {
2160 VLOG_WARN_RL(&rl
, "%s port %d: cannot send RSTP BPDU on a port which "
2161 "does not have a configured source MAC address.",
2162 ofproto
->up
.name
, ofp_to_u16(ofport
->up
.ofp_port
));
2164 ofproto_dpif_send_packet(ofport
, pkt
);
2166 dp_packet_delete(pkt
);
2170 send_bpdu_cb(struct dp_packet
*pkt
, int port_num
, void *ofproto_
)
2172 struct ofproto_dpif
*ofproto
= ofproto_
;
2173 struct stp_port
*sp
= stp_get_port(ofproto
->stp
, port_num
);
2174 struct ofport_dpif
*ofport
;
2176 ofport
= stp_port_get_aux(sp
);
2178 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on unknown port %d",
2179 ofproto
->up
.name
, port_num
);
2181 struct eth_header
*eth
= dp_packet_l2(pkt
);
2183 netdev_get_etheraddr(ofport
->up
.netdev
, ð
->eth_src
);
2184 if (eth_addr_is_zero(eth
->eth_src
)) {
2185 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on port %d "
2186 "with unknown MAC", ofproto
->up
.name
, port_num
);
2188 ofproto_dpif_send_packet(ofport
, pkt
);
2191 dp_packet_delete(pkt
);
2194 /* Configure RSTP on 'ofproto_' using the settings defined in 's'. */
2196 set_rstp(struct ofproto
*ofproto_
, const struct ofproto_rstp_settings
*s
)
2198 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2200 /* Only revalidate flows if the configuration changed. */
2201 if (!s
!= !ofproto
->rstp
) {
2202 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2206 if (!ofproto
->rstp
) {
2207 ofproto
->rstp
= rstp_create(ofproto_
->name
, s
->address
,
2208 rstp_send_bpdu_cb
, ofproto
);
2209 ofproto
->rstp_last_tick
= time_msec();
2211 rstp_set_bridge_address(ofproto
->rstp
, s
->address
);
2212 rstp_set_bridge_priority(ofproto
->rstp
, s
->priority
);
2213 rstp_set_bridge_ageing_time(ofproto
->rstp
, s
->ageing_time
);
2214 rstp_set_bridge_force_protocol_version(ofproto
->rstp
,
2215 s
->force_protocol_version
);
2216 rstp_set_bridge_max_age(ofproto
->rstp
, s
->bridge_max_age
);
2217 rstp_set_bridge_forward_delay(ofproto
->rstp
, s
->bridge_forward_delay
);
2218 rstp_set_bridge_transmit_hold_count(ofproto
->rstp
,
2219 s
->transmit_hold_count
);
2221 struct ofport
*ofport
;
2222 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2223 set_rstp_port(ofport
, NULL
);
2225 rstp_unref(ofproto
->rstp
);
2226 ofproto
->rstp
= NULL
;
2231 get_rstp_status(struct ofproto
*ofproto_
, struct ofproto_rstp_status
*s
)
2233 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2235 if (ofproto
->rstp
) {
2237 s
->root_id
= rstp_get_root_id(ofproto
->rstp
);
2238 s
->bridge_id
= rstp_get_bridge_id(ofproto
->rstp
);
2239 s
->designated_id
= rstp_get_designated_id(ofproto
->rstp
);
2240 s
->root_path_cost
= rstp_get_root_path_cost(ofproto
->rstp
);
2241 s
->designated_port_id
= rstp_get_designated_port_id(ofproto
->rstp
);
2242 s
->bridge_port_id
= rstp_get_bridge_port_id(ofproto
->rstp
);
2249 update_rstp_port_state(struct ofport_dpif
*ofport
)
2251 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2252 enum rstp_state state
;
2254 /* Figure out new state. */
2255 state
= ofport
->rstp_port
? rstp_port_get_state(ofport
->rstp_port
)
2259 if (ofport
->rstp_state
!= state
) {
2260 enum ofputil_port_state of_state
;
2263 VLOG_DBG("port %s: RSTP state changed from %s to %s",
2264 netdev_get_name(ofport
->up
.netdev
),
2265 rstp_state_name(ofport
->rstp_state
),
2266 rstp_state_name(state
));
2268 if (rstp_learn_in_state(ofport
->rstp_state
)
2269 != rstp_learn_in_state(state
)) {
2270 /* XXX: Learning action flows should also be flushed. */
2271 if (ofport
->bundle
) {
2272 if (!rstp_shift_root_learned_address(ofproto
->rstp
)
2273 || rstp_get_old_root_aux(ofproto
->rstp
) != ofport
) {
2274 bundle_flush_macs(ofport
->bundle
, false);
2278 fwd_change
= rstp_forward_in_state(ofport
->rstp_state
)
2279 != rstp_forward_in_state(state
);
2281 ofproto
->backer
->need_revalidate
= REV_RSTP
;
2282 ofport
->rstp_state
= state
;
2284 if (fwd_change
&& ofport
->bundle
) {
2285 bundle_update(ofport
->bundle
);
2288 /* Update the RSTP state bits in the OpenFlow port description. */
2289 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2290 of_state
|= (state
== RSTP_LEARNING
? OFPUTIL_PS_STP_LEARN
2291 : state
== RSTP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2292 : state
== RSTP_DISCARDING
? OFPUTIL_PS_STP_LISTEN
2294 ofproto_port_set_state(&ofport
->up
, of_state
);
2299 rstp_run(struct ofproto_dpif
*ofproto
)
2301 if (ofproto
->rstp
) {
2302 long long int now
= time_msec();
2303 long long int elapsed
= now
- ofproto
->rstp_last_tick
;
2304 struct rstp_port
*rp
;
2305 struct ofport_dpif
*ofport
;
2307 /* Every second, decrease the values of the timers. */
2308 if (elapsed
>= 1000) {
2309 rstp_tick_timers(ofproto
->rstp
);
2310 ofproto
->rstp_last_tick
= now
;
2313 while ((ofport
= rstp_get_next_changed_port_aux(ofproto
->rstp
, &rp
))) {
2314 update_rstp_port_state(ofport
);
2318 /* FIXME: This check should be done on-event (i.e., when setting
2319 * p->fdb_flush) and not periodically.
2321 while ((ofport
= rstp_check_and_reset_fdb_flush(ofproto
->rstp
, &rp
))) {
2322 if (!rstp_shift_root_learned_address(ofproto
->rstp
)
2323 || rstp_get_old_root_aux(ofproto
->rstp
) != ofport
) {
2324 bundle_flush_macs(ofport
->bundle
, false);
2328 if (rstp_shift_root_learned_address(ofproto
->rstp
)) {
2329 struct ofport_dpif
*old_root_aux
=
2330 (struct ofport_dpif
*)rstp_get_old_root_aux(ofproto
->rstp
);
2331 struct ofport_dpif
*new_root_aux
=
2332 (struct ofport_dpif
*)rstp_get_new_root_aux(ofproto
->rstp
);
2333 if (old_root_aux
!= NULL
&& new_root_aux
!= NULL
) {
2334 bundle_move(old_root_aux
->bundle
, new_root_aux
->bundle
);
2335 rstp_reset_root_changed(ofproto
->rstp
);
2341 /* Configures STP on 'ofproto_' using the settings defined in 's'. */
2343 set_stp(struct ofproto
*ofproto_
, const struct ofproto_stp_settings
*s
)
2345 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2347 /* Only revalidate flows if the configuration changed. */
2348 if (!s
!= !ofproto
->stp
) {
2349 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2353 if (!ofproto
->stp
) {
2354 ofproto
->stp
= stp_create(ofproto_
->name
, s
->system_id
,
2355 send_bpdu_cb
, ofproto
);
2356 ofproto
->stp_last_tick
= time_msec();
2359 stp_set_bridge_id(ofproto
->stp
, s
->system_id
);
2360 stp_set_bridge_priority(ofproto
->stp
, s
->priority
);
2361 stp_set_hello_time(ofproto
->stp
, s
->hello_time
);
2362 stp_set_max_age(ofproto
->stp
, s
->max_age
);
2363 stp_set_forward_delay(ofproto
->stp
, s
->fwd_delay
);
2365 struct ofport
*ofport
;
2367 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2368 set_stp_port(ofport
, NULL
);
2371 stp_unref(ofproto
->stp
);
2372 ofproto
->stp
= NULL
;
2379 get_stp_status(struct ofproto
*ofproto_
, struct ofproto_stp_status
*s
)
2381 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2385 s
->bridge_id
= stp_get_bridge_id(ofproto
->stp
);
2386 s
->designated_root
= stp_get_designated_root(ofproto
->stp
);
2387 s
->root_path_cost
= stp_get_root_path_cost(ofproto
->stp
);
2396 update_stp_port_state(struct ofport_dpif
*ofport
)
2398 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2399 enum stp_state state
;
2401 /* Figure out new state. */
2402 state
= ofport
->stp_port
? stp_port_get_state(ofport
->stp_port
)
2406 if (ofport
->stp_state
!= state
) {
2407 enum ofputil_port_state of_state
;
2410 VLOG_DBG("port %s: STP state changed from %s to %s",
2411 netdev_get_name(ofport
->up
.netdev
),
2412 stp_state_name(ofport
->stp_state
),
2413 stp_state_name(state
));
2414 if (stp_learn_in_state(ofport
->stp_state
)
2415 != stp_learn_in_state(state
)) {
2416 /* xxx Learning action flows should also be flushed. */
2417 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2418 mac_learning_flush(ofproto
->ml
);
2419 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2420 mcast_snooping_mdb_flush(ofproto
->ms
);
2422 fwd_change
= stp_forward_in_state(ofport
->stp_state
)
2423 != stp_forward_in_state(state
);
2425 ofproto
->backer
->need_revalidate
= REV_STP
;
2426 ofport
->stp_state
= state
;
2427 ofport
->stp_state_entered
= time_msec();
2429 if (fwd_change
&& ofport
->bundle
) {
2430 bundle_update(ofport
->bundle
);
2433 /* Update the STP state bits in the OpenFlow port description. */
2434 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2435 of_state
|= (state
== STP_LISTENING
? OFPUTIL_PS_STP_LISTEN
2436 : state
== STP_LEARNING
? OFPUTIL_PS_STP_LEARN
2437 : state
== STP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2438 : state
== STP_BLOCKING
? OFPUTIL_PS_STP_BLOCK
2440 ofproto_port_set_state(&ofport
->up
, of_state
);
2444 /* Configures STP on 'ofport_' using the settings defined in 's'. The
2445 * caller is responsible for assigning STP port numbers and ensuring
2446 * there are no duplicates. */
2448 set_stp_port(struct ofport
*ofport_
,
2449 const struct ofproto_port_stp_settings
*s
)
2451 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2452 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2453 struct stp_port
*sp
= ofport
->stp_port
;
2455 if (!s
|| !s
->enable
) {
2457 ofport
->stp_port
= NULL
;
2458 stp_port_disable(sp
);
2459 update_stp_port_state(ofport
);
2462 } else if (sp
&& stp_port_no(sp
) != s
->port_num
2463 && ofport
== stp_port_get_aux(sp
)) {
2464 /* The port-id changed, so disable the old one if it's not
2465 * already in use by another port. */
2466 stp_port_disable(sp
);
2469 sp
= ofport
->stp_port
= stp_get_port(ofproto
->stp
, s
->port_num
);
2471 /* Set name before enabling the port so that debugging messages can print
2473 stp_port_set_name(sp
, netdev_get_name(ofport
->up
.netdev
));
2474 stp_port_enable(sp
);
2476 stp_port_set_aux(sp
, ofport
);
2477 stp_port_set_priority(sp
, s
->priority
);
2478 stp_port_set_path_cost(sp
, s
->path_cost
);
2480 update_stp_port_state(ofport
);
2486 get_stp_port_status(struct ofport
*ofport_
,
2487 struct ofproto_port_stp_status
*s
)
2489 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2490 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2491 struct stp_port
*sp
= ofport
->stp_port
;
2493 if (!ofproto
->stp
|| !sp
) {
2499 s
->port_id
= stp_port_get_id(sp
);
2500 s
->state
= stp_port_get_state(sp
);
2501 s
->sec_in_state
= (time_msec() - ofport
->stp_state_entered
) / 1000;
2502 s
->role
= stp_port_get_role(sp
);
2508 get_stp_port_stats(struct ofport
*ofport_
,
2509 struct ofproto_port_stp_stats
*s
)
2511 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2512 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2513 struct stp_port
*sp
= ofport
->stp_port
;
2515 if (!ofproto
->stp
|| !sp
) {
2521 stp_port_get_counts(sp
, &s
->tx_count
, &s
->rx_count
, &s
->error_count
);
2527 stp_run(struct ofproto_dpif
*ofproto
)
2530 long long int now
= time_msec();
2531 long long int elapsed
= now
- ofproto
->stp_last_tick
;
2532 struct stp_port
*sp
;
2535 stp_tick(ofproto
->stp
, MIN(INT_MAX
, elapsed
));
2536 ofproto
->stp_last_tick
= now
;
2538 while (stp_get_changed_port(ofproto
->stp
, &sp
)) {
2539 struct ofport_dpif
*ofport
= stp_port_get_aux(sp
);
2542 update_stp_port_state(ofport
);
2546 if (stp_check_and_reset_fdb_flush(ofproto
->stp
)) {
2547 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2548 mac_learning_flush(ofproto
->ml
);
2549 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2550 mcast_snooping_mdb_flush(ofproto
->ms
);
2556 stp_wait(struct ofproto_dpif
*ofproto
)
2559 poll_timer_wait(1000);
2563 /* Configures RSTP on 'ofport_' using the settings defined in 's'. The
2564 * caller is responsible for assigning RSTP port numbers and ensuring
2565 * there are no duplicates. */
2567 set_rstp_port(struct ofport
*ofport_
,
2568 const struct ofproto_port_rstp_settings
*s
)
2570 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2571 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2572 struct rstp_port
*rp
= ofport
->rstp_port
;
2574 if (!s
|| !s
->enable
) {
2576 rstp_port_set_aux(rp
, NULL
);
2577 rstp_port_set_state(rp
, RSTP_DISABLED
);
2578 rstp_port_set_mac_operational(rp
, false);
2579 ofport
->rstp_port
= NULL
;
2580 rstp_port_unref(rp
);
2581 update_rstp_port_state(ofport
);
2586 /* Check if need to add a new port. */
2588 rp
= ofport
->rstp_port
= rstp_add_port(ofproto
->rstp
);
2591 rstp_port_set(rp
, s
->port_num
, s
->priority
, s
->path_cost
,
2592 s
->admin_edge_port
, s
->auto_edge
,
2593 s
->admin_p2p_mac_state
, s
->admin_port_state
, s
->mcheck
,
2595 update_rstp_port_state(ofport
);
2596 /* Synchronize operational status. */
2597 rstp_port_set_mac_operational(rp
, ofport
->may_enable
);
2601 get_rstp_port_status(struct ofport
*ofport_
,
2602 struct ofproto_port_rstp_status
*s
)
2604 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2605 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2606 struct rstp_port
*rp
= ofport
->rstp_port
;
2608 if (!ofproto
->rstp
|| !rp
) {
2614 rstp_port_get_status(rp
, &s
->port_id
, &s
->state
, &s
->role
,
2615 &s
->designated_bridge_id
, &s
->designated_port_id
,
2616 &s
->designated_path_cost
, &s
->tx_count
,
2617 &s
->rx_count
, &s
->error_count
, &s
->uptime
);
2622 set_queues(struct ofport
*ofport_
, const struct ofproto_port_queue
*qdscp
,
2625 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2626 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2628 if (ofport
->n_qdscp
!= n_qdscp
2629 || (n_qdscp
&& memcmp(ofport
->qdscp
, qdscp
,
2630 n_qdscp
* sizeof *qdscp
))) {
2631 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2632 free(ofport
->qdscp
);
2633 ofport
->qdscp
= n_qdscp
2634 ? xmemdup(qdscp
, n_qdscp
* sizeof *qdscp
)
2636 ofport
->n_qdscp
= n_qdscp
;
2644 /* Expires all MAC learning entries associated with 'bundle' and forces its
2645 * ofproto to revalidate every flow.
