2 * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014 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.
19 #include "ofproto/ofproto-dpif.h"
20 #include "ofproto/ofproto-provider.h"
27 #include "byte-order.h"
28 #include "connectivity.h"
33 #include "dynamic-string.h"
34 #include "fail-open.h"
35 #include "guarded-list.h"
39 #include "mac-learning.h"
40 #include "mcast-snooping.h"
41 #include "meta-flow.h"
42 #include "multipath.h"
43 #include "netdev-vport.h"
48 #include "odp-execute.h"
51 #include "ofp-actions.h"
52 #include "ofp-parse.h"
53 #include "ofp-print.h"
54 #include "ofproto-dpif-ipfix.h"
55 #include "ofproto-dpif-mirror.h"
56 #include "ofproto-dpif-monitor.h"
57 #include "ofproto-dpif-rid.h"
58 #include "ofproto-dpif-sflow.h"
59 #include "ofproto-dpif-upcall.h"
60 #include "ofproto-dpif-xlate.h"
61 #include "poll-loop.h"
67 #include "unaligned.h"
69 #include "vlan-bitmap.h"
72 VLOG_DEFINE_THIS_MODULE(ofproto_dpif
);
74 COVERAGE_DEFINE(ofproto_dpif_expired
);
75 COVERAGE_DEFINE(packet_in_overflow
);
84 * - Do include packets and bytes from datapath flows which have not
85 * recently been processed by a revalidator. */
86 struct ovs_mutex stats_mutex
;
87 struct dpif_flow_stats stats OVS_GUARDED
;
89 /* If non-zero then the recirculation id that has
90 * been allocated for use with this rule.
91 * The recirculation id and associated internal flow should
92 * be freed when the rule is freed */
96 /* RULE_CAST() depends on this. */
97 BUILD_ASSERT_DECL(offsetof(struct rule_dpif
, up
) == 0);
99 static void rule_get_stats(struct rule
*, uint64_t *packets
, uint64_t *bytes
,
100 long long int *used
);
101 static struct rule_dpif
*rule_dpif_cast(const struct rule
*);
102 static void rule_expire(struct rule_dpif
*);
109 * - Do include packets and bytes from datapath flows which have not
110 * recently been processed by a revalidator. */
111 struct ovs_mutex stats_mutex
;
112 uint64_t packet_count OVS_GUARDED
; /* Number of packets received. */
113 uint64_t byte_count OVS_GUARDED
; /* Number of bytes received. */
117 struct hmap_node hmap_node
; /* In struct ofproto's "bundles" hmap. */
118 struct ofproto_dpif
*ofproto
; /* Owning ofproto. */
119 void *aux
; /* Key supplied by ofproto's client. */
120 char *name
; /* Identifier for log messages. */
123 struct list ports
; /* Contains "struct ofport"s. */
124 enum port_vlan_mode vlan_mode
; /* VLAN mode */
125 int vlan
; /* -1=trunk port, else a 12-bit VLAN ID. */
126 unsigned long *trunks
; /* Bitmap of trunked VLANs, if 'vlan' == -1.
127 * NULL if all VLANs are trunked. */
128 struct lacp
*lacp
; /* LACP if LACP is enabled, otherwise NULL. */
129 struct bond
*bond
; /* Nonnull iff more than one port. */
130 bool use_priority_tags
; /* Use 802.1p tag for frames in VLAN 0? */
133 bool floodable
; /* True if no port has OFPUTIL_PC_NO_FLOOD set. */
136 static void bundle_remove(struct ofport
*);
137 static void bundle_update(struct ofbundle
*);
138 static void bundle_destroy(struct ofbundle
*);
139 static void bundle_del_port(struct ofport_dpif
*);
140 static void bundle_run(struct ofbundle
*);
141 static void bundle_wait(struct ofbundle
*);
143 static void stp_run(struct ofproto_dpif
*ofproto
);
144 static void stp_wait(struct ofproto_dpif
*ofproto
);
145 static int set_stp_port(struct ofport
*,
146 const struct ofproto_port_stp_settings
*);
148 static void rstp_run(struct ofproto_dpif
*ofproto
);
149 static void set_rstp_port(struct ofport
*,
150 const struct ofproto_port_rstp_settings
*);
153 struct hmap_node odp_port_node
; /* In dpif_backer's "odp_to_ofport_map". */
157 struct ofbundle
*bundle
; /* Bundle that contains this port, if any. */
158 struct list bundle_node
; /* In struct ofbundle's "ports" list. */
159 struct cfm
*cfm
; /* Connectivity Fault Management, if any. */
160 struct bfd
*bfd
; /* BFD, if any. */
161 bool may_enable
; /* May be enabled in bonds. */
162 bool is_tunnel
; /* This port is a tunnel. */
163 bool is_layer3
; /* This is a layer 3 port. */
164 long long int carrier_seq
; /* Carrier status changes. */
165 struct ofport_dpif
*peer
; /* Peer if patch port. */
168 struct stp_port
*stp_port
; /* Spanning Tree Protocol, if any. */
169 enum stp_state stp_state
; /* Always STP_DISABLED if STP not in use. */
170 long long int stp_state_entered
;
172 /* Rapid Spanning Tree. */
173 struct rstp_port
*rstp_port
; /* Rapid Spanning Tree Protocol, if any. */
174 enum rstp_state rstp_state
; /* Always RSTP_DISABLED if RSTP not in use. */
176 /* Queue to DSCP mapping. */
177 struct ofproto_port_queue
*qdscp
;
180 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
182 * This is deprecated. It is only for compatibility with broken device
183 * drivers in old versions of Linux that do not properly support VLANs when
184 * VLAN devices are not used. When broken device drivers are no longer in
185 * widespread use, we will delete these interfaces. */
186 ofp_port_t realdev_ofp_port
;
190 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
192 * This is deprecated. It is only for compatibility with broken device drivers
193 * in old versions of Linux that do not properly support VLANs when VLAN
194 * devices are not used. When broken device drivers are no longer in
195 * widespread use, we will delete these interfaces. */
196 struct vlan_splinter
{
197 struct hmap_node realdev_vid_node
;
198 struct hmap_node vlandev_node
;
199 ofp_port_t realdev_ofp_port
;
200 ofp_port_t vlandev_ofp_port
;
204 static void vsp_remove(struct ofport_dpif
*);
205 static void vsp_add(struct ofport_dpif
*, ofp_port_t realdev_ofp_port
, int vid
);
207 static odp_port_t
ofp_port_to_odp_port(const struct ofproto_dpif
*,
210 static ofp_port_t
odp_port_to_ofp_port(const struct ofproto_dpif
*,
213 static struct ofport_dpif
*
214 ofport_dpif_cast(const struct ofport
*ofport
)
216 return ofport
? CONTAINER_OF(ofport
, struct ofport_dpif
, up
) : NULL
;
219 static void port_run(struct ofport_dpif
*);
220 static int set_bfd(struct ofport
*, const struct smap
*);
221 static int set_cfm(struct ofport
*, const struct cfm_settings
*);
222 static void ofport_update_peer(struct ofport_dpif
*);
224 /* Reasons that we might need to revalidate every datapath flow, and
225 * corresponding coverage counters.
227 * A value of 0 means that there is no need to revalidate.
229 * It would be nice to have some cleaner way to integrate with coverage
230 * counters, but with only a few reasons I guess this is good enough for
232 enum revalidate_reason
{
233 REV_RECONFIGURE
= 1, /* Switch configuration changed. */
234 REV_STP
, /* Spanning tree protocol port status change. */
235 REV_RSTP
, /* RSTP port status change. */
236 REV_BOND
, /* Bonding changed. */
237 REV_PORT_TOGGLED
, /* Port enabled or disabled by CFM, LACP, ...*/
238 REV_FLOW_TABLE
, /* Flow table changed. */
239 REV_MAC_LEARNING
, /* Mac learning changed. */
240 REV_MCAST_SNOOPING
, /* Multicast snooping changed. */
242 COVERAGE_DEFINE(rev_reconfigure
);
243 COVERAGE_DEFINE(rev_stp
);
244 COVERAGE_DEFINE(rev_rstp
);
245 COVERAGE_DEFINE(rev_bond
);
246 COVERAGE_DEFINE(rev_port_toggled
);
247 COVERAGE_DEFINE(rev_flow_table
);
248 COVERAGE_DEFINE(rev_mac_learning
);
249 COVERAGE_DEFINE(rev_mcast_snooping
);
251 /* All datapaths of a given type share a single dpif backer instance. */
258 struct ovs_rwlock odp_to_ofport_lock
;
259 struct hmap odp_to_ofport_map OVS_GUARDED
; /* Contains "struct ofport"s. */
261 struct simap tnl_backers
; /* Set of dpif ports backing tunnels. */
263 enum revalidate_reason need_revalidate
; /* Revalidate all flows. */
265 bool recv_set_enable
; /* Enables or disables receiving packets. */
268 struct recirc_id_pool
*rid_pool
; /* Recirculation ID pool. */
269 bool enable_recirc
; /* True if the datapath supports recirculation */
271 /* True if the datapath supports variable-length
272 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions.
273 * False if the datapath supports only 8-byte (or shorter) userdata. */
274 bool variable_length_userdata
;
276 /* Maximum number of MPLS label stack entries that the datapath supports
278 size_t max_mpls_depth
;
281 /* All existing ofproto_backer instances, indexed by ofproto->up.type. */
282 static struct shash all_dpif_backers
= SHASH_INITIALIZER(&all_dpif_backers
);
284 struct ofproto_dpif
{
285 struct hmap_node all_ofproto_dpifs_node
; /* In 'all_ofproto_dpifs'. */
287 struct dpif_backer
*backer
;
289 uint64_t dump_seq
; /* Last read of udpif_dump_seq(). */
291 /* Special OpenFlow rules. */
292 struct rule_dpif
*miss_rule
; /* Sends flow table misses to controller. */
293 struct rule_dpif
*no_packet_in_rule
; /* Drops flow table misses. */
294 struct rule_dpif
*drop_frags_rule
; /* Used in OFPC_FRAG_DROP mode. */
297 struct netflow
*netflow
;
298 struct dpif_sflow
*sflow
;
299 struct dpif_ipfix
*ipfix
;
300 struct hmap bundles
; /* Contains "struct ofbundle"s. */
301 struct mac_learning
*ml
;
302 struct mcast_snooping
*ms
;
303 bool has_bonded_bundles
;
305 struct mbridge
*mbridge
;
307 struct ovs_mutex stats_mutex
;
308 struct netdev_stats stats OVS_GUARDED
; /* To account packets generated and
309 * consumed in userspace. */
313 long long int stp_last_tick
;
315 /* Rapid Spanning Tree. */
317 long long int rstp_last_tick
;
319 /* VLAN splinters. */
320 struct ovs_mutex vsp_mutex
;
321 struct hmap realdev_vid_map OVS_GUARDED
; /* (realdev,vid) -> vlandev. */
322 struct hmap vlandev_map OVS_GUARDED
; /* vlandev -> (realdev,vid). */
325 struct sset ports
; /* Set of standard port names. */
326 struct sset ghost_ports
; /* Ports with no datapath port. */
327 struct sset port_poll_set
; /* Queued names for port_poll() reply. */
328 int port_poll_errno
; /* Last errno for port_poll() reply. */
329 uint64_t change_seq
; /* Connectivity status changes. */
332 struct guarded_list pins
; /* Contains "struct ofputil_packet_in"s. */
333 struct seq
*pins_seq
; /* For notifying 'pins' reception. */
337 /* All existing ofproto_dpif instances, indexed by ->up.name. */
338 static struct hmap all_ofproto_dpifs
= HMAP_INITIALIZER(&all_ofproto_dpifs
);
340 static void ofproto_dpif_unixctl_init(void);
342 static inline struct ofproto_dpif
*
343 ofproto_dpif_cast(const struct ofproto
*ofproto
)
345 ovs_assert(ofproto
->ofproto_class
== &ofproto_dpif_class
);
346 return CONTAINER_OF(ofproto
, struct ofproto_dpif
, up
);
350 ofproto_dpif_get_max_mpls_depth(const struct ofproto_dpif
*ofproto
)
352 return ofproto
->backer
->max_mpls_depth
;
356 ofproto_dpif_get_enable_recirc(const struct ofproto_dpif
*ofproto
)
358 return ofproto
->backer
->enable_recirc
;
361 static struct ofport_dpif
*get_ofp_port(const struct ofproto_dpif
*ofproto
,
362 ofp_port_t ofp_port
);
363 static void ofproto_trace(struct ofproto_dpif
*, struct flow
*,
364 const struct ofpbuf
*packet
,
365 const struct ofpact
[], size_t ofpacts_len
,
368 /* Global variables. */
369 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
371 /* Initial mappings of port to bridge mappings. */
372 static struct shash init_ofp_ports
= SHASH_INITIALIZER(&init_ofp_ports
);
374 /* Executes 'fm'. The caller retains ownership of 'fm' and everything in
377 ofproto_dpif_flow_mod(struct ofproto_dpif
*ofproto
,
378 struct ofputil_flow_mod
*fm
)
380 ofproto_flow_mod(&ofproto
->up
, fm
);
383 /* Appends 'pin' to the queue of "packet ins" to be sent to the controller.
384 * Takes ownership of 'pin' and pin->packet. */
386 ofproto_dpif_send_packet_in(struct ofproto_dpif
*ofproto
,
387 struct ofproto_packet_in
*pin
)
389 if (!guarded_list_push_back(&ofproto
->pins
, &pin
->list_node
, 1024)) {
390 COVERAGE_INC(packet_in_overflow
);
391 free(CONST_CAST(void *, pin
->up
.packet
));
395 /* Wakes up main thread for packet-in I/O. */
396 seq_change(ofproto
->pins_seq
);
399 /* The default "table-miss" behaviour for OpenFlow1.3+ is to drop the
400 * packet rather than to send the packet to the controller.
402 * This function returns false to indicate that a packet_in message
403 * for a "table-miss" should be sent to at least one controller.
404 * False otherwise. */
406 ofproto_dpif_wants_packet_in_on_miss(struct ofproto_dpif
*ofproto
)
408 return connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
);
411 /* Factory functions. */
414 init(const struct shash
*iface_hints
)
416 struct shash_node
*node
;
418 /* Make a local copy, since we don't own 'iface_hints' elements. */
419 SHASH_FOR_EACH(node
, iface_hints
) {
420 const struct iface_hint
*orig_hint
= node
->data
;
421 struct iface_hint
*new_hint
= xmalloc(sizeof *new_hint
);
423 new_hint
->br_name
= xstrdup(orig_hint
->br_name
);
424 new_hint
->br_type
= xstrdup(orig_hint
->br_type
);
425 new_hint
->ofp_port
= orig_hint
->ofp_port
;
427 shash_add(&init_ofp_ports
, node
->name
, new_hint
);
432 enumerate_types(struct sset
*types
)
434 dp_enumerate_types(types
);
438 enumerate_names(const char *type
, struct sset
*names
)
440 struct ofproto_dpif
*ofproto
;
443 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
444 if (strcmp(type
, ofproto
->up
.type
)) {
447 sset_add(names
, ofproto
->up
.name
);
454 del(const char *type
, const char *name
)
459 error
= dpif_open(name
, type
, &dpif
);
461 error
= dpif_delete(dpif
);
468 port_open_type(const char *datapath_type
, const char *port_type
)
470 return dpif_port_open_type(datapath_type
, port_type
);
473 /* Type functions. */
475 static void process_dpif_port_changes(struct dpif_backer
*);
476 static void process_dpif_all_ports_changed(struct dpif_backer
*);
477 static void process_dpif_port_change(struct dpif_backer
*,
478 const char *devname
);
479 static void process_dpif_port_error(struct dpif_backer
*, int error
);
481 static struct ofproto_dpif
*
482 lookup_ofproto_dpif_by_port_name(const char *name
)
484 struct ofproto_dpif
*ofproto
;
486 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
487 if (sset_contains(&ofproto
->ports
, name
)) {
496 type_run(const char *type
)
498 struct dpif_backer
*backer
;
500 backer
= shash_find_data(&all_dpif_backers
, type
);
502 /* This is not necessarily a problem, since backers are only
503 * created on demand. */
507 dpif_run(backer
->dpif
);
508 udpif_run(backer
->udpif
);
510 /* If vswitchd started with other_config:flow_restore_wait set as "true",
511 * and the configuration has now changed to "false", enable receiving
512 * packets from the datapath. */
513 if (!backer
->recv_set_enable
&& !ofproto_get_flow_restore_wait()) {
516 backer
->recv_set_enable
= true;
518 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
520 VLOG_ERR("Failed to enable receiving packets in dpif.");
523 dpif_flow_flush(backer
->dpif
);
524 backer
->need_revalidate
= REV_RECONFIGURE
;
527 if (backer
->recv_set_enable
) {
528 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
531 if (backer
->need_revalidate
) {
532 struct ofproto_dpif
*ofproto
;
533 struct simap_node
*node
;
534 struct simap tmp_backers
;
536 /* Handle tunnel garbage collection. */
537 simap_init(&tmp_backers
);
538 simap_swap(&backer
->tnl_backers
, &tmp_backers
);
540 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
541 struct ofport_dpif
*iter
;
543 if (backer
!= ofproto
->backer
) {
547 HMAP_FOR_EACH (iter
, up
.hmap_node
, &ofproto
->up
.ports
) {
548 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
551 if (!iter
->is_tunnel
) {
555 dp_port
= netdev_vport_get_dpif_port(iter
->up
.netdev
,
556 namebuf
, sizeof namebuf
);
557 node
= simap_find(&tmp_backers
, dp_port
);
559 simap_put(&backer
->tnl_backers
, dp_port
, node
->data
);
560 simap_delete(&tmp_backers
, node
);
561 node
= simap_find(&backer
->tnl_backers
, dp_port
);
563 node
= simap_find(&backer
->tnl_backers
, dp_port
);
565 odp_port_t odp_port
= ODPP_NONE
;
567 if (!dpif_port_add(backer
->dpif
, iter
->up
.netdev
,
569 simap_put(&backer
->tnl_backers
, dp_port
,
570 odp_to_u32(odp_port
));
571 node
= simap_find(&backer
->tnl_backers
, dp_port
);
576 iter
->odp_port
= node
? u32_to_odp(node
->data
) : ODPP_NONE
;
577 if (tnl_port_reconfigure(iter
, iter
->up
.netdev
,
579 backer
->need_revalidate
= REV_RECONFIGURE
;
584 SIMAP_FOR_EACH (node
, &tmp_backers
) {
585 dpif_port_del(backer
->dpif
, u32_to_odp(node
->data
));
587 simap_destroy(&tmp_backers
);
589 switch (backer
->need_revalidate
) {
590 case REV_RECONFIGURE
: COVERAGE_INC(rev_reconfigure
); break;
591 case REV_STP
: COVERAGE_INC(rev_stp
); break;
592 case REV_RSTP
: COVERAGE_INC(rev_rstp
); break;
593 case REV_BOND
: COVERAGE_INC(rev_bond
); break;
594 case REV_PORT_TOGGLED
: COVERAGE_INC(rev_port_toggled
); break;
595 case REV_FLOW_TABLE
: COVERAGE_INC(rev_flow_table
); break;
596 case REV_MAC_LEARNING
: COVERAGE_INC(rev_mac_learning
); break;
597 case REV_MCAST_SNOOPING
: COVERAGE_INC(rev_mcast_snooping
); break;
599 backer
->need_revalidate
= 0;
601 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
602 struct ofport_dpif
*ofport
;
603 struct ofbundle
*bundle
;
605 if (ofproto
->backer
!= backer
) {
610 xlate_ofproto_set(ofproto
, ofproto
->up
.name
,
611 ofproto
->backer
->dpif
, ofproto
->miss_rule
,
612 ofproto
->no_packet_in_rule
, ofproto
->ml
,
613 ofproto
->stp
, ofproto
->rstp
, ofproto
->ms
,
614 ofproto
->mbridge
, ofproto
->sflow
, ofproto
->ipfix
,
615 ofproto
->netflow
, ofproto
->up
.frag_handling
,
616 ofproto
->up
.forward_bpdu
,
617 connmgr_has_in_band(ofproto
->up
.connmgr
),
618 ofproto
->backer
->enable_recirc
,
619 ofproto
->backer
->variable_length_userdata
,
620 ofproto
->backer
->max_mpls_depth
);
622 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
623 xlate_bundle_set(ofproto
, bundle
, bundle
->name
,
624 bundle
->vlan_mode
, bundle
->vlan
,
625 bundle
->trunks
, bundle
->use_priority_tags
,
626 bundle
->bond
, bundle
->lacp
,
630 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
631 int stp_port
= ofport
->stp_port
632 ? stp_port_no(ofport
->stp_port
)
634 int rstp_port
= ofport
->rstp_port
635 ? rstp_port_number(ofport
->rstp_port
)
637 xlate_ofport_set(ofproto
, ofport
->bundle
, ofport
,
638 ofport
->up
.ofp_port
, ofport
->odp_port
,
639 ofport
->up
.netdev
, ofport
->cfm
,
640 ofport
->bfd
, ofport
->peer
, stp_port
,
641 rstp_port
, ofport
->qdscp
, ofport
->n_qdscp
,
642 ofport
->up
.pp
.config
, ofport
->up
.pp
.state
,
643 ofport
->is_tunnel
, ofport
->may_enable
);
648 udpif_revalidate(backer
->udpif
);
651 process_dpif_port_changes(backer
);
656 /* Check for and handle port changes in 'backer''s dpif. */
658 process_dpif_port_changes(struct dpif_backer
*backer
)
664 error
= dpif_port_poll(backer
->dpif
, &devname
);
670 process_dpif_all_ports_changed(backer
);
674 process_dpif_port_change(backer
, devname
);
679 process_dpif_port_error(backer
, error
);
686 process_dpif_all_ports_changed(struct dpif_backer
*backer
)
688 struct ofproto_dpif
*ofproto
;
689 struct dpif_port dpif_port
;
690 struct dpif_port_dump dump
;
691 struct sset devnames
;
694 sset_init(&devnames
);
695 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
696 if (ofproto
->backer
== backer
) {
697 struct ofport
*ofport
;
699 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
700 sset_add(&devnames
, netdev_get_name(ofport
->netdev
));
704 DPIF_PORT_FOR_EACH (&dpif_port
, &dump
, backer
->dpif
) {
705 sset_add(&devnames
, dpif_port
.name
);
708 SSET_FOR_EACH (devname
, &devnames
) {
709 process_dpif_port_change(backer
, devname
);
711 sset_destroy(&devnames
);
715 process_dpif_port_change(struct dpif_backer
*backer
, const char *devname
)
717 struct ofproto_dpif
*ofproto
;
718 struct dpif_port port
;
720 /* Don't report on the datapath's device. */
721 if (!strcmp(devname
, dpif_base_name(backer
->dpif
))) {
725 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
,
726 &all_ofproto_dpifs
) {
727 if (simap_contains(&ofproto
->backer
->tnl_backers
, devname
)) {
732 ofproto
= lookup_ofproto_dpif_by_port_name(devname
);
733 if (dpif_port_query_by_name(backer
->dpif
, devname
, &port
)) {
734 /* The port was removed. If we know the datapath,
735 * report it through poll_set(). If we don't, it may be
736 * notifying us of a removal we initiated, so ignore it.