2647 * Normally MAC learning entries are removed only from the ofproto associated
2648 * with 'bundle', but if 'all_ofprotos' is true, then the MAC learning entries
2649 * are removed from every ofproto. When patch ports and SLB bonds are in use
2650 * and a VM migration happens and the gratuitous ARPs are somehow lost, this
2651 * avoids a MAC_ENTRY_IDLE_TIME delay before the migrated VM can communicate
2652 * with the host from which it migrated. */
2654 bundle_flush_macs(struct ofbundle
*bundle
, bool all_ofprotos
)
2656 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2657 struct mac_learning
*ml
= ofproto
->ml
;
2658 struct mac_entry
*mac
, *next_mac
;
2660 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2661 ovs_rwlock_wrlock(&ml
->rwlock
);
2662 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2663 if (mac_entry_get_port(ml
, mac
) == bundle
) {
2665 struct ofproto_dpif
*o
;
2667 HMAP_FOR_EACH (o
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
2669 struct mac_entry
*e
;
2671 ovs_rwlock_wrlock(&o
->ml
->rwlock
);
2672 e
= mac_learning_lookup(o
->ml
, mac
->mac
, mac
->vlan
);
2674 mac_learning_expire(o
->ml
, e
);
2676 ovs_rwlock_unlock(&o
->ml
->rwlock
);
2681 mac_learning_expire(ml
, mac
);
2684 ovs_rwlock_unlock(&ml
->rwlock
);
2688 bundle_move(struct ofbundle
*old
, struct ofbundle
*new)
2690 struct ofproto_dpif
*ofproto
= old
->ofproto
;
2691 struct mac_learning
*ml
= ofproto
->ml
;
2692 struct mac_entry
*mac
, *next_mac
;
2694 ovs_assert(new->ofproto
== old
->ofproto
);
2696 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2697 ovs_rwlock_wrlock(&ml
->rwlock
);
2698 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2699 if (mac_entry_get_port(ml
, mac
) == old
) {
2700 mac_entry_set_port(ml
, mac
, new);
2703 ovs_rwlock_unlock(&ml
->rwlock
);
2706 static struct ofbundle
*
2707 bundle_lookup(const struct ofproto_dpif
*ofproto
, void *aux
)
2709 struct ofbundle
*bundle
;
2711 HMAP_FOR_EACH_IN_BUCKET (bundle
, hmap_node
, hash_pointer(aux
, 0),
2712 &ofproto
->bundles
) {
2713 if (bundle
->aux
== aux
) {
2721 bundle_update(struct ofbundle
*bundle
)
2723 struct ofport_dpif
*port
;
2725 bundle
->floodable
= true;
2726 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2727 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2729 || (bundle
->ofproto
->stp
&& !stp_forward_in_state(port
->stp_state
))
2730 || (bundle
->ofproto
->rstp
&& !rstp_forward_in_state(port
->rstp_state
))) {
2731 bundle
->floodable
= false;
2738 bundle_del_port(struct ofport_dpif
*port
)
2740 struct ofbundle
*bundle
= port
->bundle
;
2742 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2744 ovs_list_remove(&port
->bundle_node
);
2745 port
->bundle
= NULL
;
2748 lacp_slave_unregister(bundle
->lacp
, port
);
2751 bond_slave_unregister(bundle
->bond
, port
);
2754 bundle_update(bundle
);
2758 bundle_add_port(struct ofbundle
*bundle
, ofp_port_t ofp_port
,
2759 struct lacp_slave_settings
*lacp
)
2761 struct ofport_dpif
*port
;
2763 port
= ofp_port_to_ofport(bundle
->ofproto
, ofp_port
);
2768 if (port
->bundle
!= bundle
) {
2769 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2771 bundle_remove(&port
->up
);
2774 port
->bundle
= bundle
;
2775 ovs_list_push_back(&bundle
->ports
, &port
->bundle_node
);
2776 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2778 || (bundle
->ofproto
->stp
&& !stp_forward_in_state(port
->stp_state
))
2779 || (bundle
->ofproto
->rstp
&& !rstp_forward_in_state(port
->rstp_state
))) {
2780 bundle
->floodable
= false;
2784 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2785 lacp_slave_register(bundle
->lacp
, port
, lacp
);
2792 bundle_destroy(struct ofbundle
*bundle
)
2794 struct ofproto_dpif
*ofproto
;
2795 struct ofport_dpif
*port
, *next_port
;
2801 ofproto
= bundle
->ofproto
;
2802 mbridge_unregister_bundle(ofproto
->mbridge
, bundle
);
2805 xlate_bundle_remove(bundle
);
2808 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2809 bundle_del_port(port
);
2812 bundle_flush_macs(bundle
, true);
2813 hmap_remove(&ofproto
->bundles
, &bundle
->hmap_node
);
2815 free(bundle
->trunks
);
2816 lacp_unref(bundle
->lacp
);
2817 bond_unref(bundle
->bond
);
2822 bundle_set(struct ofproto
*ofproto_
, void *aux
,
2823 const struct ofproto_bundle_settings
*s
)
2825 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2826 bool need_flush
= false;
2827 struct ofport_dpif
*port
;
2828 struct ofbundle
*bundle
;
2829 unsigned long *trunks
;
2835 bundle_destroy(bundle_lookup(ofproto
, aux
));
2839 ovs_assert(s
->n_slaves
== 1 || s
->bond
!= NULL
);
2840 ovs_assert((s
->lacp
!= NULL
) == (s
->lacp_slaves
!= NULL
));
2842 bundle
= bundle_lookup(ofproto
, aux
);
2844 bundle
= xmalloc(sizeof *bundle
);
2846 bundle
->ofproto
= ofproto
;
2847 hmap_insert(&ofproto
->bundles
, &bundle
->hmap_node
,
2848 hash_pointer(aux
, 0));
2850 bundle
->name
= NULL
;
2852 ovs_list_init(&bundle
->ports
);
2853 bundle
->vlan_mode
= PORT_VLAN_TRUNK
;
2855 bundle
->trunks
= NULL
;
2856 bundle
->use_priority_tags
= s
->use_priority_tags
;
2857 bundle
->lacp
= NULL
;
2858 bundle
->bond
= NULL
;
2860 bundle
->floodable
= true;
2861 mbridge_register_bundle(ofproto
->mbridge
, bundle
);
2864 if (!bundle
->name
|| strcmp(s
->name
, bundle
->name
)) {
2866 bundle
->name
= xstrdup(s
->name
);
2871 ofproto
->lacp_enabled
= true;
2872 if (!bundle
->lacp
) {
2873 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2874 bundle
->lacp
= lacp_create();
2876 lacp_configure(bundle
->lacp
, s
->lacp
);
2878 lacp_unref(bundle
->lacp
);
2879 bundle
->lacp
= NULL
;
2882 /* Update set of ports. */
2884 for (i
= 0; i
< s
->n_slaves
; i
++) {
2885 if (!bundle_add_port(bundle
, s
->slaves
[i
],
2886 s
->lacp
? &s
->lacp_slaves
[i
] : NULL
)) {
2890 if (!ok
|| ovs_list_size(&bundle
->ports
) != s
->n_slaves
) {
2891 struct ofport_dpif
*next_port
;
2893 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2894 for (i
= 0; i
< s
->n_slaves
; i
++) {
2895 if (s
->slaves
[i
] == port
->up
.ofp_port
) {
2900 bundle_del_port(port
);
2904 ovs_assert(ovs_list_size(&bundle
->ports
) <= s
->n_slaves
);
2906 if (ovs_list_is_empty(&bundle
->ports
)) {
2907 bundle_destroy(bundle
);
2911 /* Set VLAN tagging mode */
2912 if (s
->vlan_mode
!= bundle
->vlan_mode
2913 || s
->use_priority_tags
!= bundle
->use_priority_tags
) {
2914 bundle
->vlan_mode
= s
->vlan_mode
;
2915 bundle
->use_priority_tags
= s
->use_priority_tags
;
2920 vlan
= (s
->vlan_mode
== PORT_VLAN_TRUNK
? -1
2921 : s
->vlan
>= 0 && s
->vlan
<= 4095 ? s
->vlan
2923 if (vlan
!= bundle
->vlan
) {
2924 bundle
->vlan
= vlan
;
2928 /* Get trunked VLANs. */
2929 switch (s
->vlan_mode
) {
2930 case PORT_VLAN_ACCESS
:
2934 case PORT_VLAN_TRUNK
:
2935 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2938 case PORT_VLAN_NATIVE_UNTAGGED
:
2939 case PORT_VLAN_NATIVE_TAGGED
:
2940 if (vlan
!= 0 && (!s
->trunks
2941 || !bitmap_is_set(s
->trunks
, vlan
)
2942 || bitmap_is_set(s
->trunks
, 0))) {
2943 /* Force trunking the native VLAN and prohibit trunking VLAN 0. */
2945 trunks
= bitmap_clone(s
->trunks
, 4096);
2947 trunks
= bitmap_allocate1(4096);
2949 bitmap_set1(trunks
, vlan
);
2950 bitmap_set0(trunks
, 0);
2952 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2959 if (!vlan_bitmap_equal(trunks
, bundle
->trunks
)) {
2960 free(bundle
->trunks
);
2961 if (trunks
== s
->trunks
) {
2962 bundle
->trunks
= vlan_bitmap_clone(trunks
);
2964 bundle
->trunks
= trunks
;
2969 if (trunks
!= s
->trunks
) {
2974 if (!ovs_list_is_short(&bundle
->ports
)) {
2975 bundle
->ofproto
->has_bonded_bundles
= true;
2977 if (bond_reconfigure(bundle
->bond
, s
->bond
)) {
2978 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2981 bundle
->bond
= bond_create(s
->bond
, ofproto
);
2982 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2985 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2986 bond_slave_register(bundle
->bond
, port
,
2987 port
->up
.ofp_port
, port
->up
.netdev
);
2990 bond_unref(bundle
->bond
);
2991 bundle
->bond
= NULL
;
2994 /* If we changed something that would affect MAC learning, un-learn
2995 * everything on this port and force flow revalidation. */
2997 bundle_flush_macs(bundle
, false);
3004 bundle_remove(struct ofport
*port_
)
3006 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
3007 struct ofbundle
*bundle
= port
->bundle
;
3010 bundle_del_port(port
);
3011 if (ovs_list_is_empty(&bundle
->ports
)) {
3012 bundle_destroy(bundle
);
3013 } else if (ovs_list_is_short(&bundle
->ports
)) {
3014 bond_unref(bundle
->bond
);
3015 bundle
->bond
= NULL
;
3021 send_pdu_cb(void *port_
, const void *pdu
, size_t pdu_size
)
3023 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 10);
3024 struct ofport_dpif
*port
= port_
;
3028 error
= netdev_get_etheraddr(port
->up
.netdev
, &ea
);
3030 struct dp_packet packet
;
3033 dp_packet_init(&packet
, 0);
3034 packet_pdu
= eth_compose(&packet
, eth_addr_lacp
, ea
, ETH_TYPE_LACP
,
3036 memcpy(packet_pdu
, pdu
, pdu_size
);
3038 ofproto_dpif_send_packet(port
, &packet
);
3039 dp_packet_uninit(&packet
);
3041 VLOG_ERR_RL(&rl
, "port %s: cannot obtain Ethernet address of iface "
3042 "%s (%s)", port
->bundle
->name
,
3043 netdev_get_name(port
->up
.netdev
), ovs_strerror(error
));
3048 bundle_send_learning_packets(struct ofbundle
*bundle
)
3050 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
3051 int error
, n_packets
, n_errors
;
3052 struct mac_entry
*e
;
3054 struct ovs_list list_node
;
3055 struct ofport_dpif
*port
;
3056 struct dp_packet
*pkt
;
3058 struct ovs_list packets
;
3060 ovs_list_init(&packets
);
3061 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
3062 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
3063 if (mac_entry_get_port(ofproto
->ml
, e
) != bundle
) {
3064 pkt_node
= xmalloc(sizeof *pkt_node
);
3065 pkt_node
->pkt
= bond_compose_learning_packet(bundle
->bond
,
3067 (void **)&pkt_node
->port
);
3068 ovs_list_push_back(&packets
, &pkt_node
->list_node
);
3071 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3073 error
= n_packets
= n_errors
= 0;
3074 LIST_FOR_EACH_POP (pkt_node
, list_node
, &packets
) {
3077 ret
= ofproto_dpif_send_packet(pkt_node
->port
, pkt_node
->pkt
);
3078 dp_packet_delete(pkt_node
->pkt
);
3088 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3089 VLOG_WARN_RL(&rl
, "bond %s: %d errors sending %d gratuitous learning "
3090 "packets, last error was: %s",
3091 bundle
->name
, n_errors
, n_packets
, ovs_strerror(error
));
3093 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