737 * If there's a pending ENOBUFS, let it stand, since
738 * everything will be reevaluated. */
739 if (ofproto
&& ofproto
->port_poll_errno
!= ENOBUFS
) {
740 sset_add(&ofproto
->port_poll_set
, devname
);
741 ofproto
->port_poll_errno
= 0;
743 } else if (!ofproto
) {
744 /* The port was added, but we don't know with which
745 * ofproto we should associate it. Delete it. */
746 dpif_port_del(backer
->dpif
, port
.port_no
);
748 struct ofport_dpif
*ofport
;
750 ofport
= ofport_dpif_cast(shash_find_data(
751 &ofproto
->up
.port_by_name
, devname
));
753 && ofport
->odp_port
!= port
.port_no
754 && !odp_port_to_ofport(backer
, port
.port_no
))
756 /* 'ofport''s datapath port number has changed from
757 * 'ofport->odp_port' to 'port.port_no'. Update our internal data
758 * structures to match. */
759 ovs_rwlock_wrlock(&backer
->odp_to_ofport_lock
);
760 hmap_remove(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
);
761 ofport
->odp_port
= port
.port_no
;
762 hmap_insert(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
,
763 hash_odp_port(port
.port_no
));
764 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
765 backer
->need_revalidate
= REV_RECONFIGURE
;
768 dpif_port_destroy(&port
);
771 /* Propagate 'error' to all ofprotos based on 'backer'. */
773 process_dpif_port_error(struct dpif_backer
*backer
, int error
)
775 struct ofproto_dpif
*ofproto
;
777 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
778 if (ofproto
->backer
== backer
) {
779 sset_clear(&ofproto
->port_poll_set
);
780 ofproto
->port_poll_errno
= error
;
786 type_wait(const char *type
)
788 struct dpif_backer
*backer
;
790 backer
= shash_find_data(&all_dpif_backers
, type
);
792 /* This is not necessarily a problem, since backers are only
793 * created on demand. */
797 dpif_wait(backer
->dpif
);
800 /* Basic life-cycle. */
802 static int add_internal_flows(struct ofproto_dpif
*);
804 static struct ofproto
*
807 struct ofproto_dpif
*ofproto
= xmalloc(sizeof *ofproto
);
812 dealloc(struct ofproto
*ofproto_
)
814 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
819 close_dpif_backer(struct dpif_backer
*backer
)
821 ovs_assert(backer
->refcount
> 0);
823 if (--backer
->refcount
) {
827 udpif_destroy(backer
->udpif
);
829 simap_destroy(&backer
->tnl_backers
);
830 ovs_rwlock_destroy(&backer
->odp_to_ofport_lock
);
831 hmap_destroy(&backer
->odp_to_ofport_map
);
832 shash_find_and_delete(&all_dpif_backers
, backer
->type
);
833 recirc_id_pool_destroy(backer
->rid_pool
);
835 dpif_close(backer
->dpif
);
839 /* Datapath port slated for removal from datapath. */
841 struct list list_node
;
845 static bool check_variable_length_userdata(struct dpif_backer
*backer
);
846 static size_t check_max_mpls_depth(struct dpif_backer
*backer
);
847 static bool check_recirc(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 list garbage_list
;
857 struct odp_garbage
*garbage
, *next
;
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 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 list_push_front(&garbage_list
, &garbage
->list_node
);
931 dpif_port_dump_done(&port_dump
);
933 LIST_FOR_EACH_SAFE (garbage
, next
, list_node
, &garbage_list
) {
934 dpif_port_del(backer
->dpif
, garbage
->odp_port
);
935 list_remove(&garbage
->list_node
);
939 shash_add(&all_dpif_backers
, type
, backer
);
941 backer
->enable_recirc
= check_recirc(backer
);
942 backer
->variable_length_userdata
= check_variable_length_userdata(backer
);
943 backer
->max_mpls_depth
= check_max_mpls_depth(backer
);
944 backer
->rid_pool
= recirc_id_pool_create();
946 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
948 VLOG_ERR("failed to listen on datapath of type %s: %s",
949 type
, ovs_strerror(error
));
950 close_dpif_backer(backer
);
954 if (backer
->recv_set_enable
) {
955 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
961 /* Tests whether 'backer''s datapath supports recirculation. Only newer
962 * datapaths support OVS_KEY_ATTR_RECIRC_ID in keys. We need to disable some
963 * features on older datapaths that don't support this feature.
965 * Returns false if 'backer' definitely does not support recirculation, true if
966 * it seems to support recirculation or if at least the error we get is
969 check_recirc(struct dpif_backer
*backer
)
972 struct odputil_keybuf keybuf
;
975 bool enable_recirc
= false;
977 memset(&flow
, 0, sizeof flow
);
981 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
982 odp_flow_key_from_flow(&key
, &flow
, NULL
, 0, true);
984 error
= dpif_flow_put(backer
->dpif
, DPIF_FP_CREATE
,
985 ofpbuf_data(&key
), ofpbuf_size(&key
), NULL
, 0, NULL
,
987 if (error
&& error
!= EEXIST
) {
988 if (error
!= EINVAL
) {
989 VLOG_WARN("%s: Reciculation flow probe failed (%s)",
990 dpif_name(backer
->dpif
), ovs_strerror(error
));
995 error
= dpif_flow_del(backer
->dpif
, ofpbuf_data(&key
), ofpbuf_size(&key
),
998 VLOG_WARN("%s: failed to delete recirculation feature probe flow",
999 dpif_name(backer
->dpif
));
1002 enable_recirc
= true;
1005 if (enable_recirc
) {
1006 VLOG_INFO("%s: Datapath supports recirculation",
1007 dpif_name(backer
->dpif
));
1009 VLOG_INFO("%s: Datapath does not support recirculation",
1010 dpif_name(backer
->dpif
));
1013 return enable_recirc
;
1016 /* Tests whether 'backer''s datapath supports variable-length
1017 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. We need
1018 * to disable some features on older datapaths that don't support this
1021 * Returns false if 'backer' definitely does not support variable-length
1022 * userdata, true if it seems to support them or if at least the error we get
1025 check_variable_length_userdata(struct dpif_backer
*backer
)
1027 struct eth_header
*eth
;
1028 struct ofpbuf actions
;
1029 struct dpif_execute execute
;
1030 struct ofpbuf packet
;
1034 /* Compose a userspace action that will cause an ERANGE error on older
1035 * datapaths that don't support variable-length userdata.
1037 * We really test for using userdata longer than 8 bytes, but older
1038 * datapaths accepted these, silently truncating the userdata to 8 bytes.
1039 * The same older datapaths rejected userdata shorter than 8 bytes, so we
1040 * test for that instead as a proxy for longer userdata support. */
1041 ofpbuf_init(&actions
, 64);
1042 start
= nl_msg_start_nested(&actions
, OVS_ACTION_ATTR_USERSPACE
);
1043 nl_msg_put_u32(&actions
, OVS_USERSPACE_ATTR_PID
,
1044 dpif_port_get_pid(backer
->dpif
, ODPP_NONE
, 0));
1045 nl_msg_put_unspec_zero(&actions
, OVS_USERSPACE_ATTR_USERDATA
, 4);
1046 nl_msg_end_nested(&actions
, start
);
1048 /* Compose a dummy ethernet packet. */
1049 ofpbuf_init(&packet
, ETH_HEADER_LEN
);
1050 eth
= ofpbuf_put_zeros(&packet
, ETH_HEADER_LEN
);
1051 eth
->eth_type
= htons(0x1234);
1053 /* Execute the actions. On older datapaths this fails with ERANGE, on
1054 * newer datapaths it succeeds. */
1055 execute
.actions
= ofpbuf_data(&actions
);
1056 execute
.actions_len
= ofpbuf_size(&actions
);
1057 execute
.packet
= &packet
;
1058 execute
.md
= PKT_METADATA_INITIALIZER(0);
1059 execute
.needs_help
= false;
1061 error
= dpif_execute(backer
->dpif
, &execute
);
1063 ofpbuf_uninit(&packet
);
1064 ofpbuf_uninit(&actions
);
1071 /* Variable-length userdata is not supported. */
1072 VLOG_WARN("%s: datapath does not support variable-length userdata "
1073 "feature (needs Linux 3.10+ or kernel module from OVS "
1074 "1..11+). The NXAST_SAMPLE action will be ignored.",
1075 dpif_name(backer
->dpif
));
1079 /* Something odd happened. We're not sure whether variable-length
1080 * userdata is supported. Default to "yes". */
1081 VLOG_WARN("%s: variable-length userdata feature probe failed (%s)",
1082 dpif_name(backer
->dpif
), ovs_strerror(error
));
1087 /* Tests the MPLS label stack depth supported by 'backer''s datapath.
1089 * Returns the number of elements in a struct flow's mpls_lse field
1090 * if the datapath supports at least that many entries in an
1092 * Otherwise returns the number of MPLS push actions supported by
1095 check_max_mpls_depth(struct dpif_backer
*backer
)
1100 for (n
= 0; n
< FLOW_MAX_MPLS_LABELS
; n
++) {
1101 struct odputil_keybuf keybuf
;
1105 memset(&flow
, 0, sizeof flow
);
1106 flow
.dl_type
= htons(ETH_TYPE_MPLS
);
1107 flow_set_mpls_bos(&flow
, n
, 1);
1109 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1110 odp_flow_key_from_flow(&key
, &flow
, NULL
, 0, false);
1112 error
= dpif_flow_put(backer
->dpif
, DPIF_FP_CREATE
,
1113 ofpbuf_data(&key
), ofpbuf_size(&key
), NULL
, 0, NULL
, 0, NULL
);
1114 if (error
&& error
!= EEXIST
) {
1115 if (error
!= EINVAL
) {
1116 VLOG_WARN("%s: MPLS stack length feature probe failed (%s)",
1117 dpif_name(backer
->dpif
), ovs_strerror(error
));
1122 error
= dpif_flow_del(backer
->dpif
, ofpbuf_data(&key
), ofpbuf_size(&key
), NULL
);
1124 VLOG_WARN("%s: failed to delete MPLS feature probe flow",
1125 dpif_name(backer
->dpif
));
1129 VLOG_INFO("%s: MPLS label stack length probed as %d",
1130 dpif_name(backer
->dpif
), n
);
1135 construct(struct ofproto
*ofproto_
)
1137 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1138 struct shash_node
*node
, *next
;
1141 error
= open_dpif_backer(ofproto
->up
.type
, &ofproto
->backer
);
1146 ofproto
->netflow
= NULL
;
1147 ofproto
->sflow
= NULL
;
1148 ofproto
->ipfix
= NULL
;
1149 ofproto
->stp
= NULL
;
1150 ofproto
->rstp
= NULL
;
1151 ofproto
->dump_seq
= 0;
1152 hmap_init(&ofproto
->bundles
);
1153 ofproto
->ml
= mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME
);
1155 ofproto
->mbridge
= mbridge_create();
1156 ofproto
->has_bonded_bundles
= false;
1157 ofproto
->lacp_enabled
= false;
1158 ovs_mutex_init_adaptive(&ofproto
->stats_mutex
);
1159 ovs_mutex_init(&ofproto
->vsp_mutex
);
1161 guarded_list_init(&ofproto
->pins
);
1163 ofproto_dpif_unixctl_init();
1165 hmap_init(&ofproto
->vlandev_map
);
1166 hmap_init(&ofproto
->realdev_vid_map
);
1168 sset_init(&ofproto
->ports
);
1169 sset_init(&ofproto
->ghost_ports
);
1170 sset_init(&ofproto
->port_poll_set
);
1171 ofproto
->port_poll_errno
= 0;
1172 ofproto
->change_seq
= 0;
1173 ofproto
->pins_seq
= seq_create();
1174 ofproto
->pins_seqno
= seq_read(ofproto
->pins_seq
);
1177 SHASH_FOR_EACH_SAFE (node
, next
, &init_ofp_ports
) {
1178 struct iface_hint
*iface_hint
= node
->data
;
1180 if (!strcmp(iface_hint
->br_name
, ofproto
->up
.name
)) {
1181 /* Check if the datapath already has this port. */
1182 if (dpif_port_exists(ofproto
->backer
->dpif
, node
->name
)) {
1183 sset_add(&ofproto
->ports
, node
->name
);
1186 free(iface_hint
->br_name
);
1187 free(iface_hint
->br_type
);
1189 shash_delete(&init_ofp_ports
, node
);
1193 hmap_insert(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
,
1194 hash_string(ofproto
->up
.name
, 0));
1195 memset(&ofproto
->stats
, 0, sizeof ofproto
->stats
);
1197 ofproto_init_tables(ofproto_
, N_TABLES
);
1198 error
= add_internal_flows(ofproto
);
1200 ofproto
->up
.tables
[TBL_INTERNAL
].flags
= OFTABLE_HIDDEN
| OFTABLE_READONLY
;
1206 add_internal_miss_flow(struct ofproto_dpif
*ofproto
, int id
,
1207 const struct ofpbuf
*ofpacts
, struct rule_dpif
**rulep
)
1213 match_init_catchall(&match
);
1214 match_set_reg(&match
, 0, id
);
1216 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 0, 0, ofpacts
,
1218 *rulep
= error
? NULL
: rule_dpif_cast(rule
);
1224 add_internal_flows(struct ofproto_dpif
*ofproto
)
1226 struct ofpact_controller
*controller
;
1227 uint64_t ofpacts_stub
[128 / 8];
1228 struct ofpbuf ofpacts
;
1229 struct rule
*unused_rulep OVS_UNUSED
;
1230 struct ofpact_resubmit
*resubmit
;
1235 ofpbuf_use_stack(&ofpacts
, ofpacts_stub
, sizeof ofpacts_stub
);
1238 controller
= ofpact_put_CONTROLLER(&ofpacts
);
1239 controller
->max_len
= UINT16_MAX
;
1240 controller
->controller_id
= 0;
1241 controller
->reason
= OFPR_NO_MATCH
;
1242 ofpact_pad(&ofpacts
);
1244 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1245 &ofproto
->miss_rule
);
1250 ofpbuf_clear(&ofpacts
);
1251 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1252 &ofproto
->no_packet_in_rule
);
1257 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1258 &ofproto
->drop_frags_rule
);
1263 /* Continue non-recirculation rule lookups from table 0.
1265 * (priority=2), recirc=0, actions=resubmit(, 0)
1267 resubmit
= ofpact_put_RESUBMIT(&ofpacts
);
1268 resubmit
->in_port
= OFPP_IN_PORT
;
1269 resubmit
->table_id
= 0;
1271 match_init_catchall(&match
);
1272 match_set_recirc_id(&match
, 0);
1274 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 2, 0, &ofpacts
,
1280 /* Drop any run away recirc rule lookups. Recirc_id has to be
1281 * non-zero when reaching this rule.
1283 * (priority=1), *, actions=drop
1285 ofpbuf_clear(&ofpacts
);
1286 match_init_catchall(&match
);
1287 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 1, 0, &ofpacts
,
1294 destruct(struct ofproto
*ofproto_
)
1296 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1297 struct ofproto_packet_in
*pin
, *next_pin
;
1298 struct rule_dpif
*rule
;
1299 struct oftable
*table
;
1302 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1304 xlate_remove_ofproto(ofproto
);
1307 /* Ensure that the upcall processing threads have no remaining references
1308 * to the ofproto or anything in it. */
1309 udpif_synchronize(ofproto
->backer
->udpif
);
1311 hmap_remove(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
);
1313 OFPROTO_FOR_EACH_TABLE (table
, &ofproto
->up
) {
1314 CLS_FOR_EACH_SAFE (rule
, up
.cr
, &table
->cls
) {
1315 ofproto_rule_delete(&ofproto
->up
, &rule
->up
);
1319 guarded_list_pop_all(&ofproto
->pins
, &pins
);
1320 LIST_FOR_EACH_SAFE (pin
, next_pin
, list_node
, &pins
) {
1321 list_remove(&pin
->list_node
);
1322 free(CONST_CAST(void *, pin
->up
.packet
));
1325 guarded_list_destroy(&ofproto
->pins
);
1327 mbridge_unref(ofproto
->mbridge
);
1329 netflow_unref(ofproto
->netflow
);
1330 dpif_sflow_unref(ofproto
->sflow
);
1331 dpif_ipfix_unref(ofproto
->ipfix
);
1332 hmap_destroy(&ofproto
->bundles
);
1333 mac_learning_unref(ofproto
->ml
);
1334 mcast_snooping_unref(ofproto
->ms
);
1336 hmap_destroy(&ofproto
->vlandev_map
);
1337 hmap_destroy(&ofproto
->realdev_vid_map
);
1339 sset_destroy(&ofproto
->ports
);
1340 sset_destroy(&ofproto
->ghost_ports
);
1341 sset_destroy(&ofproto
->port_poll_set
);
1343 ovs_mutex_destroy(&ofproto
->stats_mutex
);
1344 ovs_mutex_destroy(&ofproto
->vsp_mutex
);
1346 seq_destroy(ofproto
->pins_seq
);
1348 close_dpif_backer(ofproto
->backer
);
1352 run(struct ofproto
*ofproto_
)
1354 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1355 uint64_t new_seq
, new_dump_seq
;
1357 if (mbridge_need_revalidate(ofproto
->mbridge
)) {
1358 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1359 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1360 mac_learning_flush(ofproto
->ml
);
1361 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1362 mcast_snooping_mdb_flush(ofproto
->ms
);
1365 /* Always updates the ofproto->pins_seqno to avoid frequent wakeup during
1366 * flow restore. Even though nothing is processed during flow restore,
1367 * all queued 'pins' will be handled immediately when flow restore
1369 ofproto
->pins_seqno
= seq_read(ofproto
->pins_seq
);
1371 /* Do not perform any periodic activity required by 'ofproto' while
1372 * waiting for flow restore to complete. */
1373 if (!ofproto_get_flow_restore_wait()) {
1374 struct ofproto_packet_in
*pin
, *next_pin
;
1377 guarded_list_pop_all(&ofproto
->pins
, &pins
);
1378 LIST_FOR_EACH_SAFE (pin
, next_pin
, list_node
, &pins
) {
1379 connmgr_send_packet_in(ofproto
->up
.connmgr
, pin
);
1380 list_remove(&pin
->list_node
);
1381 free(CONST_CAST(void *, pin
->up
.packet
));
1386 if (ofproto
->netflow
) {
1387 netflow_run(ofproto
->netflow
);
1389 if (ofproto
->sflow
) {
1390 dpif_sflow_run(ofproto
->sflow
);
1392 if (ofproto
->ipfix
) {
1393 dpif_ipfix_run(ofproto
->ipfix
);
1396 new_seq
= seq_read(connectivity_seq_get());
1397 if (ofproto
->change_seq
!= new_seq
) {
1398 struct ofport_dpif
*ofport
;
1400 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1404 ofproto
->change_seq
= new_seq
;
1406 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1407 struct ofbundle
*bundle
;
1409 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1416 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1417 if (mac_learning_run(ofproto
->ml
)) {
1418 ofproto
->backer
->need_revalidate
= REV_MAC_LEARNING
;
1420 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1422 if (mcast_snooping_run(ofproto