3094 bundle
->name
, n_packets
);
3099 bundle_run(struct ofbundle
*bundle
)
3102 lacp_run(bundle
->lacp
, send_pdu_cb
);
3105 struct ofport_dpif
*port
;
3107 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
3108 bond_slave_set_may_enable(bundle
->bond
, port
, port
->may_enable
);
3111 if (bond_run(bundle
->bond
, lacp_status(bundle
->lacp
))) {
3112 bundle
->ofproto
->backer
->need_revalidate
= REV_BOND
;
3115 if (bond_should_send_learning_packets(bundle
->bond
)) {
3116 bundle_send_learning_packets(bundle
);
3122 bundle_wait(struct ofbundle
*bundle
)
3125 lacp_wait(bundle
->lacp
);
3128 bond_wait(bundle
->bond
);
3135 mirror_set__(struct ofproto
*ofproto_
, void *aux
,
3136 const struct ofproto_mirror_settings
*s
)
3138 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3139 struct ofbundle
**srcs
, **dsts
;
3144 mirror_destroy(ofproto
->mbridge
, aux
);
3148 srcs
= xmalloc(s
->n_srcs
* sizeof *srcs
);
3149 dsts
= xmalloc(s
->n_dsts
* sizeof *dsts
);
3151 for (i
= 0; i
< s
->n_srcs
; i
++) {
3152 srcs
[i
] = bundle_lookup(ofproto
, s
->srcs
[i
]);
3155 for (i
= 0; i
< s
->n_dsts
; i
++) {
3156 dsts
[i
] = bundle_lookup(ofproto
, s
->dsts
[i
]);
3159 error
= mirror_set(ofproto
->mbridge
, aux
, s
->name
, srcs
, s
->n_srcs
, dsts
,
3160 s
->n_dsts
, s
->src_vlans
,
3161 bundle_lookup(ofproto
, s
->out_bundle
), s
->out_vlan
);
3168 mirror_get_stats__(struct ofproto
*ofproto
, void *aux
,
3169 uint64_t *packets
, uint64_t *bytes
)
3171 return mirror_get_stats(ofproto_dpif_cast(ofproto
)->mbridge
, aux
, packets
,
3176 set_flood_vlans(struct ofproto
*ofproto_
, unsigned long *flood_vlans
)
3178 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3179 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
3180 if (mac_learning_set_flood_vlans(ofproto
->ml
, flood_vlans
)) {
3181 mac_learning_flush(ofproto
->ml
);
3183 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3188 is_mirror_output_bundle(const struct ofproto
*ofproto_
, void *aux
)
3190 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3191 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3192 return bundle
&& mirror_bundle_out(ofproto
->mbridge
, bundle
) != 0;
3196 forward_bpdu_changed(struct ofproto
*ofproto_
)
3198 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3199 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3203 set_mac_table_config(struct ofproto
*ofproto_
, unsigned int idle_time
,
3206 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3207 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
3208 mac_learning_set_idle_time(ofproto
->ml
, idle_time
);
3209 mac_learning_set_max_entries(ofproto
->ml
, max_entries
);
3210 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3213 /* Configures multicast snooping on 'ofport' using the settings
3214 * defined in 's'. */
3216 set_mcast_snooping(struct ofproto
*ofproto_
,
3217 const struct ofproto_mcast_snooping_settings
*s
)
3219 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3221 /* Only revalidate flows if the configuration changed. */
3222 if (!s
!= !ofproto
->ms
) {
3223 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3228 ofproto
->ms
= mcast_snooping_create();
3231 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3232 mcast_snooping_set_idle_time(ofproto
->ms
, s
->idle_time
);
3233 mcast_snooping_set_max_entries(ofproto
->ms
, s
->max_entries
);
3234 if (mcast_snooping_set_flood_unreg(ofproto
->ms
, s
->flood_unreg
)) {
3235 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3237 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3239 mcast_snooping_unref(ofproto
->ms
);
3246 /* Configures multicast snooping port's flood settings on 'ofproto'. */
3248 set_mcast_snooping_port(struct ofproto
*ofproto_
, void *aux
,
3249 const struct ofproto_mcast_snooping_port_settings
*s
)
3251 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3252 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3254 if (ofproto
->ms
&& s
) {
3255 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3256 mcast_snooping_set_port_flood(ofproto
->ms
, bundle
, s
->flood
);
3257 mcast_snooping_set_port_flood_reports(ofproto
->ms
, bundle
,
3259 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3267 struct ofport_dpif
*
3268 ofp_port_to_ofport(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
3270 struct ofport
*ofport
= ofproto_get_port(&ofproto
->up
, ofp_port
);
3271 return ofport
? ofport_dpif_cast(ofport
) : NULL
;
3275 ofproto_port_from_dpif_port(struct ofproto_dpif
*ofproto
,
3276 struct ofproto_port
*ofproto_port
,
3277 struct dpif_port
*dpif_port
)
3279 ofproto_port
->name
= dpif_port
->name
;
3280 ofproto_port
->type
= dpif_port
->type
;
3281 ofproto_port
->ofp_port
= odp_port_to_ofp_port(ofproto
, dpif_port
->port_no
);
3285 ofport_update_peer(struct ofport_dpif
*ofport
)
3287 const struct ofproto_dpif
*ofproto
;
3288 struct dpif_backer
*backer
;
3291 if (!netdev_vport_is_patch(ofport
->up
.netdev
)) {
3295 backer
= ofproto_dpif_cast(ofport
->up
.ofproto
)->backer
;
3296 backer
->need_revalidate
= REV_RECONFIGURE
;
3299 ofport
->peer
->peer
= NULL
;
3300 ofport
->peer
= NULL
;
3303 peer_name
= netdev_vport_patch_peer(ofport
->up
.netdev
);
3308 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
3309 struct ofport
*peer_ofport
;
3310 struct ofport_dpif
*peer
;
3313 if (ofproto
->backer
!= backer
) {
3317 peer_ofport
= shash_find_data(&ofproto
->up
.port_by_name
, peer_name
);
3322 peer
= ofport_dpif_cast(peer_ofport
);
3323 peer_peer
= netdev_vport_patch_peer(peer
->up
.netdev
);
3324 if (peer_peer
&& !strcmp(netdev_get_name(ofport
->up
.netdev
),
3326 ofport
->peer
= peer
;
3327 ofport
->peer
->peer
= ofport
;
3337 port_run(struct ofport_dpif
*ofport
)
3339 long long int carrier_seq
= netdev_get_carrier_resets(ofport
->up
.netdev
);
3340 bool carrier_changed
= carrier_seq
!= ofport
->carrier_seq
;
3341 bool enable
= netdev_get_carrier(ofport
->up
.netdev
);
3342 bool cfm_enable
= false;
3343 bool bfd_enable
= false;
3345 ofport
->carrier_seq
= carrier_seq
;
3348 int cfm_opup
= cfm_get_opup(ofport
->cfm
);
3350 cfm_enable
= !cfm_get_fault(ofport
->cfm
);
3352 if (cfm_opup
>= 0) {
3353 cfm_enable
= cfm_enable
&& cfm_opup
;
3358 bfd_enable
= bfd_forwarding(ofport
->bfd
);
3361 if (ofport
->bfd
|| ofport
->cfm
) {
3362 enable
= enable
&& (cfm_enable
|| bfd_enable
);
3365 if (ofport
->bundle
) {
3366 enable
= enable
&& lacp_slave_may_enable(ofport
->bundle
->lacp
, ofport
);
3367 if (carrier_changed
) {
3368 lacp_slave_carrier_changed(ofport
->bundle
->lacp
, ofport
);
3372 if (ofport
->may_enable
!= enable
) {
3373 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3375 ofproto
->backer
->need_revalidate
= REV_PORT_TOGGLED
;
3377 if (ofport
->rstp_port
) {
3378 rstp_port_set_mac_operational(ofport
->rstp_port
, enable
);
3382 ofport
->may_enable
= enable
;
3386 port_query_by_name(const struct ofproto
*ofproto_
, const char *devname
,
3387 struct ofproto_port
*ofproto_port
)
3389 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3390 struct dpif_port dpif_port
;
3393 if (sset_contains(&ofproto
->ghost_ports
, devname
)) {
3394 const char *type
= netdev_get_type_from_name(devname
);
3396 /* We may be called before ofproto->up.port_by_name is populated with
3397 * the appropriate ofport. For this reason, we must get the name and
3398 * type from the netdev layer directly. */
3400 const struct ofport
*ofport
;
3402 ofport
= shash_find_data(&ofproto
->up
.port_by_name
, devname
);
3403 ofproto_port
->ofp_port
= ofport
? ofport
->ofp_port
: OFPP_NONE
;
3404 ofproto_port
->name
= xstrdup(devname
);
3405 ofproto_port
->type
= xstrdup(type
);
3411 if (!sset_contains(&ofproto
->ports
, devname
)) {
3414 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
3415 devname
, &dpif_port
);
3417 ofproto_port_from_dpif_port(ofproto
, ofproto_port
, &dpif_port
);
3423 port_add(struct ofproto
*ofproto_
, struct netdev
*netdev
)
3425 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3426 const char *devname
= netdev_get_name(netdev
);
3427 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
3428 const char *dp_port_name
;
3430 if (netdev_vport_is_patch(netdev
)) {
3431 sset_add(&ofproto
->ghost_ports
, netdev_get_name(netdev
));
3435 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
3436 if (!dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
3437 odp_port_t port_no
= ODPP_NONE
;
3440 error
= dpif_port_add(ofproto
->backer
->dpif
, netdev
, &port_no
);
3444 if (netdev_get_tunnel_config(netdev
)) {
3445 simap_put(&ofproto
->backer
->tnl_backers
,
3446 dp_port_name
, odp_to_u32(port_no
));
3450 if (netdev_get_tunnel_config(netdev
)) {
3451 sset_add(&ofproto
->ghost_ports
, devname
);
3453 sset_add(&ofproto
->ports
, devname
);
3459 port_del(struct ofproto
*ofproto_
, ofp_port_t ofp_port
)
3461 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3462 struct ofport_dpif
*ofport
= ofp_port_to_ofport(ofproto
, ofp_port
);
3469 sset_find_and_delete(&ofproto
->ghost_ports
,
3470 netdev_get_name(ofport
->up
.netdev
));
3471 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3472 if (!ofport
->is_tunnel
&& !netdev_vport_is_patch(ofport
->up
.netdev
)) {
3473 error
= dpif_port_del(ofproto
->backer
->dpif
, ofport
->odp_port
);
3475 /* The caller is going to close ofport->up.netdev. If this is a
3476 * bonded port, then the bond is using that netdev, so remove it
3477 * from the bond. The client will need to reconfigure everything
3478 * after deleting ports, so then the slave will get re-added. */
3479 bundle_remove(&ofport
->up
);
3486 port_get_stats(const struct ofport
*ofport_
, struct netdev_stats
*stats
)
3488 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3491 error
= netdev_get_stats(ofport
->up
.netdev
, stats
);
3493 if (!error
&& ofport_
->ofp_port
== OFPP_LOCAL
) {
3494 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3496 ovs_mutex_lock(&ofproto
->stats_mutex
);
3497 /* ofproto->stats.tx_packets represents packets that we created
3498 * internally and sent to some port (e.g. packets sent with
3499 * ofproto_dpif_send_packet()). Account for them as if they had
3500 * come from OFPP_LOCAL and got forwarded. */
3502 if (stats
->rx_packets
!= UINT64_MAX
) {
3503 stats
->rx_packets
+= ofproto
->stats
.tx_packets
;
3506 if (stats
->rx_bytes
!= UINT64_MAX
) {
3507 stats
->rx_bytes
+= ofproto
->stats
.tx_bytes
;
3510 /* ofproto->stats.rx_packets represents packets that were received on
3511 * some port and we processed internally and dropped (e.g. STP).
3512 * Account for them as if they had been forwarded to OFPP_LOCAL. */
3514 if (stats
->tx_packets
!= UINT64_MAX
) {
3515 stats
->tx_packets
+= ofproto
->stats
.rx_packets
;
3518 if (stats
->tx_bytes
!= UINT64_MAX
) {
3519 stats
->tx_bytes
+= ofproto
->stats
.rx_bytes
;
3521 ovs_mutex_unlock(&ofproto
->stats_mutex
);
3528 port_get_lacp_stats(const struct ofport
*ofport_
, struct lacp_slave_stats
*stats
)
3530 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3531 if (ofport
->bundle
&& ofport
->bundle
->lacp
) {
3532 if (lacp_get_slave_stats(ofport
->bundle
->lacp
, ofport
, stats
)) {
3539 struct port_dump_state
{
3540 struct sset_position pos
;
3543 struct ofproto_port port
;
3548 port_dump_start(const struct ofproto
*ofproto_ OVS_UNUSED
, void **statep
)
3550 *statep
= xzalloc(sizeof(struct port_dump_state
));
3555 port_dump_next(const struct ofproto
*ofproto_
, void *state_
,
3556 struct ofproto_port
*port
)
3558 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3559 struct port_dump_state
*state
= state_
;
3560 const struct sset
*sset
;
3561 struct sset_node
*node
;
3563 if (state
->has_port
) {
3564 ofproto_port_destroy(&state
->port
);
3565 state
->has_port
= false;
3567 sset
= state
->ghost
? &ofproto
->ghost_ports
: &ofproto
->ports
;
3568 while ((node
= sset_at_position(sset
, &state
->pos
))) {
3571 error
= port_query_by_name(ofproto_
, node
->name
, &state
->port
);
3573 *port
= state
->port
;
3574 state
->has_port
= true;
3576 } else if (error
!= ENODEV
) {
3581 if (!state
->ghost
) {
3582 state
->ghost
= true;
3583 memset(&state
->pos
, 0, sizeof state
->pos
);
3584 return port_dump_next(ofproto_
, state_
, port
);
3591 port_dump_done(const struct ofproto
*ofproto_ OVS_UNUSED
, void *state_
)
3593 struct port_dump_state
*state
= state_
;
3595 if (state
->has_port
) {
3596 ofproto_port_destroy(&state
->port
);
3603 port_poll(const struct ofproto
*ofproto_
, char **devnamep
)
3605 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3607 if (ofproto
->port_poll_errno
) {
3608 int error
= ofproto
->port_poll_errno
;
3609 ofproto
->port_poll_errno
= 0;
3613 if (sset_is_empty(&ofproto
->port_poll_set
)) {
3617 *devnamep
= sset_pop(&ofproto
->port_poll_set
);
3622 port_poll_wait(const struct ofproto
*ofproto_
)
3624 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3625 dpif_port_poll_wait(ofproto
->backer
->dpif
);
3629 port_is_lacp_current(const struct ofport
*ofport_
)
3631 const struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3632 return (ofport
->bundle
&& ofport
->bundle
->lacp
3633 ? lacp_slave_is_current(ofport
->bundle
->lacp
, ofport
)
3637 /* If 'rule' is an OpenFlow rule, that has expired according to OpenFlow rules,
3638 * then delete it entirely. */
3640 rule_expire(struct rule_dpif
*rule
, long long now
)
3641 OVS_REQUIRES(ofproto_mutex
)
3643 uint16_t hard_timeout
, idle_timeout
;
3646 hard_timeout
= rule
->up
.hard_timeout
;
3647 idle_timeout
= rule
->up
.idle_timeout
;
3649 /* Has 'rule' expired? */
3651 long long int modified
;
3653 ovs_mutex_lock(&rule
->up
.mutex
);
3654 modified
= rule
->up
.modified
;
3655 ovs_mutex_unlock(&rule
->up
.mutex
);
3657 if (now
> modified
+ hard_timeout
* 1000) {
3658 reason
= OFPRR_HARD_TIMEOUT
;
3662 if (reason
< 0 && idle_timeout
) {
3665 ovs_mutex_lock(&rule
->stats_mutex
);
3666 used
= rule
->stats
.used
;
3667 ovs_mutex_unlock(&rule
->stats_mutex
);
3669 if (now
> used
+ idle_timeout
* 1000) {
3670 reason
= OFPRR_IDLE_TIMEOUT
;
3675 COVERAGE_INC(ofproto_dpif_expired
);
3676 ofproto_rule_expire(&rule
->up
, reason
);
3681 ofproto_dpif_set_packet_odp_port(const struct ofproto_dpif
*ofproto
,
3682 ofp_port_t in_port
, struct dp_packet
*packet
)
3684 if (in_port
== OFPP_NONE
) {
3685 in_port
= OFPP_LOCAL
;
3687 packet
->md
.in_port
.odp_port
= ofp_port_to_odp_port(ofproto
, in_port
);
3691 ofproto_dpif_execute_actions__(struct ofproto_dpif
*ofproto
,
3692 const struct flow
*flow
,
3693 struct rule_dpif
*rule
,
3694 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3695 int indentation
, int depth
, int resubmits
,
3696 struct dp_packet
*packet
)
3698 struct dpif_flow_stats stats
;
3699 struct xlate_out xout
;
3700 struct xlate_in xin
;
3701 struct dpif_execute execute
;
3704 ovs_assert((rule
!= NULL
) != (ofpacts
!= NULL
));
3706 dpif_flow_stats_extract(flow
, packet
, time_msec(), &stats
);
3709 rule_dpif_credit_stats(rule
, &stats
);
3712 uint64_t odp_actions_stub
[1024 / 8];
3713 struct ofpbuf odp_actions
= OFPBUF_STUB_INITIALIZER(odp_actions_stub
);
3714 xlate_in_init(&xin
, ofproto
, flow
, flow
->in_port
.ofp_port
, rule
,
3715 stats
.tcp_flags
, packet
, NULL
, &odp_actions
);
3716 xin
.ofpacts
= ofpacts
;
3717 xin
.ofpacts_len
= ofpacts_len
;
3718 xin
.resubmit_stats
= &stats
;
3719 xin
.indentation
= indentation
;
3721 xin
.resubmits
= resubmits
;
3722 if (xlate_actions(&xin
, &xout
) != XLATE_OK
) {
3727 execute
.actions
= odp_actions
.data
;
3728 execute
.actions_len
= odp_actions
.size
;
3730 pkt_metadata_from_flow(&packet
->md
, flow
);
3731 execute
.packet
= packet
;
3732 execute
.flow
= flow
;
3733 execute
.needs_help
= (xout
.slow
& SLOW_ACTION
) != 0;
3734 execute
.probe
= false;
3737 /* Fix up in_port. */
3738 ofproto_dpif_set_packet_odp_port(ofproto
, flow
->in_port
.ofp_port
, packet
);
3740 error
= dpif_execute(ofproto
->backer
->dpif
, &execute
);
3742 xlate_out_uninit(&xout
);
3743 ofpbuf_uninit(&odp_actions
);
3748 /* Executes, within 'ofproto', the actions in 'rule' or 'ofpacts' on 'packet'.
3749 * 'flow' must reflect the data in 'packet'. */
3751 ofproto_dpif_execute_actions(struct ofproto_dpif
*ofproto
,
3752 const struct flow
*flow
,
3753 struct rule_dpif
*rule
,
3754 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3755 struct dp_packet
*packet
)
3757 return ofproto_dpif_execute_actions__(ofproto
, flow
, rule
, ofpacts
,
3758 ofpacts_len
, 0, 0, 0, packet
);
3762 rule_dpif_credit_stats(struct rule_dpif
*rule
,
3763 const struct dpif_flow_stats
*stats
)
3765 ovs_mutex_lock(&rule
->stats_mutex
);
3766 if (OVS_UNLIKELY(rule
->new_rule
)) {
3767 rule_dpif_credit_stats(rule
->new_rule
, stats
);
3769 rule
->stats
.n_packets
+= stats
->n_packets
;
3770 rule
->stats
.n_bytes
+= stats
->n_bytes
;
3771 rule
->stats
.used
= MAX(rule
->stats
.used
, stats
->used
);
3773 ovs_mutex_unlock(&rule
->stats_mutex
);
3777 rule_dpif_get_flow_cookie(const struct rule_dpif
*rule
)
3778 OVS_REQUIRES(rule
->up
.mutex
)
3780 return rule
->up
.flow_cookie
;
3784 rule_dpif_reduce_timeouts(struct rule_dpif
*rule
, uint16_t idle_timeout
,
3785 uint16_t hard_timeout
)
3787 ofproto_rule_reduce_timeouts(&rule
->up
, idle_timeout
, hard_timeout
);
3790 /* Returns 'rule''s actions. The returned actions are RCU-protected, and can
3791 * be read until the calling thread quiesces. */
3792 const struct rule_actions
*
3793 rule_dpif_get_actions(const struct rule_dpif
*rule
)
3795 return rule_get_actions(&rule
->up
);
3798 /* Sets 'rule''s recirculation id. */
3800 rule_dpif_set_recirc_id(struct rule_dpif
*rule
, uint32_t id
)
3801 OVS_REQUIRES(rule
->up
.mutex
)
3803 ovs_assert(!rule
->recirc_id
|| rule
->recirc_id
== id
);
3804 if (rule
->recirc_id
== id
) {
3805 /* Release the new reference to the same id. */
3808 rule
->recirc_id
= id
;
3812 /* Sets 'rule''s recirculation id. */
3814 rule_set_recirc_id(struct rule
*rule_
, uint32_t id
)
3816 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3818 ovs_mutex_lock(&rule
->up
.mutex
);
3819 rule_dpif_set_recirc_id(rule
, id
);
3820 ovs_mutex_unlock(&rule
->up
.mutex
);
3824 ofproto_dpif_get_tables_version(struct ofproto_dpif
*ofproto OVS_UNUSED
)
3826 cls_version_t version
;
3828 atomic_read_relaxed(&ofproto
->tables_version
, &version
);
3833 /* The returned rule (if any) is valid at least until the next RCU quiescent
3834 * period. If the rule needs to stay around longer, the caller should take
3837 * 'flow' is non-const to allow for temporary modifications during the lookup.
3838 * Any changes are restored before returning. */
3839 static struct rule_dpif
*
3840 rule_dpif_lookup_in_table(struct ofproto_dpif
*ofproto
, cls_version_t version
,
3841 uint8_t table_id
, struct flow
*flow
,
3842 struct flow_wildcards
*wc
)
3844 struct classifier
*cls
= &ofproto
->up
.tables
[table_id
].cls
;
3845 return rule_dpif_cast(rule_from_cls_rule(classifier_lookup(cls
, version
,
3849 /* Look up 'flow' in 'ofproto''s classifier version 'version', starting from
3850 * table '*table_id'. Returns the rule that was found, which may be one of the
3851 * special rules according to packet miss hadling. If 'may_packet_in' is
3852 * false, returning of the miss_rule (which issues packet ins for the
3853 * controller) is avoided. Updates 'wc', if nonnull, to reflect the fields
3854 * that were used during the lookup.
3856 * If 'honor_table_miss' is true, the first lookup occurs in '*table_id', but
3857 * if none is found then the table miss configuration for that table is
3858 * honored, which can result in additional lookups in other OpenFlow tables.
3859 * In this case the function updates '*table_id' to reflect the final OpenFlow
3860 * table that was searched.
3862 * If 'honor_table_miss' is false, then only one table lookup occurs, in
3865 * The rule is returned in '*rule', which is valid at least until the next
3866 * RCU quiescent period. If the '*rule' needs to stay around longer, the
3867 * caller must take a reference.
3869 * 'in_port' allows the lookup to take place as if the in port had the value
3870 * 'in_port'. This is needed for resubmit action support.
3872 * 'flow' is non-const to allow for temporary modifications during the lookup.
3873 * Any changes are restored before returning. */
3875 rule_dpif_lookup_from_table(struct ofproto_dpif
*ofproto
,
3876 cls_version_t version
, struct flow
*flow
,
3877 struct flow_wildcards
*wc
,
3878 const struct dpif_flow_stats
*stats
,
3879 uint8_t *table_id
, ofp_port_t in_port
,
3880 bool may_packet_in
, bool honor_table_miss
)
3882 ovs_be16 old_tp_src
= flow
->tp_src
, old_tp_dst
= flow
->tp_dst
;
3883 ofp_port_t old_in_port
= flow
->in_port
.ofp_port
;
3884 enum ofputil_table_miss miss_config
;
3885 struct rule_dpif
*rule
;
3888 /* We always unwildcard nw_frag (for IP), so they
3889 * need not be unwildcarded here. */
3890 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
3891 && ofproto
->up
.frag_handling
!= OFPUTIL_FRAG_NX_MATCH
) {
3892 if (ofproto
->up
.frag_handling
== OFPUTIL_FRAG_NORMAL
) {
3893 /* We must pretend that transport ports are unavailable. */
3894 flow
->tp_src
= htons(0);
3895 flow
->tp_dst
= htons(0);
3897 /* Must be OFPUTIL_FRAG_DROP (we don't have OFPUTIL_FRAG_REASM).