->ms
)) {
1423 ofproto
->backer
->need_revalidate
= REV_MCAST_SNOOPING
;
1426 new_dump_seq
= seq_read(udpif_dump_seq(ofproto
->backer
->udpif
));
1427 if (ofproto
->dump_seq
!= new_dump_seq
) {
1428 struct rule
*rule
, *next_rule
;
1430 /* We know stats are relatively fresh, so now is a good time to do some
1432 ofproto
->dump_seq
= new_dump_seq
;
1434 /* Expire OpenFlow flows whose idle_timeout or hard_timeout
1436 ovs_mutex_lock(&ofproto_mutex
);
1437 LIST_FOR_EACH_SAFE (rule
, next_rule
, expirable
,
1438 &ofproto
->up
.expirable
) {
1439 rule_expire(rule_dpif_cast(rule
));
1441 ovs_mutex_unlock(&ofproto_mutex
);
1443 /* All outstanding data in existing flows has been accounted, so it's a
1444 * good time to do bond rebalancing. */
1445 if (ofproto
->has_bonded_bundles
) {
1446 struct ofbundle
*bundle
;
1448 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1450 bond_rebalance(bundle
->bond
);
1460 wait(struct ofproto
*ofproto_
)
1462 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1464 if (ofproto_get_flow_restore_wait()) {
1468 if (ofproto
->sflow
) {
1469 dpif_sflow_wait(ofproto
->sflow
);
1471 if (ofproto
->ipfix
) {
1472 dpif_ipfix_wait(ofproto
->ipfix
);
1474 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1475 struct ofbundle
*bundle
;
1477 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1478 bundle_wait(bundle
);
1481 if (ofproto
->netflow
) {
1482 netflow_wait(ofproto
->netflow
);
1484 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
1485 mac_learning_wait(ofproto
->ml
);
1486 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1487 mcast_snooping_wait(ofproto
->ms
);
1489 if (ofproto
->backer
->need_revalidate
) {
1490 /* Shouldn't happen, but if it does just go around again. */
1491 VLOG_DBG_RL(&rl
, "need revalidate in ofproto_wait_cb()");
1492 poll_immediate_wake();
1495 seq_wait(udpif_dump_seq(ofproto
->backer
->udpif
), ofproto
->dump_seq
);
1496 seq_wait(ofproto
->pins_seq
, ofproto
->pins_seqno
);
1500 type_get_memory_usage(const char *type
, struct simap
*usage
)
1502 struct dpif_backer
*backer
;
1504 backer
= shash_find_data(&all_dpif_backers
, type
);
1506 udpif_get_memory_usage(backer
->udpif
, usage
);
1511 flush(struct ofproto
*ofproto_
)
1513 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1514 struct dpif_backer
*backer
= ofproto
->backer
;
1517 udpif_flush(backer
->udpif
);
1522 query_tables(struct ofproto
*ofproto
,
1523 struct ofputil_table_features
*features
,
1524 struct ofputil_table_stats
*stats
)
1526 strcpy(features
->name
, "classifier");
1531 for (i
= 0; i
< ofproto
->n_tables
; i
++) {
1532 unsigned long missed
, matched
;
1534 atomic_read_relaxed(&ofproto
->tables
[i
].n_matched
, &matched
);
1535 atomic_read_relaxed(&ofproto
->tables
[i
].n_missed
, &missed
);
1537 stats
[i
].matched_count
= matched
;
1538 stats
[i
].lookup_count
= matched
+ missed
;
1543 static struct ofport
*
1546 struct ofport_dpif
*port
= xmalloc(sizeof *port
);
1551 port_dealloc(struct ofport
*port_
)
1553 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1558 port_construct(struct ofport
*port_
)
1560 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1561 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1562 const struct netdev
*netdev
= port
->up
.netdev
;
1563 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1564 struct dpif_port dpif_port
;
1567 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1568 port
->bundle
= NULL
;
1571 port
->may_enable
= true;
1572 port
->stp_port
= NULL
;
1573 port
->stp_state
= STP_DISABLED
;
1574 port
->rstp_port
= NULL
;
1575 port
->rstp_state
= RSTP_DISABLED
;
1576 port
->is_tunnel
= false;
1580 port
->realdev_ofp_port
= 0;
1581 port
->vlandev_vid
= 0;
1582 port
->carrier_seq
= netdev_get_carrier_resets(netdev
);
1583 port
->is_layer3
= netdev_vport_is_layer3(netdev
);
1585 if (netdev_vport_is_patch(netdev
)) {
1586 /* By bailing out here, we don't submit the port to the sFlow module
1587 * to be considered for counter polling export. This is correct
1588 * because the patch port represents an interface that sFlow considers
1589 * to be "internal" to the switch as a whole, and therefore not an
1590 * candidate for counter polling. */
1591 port
->odp_port
= ODPP_NONE
;
1592 ofport_update_peer(port
);
1596 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
1597 netdev_vport_get_dpif_port(netdev
, namebuf
,
1604 port
->odp_port
= dpif_port
.port_no
;
1606 if (netdev_get_tunnel_config(netdev
)) {
1607 tnl_port_add(port
, port
->up
.netdev
, port
->odp_port
);
1608 port
->is_tunnel
= true;
1609 if (ofproto
->ipfix
) {
1610 dpif_ipfix_add_tunnel_port(ofproto
->ipfix
, port_
, port
->odp_port
);
1613 /* Sanity-check that a mapping doesn't already exist. This
1614 * shouldn't happen for non-tunnel ports. */
1615 if (odp_port_to_ofp_port(ofproto
, port
->odp_port
) != OFPP_NONE
) {
1616 VLOG_ERR("port %s already has an OpenFlow port number",
1618 dpif_port_destroy(&dpif_port
);
1622 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1623 hmap_insert(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
,
1624 hash_odp_port(port
->odp_port
));
1625 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1627 dpif_port_destroy(&dpif_port
);
1629 if (ofproto
->sflow
) {
1630 dpif_sflow_add_port(ofproto
->sflow
, port_
, port
->odp_port
);
1637 port_destruct(struct ofport
*port_
)
1639 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1640 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1641 const char *devname
= netdev_get_name(port
->up
.netdev
);
1642 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1643 const char *dp_port_name
;
1645 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1647 xlate_ofport_remove(port
);
1650 dp_port_name
= netdev_vport_get_dpif_port(port
->up
.netdev
, namebuf
,
1652 if (dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
1653 /* The underlying device is still there, so delete it. This
1654 * happens when the ofproto is being destroyed, since the caller
1655 * assumes that removal of attached ports will happen as part of
1657 if (!port
->is_tunnel
) {
1658 dpif_port_del(ofproto
->backer
->dpif
, port
->odp_port
);
1663 port
->peer
->peer
= NULL
;
1667 if (port
->odp_port
!= ODPP_NONE
&& !port
->is_tunnel
) {
1668 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1669 hmap_remove(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
);
1670 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1673 if (port
->is_tunnel
&& ofproto
->ipfix
) {
1674 dpif_ipfix_del_tunnel_port(ofproto
->ipfix
, port
->odp_port
);
1678 sset_find_and_delete(&ofproto
->ports
, devname
);
1679 sset_find_and_delete(&ofproto
->ghost_ports
, devname
);
1680 bundle_remove(port_
);
1681 set_cfm(port_
, NULL
);
1682 set_bfd(port_
, NULL
);
1683 if (port
->stp_port
) {
1684 stp_port_disable(port
->stp_port
);
1686 if (port
->rstp_port
) {
1687 rstp_delete_port(port
->rstp_port
);
1689 if (ofproto
->sflow
) {
1690 dpif_sflow_del_port(ofproto
->sflow
, port
->odp_port
);
1697 port_modified(struct ofport
*port_
)
1699 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1701 if (port
->bundle
&& port
->bundle
->bond
) {
1702 bond_slave_set_netdev(port
->bundle
->bond
, port
, port
->up
.netdev
);
1706 cfm_set_netdev(port
->cfm
, port
->up
.netdev
);
1710 bfd_set_netdev(port
->bfd
, port
->up
.netdev
);
1713 ofproto_dpif_monitor_port_update(port
, port
->bfd
, port
->cfm
,
1714 port
->up
.pp
.hw_addr
);
1716 if (port
->is_tunnel
&& tnl_port_reconfigure(port
, port
->up
.netdev
,
1718 ofproto_dpif_cast(port
->up
.ofproto
)->backer
->need_revalidate
=
1722 ofport_update_peer(port
);
1726 port_reconfigured(struct ofport
*port_
, enum ofputil_port_config old_config
)
1728 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1729 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1730 enum ofputil_port_config changed
= old_config
^ port
->up
.pp
.config
;
1732 if (changed
& (OFPUTIL_PC_NO_RECV
| OFPUTIL_PC_NO_RECV_STP
|
1733 OFPUTIL_PC_NO_FWD
| OFPUTIL_PC_NO_FLOOD
|
1734 OFPUTIL_PC_NO_PACKET_IN
)) {
1735 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1737 if (changed
& OFPUTIL_PC_NO_FLOOD
&& port
->bundle
) {
1738 bundle_update(port
->bundle
);
1744 set_sflow(struct ofproto
*ofproto_
,
1745 const struct ofproto_sflow_options
*sflow_options
)
1747 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1748 struct dpif_sflow
*ds
= ofproto
->sflow
;
1750 if (sflow_options
) {
1752 struct ofport_dpif
*ofport
;
1754 ds
= ofproto
->sflow
= dpif_sflow_create();
1755 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1756 dpif_sflow_add_port(ds
, &ofport
->up
, ofport
->odp_port
);
1758 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1760 dpif_sflow_set_options(ds
, sflow_options
);
1763 dpif_sflow_unref(ds
);
1764 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1765 ofproto
->sflow
= NULL
;
1773 struct ofproto
*ofproto_
,
1774 const struct ofproto_ipfix_bridge_exporter_options
*bridge_exporter_options
,
1775 const struct ofproto_ipfix_flow_exporter_options
*flow_exporters_options
,
1776 size_t n_flow_exporters_options
)
1778 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1779 struct dpif_ipfix
*di
= ofproto
->ipfix
;
1780 bool has_options
= bridge_exporter_options
|| flow_exporters_options
;
1781 bool new_di
= false;
1783 if (has_options
&& !di
) {
1784 di
= ofproto
->ipfix
= dpif_ipfix_create();
1789 /* Call set_options in any case to cleanly flush the flow
1790 * caches in the last exporters that are to be destroyed. */
1791 dpif_ipfix_set_options(
1792 di
, bridge_exporter_options
, flow_exporters_options
,
1793 n_flow_exporters_options
);
1795 /* Add tunnel ports only when a new ipfix created */
1796 if (new_di
== true) {
1797 struct ofport_dpif
*ofport
;
1798 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1799 if (ofport
->is_tunnel
== true) {
1800 dpif_ipfix_add_tunnel_port(di
, &ofport
->up
, ofport
->odp_port
);
1806 dpif_ipfix_unref(di
);
1807 ofproto
->ipfix
= NULL
;
1815 set_cfm(struct ofport
*ofport_
, const struct cfm_settings
*s
)
1817 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1822 struct ofproto_dpif
*ofproto
;
1824 ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
1825 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1826 ofport
->cfm
= cfm_create(ofport
->up
.netdev
);
1829 if (cfm_configure(ofport
->cfm
, s
)) {
1836 cfm_unref(ofport
->cfm
);
1839 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
1840 ofport
->up
.pp
.hw_addr
);
1845 cfm_status_changed(struct ofport
*ofport_
)
1847 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1849 return ofport
->cfm
? cfm_check_status_change(ofport
->cfm
) : true;
1853 get_cfm_status(const struct ofport
*ofport_
,
1854 struct cfm_status
*status
)
1856 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1860 cfm_get_status(ofport
->cfm
, status
);
1869 set_bfd(struct ofport
*ofport_
, const struct smap
*cfg
)
1871 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
1872 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1876 ofport
->bfd
= bfd_configure(old
, netdev_get_name(ofport
->up
.netdev
),
1877 cfg
, ofport
->up
.netdev
);
1878 if (ofport
->bfd
!= old
) {
1879 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1881 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
1882 ofport
->up
.pp
.hw_addr
);
1887 bfd_status_changed(struct ofport
*ofport_
)
1889 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1891 return ofport
->bfd
? bfd_check_status_change(ofport
->bfd
) : true;
1895 get_bfd_status(struct ofport
*ofport_
, struct smap
*smap
)
1897 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1901 bfd_get_status(ofport
->bfd
, smap
);
1909 /* Spanning Tree. */
1912 rstp_send_bpdu_cb(struct ofpbuf
*pkt
, int port_num
, void *ofproto_
)
1914 struct ofproto_dpif
*ofproto
= ofproto_
;
1915 struct rstp_port
*rp
= rstp_get_port(ofproto
->rstp
, port_num
);
1916 struct ofport_dpif
*ofport
;
1918 ofport
= rstp_port_get_aux(rp
);
1920 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on unknown port %d",
1921 ofproto
->up
.name
, port_num
);
1923 struct eth_header
*eth
= ofpbuf_l2(pkt
);
1925 netdev_get_etheraddr(ofport
->up
.netdev
, eth
->eth_src
);
1926 if (eth_addr_is_zero(eth
->eth_src
)) {
1927 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on port %d "
1928 "with unknown MAC", ofproto
->up
.name
, port_num
);
1930 ofproto_dpif_send_packet(ofport
, pkt
);
1937 send_bpdu_cb(struct ofpbuf
*pkt
, int port_num
, void *ofproto_
)
1939 struct ofproto_dpif
*ofproto
= ofproto_
;
1940 struct stp_port
*sp
= stp_get_port(ofproto
->stp
, port_num
);
1941 struct ofport_dpif
*ofport
;
1943 ofport
= stp_port_get_aux(sp
);
1945 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on unknown port %d",
1946 ofproto
->up
.name
, port_num
);
1948 struct eth_header
*eth
= ofpbuf_l2(pkt
);
1950 netdev_get_etheraddr(ofport
->up
.netdev
, eth
->eth_src
);
1951 if (eth_addr_is_zero(eth
->eth_src
)) {
1952 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on port %d "
1953 "with unknown MAC", ofproto
->up
.name
, port_num
);
1955 ofproto_dpif_send_packet(ofport
, pkt
);
1961 /* Configure RSTP on 'ofproto_' using the settings defined in 's'. */
1963 set_rstp(struct ofproto
*ofproto_
, const struct ofproto_rstp_settings
*s
)
1965 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1967 /* Only revalidate flows if the configuration changed. */
1968 if (!s
!= !ofproto
->rstp
) {
1969 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1973 if (!ofproto
->rstp
) {
1974 ofproto
->rstp
= rstp_create(ofproto_
->name
, s
->address
,
1975 rstp_send_bpdu_cb
, ofproto
);
1976 ofproto
->rstp_last_tick
= time_msec();
1978 rstp_set_bridge_address(ofproto
->rstp
, s
->address
);
1979 rstp_set_bridge_priority(ofproto
->rstp
, s
->priority
);
1980 rstp_set_bridge_ageing_time(ofproto
->rstp
, s
->ageing_time
);
1981 rstp_set_bridge_force_protocol_version(ofproto
->rstp
,
1982 s
->force_protocol_version
);
1983 rstp_set_bridge_max_age(ofproto
->rstp
, s
->bridge_max_age
);
1984 rstp_set_bridge_forward_delay(ofproto
->rstp
, s
->bridge_forward_delay
);
1985 rstp_set_bridge_transmit_hold_count(ofproto
->rstp
,
1986 s
->transmit_hold_count
);
1988 struct ofport
*ofport
;
1989 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
1990 set_rstp_port(ofport
, NULL
);
1992 rstp_unref(ofproto
->rstp
);
1993 ofproto
->rstp
= NULL
;
1998 get_rstp_status(struct ofproto
*ofproto_
, struct ofproto_rstp_status
*s
)
2000 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2002 if (ofproto
->rstp
) {
2004 s
->root_id
= rstp_get_root_id(ofproto
->rstp
);
2005 s
->bridge_id
= rstp_get_bridge_id(ofproto
->rstp
);
2006 s
->designated_id
= rstp_get_designated_id(ofproto
->rstp
);
2007 s
->root_path_cost
= rstp_get_root_path_cost(ofproto
->rstp
);
2008 s
->designated_port_id
= rstp_get_designated_port_id(ofproto
->rstp
);
2009 s
->bridge_port_id
= rstp_get_bridge_port_id(ofproto
->rstp
);
2016 update_rstp_port_state(struct ofport_dpif
*ofport
)
2018 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2019 enum rstp_state state
;
2021 /* Figure out new state. */
2022 state
= ofport
->rstp_port
? rstp_port_get_state(ofport
->rstp_port
)
2026 if (ofport
->rstp_state
!= state
) {
2027 enum ofputil_port_state of_state
;
2030 VLOG_DBG_RL(&rl
, "port %s: RSTP state changed from %s to %s",
2031 netdev_get_name(ofport
->up
.netdev
),
2032 rstp_state_name(ofport
->rstp_state
),
2033 rstp_state_name(state
));
2034 if (rstp_learn_in_state(ofport
->rstp_state
)
2035 != rstp_learn_in_state(state
)) {
2036 /* xxx Learning action flows should also be flushed. */
2037 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2038 mac_learning_flush(ofproto
->ml
);
2039 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2041 fwd_change
= rstp_forward_in_state(ofport
->rstp_state
)
2042 != rstp_forward_in_state(state
);
2044 ofproto
->backer
->need_revalidate
= REV_RSTP
;
2045 ofport
->rstp_state
= state
;
2047 if (fwd_change
&& ofport
->bundle
) {
2048 bundle_update(ofport
->bundle
);
2051 /* Update the RSTP state bits in the OpenFlow port description. */
2052 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2053 of_state
|= (state
== RSTP_LEARNING
? OFPUTIL_PS_STP_LEARN
2054 : state
== RSTP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2055 : state
== RSTP_DISCARDING
? OFPUTIL_PS_STP_LISTEN
2057 ofproto_port_set_state(&ofport
->up
, of_state
);
2062 rstp_run(struct ofproto_dpif
*ofproto
)
2064 if (ofproto
->rstp
) {
2065 long long int now
= time_msec();
2066 long long int elapsed
= now
- ofproto
->rstp_last_tick
;
2067 struct rstp_port
*rp
;
2068 /* Every second, decrease the values of the timers. */
2069 if (elapsed
>= 1000) {
2070 rstp_tick_timers(ofproto
->rstp
);
2071 ofproto
->rstp_last_tick
= now
;
2073 while (rstp_get_changed_port(ofproto
->rstp
, &rp
)) {
2074 struct ofport_dpif
*ofport
= rstp_port_get_aux(rp
);
2076 update_rstp_port_state(ofport
);
2079 /* FIXME: This check should be done on-event (i.e., when setting
2080 * p->fdb_flush) and not periodically.
2082 if (rstp_check_and_reset_fdb_flush(ofproto
->rstp
)) {
2083 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2084 /* FIXME: RSTP should be able to flush the entries pertaining to a
2085 * single port, not the whole table.