3898 * Use the drop_frags_rule (which cannot disappear). */
3899 rule
= ofproto
->drop_frags_rule
;
3901 struct oftable
*tbl
= &ofproto
->up
.tables
[*table_id
];
3904 atomic_add_relaxed(&tbl
->n_matched
, stats
->n_packets
, &orig
);
3910 /* Look up a flow with 'in_port' as the input port. Then restore the
3911 * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
3912 * have surprising behavior). */
3913 flow
->in_port
.ofp_port
= in_port
;
3915 /* Our current implementation depends on n_tables == N_TABLES, and
3916 * TBL_INTERNAL being the last table. */
3917 BUILD_ASSERT_DECL(N_TABLES
== TBL_INTERNAL
+ 1);
3919 miss_config
= OFPUTIL_TABLE_MISS_CONTINUE
;
3921 for (next_id
= *table_id
;
3922 next_id
< ofproto
->up
.n_tables
;
3923 next_id
++, next_id
+= (next_id
== TBL_INTERNAL
))
3925 *table_id
= next_id
;
3926 rule
= rule_dpif_lookup_in_table(ofproto
, version
, next_id
, flow
, wc
);
3928 struct oftable
*tbl
= &ofproto
->up
.tables
[next_id
];
3931 atomic_add_relaxed(rule
? &tbl
->n_matched
: &tbl
->n_missed
,
3932 stats
->n_packets
, &orig
);
3935 goto out
; /* Match. */
3937 if (honor_table_miss
) {
3938 miss_config
= ofproto_table_get_miss_config(&ofproto
->up
,
3940 if (miss_config
== OFPUTIL_TABLE_MISS_CONTINUE
) {
3947 rule
= ofproto
->no_packet_in_rule
;
3948 if (may_packet_in
) {
3949 if (miss_config
== OFPUTIL_TABLE_MISS_CONTINUE
3950 || miss_config
== OFPUTIL_TABLE_MISS_CONTROLLER
) {
3951 struct ofport_dpif
*port
;
3953 port
= ofp_port_to_ofport(ofproto
, old_in_port
);
3955 VLOG_WARN_RL(&rl
, "packet-in on unknown OpenFlow port %"PRIu16
,
3957 } else if (!(port
->up
.pp
.config
& OFPUTIL_PC_NO_PACKET_IN
)) {
3958 rule
= ofproto
->miss_rule
;
3960 } else if (miss_config
== OFPUTIL_TABLE_MISS_DEFAULT
&&
3961 connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
)) {
3962 rule
= ofproto
->miss_rule
;
3966 /* Restore port numbers, as they may have been modified above. */
3967 flow
->tp_src
= old_tp_src
;
3968 flow
->tp_dst
= old_tp_dst
;
3969 /* Restore the old in port. */
3970 flow
->in_port
.ofp_port
= old_in_port
;
3975 static struct rule_dpif
*rule_dpif_cast(const struct rule
*rule
)
3977 return rule
? CONTAINER_OF(rule
, struct rule_dpif
, up
) : NULL
;
3980 static struct rule
*
3983 struct rule_dpif
*rule
= xzalloc(sizeof *rule
);
3988 rule_dealloc(struct rule
*rule_
)
3990 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3995 check_mask(struct ofproto_dpif
*ofproto
, const struct miniflow
*flow
)
3997 const struct odp_support
*support
;
3998 uint16_t ct_state
, ct_zone
;
4002 support
= &ofproto_dpif_get_support(ofproto
)->odp
;
4003 ct_state
= MINIFLOW_GET_U16(flow
, ct_state
);
4004 if (support
->ct_state
&& support
->ct_zone
&& support
->ct_mark
4005 && support
->ct_label
&& support
->ct_state_nat
) {
4006 return ct_state
& CS_UNSUPPORTED_MASK
? OFPERR_OFPBMC_BAD_MASK
: 0;
4009 ct_zone
= MINIFLOW_GET_U16(flow
, ct_zone
);
4010 ct_mark
= MINIFLOW_GET_U32(flow
, ct_mark
);
4011 ct_label
= MINIFLOW_GET_U128(flow
, ct_label
);
4013 if ((ct_state
&& !support
->ct_state
)
4014 || (ct_state
& CS_UNSUPPORTED_MASK
)
4015 || ((ct_state
& (CS_SRC_NAT
| CS_DST_NAT
)) && !support
->ct_state_nat
)
4016 || (ct_zone
&& !support
->ct_zone
)
4017 || (ct_mark
&& !support
->ct_mark
)
4018 || (!ovs_u128_is_zero(ct_label
) && !support
->ct_label
)) {
4019 return OFPERR_OFPBMC_BAD_MASK
;
4026 check_actions(const struct ofproto_dpif
*ofproto
,
4027 const struct rule_actions
*const actions
)
4029 const struct ofpact
*ofpact
;
4031 OFPACT_FOR_EACH (ofpact
, actions
->ofpacts
, actions
->ofpacts_len
) {
4032 const struct odp_support
*support
;
4033 const struct ofpact_conntrack
*ct
;
4034 const struct ofpact
*a
;
4036 if (ofpact
->type
!= OFPACT_CT
) {
4040 ct
= CONTAINER_OF(ofpact
, struct ofpact_conntrack
, ofpact
);
4041 support
= &ofproto_dpif_get_support(ofproto
)->odp
;
4043 if (!support
->ct_state
) {
4044 return OFPERR_OFPBAC_BAD_TYPE
;
4046 if ((ct
->zone_imm
|| ct
->zone_src
.field
) && !support
->ct_zone
) {
4047 return OFPERR_OFPBAC_BAD_ARGUMENT
;
4050 OFPACT_FOR_EACH(a
, ct
->actions
, ofpact_ct_get_action_len(ct
)) {
4051 const struct mf_field
*dst
= ofpact_get_mf_dst(a
);
4053 if (a
->type
== OFPACT_NAT
&& !support
->ct_state_nat
) {
4054 /* The backer doesn't seem to support the NAT bits in
4055 * 'ct_state': assume that it doesn't support the NAT
4057 return OFPERR_OFPBAC_BAD_TYPE
;
4059 if (dst
&& ((dst
->id
== MFF_CT_MARK
&& !support
->ct_mark
)
4060 || (dst
->id
== MFF_CT_LABEL
&& !support
->ct_label
))) {
4061 return OFPERR_OFPBAC_BAD_SET_ARGUMENT
;
4070 rule_check(struct rule
*rule
)
4072 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->ofproto
);
4075 err
= check_mask(ofproto
, &rule
->cr
.match
.mask
->masks
);
4079 return check_actions(ofproto
, rule
->actions
);
4083 rule_construct(struct rule
*rule_
)
4084 OVS_NO_THREAD_SAFETY_ANALYSIS
4086 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4089 error
= rule_check(rule_
);
4094 ovs_mutex_init_adaptive(&rule
->stats_mutex
);
4095 rule
->stats
.n_packets
= 0;
4096 rule
->stats
.n_bytes
= 0;
4097 rule
->stats
.used
= rule
->up
.modified
;
4098 rule
->recirc_id
= 0;
4099 rule
->new_rule
= NULL
;
4105 rule_insert(struct rule
*rule_
, struct rule
*old_rule_
, bool forward_stats
)
4106 OVS_REQUIRES(ofproto_mutex
)
4108 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4110 if (old_rule_
&& forward_stats
) {
4111 struct rule_dpif
*old_rule
= rule_dpif_cast(old_rule_
);
4113 ovs_assert(!old_rule
->new_rule
);
4115 /* Take a reference to the new rule, and refer all stats updates from
4116 * the old rule to the new rule. */
4117 rule_dpif_ref(rule
);
4119 ovs_mutex_lock(&old_rule
->stats_mutex
);
4120 ovs_mutex_lock(&rule
->stats_mutex
);
4121 old_rule
->new_rule
= rule
; /* Forward future stats. */
4122 rule
->stats
= old_rule
->stats
; /* Transfer stats to the new rule. */
4123 ovs_mutex_unlock(&rule
->stats_mutex
);
4124 ovs_mutex_unlock(&old_rule
->stats_mutex
);
4129 rule_destruct(struct rule
*rule_
)
4130 OVS_NO_THREAD_SAFETY_ANALYSIS
4132 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4134 ovs_mutex_destroy(&rule
->stats_mutex
);
4135 /* Release reference to the new rule, if any. */
4136 if (rule
->new_rule
) {
4137 rule_dpif_unref(rule
->new_rule
);
4139 if (rule
->recirc_id
) {
4140 recirc_free_id(rule
->recirc_id
);
4145 rule_get_stats(struct rule
*rule_
, uint64_t *packets
, uint64_t *bytes
,
4146 long long int *used
)
4148 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
4150 ovs_mutex_lock(&rule
->stats_mutex
);
4151 if (OVS_UNLIKELY(rule
->new_rule
)) {
4152 rule_get_stats(&rule
->new_rule
->up
, packets
, bytes
, used
);
4154 *packets
= rule
->stats
.n_packets
;
4155 *bytes
= rule
->stats
.n_bytes
;
4156 *used
= rule
->stats
.used
;
4158 ovs_mutex_unlock(&rule
->stats_mutex
);
4162 rule_dpif_execute(struct rule_dpif
*rule
, const struct flow
*flow
,
4163 struct dp_packet
*packet
)
4165 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
4167 ofproto_dpif_execute_actions(ofproto
, flow
, rule
, NULL
, 0, packet
);
4171 rule_execute(struct rule
*rule
, const struct flow
*flow
,
4172 struct dp_packet
*packet
)
4174 rule_dpif_execute(rule_dpif_cast(rule
), flow
, packet
);
4175 dp_packet_delete(packet
);
4179 static struct group_dpif
*group_dpif_cast(const struct ofgroup
*group
)
4181 return group
? CONTAINER_OF(group
, struct group_dpif
, up
) : NULL
;
4184 static struct ofgroup
*
4187 struct group_dpif
*group
= xzalloc(sizeof *group
);
4192 group_dealloc(struct ofgroup
*group_
)
4194 struct group_dpif
*group
= group_dpif_cast(group_
);
4199 group_construct_stats(struct group_dpif
*group
)
4200 OVS_REQUIRES(group
->stats_mutex
)
4202 struct ofputil_bucket
*bucket
;
4203 const struct ovs_list
*buckets
;
4205 group
->packet_count
= 0;
4206 group
->byte_count
= 0;
4208 group_dpif_get_buckets(group
, &buckets
);
4209 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
4210 bucket
->stats
.packet_count
= 0;
4211 bucket
->stats
.byte_count
= 0;
4216 group_dpif_credit_stats(struct group_dpif
*group
,
4217 struct ofputil_bucket
*bucket
,
4218 const struct dpif_flow_stats
*stats
)
4220 ovs_mutex_lock(&group
->stats_mutex
);
4221 group
->packet_count
+= stats
->n_packets
;
4222 group
->byte_count
+= stats
->n_bytes
;
4224 bucket
->stats
.packet_count
+= stats
->n_packets
;
4225 bucket
->stats
.byte_count
+= stats
->n_bytes
;
4226 } else { /* Credit to all buckets */
4227 const struct ovs_list
*buckets
;
4229 group_dpif_get_buckets(group
, &buckets
);
4230 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
4231 bucket
->stats
.packet_count
+= stats
->n_packets
;
4232 bucket
->stats
.byte_count
+= stats
->n_bytes
;
4235 ovs_mutex_unlock(&group
->stats_mutex
);
4239 group_construct(struct ofgroup
*group_
)
4241 struct group_dpif
*group
= group_dpif_cast(group_
);
4243 ovs_mutex_init_adaptive(&group
->stats_mutex
);
4244 ovs_mutex_lock(&group
->stats_mutex
);
4245 group_construct_stats(group
);
4246 ovs_mutex_unlock(&group
->stats_mutex
);
4251 group_destruct(struct ofgroup
*group_
)
4253 struct group_dpif
*group
= group_dpif_cast(group_
);
4254 ovs_mutex_destroy(&group
->stats_mutex
);
4258 group_modify(struct ofgroup
*group_
)
4260 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(group_
->ofproto
);
4262 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
4268 group_get_stats(const struct ofgroup
*group_
, struct ofputil_group_stats
*ogs
)
4270 struct group_dpif
*group
= group_dpif_cast(group_
);
4271 struct ofputil_bucket
*bucket
;
4272 const struct ovs_list
*buckets
;
4273 struct bucket_counter
*bucket_stats
;
4275 ovs_mutex_lock(&group
->stats_mutex
);
4276 ogs
->packet_count
= group
->packet_count
;
4277 ogs
->byte_count
= group
->byte_count
;
4279 group_dpif_get_buckets(group
, &buckets
);
4280 bucket_stats
= ogs
->bucket_stats
;
4281 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
4282 bucket_stats
->packet_count
= bucket
->stats
.packet_count
;
4283 bucket_stats
->byte_count
= bucket
->stats
.byte_count
;
4286 ovs_mutex_unlock(&group
->stats_mutex
);
4291 /* If the group exists, this function increments the groups's reference count.