2087 mac_learning_flush(ofproto
->ml
);
2088 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2093 /* Configures STP on 'ofproto_' using the settings defined in 's'. */
2095 set_stp(struct ofproto
*ofproto_
, const struct ofproto_stp_settings
*s
)
2097 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2099 /* Only revalidate flows if the configuration changed. */
2100 if (!s
!= !ofproto
->stp
) {
2101 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2105 if (!ofproto
->stp
) {
2106 ofproto
->stp
= stp_create(ofproto_
->name
, s
->system_id
,
2107 send_bpdu_cb
, ofproto
);
2108 ofproto
->stp_last_tick
= time_msec();
2111 stp_set_bridge_id(ofproto
->stp
, s
->system_id
);
2112 stp_set_bridge_priority(ofproto
->stp
, s
->priority
);
2113 stp_set_hello_time(ofproto
->stp
, s
->hello_time
);
2114 stp_set_max_age(ofproto
->stp
, s
->max_age
);
2115 stp_set_forward_delay(ofproto
->stp
, s
->fwd_delay
);
2117 struct ofport
*ofport
;
2119 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2120 set_stp_port(ofport
, NULL
);
2123 stp_unref(ofproto
->stp
);
2124 ofproto
->stp
= NULL
;
2131 get_stp_status(struct ofproto
*ofproto_
, struct ofproto_stp_status
*s
)
2133 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2137 s
->bridge_id
= stp_get_bridge_id(ofproto
->stp
);
2138 s
->designated_root
= stp_get_designated_root(ofproto
->stp
);
2139 s
->root_path_cost
= stp_get_root_path_cost(ofproto
->stp
);
2148 update_stp_port_state(struct ofport_dpif
*ofport
)
2150 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2151 enum stp_state state
;
2153 /* Figure out new state. */
2154 state
= ofport
->stp_port
? stp_port_get_state(ofport
->stp_port
)
2158 if (ofport
->stp_state
!= state
) {
2159 enum ofputil_port_state of_state
;
2162 VLOG_DBG_RL(&rl
, "port %s: STP state changed from %s to %s",
2163 netdev_get_name(ofport
->up
.netdev
),
2164 stp_state_name(ofport
->stp_state
),
2165 stp_state_name(state
));
2166 if (stp_learn_in_state(ofport
->stp_state
)
2167 != stp_learn_in_state(state
)) {
2168 /* xxx Learning action flows should also be flushed. */
2169 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2170 mac_learning_flush(ofproto
->ml
);
2171 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2172 mcast_snooping_mdb_flush(ofproto
->ms
);
2174 fwd_change
= stp_forward_in_state(ofport
->stp_state
)
2175 != stp_forward_in_state(state
);
2177 ofproto
->backer
->need_revalidate
= REV_STP
;
2178 ofport
->stp_state
= state
;
2179 ofport
->stp_state_entered
= time_msec();
2181 if (fwd_change
&& ofport
->bundle
) {
2182 bundle_update(ofport
->bundle
);
2185 /* Update the STP state bits in the OpenFlow port description. */
2186 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2187 of_state
|= (state
== STP_LISTENING
? OFPUTIL_PS_STP_LISTEN
2188 : state
== STP_LEARNING
? OFPUTIL_PS_STP_LEARN
2189 : state
== STP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2190 : state
== STP_BLOCKING
? OFPUTIL_PS_STP_BLOCK
2192 ofproto_port_set_state(&ofport
->up
, of_state
);
2196 /* Configures STP on 'ofport_' using the settings defined in 's'. The
2197 * caller is responsible for assigning STP port numbers and ensuring
2198 * there are no duplicates. */
2200 set_stp_port(struct ofport
*ofport_
,
2201 const struct ofproto_port_stp_settings
*s
)
2203 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2204 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2205 struct stp_port
*sp
= ofport
->stp_port
;
2207 if (!s
|| !s
->enable
) {
2209 ofport
->stp_port
= NULL
;
2210 stp_port_disable(sp
);
2211 update_stp_port_state(ofport
);
2214 } else if (sp
&& stp_port_no(sp
) != s
->port_num
2215 && ofport
== stp_port_get_aux(sp
)) {
2216 /* The port-id changed, so disable the old one if it's not
2217 * already in use by another port. */
2218 stp_port_disable(sp
);
2221 sp
= ofport
->stp_port
= stp_get_port(ofproto
->stp
, s
->port_num
);
2222 stp_port_enable(sp
);
2224 stp_port_set_name(sp
, netdev_get_name(ofport
->up
.netdev
));
2225 stp_port_set_aux(sp
, ofport
);
2226 stp_port_set_priority(sp
, s
->priority
);
2227 stp_port_set_path_cost(sp
, s
->path_cost
);
2229 update_stp_port_state(ofport
);
2235 get_stp_port_status(struct ofport
*ofport_
,
2236 struct ofproto_port_stp_status
*s
)
2238 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2239 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2240 struct stp_port
*sp
= ofport
->stp_port
;
2242 if (!ofproto
->stp
|| !sp
) {
2248 s
->port_id
= stp_port_get_id(sp
);
2249 s
->state
= stp_port_get_state(sp
);
2250 s
->sec_in_state
= (time_msec() - ofport
->stp_state_entered
) / 1000;
2251 s
->role
= stp_port_get_role(sp
);
2257 get_stp_port_stats(struct ofport
*ofport_
,
2258 struct ofproto_port_stp_stats
*s
)
2260 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2261 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2262 struct stp_port
*sp
= ofport
->stp_port
;
2264 if (!ofproto
->stp
|| !sp
) {
2270 stp_port_get_counts(sp
, &s
->tx_count
, &s
->rx_count
, &s
->error_count
);
2276 stp_run(struct ofproto_dpif
*ofproto
)
2279 long long int now
= time_msec();
2280 long long int elapsed
= now
- ofproto
->stp_last_tick
;
2281 struct stp_port
*sp
;
2284 stp_tick(ofproto
->stp
, MIN(INT_MAX
, elapsed
));
2285 ofproto
->stp_last_tick
= now
;
2287 while (stp_get_changed_port(ofproto
->stp
, &sp
)) {
2288 struct ofport_dpif
*ofport
= stp_port_get_aux(sp
);
2291 update_stp_port_state(ofport
);
2295 if (stp_check_and_reset_fdb_flush(ofproto
->stp
)) {
2296 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2297 mac_learning_flush(ofproto
->ml
);
2298 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2299 mcast_snooping_mdb_flush(ofproto
->ms
);
2305 stp_wait(struct ofproto_dpif
*ofproto
)
2308 poll_timer_wait(1000);
2312 /* Configures RSTP on 'ofport_' using the settings defined in 's'. The
2313 * caller is responsible for assigning RSTP port numbers and ensuring
2314 * there are no duplicates. */
2316 set_rstp_port(struct ofport
*ofport_
,
2317 const struct ofproto_port_rstp_settings
*s
)
2319 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2320 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2321 struct rstp_port
*rp
= ofport
->rstp_port
;
2324 if (!s
|| !s
->enable
) {
2326 ofport
->rstp_port
= NULL
;
2327 rstp_delete_port(rp
);
2328 update_rstp_port_state(ofport
);
2331 } else if (rp
&& rstp_port_number(rp
) != s
->port_num
2332 && ofport
== rstp_port_get_aux(rp
)) {
2333 /* The port-id changed, so disable the old one if it's not
2334 * already in use by another port. */
2335 if (s
->port_num
!= 0) {
2337 stp_port
= ofport
->stp_port
? stp_port_no(ofport
->stp_port
) : -1;
2338 xlate_ofport_set(ofproto
, ofport
->bundle
, ofport
,
2339 ofport
->up
.ofp_port
, ofport
->odp_port
,
2340 ofport
->up
.netdev
, ofport
->cfm
,
2341 ofport
->bfd
, ofport
->peer
, stp_port
,
2343 ofport
->qdscp
, ofport
->n_qdscp
,
2344 ofport
->up
.pp
.config
, ofport
->up
.pp
.state
,
2345 ofport
->is_tunnel
, ofport
->may_enable
);
2349 rstp_port_set_aux(rp
, ofport
);
2350 rstp_port_set_priority(rp
, s
->priority
);
2351 rstp_port_set_port_number(rp
, s
->port_num
);
2352 rstp_port_set_path_cost(rp
, s
->path_cost
);
2353 rstp_port_set_admin_edge(rp
, s
->admin_edge_port
);
2354 rstp_port_set_auto_edge(rp
, s
->auto_edge
);
2355 rstp_port_set_mcheck(rp
, s
->mcheck
);
2357 update_rstp_port_state(ofport
);
2361 rp
= ofport
->rstp_port
= rstp_get_port(ofproto
->rstp
, s
->port_num
);
2362 /* Enable RSTP on port */
2364 rp
= ofport
->rstp_port
= rstp_add_port(ofproto
->rstp
);
2367 rstp_port_set_aux(rp
, ofport
);
2368 rstp_port_set_priority(rp
, s
->priority
);
2369 rstp_port_set_port_number(rp
, s
->port_num
);
2370 rstp_port_set_path_cost(rp
, s
->path_cost
);
2371 rstp_port_set_admin_edge(rp
, s
->admin_edge_port
);
2372 rstp_port_set_auto_edge(rp
, s
->auto_edge
);
2373 rstp_port_set_mcheck(rp
, s
->mcheck
);
2375 update_rstp_port_state(ofport
);
2379 get_rstp_port_status(struct ofport
*ofport_
,
2380 struct ofproto_port_rstp_status
*s
)
2382 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2383 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2384 struct rstp_port
*rp
= ofport
->rstp_port
;
2386 if (!ofproto
->rstp
|| !rp
) {
2392 s
->port_id
= rstp_port_get_id(rp
);
2393 s
->state
= rstp_port_get_state(rp
);
2394 s
->role
= rstp_port_get_role(rp
);
2395 rstp_port_get_counts(rp
, &s
->tx_count
, &s
->rx_count
,
2396 &s
->error_count
, &s
->uptime
);
2401 set_queues(struct ofport
*ofport_
, const struct ofproto_port_queue
*qdscp
,
2404 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2405 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2407 if (ofport
->n_qdscp
!= n_qdscp
2408 || (n_qdscp
&& memcmp(ofport
->qdscp
, qdscp
,
2409 n_qdscp
* sizeof *qdscp
))) {
2410 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2411 free(ofport
->qdscp
);
2412 ofport
->qdscp
= n_qdscp
2413 ? xmemdup(qdscp
, n_qdscp
* sizeof *qdscp
)
2415 ofport
->n_qdscp
= n_qdscp
;
2423 /* Expires all MAC learning entries associated with 'bundle' and forces its
2424 * ofproto to revalidate every flow.
2426 * Normally MAC learning entries are removed only from the ofproto associated
2427 * with 'bundle', but if 'all_ofprotos' is true, then the MAC learning entries
2428 * are removed from every ofproto. When patch ports and SLB bonds are in use
2429 * and a VM migration happens and the gratuitous ARPs are somehow lost, this
2430 * avoids a MAC_ENTRY_IDLE_TIME delay before the migrated VM can communicate
2431 * with the host from which it migrated. */
2433 bundle_flush_macs(struct ofbundle
*bundle
, bool all_ofprotos
)
2435 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2436 struct mac_learning
*ml
= ofproto
->ml
;
2437 struct mac_entry
*mac
, *next_mac
;
2439 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2440 ovs_rwlock_wrlock(&ml
->rwlock
);
2441 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2442 if (mac
->port
.p
== bundle
) {
2444 struct ofproto_dpif
*o
;
2446 HMAP_FOR_EACH (o
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
2448 struct mac_entry
*e
;
2450 ovs_rwlock_wrlock(&o
->ml
->rwlock
);
2451 e
= mac_learning_lookup(o
->ml
, mac
->mac
, mac
->vlan
);
2453 mac_learning_expire(o
->ml
, e
);
2455 ovs_rwlock_unlock(&o
->ml
->rwlock
);
2460 mac_learning_expire(ml
, mac
);
2463 ovs_rwlock_unlock(&ml
->rwlock
);