4293 * Make sure to call group_dpif_unref() after no longer needing to maintain
4294 * a reference to the group. */
4296 group_dpif_lookup(struct ofproto_dpif
*ofproto
, uint32_t group_id
,
4297 struct group_dpif
**group
)
4299 struct ofgroup
*ofgroup
;
4302 found
= ofproto_group_lookup(&ofproto
->up
, group_id
, &ofgroup
);
4303 *group
= found
? group_dpif_cast(ofgroup
) : NULL
;
4309 group_dpif_get_buckets(const struct group_dpif
*group
,
4310 const struct ovs_list
**buckets
)
4312 *buckets
= &group
->up
.buckets
;
4315 enum ofp11_group_type
4316 group_dpif_get_type(const struct group_dpif
*group
)
4318 return group
->up
.type
;
4322 group_dpif_get_selection_method(const struct group_dpif
*group
)
4324 return group
->up
.props
.selection_method
;
4327 /* Sends 'packet' out 'ofport'.
4328 * May modify 'packet'.
4329 * Returns 0 if successful, otherwise a positive errno value. */
4331 ofproto_dpif_send_packet(const struct ofport_dpif
*ofport
, struct dp_packet
*packet
)
4333 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
4336 error
= xlate_send_packet(ofport
, packet
);
4338 ovs_mutex_lock(&ofproto
->stats_mutex
);
4339 ofproto
->stats
.tx_packets
++;
4340 ofproto
->stats
.tx_bytes
+= dp_packet_size(packet
);
4341 ovs_mutex_unlock(&ofproto
->stats_mutex
);
4346 group_dpif_get_selection_method_param(const struct group_dpif
*group
)
4348 return group
->up
.props
.selection_method_param
;
4351 const struct field_array
*
4352 group_dpif_get_fields(const struct group_dpif
*group
)
4354 return &group
->up
.props
.fields
;
4357 /* Return the version string of the datapath that backs up
4361 get_datapath_version(const struct ofproto
*ofproto_
)
4363 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4365 return ofproto
->backer
->dp_version_string
;
4369 set_frag_handling(struct ofproto
*ofproto_
,
4370 enum ofputil_frag_handling frag_handling
)
4372 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4373 if (frag_handling
!= OFPUTIL_FRAG_REASM
) {
4374 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4382 packet_out(struct ofproto
*ofproto_
, struct dp_packet
*packet
,
4383 const struct flow
*flow
,
4384 const struct ofpact
*ofpacts
, size_t ofpacts_len
)
4386 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4388 ofproto_dpif_execute_actions(ofproto
, flow
, NULL
, ofpacts
,
4389 ofpacts_len
, packet
);
4394 nxt_resume(struct ofproto
*ofproto_
,
4395 const struct ofputil_packet_in_private
*pin
)
4397 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4399 /* Translate pin into datapath actions. */
4400 uint64_t odp_actions_stub
[1024 / 8];
4401 struct ofpbuf odp_actions
= OFPBUF_STUB_INITIALIZER(odp_actions_stub
);
4402 enum slow_path_reason slow
;
4403 enum ofperr error
= xlate_resume(ofproto
, pin
, &odp_actions
, &slow
);
4405 /* Steal 'pin->packet' and put it into a dp_packet. */
4406 struct dp_packet packet
;
4407 dp_packet_init(&packet
, pin
->public.packet_len
);
4408 dp_packet_put(&packet
, pin
->public.packet
, pin
->public.packet_len
);
4410 pkt_metadata_from_flow(&packet
.md
, &pin
->public.flow_metadata
.flow
);
4412 /* Fix up in_port. */
4413 ofproto_dpif_set_packet_odp_port(ofproto
,
4414 pin
->public.flow_metadata
.flow
.in_port
.ofp_port
,
4417 struct flow headers
;
4418 flow_extract(&packet
, &headers
);
4420 /* Execute the datapath actions on the packet. */
4421 struct dpif_execute execute
= {
4422 .actions
= odp_actions
.data
,
4423 .actions_len
= odp_actions
.size
,
4424 .needs_help
= (slow
& SLOW_ACTION
) != 0,
4428 dpif_execute(ofproto
->backer
->dpif
, &execute
);
4431 ofpbuf_uninit(&odp_actions
);
4432 dp_packet_uninit(&packet
);
4440 set_netflow(struct ofproto
*ofproto_
,
4441 const struct netflow_options
*netflow_options
)
4443 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4445 if (netflow_options
) {
4446 if (!ofproto
->netflow
) {
4447 ofproto
->netflow
= netflow_create();
4448 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4450 return netflow_set_options(ofproto
->netflow
, netflow_options
);
4451 } else if (ofproto
->netflow
) {
4452 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4453 netflow_unref(ofproto
->netflow
);
4454 ofproto
->netflow
= NULL
;
4461 get_netflow_ids(const struct ofproto
*ofproto_
,
4462 uint8_t *engine_type
, uint8_t *engine_id
)
4464 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4466 dpif_get_netflow_ids(ofproto
->backer
->dpif
, engine_type
, engine_id
);
4469 static struct ofproto_dpif
*
4470 ofproto_dpif_lookup(const char *name
)
4472 struct ofproto_dpif
*ofproto
;
4474 HMAP_FOR_EACH_WITH_HASH (ofproto
, all_ofproto_dpifs_node
,
4475 hash_string(name
, 0), &all_ofproto_dpifs
) {
4476 if (!strcmp(ofproto
->up
.name
, name
)) {
4484 ofproto_unixctl_fdb_flush(struct unixctl_conn
*conn
, int argc
,
4485 const char *argv
[], void *aux OVS_UNUSED
)
4487 struct ofproto_dpif
*ofproto
;
4490 ofproto
= ofproto_dpif_lookup(argv
[1]);
4492 unixctl_command_reply_error(conn
, "no such bridge");
4495 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4496 mac_learning_flush(ofproto
->ml
);
4497 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4499 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4500 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4501 mac_learning_flush(ofproto
->ml
);
4502 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4506 unixctl_command_reply(conn
, "table successfully flushed");
4510 ofproto_unixctl_mcast_snooping_flush(struct unixctl_conn
*conn
, int argc
,
4511 const char *argv
[], void *aux OVS_UNUSED
)
4513 struct ofproto_dpif
*ofproto
;
4516 ofproto
= ofproto_dpif_lookup(argv
[1]);
4518 unixctl_command_reply_error(conn
, "no such bridge");
4522 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4523 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4526 mcast_snooping_mdb_flush(ofproto
->ms
);
4528 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4529 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4532 mcast_snooping_mdb_flush(ofproto
->ms
);
4536 unixctl_command_reply(conn
, "table successfully flushed");
4539 static struct ofport_dpif
*
4540 ofbundle_get_a_port(const struct ofbundle
*bundle
)
4542 return CONTAINER_OF(ovs_list_front(&bundle
->ports
), struct ofport_dpif
,
4547 ofproto_unixctl_fdb_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4548 const char *argv
[], void *aux OVS_UNUSED
)
4550 struct ds ds
= DS_EMPTY_INITIALIZER
;
4551 const struct ofproto_dpif
*ofproto
;
4552 const struct mac_entry
*e
;
4554 ofproto
= ofproto_dpif_lookup(argv
[1]);
4556 unixctl_command_reply_error(conn
, "no such bridge");
4560 ds_put_cstr(&ds
, " port VLAN MAC Age\n");
4561 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
4562 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
4563 struct ofbundle
*bundle
= mac_entry_get_port(ofproto
->ml
, e
);
4564 char name
[OFP_MAX_PORT_NAME_LEN
];
4566 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4568 ds_put_format(&ds
, "%5s %4d "ETH_ADDR_FMT
" %3d\n",
4569 name
, e
->vlan
, ETH_ADDR_ARGS(e
->mac
),
4570 mac_entry_age(ofproto
->ml
, e
));
4572 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4573 unixctl_command_reply(conn
, ds_cstr(&ds
));
4578 ofproto_unixctl_mcast_snooping_show(struct unixctl_conn
*conn
,
4579 int argc OVS_UNUSED
,
4581 void *aux OVS_UNUSED
)
4583 struct ds ds
= DS_EMPTY_INITIALIZER
;
4584 const struct ofproto_dpif
*ofproto
;
4585 const struct ofbundle
*bundle
;
4586 const struct mcast_group
*grp
;
4587 struct mcast_group_bundle
*b
;
4588 struct mcast_mrouter_bundle
*mrouter
;
4590 ofproto
= ofproto_dpif_lookup(argv
[1]);
4592 unixctl_command_reply_error(conn
, "no such bridge");
4596 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4597 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4601 ds_put_cstr(&ds
, " port VLAN GROUP Age\n");
4602 ovs_rwlock_rdlock(&ofproto
->ms
->rwlock
);
4603 LIST_FOR_EACH (grp
, group_node
, &ofproto
->ms
->group_lru
) {
4604 LIST_FOR_EACH(b
, bundle_node
, &grp
->bundle_lru
) {
4605 char name
[OFP_MAX_PORT_NAME_LEN
];
4608 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4610 ds_put_format(&ds
, "%5s %4d ", name
, grp
->vlan
);
4611 ipv6_format_mapped(&grp
->addr
, &ds
);
4612 ds_put_format(&ds
, " %3d\n",
4613 mcast_bundle_age(ofproto
->ms
, b
));
4617 /* ports connected to multicast routers */
4618 LIST_FOR_EACH(mrouter
, mrouter_node
, &ofproto
->ms
->mrouter_lru
) {
4619 char name
[OFP_MAX_PORT_NAME_LEN
];
4621 bundle
= mrouter
->port
;
4622 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4624 ds_put_format(&ds
, "%5s %4d querier %3d\n",
4625 name
, mrouter
->vlan
,
4626 mcast_mrouter_age(ofproto
->ms
, mrouter
));
4628 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
4629 unixctl_command_reply(conn
, ds_cstr(&ds
));
4634 struct xlate_out xout
;
4635 struct xlate_in xin
;
4636 const struct flow
*key
;
4639 struct flow_wildcards wc
;
4640 struct ofpbuf odp_actions
;
4644 trace_format_rule(struct ds
*result
, int level
, const struct rule_dpif
*rule
)
4646 const struct rule_actions
*actions
;
4649 ds_put_char_multiple(result
, '\t', level
);
4651 ds_put_cstr(result
, "No match\n");
4655 ovs_mutex_lock(&rule
->up
.mutex
);
4656 cookie
= rule
->up
.flow_cookie
;
4657 ovs_mutex_unlock(&rule
->up
.mutex
);
4659 ds_put_format(result
, "Rule: table=%"PRIu8
" cookie=%#"PRIx64
" ",
4660 rule
? rule
->up
.table_id
: 0, ntohll(cookie
));
4661 cls_rule_format(&rule
->up
.cr
, result
);
4662 ds_put_char(result
, '\n');
4664 actions
= rule_dpif_get_actions(rule
);
4666 ds_put_char_multiple(result
, '\t', level
);
4667 ds_put_cstr(result
, "OpenFlow actions=");
4668 ofpacts_format(actions
->ofpacts
, actions
->ofpacts_len
, result
);
4669 ds_put_char(result
, '\n');
4673 trace_format_flow(struct ds
*result
, int level
, const char *title
,
4674 struct trace_ctx
*trace
)
4676 ds_put_char_multiple(result
, '\t', level
);
4677 ds_put_format(result
, "%s: ", title
);
4678 /* Do not report unchanged flows for resubmits. */
4679 if ((level
> 0 && flow_equal(&trace
->xin
.flow
, &trace
->flow
))
4680 || (level
== 0 && flow_equal(&trace
->xin
.flow
, trace
->key
))) {
4681 ds_put_cstr(result
, "unchanged");
4683 flow_format(result
, &trace
->xin
.flow
);
4684 trace
->flow
= trace
->xin
.flow
;
4686 ds_put_char(result
, '\n');
4690 trace_format_regs(struct ds
*result
, int level
, const char *title
,
4691 struct trace_ctx
*trace
)
4695 ds_put_char_multiple(result
, '\t', level
);
4696 ds_put_format(result
, "%s:", title
);
4697 for (i
= 0; i
< FLOW_N_REGS
; i
++) {
4698 ds_put_format(result
, " reg%"PRIuSIZE
"=0x%"PRIx32
, i
, trace
->flow
.regs
[i
]);
4700 ds_put_char(result
, '\n');
4704 trace_format_odp(struct ds
*result
, int level
, const char *title
,
4705 struct trace_ctx
*trace
)
4707 struct ofpbuf
*odp_actions
= &trace
->odp_actions
;
4709 ds_put_char_multiple(result
, '\t', level
);
4710 ds_put_format(result
, "%s: ", title
);
4711 format_odp_actions(result
, odp_actions
->data
, odp_actions
->size
);
4712 ds_put_char(result
, '\n');
4716 trace_format_megaflow(struct ds
*result
, int level
, const char *title
,
4717 struct trace_ctx
*trace
)
4721 ds_put_char_multiple(result
, '\t', level
);
4722 ds_put_format(result
, "%s: ", title
);
4723 match_init(&match
, trace
->key
, &trace
->wc
);
4724 match_format(&match
, result
, OFP_DEFAULT_PRIORITY
);
4725 ds_put_char(result
, '\n');
4728 static void trace_report(struct xlate_in
*, int indentation
,
4729 const char *format
, ...)
4730 OVS_PRINTF_FORMAT(3, 4);
4731 static void trace_report_valist(struct xlate_in
*, int indentation
,
4732 const char *format
, va_list args
)
4733 OVS_PRINTF_FORMAT(3, 0);
4736 trace_resubmit(struct xlate_in
*xin
, struct rule_dpif
*rule
, int indentation
)
4738 struct trace_ctx
*trace
= CONTAINER_OF(xin
, struct trace_ctx
, xin
);
4739 struct ds
*result
= trace
->result
;
4742 if (rule
== xin
->ofproto
->miss_rule
) {
4743 trace_report(xin
, indentation
,
4744 "No match, flow generates \"packet in\"s.");
4745 } else if (rule
== xin
->ofproto
->no_packet_in_rule
) {
4746 trace_report(xin
, indentation
, "No match, packets dropped because "
4747 "OFPPC_NO_PACKET_IN is set on in_port.");
4748 } else if (rule
== xin
->ofproto
->drop_frags_rule
) {
4749 trace_report(xin
, indentation
,
4750 "Packets dropped because they are IP fragments and "
4751 "the fragment handling mode is \"drop\".");
4755 ds_put_char(result
, '\n');
4757 trace_format_flow(result
, indentation
, "Resubmitted flow", trace
);
4758 trace_format_regs(result
, indentation
, "Resubmitted regs", trace
);
4759 trace_format_odp(result
, indentation
, "Resubmitted odp", trace
);
4760 trace_format_megaflow(result
, indentation
, "Resubmitted megaflow",
4763 trace_format_rule(result
, indentation
, rule
);
4767 trace_report_valist(struct xlate_in
*xin
, int indentation
,
4768 const char *format
, va_list args
)
4770 struct trace_ctx
*trace
= CONTAINER_OF(xin
, struct trace_ctx
, xin
);
4771 struct ds
*result
= trace
->result
;
4773 ds_put_char_multiple(result
, '\t', indentation
);
4774 ds_put_format_valist(result
, format
, args
);
4775 ds_put_char(result
, '\n');
4779 trace_report(struct xlate_in
*xin
, int indentation
, const char *format
, ...)
4783 va_start(args
, format
);
4784 trace_report_valist(xin
, indentation
, format
, args
);
4788 /* Parses the 'argc' elements of 'argv', ignoring argv[0]. The following
4789 * forms are supported:
4791 * - [dpname] odp_flow [-generate | packet]
4792 * - bridge br_flow [-generate | packet]
4794 * On success, initializes '*ofprotop' and 'flow' and returns NULL. On failure
4795 * returns a nonnull malloced error message. */
4796 static char * OVS_WARN_UNUSED_RESULT
4797 parse_flow_and_packet(int argc
, const char *argv
[],
4798 struct ofproto_dpif
**ofprotop
, struct flow
*flow
,
4799 struct dp_packet
**packetp
)
4801 const struct dpif_backer
*backer
= NULL
;
4802 const char *error
= NULL
;
4804 struct simap port_names
= SIMAP_INITIALIZER(&port_names
);
4805 struct dp_packet
*packet
;
4806 struct ofpbuf odp_key
;
4807 struct ofpbuf odp_mask
;
4809 ofpbuf_init(&odp_key
, 0);
4810 ofpbuf_init(&odp_mask
, 0);
4812 /* Handle "-generate" or a hex string as the last argument. */
4813 if (!strcmp(argv
[argc
- 1], "-generate")) {
4814 packet
= dp_packet_new(0);
4817 error
= eth_from_hex(argv
[argc
- 1], &packet
);
4820 } else if (argc
== 4) {
4821 /* The 3-argument form must end in "-generate' or a hex string. */
4827 /* odp_flow can have its in_port specified as a name instead of port no.
4828 * We do not yet know whether a given flow is a odp_flow or a br_flow.
4829 * But, to know whether a flow is odp_flow through odp_flow_from_string(),
4830 * we need to create a simap of name to port no. */
4832 const char *dp_type
;
4833 if (!strncmp(argv
[1], "ovs-", 4)) {
4834 dp_type
= argv
[1] + 4;
4838 backer
= shash_find_data(&all_dpif_backers
, dp_type
);
4839 } else if (argc
== 2) {
4840 struct shash_node
*node
;
4841 if (shash_count(&all_dpif_backers
) == 1) {
4842 node
= shash_first(&all_dpif_backers
);
4843 backer
= node
->data
;
4846 error
= "Syntax error";
4849 if (backer
&& backer
->dpif
) {
4850 struct dpif_port dpif_port
;
4851 struct dpif_port_dump port_dump
;
4852 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, backer
->dpif
) {
4853 simap_put(&port_names
, dpif_port
.name
,
4854 odp_to_u32(dpif_port
.port_no
));
4858 /* Parse the flow and determine whether a datapath or
4859 * bridge is specified. If function odp_flow_key_from_string()
4860 * returns 0, the flow is a odp_flow. If function
4861 * parse_ofp_exact_flow() returns NULL, the flow is a br_flow. */
4862 if (!odp_flow_from_string(argv
[argc
- 1], &port_names
,
4863 &odp_key
, &odp_mask
)) {
4865 error
= "Cannot find the datapath";
4869 if (odp_flow_key_to_flow(odp_key
.data
, odp_key
.size
, flow
) == ODP_FIT_ERROR
) {
4870 error
= "Failed to parse datapath flow key";
4874 *ofprotop
= xlate_lookup_ofproto(backer
, flow
,
4875 &flow
->in_port
.ofp_port
);
4876 if (*ofprotop
== NULL
) {
4877 error
= "Invalid datapath flow";
4881 char *err
= parse_ofp_exact_flow(flow
, NULL
, argv
[argc
- 1], NULL
);
4884 m_err
= xasprintf("Bad openflow flow syntax: %s", err
);
4889 error
= "Must specify bridge name";
4893 *ofprotop
= ofproto_dpif_lookup(argv
[1]);
4895 error
= "Unknown bridge name";
4901 /* Generate a packet, if requested. */
4903 if (!dp_packet_size(packet
)) {
4904 flow_compose(packet
, flow
);
4906 /* Use the metadata from the flow and the packet argument
4907 * to reconstruct the flow. */
4908 pkt_metadata_from_flow(&packet
->md
, flow
);
4909 flow_extract(packet
, flow
);
4914 if (error
&& !m_err
) {
4915 m_err
= xstrdup(error
);
4918 dp_packet_delete(packet
);
4922 ofpbuf_uninit(&odp_key
);
4923 ofpbuf_uninit(&odp_mask
);
4924 simap_destroy(&port_names
);
4929 ofproto_unixctl_trace(struct unixctl_conn
*conn
, int argc
, const char *argv
[],
4930 void *aux OVS_UNUSED
)
4932 struct ofproto_dpif
*ofproto
;
4933 struct dp_packet
*packet
;
4937 error
= parse_flow_and_packet(argc
, argv
, &ofproto
, &flow
, &packet
);
4942 ofproto_trace(ofproto
, &flow
, packet
, NULL
, 0, &result
);
4943 unixctl_command_reply(conn
, ds_cstr(&result
));
4944 ds_destroy(&result
);
4945 dp_packet_delete(packet
);
4947 unixctl_command_reply_error(conn
, error
);
4953 ofproto_unixctl_trace_actions(struct unixctl_conn
*conn
, int argc
,
4954 const char *argv
[], void *aux OVS_UNUSED
)
4956 enum ofputil_protocol usable_protocols
;
4957 struct ofproto_dpif
*ofproto
;
4958 bool enforce_consistency
;
4959 struct ofpbuf ofpacts
;
4960 struct dp_packet
*packet
;
4965 /* Three kinds of error return values! */
4971 ofpbuf_init(&ofpacts
, 0);
4973 /* Parse actions. */
4974 error
= ofpacts_parse_actions(argv
[--argc
], &ofpacts
, &usable_protocols
);
4976 unixctl_command_reply_error(conn
, error
);
4981 /* OpenFlow 1.1 and later suggest that the switch enforces certain forms of
4982 * consistency between the flow and the actions. With -consistent, we
4983 * enforce consistency even for a flow supported in OpenFlow 1.0. */
4984 if (!strcmp(argv
[1], "-consistent")) {
4985 enforce_consistency
= true;
4989 enforce_consistency
= false;
4992 error
= parse_flow_and_packet(argc
, argv
, &ofproto
, &flow
, &packet
);
4994 unixctl_command_reply_error(conn
, error
);
4999 /* Do the same checks as handle_packet_out() in ofproto.c.
5001 * We pass a 'table_id' of 0 to ofpacts_check(), which isn't
5002 * strictly correct because these actions aren't in any table, but it's OK
5003 * because it 'table_id' is used only to check goto_table instructions, but
5004 * packet-outs take a list of actions and therefore it can't include
5007 * We skip the "meter" check here because meter is an instruction, not an
5008 * action, and thus cannot appear in ofpacts. */
5009 in_port
= ofp_to_u16(flow
.in_port
.ofp_port
);
5010 if (in_port
>= ofproto
->up
.max_ports
&& in_port
< ofp_to_u16(OFPP_MAX
)) {
5011 unixctl_command_reply_error(conn
, "invalid in_port");
5014 if (enforce_consistency
) {
5015 retval
= ofpacts_check_consistency(ofpacts
.data
, ofpacts
.size
, &flow
,
5016 u16_to_ofp(ofproto
->up
.max_ports
),
5017 0, ofproto
->up
.n_tables
,
5020 retval
= ofpacts_check(ofpacts
.data
, ofpacts
.size
, &flow
,
5021 u16_to_ofp(ofproto
->up
.max_ports
), 0,
5022 ofproto
->up
.n_tables
, &usable_protocols
);
5025 retval
= ofproto_check_ofpacts(&ofproto
->up
, ofpacts
.data
,
5031 ds_put_format(&result
, "Bad actions: %s", ofperr_to_string(retval
));
5032 unixctl_command_reply_error(conn
, ds_cstr(&result
));
5036 ofproto_trace(ofproto
, &flow
, packet
,
5037 ofpacts
.data
, ofpacts
.size
, &result
);
5038 unixctl_command_reply(conn
, ds_cstr(&result
));
5041 ds_destroy(&result
);
5042 dp_packet_delete(packet
);
5043 ofpbuf_uninit(&ofpacts
);
5046 /* Implements a "trace" through 'ofproto''s flow table, appending a textual
5047 * description of the results to 'ds'.
5049 * The trace follows a packet with the specified 'flow' through the flow
5050 * table. 'packet' may be nonnull to trace an actual packet, with consequent
5051 * side effects (if it is nonnull then its flow must be 'flow').
5053 * If 'ofpacts' is nonnull then its 'ofpacts_len' bytes specify the actions to
5054 * trace, otherwise the actions are determined by a flow table lookup. */
5056 ofproto_trace(struct ofproto_dpif
*ofproto
, struct flow
*flow
,
5057 const struct dp_packet
*packet
,
5058 const struct ofpact ofpacts
[], size_t ofpacts_len
,
5061 struct trace_ctx trace
;
5062 enum xlate_error error
;
5064 ds_put_format(ds
, "Bridge: %s\n", ofproto
->up
.name
);
5065 ds_put_cstr(ds
, "Flow: ");
5066 flow_format(ds
, flow
);
5067 ds_put_char(ds
, '\n');
5069 ofpbuf_init(&trace
.odp_actions
, 0);
5072 trace
.key
= flow
; /* Original flow key, used for megaflow. */
5073 trace
.flow
= *flow
; /* May be modified by actions. */
5074 xlate_in_init(&trace
.xin
, ofproto
, flow
, flow
->in_port
.ofp_port
, NULL
,
5075 ntohs(flow
->tcp_flags
), packet
, &trace
.wc
,
5076 &trace
.odp_actions
);
5077 trace
.xin
.ofpacts
= ofpacts
;
5078 trace
.xin
.ofpacts_len
= ofpacts_len
;
5079 trace
.xin
.resubmit_hook
= trace_resubmit
;
5080 trace
.xin
.report_hook
= trace_report_valist
;
5082 error
= xlate_actions(&trace
.xin
, &trace
.xout
);
5083 ds_put_char(ds
, '\n');
5084 trace
.xin
.flow
.actset_output
= 0;
5085 trace_format_flow(ds
, 0, "Final flow", &trace
);
5086 trace_format_megaflow(ds
, 0, "Megaflow", &trace
);
5088 ds_put_cstr(ds
, "Datapath actions: ");
5089 format_odp_actions(ds
, trace
.odp_actions
.data
, trace
.odp_actions
.size
);
5091 if (error
!= XLATE_OK
) {
5092 ds_put_format(ds
, "\nTranslation failed (%s), packet is dropped.\n",
5093 xlate_strerror(error
));
5094 } else if (trace
.xout
.slow
) {
5095 enum slow_path_reason slow
;
5097 ds_put_cstr(ds
, "\nThis flow is handled by the userspace "
5098 "slow path because it:");
5100 slow
= trace
.xout
.slow
;
5102 enum slow_path_reason bit
= rightmost_1bit(slow
);
5104 ds_put_format(ds
, "\n\t- %s.",
5105 slow_path_reason_to_explanation(bit
));
5111 xlate_out_uninit(&trace
.xout
);
5112 ofpbuf_uninit(&trace
.odp_actions
);
5115 /* Store the current ofprotos in 'ofproto_shash'. Returns a sorted list
5116 * of the 'ofproto_shash' nodes. It is the responsibility of the caller
5117 * to destroy 'ofproto_shash' and free the returned value. */
5118 static const struct shash_node
**
5119 get_ofprotos(struct shash
*ofproto_shash
)
5121 const struct ofproto_dpif
*ofproto
;
5123 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
5124 char *name
= xasprintf("%s@%s", ofproto
->up
.type
, ofproto
->up
.name
);
5125 shash_add_nocopy(ofproto_shash
, name
, ofproto
);
5128 return shash_sort(ofproto_shash
);
5132 ofproto_unixctl_dpif_dump_dps(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
5133 const char *argv
[] OVS_UNUSED
,
5134 void *aux OVS_UNUSED
)
5136 struct ds ds
= DS_EMPTY_INITIALIZER
;
5137 struct shash ofproto_shash
;
5138 const struct shash_node
**sorted_ofprotos
;
5141 shash_init(&ofproto_shash
);
5142 sorted_ofprotos
= get_ofprotos(&ofproto_shash
);
5143 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
5144 const struct shash_node
*node
= sorted_ofprotos
[i
];
5145 ds_put_format(&ds
, "%s\n", node
->name
);
5148 shash_destroy(&ofproto_shash
);
5149 free(sorted_ofprotos
);
5151 unixctl_command_reply(conn
, ds_cstr(&ds
));
5156 dpif_show_backer(const struct dpif_backer
*backer
, struct ds
*ds
)
5158 const struct shash_node
**ofprotos
;
5159 struct dpif_dp_stats dp_stats
;
5160 struct shash ofproto_shash
;
5163 dpif_get_dp_stats(backer
->dpif
, &dp_stats
);
5165 ds_put_format(ds
, "%s: hit:%"PRIu64
" missed:%"PRIu64
"\n",
5166 dpif_name(backer
->dpif
), dp_stats
.n_hit
, dp_stats
.n_missed
);
5168 shash_init(&ofproto_shash
);
5169 ofprotos
= get_ofprotos(&ofproto_shash
);
5170 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
5171 struct ofproto_dpif
*ofproto
= ofprotos
[i
]->data
;
5172 const struct shash_node
**ports
;
5175 if (ofproto
->backer
!= backer
) {
5179 ds_put_format(ds
, "\t%s:\n", ofproto
->up
.name
);
5181 ports
= shash_sort(&ofproto
->up
.port_by_name
);
5182 for (j
= 0; j
< shash_count(&ofproto
->up
.port_by_name
); j
++) {
5183 const struct shash_node
*node
= ports
[j
];
5184 struct ofport
*ofport
= node
->data
;
5186 odp_port_t odp_port
;
5188 ds_put_format(ds
, "\t\t%s %u/", netdev_get_name(ofport
->netdev
),
5191 odp_port
= ofp_port_to_odp_port(ofproto
, ofport
->ofp_port
);
5192 if (odp_port
!= ODPP_NONE
) {
5193 ds_put_format(ds
, "%"PRIu32
":", odp_port
);
5195 ds_put_cstr(ds
, "none:");
5198 ds_put_format(ds
, " (%s", netdev_get_type(ofport
->netdev
));
5201 if (!netdev_get_config(ofport
->netdev
, &config
)) {
5202 const struct smap_node
**nodes
;
5205 nodes
= smap_sort(&config
);
5206 for (i
= 0; i
< smap_count(&config
); i
++) {
5207 const struct smap_node
*node
= nodes
[i
];
5208 ds_put_format(ds
, "%c %s=%s", i
? ',' : ':',
5209 node
->key
, node
->value
);
5213 smap_destroy(&config
);
5215 ds_put_char(ds
, ')');
5216 ds_put_char(ds
, '\n');
5220 shash_destroy(&ofproto_shash
);
5225 ofproto_unixctl_dpif_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
5226 const char *argv
[] OVS_UNUSED
, void *aux OVS_UNUSED
)
5228 struct ds ds
= DS_EMPTY_INITIALIZER
;
5229 const struct shash_node
**backers
;
5232 backers
= shash_sort(&all_dpif_backers
);
5233 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
5234 dpif_show_backer(backers
[i
]->data
, &ds
);
5238 unixctl_command_reply(conn
, ds_cstr(&ds
));
5243 ofproto_unixctl_dpif_dump_flows(struct unixctl_conn
*conn
,
5244 int argc OVS_UNUSED
, const char *argv
[],
5245 void *aux OVS_UNUSED
)
5247 const struct ofproto_dpif
*ofproto
;
5249 struct ds ds
= DS_EMPTY_INITIALIZER
;
5250 bool verbosity
= false;
5252 struct dpif_port dpif_port
;
5253 struct dpif_port_dump port_dump
;
5254 struct hmap portno_names
;
5256 struct dpif_flow_dump
*flow_dump
;
5257 struct dpif_flow_dump_thread
*flow_dump_thread
;
5261 ofproto
= ofproto_dpif_lookup(argv
[argc
- 1]);
5263 unixctl_command_reply_error(conn
, "no such bridge");
5267 if (argc
> 2 && !strcmp(argv
[1], "-m")) {
5271 hmap_init(&portno_names
);
5272 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, ofproto
->backer
->dpif
) {
5273 odp_portno_names_set(&portno_names
, dpif_port
.port_no
, dpif_port
.name
);
5277 flow_dump
= dpif_flow_dump_create(ofproto
->backer
->dpif
, false);
5278 flow_dump_thread
= dpif_flow_dump_thread_create(flow_dump
);
5279 while (dpif_flow_dump_next(flow_dump_thread
, &f
, 1)) {
5282 if (odp_flow_key_to_flow(f
.key
, f
.key_len
, &flow
) == ODP_FIT_ERROR
5283 || xlate_lookup_ofproto(ofproto
->backer
, &flow
, NULL
) != ofproto
) {
5288 odp_format_ufid(&f
.ufid
, &ds
);
5289 ds_put_cstr(&ds
, " ");
5291 odp_flow_format(f
.key
, f
.key_len
, f
.mask
, f
.mask_len
,
5292 &portno_names
, &ds
, verbosity
);
5293 ds_put_cstr(&ds
, ", ");
5294 dpif_flow_stats_format(&f
.stats
, &ds
);
5295 ds_put_cstr(&ds
, ", actions:");
5296 format_odp_actions(&ds
, f
.actions
, f
.actions_len
);
5297 ds_put_char(&ds
, '\n');
5299 dpif_flow_dump_thread_destroy(flow_dump_thread
);
5300 error
= dpif_flow_dump_destroy(flow_dump
);
5304 ds_put_format(&ds
, "dpif/dump_flows failed: %s", ovs_strerror(errno
));
5305 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
5307 unixctl_command_reply(conn
, ds_cstr(&ds
));
5309 odp_portno_names_destroy(&portno_names
);
5310 hmap_destroy(&portno_names
);
5315 ofproto_revalidate_all_backers(void)
5317 const struct shash_node
**backers
;
5320 backers
= shash_sort(&all_dpif_backers
);
5321 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
5322 struct dpif_backer
*backer
= backers
[i
]->data
;
5323 backer
->need_revalidate
= REV_RECONFIGURE
;
5329 disable_tnl_push_pop(struct unixctl_conn
*conn OVS_UNUSED
, int argc OVS_UNUSED
,
5330 const char *argv
[], void *aux OVS_UNUSED
)
5332 if (!strcasecmp(argv
[1], "off")) {
5333 ofproto_use_tnl_push_pop
= false;
5334 unixctl_command_reply(conn
, "Tunnel push-pop off");
5335 ofproto_revalidate_all_backers();
5336 } else if (!strcasecmp(argv
[1], "on")) {
5337 ofproto_use_tnl_push_pop
= true;
5338 unixctl_command_reply(conn
, "Tunnel push-pop on");
5339 ofproto_revalidate_all_backers();
5341 unixctl_command_reply_error(conn
, "Invalid argument");
5346 ofproto_unixctl_init(void)
5348 static bool registered
;
5354 unixctl_command_register(
5356 "{[dp_name] odp_flow | bridge br_flow} [-generate|packet]",
5357 1, 3, ofproto_unixctl_trace
, NULL
);
5358 unixctl_command_register(
5359 "ofproto/trace-packet-out",
5360 "[-consistent] {[dp_name] odp_flow | bridge br_flow} [-generate|packet] actions",
5361 2, 6, ofproto_unixctl_trace_actions
, NULL
);
5362 unixctl_command_register("fdb/flush", "[bridge]", 0, 1,
5363 ofproto_unixctl_fdb_flush
, NULL
);
5364 unixctl_command_register("fdb/show", "bridge", 1, 1,
5365 ofproto_unixctl_fdb_show
, NULL
);
5366 unixctl_command_register("mdb/flush", "[bridge]", 0, 1,
5367 ofproto_unixctl_mcast_snooping_flush
, NULL
);
5368 unixctl_command_register("mdb/show", "bridge", 1, 1,
5369 ofproto_unixctl_mcast_snooping_show
, NULL
);
5370 unixctl_command_register("dpif/dump-dps", "", 0, 0,
5371 ofproto_unixctl_dpif_dump_dps
, NULL
);
5372 unixctl_command_register("dpif/show", "", 0, 0, ofproto_unixctl_dpif_show
,
5374 unixctl_command_register("dpif/dump-flows", "[-m] bridge", 1, 2,
5375 ofproto_unixctl_dpif_dump_flows
, NULL
);
5377 unixctl_command_register("ofproto/tnl-push-pop", "[on]|[off]", 1, 1,
5378 disable_tnl_push_pop
, NULL
);
5381 /* Returns true if 'table' is the table used for internal rules,
5382 * false otherwise. */
5384 table_is_internal(uint8_t table_id
)
5386 return table_id
== TBL_INTERNAL
;
5391 ofp_port_to_odp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
5393 const struct ofport_dpif
*ofport
= ofp_port_to_ofport(ofproto
, ofp_port
);
5394 return ofport
? ofport
->odp_port
: ODPP_NONE
;
5397 struct ofport_dpif
*
5398 odp_port_to_ofport(const struct dpif_backer
*backer
, odp_port_t odp_port
)
5400 struct ofport_dpif
*port
;
5402 ovs_rwlock_rdlock(&backer
->odp_to_ofport_lock
);
5403 HMAP_FOR_EACH_IN_BUCKET (port
, odp_port_node
, hash_odp_port(odp_port
),
5404 &backer
->odp_to_ofport_map
) {
5405 if (port
->odp_port
== odp_port
) {
5406 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5411 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5416 odp_port_to_ofp_port(const struct ofproto_dpif
*ofproto
, odp_port_t odp_port
)
5418 struct ofport_dpif
*port
;
5420 port
= odp_port_to_ofport(ofproto
->backer
, odp_port
);
5421 if (port
&& &ofproto
->up
== port
->up
.ofproto
) {
5422 return port
->up
.ofp_port
;
5429 ofproto_dpif_add_internal_flow(struct ofproto_dpif
*ofproto
,
5430 const struct match
*match
, int priority
,
5431 uint16_t idle_timeout
,
5432 const struct ofpbuf
*ofpacts
,
5433 struct rule
**rulep
)
5435 struct ofproto_flow_mod ofm
;
5436 struct rule_dpif
*rule
;
5439 ofm
.fm
= (struct ofputil_flow_mod
) {
5441 .priority
= priority
,
5442 .table_id
= TBL_INTERNAL
,
5443 .command
= OFPFC_ADD
,
5444 .idle_timeout
= idle_timeout
,
5445 .flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
,
5446 .ofpacts
= ofpacts
->data
,
5447 .ofpacts_len
= ofpacts
->size
,
5448 .delete_reason
= OVS_OFPRR_NONE
,
5451 error
= ofproto_flow_mod(&ofproto
->up
, &ofm
);
5453 VLOG_ERR_RL(&rl
, "failed to add internal flow (%s)",
5454 ofperr_to_string(error
));
5459 rule
= rule_dpif_lookup_in_table(ofproto
,
5460 ofproto_dpif_get_tables_version(ofproto
),
5461 TBL_INTERNAL
, &ofm
.fm
.match
.flow
,
5472 ofproto_dpif_delete_internal_flow(struct ofproto_dpif
*ofproto
,
5473 struct match
*match
, int priority
)
5475 struct ofproto_flow_mod ofm
;
5478 ofm
.fm
= (struct ofputil_flow_mod
) {
5480 .priority
= priority
,
5481 .table_id
= TBL_INTERNAL
,
5482 .flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
,
5483 .command
= OFPFC_DELETE_STRICT
,
5486 error
= ofproto_flow_mod(&ofproto
->up
, &ofm
);
5488 VLOG_ERR_RL(&rl
, "failed to delete internal flow (%s)",
5489 ofperr_to_string(error
));
5497 ofproto_dpif_get_uuid(const struct ofproto_dpif
*ofproto
)
5499 return &ofproto
->uuid
;
5502 const struct ofproto_class ofproto_dpif_class
= {
5516 NULL
, /* get_memory_usage. */
5517 type_get_memory_usage
,
5536 port_is_lacp_current
,
5537 port_get_lacp_stats
,
5538 NULL
, /* rule_choose_table */
5542 NULL
, /* rule_delete */
5564 aa_vlan_get_queue_size
,
5571 get_stp_port_status
,
5576 get_rstp_port_status
,
5583 is_mirror_output_bundle
,
5584 forward_bpdu_changed
,
5585 set_mac_table_config
,
5587 set_mcast_snooping_port
,
5588 NULL
, /* meter_get_features */
5589 NULL
, /* meter_set */
5590 NULL
, /* meter_get */
5591 NULL
, /* meter_del */
5592 group_alloc
, /* group_alloc */
5593 group_construct
, /* group_construct */
5594 group_destruct
, /* group_destruct */
5595 group_dealloc
, /* group_dealloc */
5596 group_modify
, /* group_modify */
5597 group_get_stats
, /* group_get_stats */
5598 get_datapath_version
, /* get_datapath_version */