2466 static struct ofbundle
*
2467 bundle_lookup(const struct ofproto_dpif
*ofproto
, void *aux
)
2469 struct ofbundle
*bundle
;
2471 HMAP_FOR_EACH_IN_BUCKET (bundle
, hmap_node
, hash_pointer(aux
, 0),
2472 &ofproto
->bundles
) {
2473 if (bundle
->aux
== aux
) {
2481 bundle_update(struct ofbundle
*bundle
)
2483 struct ofport_dpif
*port
;
2485 bundle
->floodable
= true;
2486 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2487 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2489 || !stp_forward_in_state(port
->stp_state
)) {
2490 bundle
->floodable
= false;
2497 bundle_del_port(struct ofport_dpif
*port
)
2499 struct ofbundle
*bundle
= port
->bundle
;
2501 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2503 list_remove(&port
->bundle_node
);
2504 port
->bundle
= NULL
;
2507 lacp_slave_unregister(bundle
->lacp
, port
);
2510 bond_slave_unregister(bundle
->bond
, port
);
2513 bundle_update(bundle
);
2517 bundle_add_port(struct ofbundle
*bundle
, ofp_port_t ofp_port
,
2518 struct lacp_slave_settings
*lacp
)
2520 struct ofport_dpif
*port
;
2522 port
= get_ofp_port(bundle
->ofproto
, ofp_port
);
2527 if (port
->bundle
!= bundle
) {
2528 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2530 bundle_remove(&port
->up
);
2533 port
->bundle
= bundle
;
2534 list_push_back(&bundle
->ports
, &port
->bundle_node
);
2535 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2537 || !stp_forward_in_state(port
->stp_state
)) {
2538 bundle
->floodable
= false;
2542 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2543 lacp_slave_register(bundle
->lacp
, port
, lacp
);
2550 bundle_destroy(struct ofbundle
*bundle
)
2552 struct ofproto_dpif
*ofproto
;
2553 struct ofport_dpif
*port
, *next_port
;
2559 ofproto
= bundle
->ofproto
;
2560 mbridge_unregister_bundle(ofproto
->mbridge
, bundle
->aux
);
2563 xlate_bundle_remove(bundle
);
2566 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2567 bundle_del_port(port
);
2570 bundle_flush_macs(bundle
, true);
2571 hmap_remove(&ofproto
->bundles
, &bundle
->hmap_node
);
2573 free(bundle
->trunks
);
2574 lacp_unref(bundle
->lacp
);
2575 bond_unref(bundle
->bond
);
2580 bundle_set(struct ofproto
*ofproto_
, void *aux
,
2581 const struct ofproto_bundle_settings
*s
)
2583 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2584 bool need_flush
= false;
2585 struct ofport_dpif
*port
;
2586 struct ofbundle
*bundle
;
2587 unsigned long *trunks
;
2593 bundle_destroy(bundle_lookup(ofproto
, aux
));
2597 ovs_assert(s
->n_slaves
== 1 || s
->bond
!= NULL
);
2598 ovs_assert((s
->lacp
!= NULL
) == (s
->lacp_slaves
!= NULL
));
2600 bundle
= bundle_lookup(ofproto
, aux
);
2602 bundle
= xmalloc(sizeof *bundle
);
2604 bundle
->ofproto
= ofproto
;
2605 hmap_insert(&ofproto
->bundles
, &bundle
->hmap_node
,
2606 hash_pointer(aux
, 0));
2608 bundle
->name
= NULL
;
2610 list_init(&bundle
->ports
);
2611 bundle
->vlan_mode
= PORT_VLAN_TRUNK
;
2613 bundle
->trunks
= NULL
;
2614 bundle
->use_priority_tags
= s
->use_priority_tags
;
2615 bundle
->lacp
= NULL
;
2616 bundle
->bond
= NULL
;
2618 bundle
->floodable
= true;
2619 mbridge_register_bundle(ofproto
->mbridge
, bundle
);
2622 if (!bundle
->name
|| strcmp(s
->name
, bundle
->name
)) {
2624 bundle
->name
= xstrdup(s
->name
);
2629 ofproto
->lacp_enabled
= true;
2630 if (!bundle
->lacp
) {
2631 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2632 bundle
->lacp
= lacp_create();
2634 lacp_configure(bundle
->lacp
, s
->lacp
);
2636 lacp_unref(bundle
->lacp
);
2637 bundle
->lacp
= NULL
;
2640 /* Update set of ports. */
2642 for (i
= 0; i
< s
->n_slaves
; i
++) {
2643 if (!bundle_add_port(bundle
, s
->slaves
[i
],
2644 s
->lacp
? &s
->lacp_slaves
[i
] : NULL
)) {
2648 if (!ok
|| list_size(&bundle
->ports
) != s
->n_slaves
) {
2649 struct ofport_dpif
*next_port
;
2651 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2652 for (i
= 0; i
< s
->n_slaves
; i
++) {
2653 if (s
->slaves
[i
] == port
->up
.ofp_port
) {
2658 bundle_del_port(port
);
2662 ovs_assert(list_size(&bundle
->ports
) <= s
->n_slaves
);
2664 if (list_is_empty(&bundle
->ports
)) {
2665 bundle_destroy(bundle
);
2669 /* Set VLAN tagging mode */
2670 if (s
->vlan_mode
!= bundle
->vlan_mode
2671 || s
->use_priority_tags
!= bundle
->use_priority_tags
) {
2672 bundle
->vlan_mode
= s
->vlan_mode
;
2673 bundle
->use_priority_tags
= s
->use_priority_tags
;
2678 vlan
= (s
->vlan_mode
== PORT_VLAN_TRUNK
? -1
2679 : s
->vlan
>= 0 && s
->vlan
<= 4095 ? s
->vlan
2681 if (vlan
!= bundle
->vlan
) {
2682 bundle
->vlan
= vlan
;
2686 /* Get trunked VLANs. */
2687 switch (s
->vlan_mode
) {
2688 case PORT_VLAN_ACCESS
:
2692 case PORT_VLAN_TRUNK
:
2693 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2696 case PORT_VLAN_NATIVE_UNTAGGED
:
2697 case PORT_VLAN_NATIVE_TAGGED
:
2698 if (vlan
!= 0 && (!s
->trunks
2699 || !bitmap_is_set(s
->trunks
, vlan
)
2700 || bitmap_is_set(s
->trunks
, 0))) {
2701 /* Force trunking the native VLAN and prohibit trunking VLAN 0. */
2703 trunks
= bitmap_clone(s
->trunks
, 4096);
2705 trunks
= bitmap_allocate1(4096);
2707 bitmap_set1(trunks
, vlan
);
2708 bitmap_set0(trunks
, 0);
2710 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2717 if (!vlan_bitmap_equal(trunks
, bundle
->trunks
)) {
2718 free(bundle
->trunks
);
2719 if (trunks
== s
->trunks
) {
2720 bundle
->trunks
= vlan_bitmap_clone(trunks
);
2722 bundle
->trunks
= trunks
;
2727 if (trunks
!= s
->trunks
) {
2732 if (!list_is_short(&bundle
->ports
)) {
2733 bundle
->ofproto
->has_bonded_bundles
= true;
2735 if (bond_reconfigure(bundle
->bond
, s
->bond
)) {
2736 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2739 bundle
->bond
= bond_create(s
->bond
, ofproto
);
2740 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2743 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2744 bond_slave_register(bundle
->bond
, port
,
2745 port
->up
.ofp_port
, port
->up
.netdev
);
2748 bond_unref(bundle
->bond
);
2749 bundle
->bond
= NULL
;
2752 /* If we changed something that would affect MAC learning, un-learn
2753 * everything on this port and force flow revalidation. */
2755 bundle_flush_macs(bundle
, false);
2762 bundle_remove(struct ofport
*port_
)
2764 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
2765 struct ofbundle
*bundle
= port
->bundle
;
2768 bundle_del_port(port
);
2769 if (list_is_empty(&bundle
->ports
)) {
2770 bundle_destroy(bundle
);
2771 } else if (list_is_short(&bundle
->ports
)) {
2772 bond_unref(bundle
->bond
);
2773 bundle
->bond
= NULL
;
2779 send_pdu_cb(void *port_
, const void *pdu
, size_t pdu_size
)
2781 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 10);
2782 struct ofport_dpif
*port
= port_
;
2783 uint8_t ea
[ETH_ADDR_LEN
];
2786 error
= netdev_get_etheraddr(port
->up
.netdev
, ea
);
2788 struct ofpbuf packet
;
2791 ofpbuf_init(&packet
, 0);
2792 packet_pdu
= eth_compose(&packet
, eth_addr_lacp
, ea
, ETH_TYPE_LACP
,
2794 memcpy(packet_pdu
, pdu
, pdu_size
);
2796 ofproto_dpif_send_packet(port
, &packet
);
2797 ofpbuf_uninit(&packet
);
2799 VLOG_ERR_RL(&rl
, "port %s: cannot obtain Ethernet address of iface "
2800 "%s (%s)", port
->bundle
->name
,
2801 netdev_get_name(port
->up
.netdev
), ovs_strerror(error
));
2806 bundle_send_learning_packets(struct ofbundle
*bundle
)
2808 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2809 struct ofpbuf
*learning_packet
;
2810 int error
, n_packets
, n_errors
;
2811 struct mac_entry
*e
;
2812 struct list packets
;
2814 list_init(&packets
);
2815 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
2816 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
2817 if (e
->port
.p
!= bundle
) {
2820 learning_packet
= bond_compose_learning_packet(bundle
->bond
,
2823 /* Temporarily use 'frame' as a private pointer (see below). */
2824 ovs_assert(learning_packet
->frame
== ofpbuf_data(learning_packet
));
2825 learning_packet
->frame
= port_void
;
2826 list_push_back(&packets
, &learning_packet
->list_node
);
2829 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2831 error
= n_packets
= n_errors
= 0;
2832 LIST_FOR_EACH (learning_packet
, list_node
, &packets
) {
2834 void *port_void
= learning_packet
->frame
;
2836 /* Restore 'frame'. */
2837 learning_packet
->frame
= ofpbuf_data(learning_packet
);
2838 ret
= ofproto_dpif_send_packet(port_void
, learning_packet
);
2845 ofpbuf_list_delete(&packets
);
2848 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2849 VLOG_WARN_RL(&rl
, "bond %s: %d errors sending %d gratuitous learning "
2850 "packets, last error was: %s",
2851 bundle
->name
, n_errors
, n_packets
, ovs_strerror(error
));
2853 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
2854 bundle
->name
, n_packets
);
2859 bundle_run(struct ofbundle
*bundle
)
2862 lacp_run(bundle
->lacp
, send_pdu_cb
);
2865 struct ofport_dpif
*port
;
2867 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2868 bond_slave_set_may_enable(bundle
->bond
, port
, port
->may_enable
);
2871 if (bond_run(bundle
->bond
, lacp_status(bundle
->lacp
))) {
2872 bundle
->ofproto
->backer
->need_revalidate
= REV_BOND
;
2875 if (bond_should_send_learning_packets(bundle
->bond
)) {
2876 bundle_send_learning_packets(bundle
);
2882 bundle_wait(struct ofbundle
*bundle
)
2885 lacp_wait(bundle
->lacp
);
2888 bond_wait(bundle
->bond
);
2895 mirror_set__(struct ofproto
*ofproto_
, void *aux
,
2896 const struct ofproto_mirror_settings
*s
)
2898 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2899 struct ofbundle
**srcs
, **dsts
;
2904 mirror_destroy(ofproto
->mbridge
, aux
);
2908 srcs
= xmalloc(s
->n_srcs
* sizeof *srcs
);
2909 dsts
= xmalloc(s
->n_dsts
* sizeof *dsts
);
2911 for (i
= 0; i
< s
->n_srcs
; i
++) {
2912 srcs
[i
] = bundle_lookup(ofproto
, s
->srcs
[i
]);
2915 for (i
= 0; i
< s
->n_dsts
; i
++) {
2916 dsts
[i
] = bundle_lookup(ofproto
, s
->dsts
[i
]);
2919 error
= mirror_set(ofproto
->mbridge
, aux
, s
->name
, srcs
, s
->n_srcs
, dsts
,
2920 s
->n_dsts
, s
->src_vlans
,
2921 bundle_lookup(ofproto
, s
->out_bundle
), s
->out_vlan
);
2928 mirror_get_stats__(struct ofproto
*ofproto
, void *aux
,
2929 uint64_t *packets
, uint64_t *bytes
)
2931 return mirror_get_stats(ofproto_dpif_cast(ofproto
)->mbridge
, aux
, packets
,
2936 set_flood_vlans(struct ofproto
*ofproto_
, unsigned long *flood_vlans
)
2938 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2939 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2940 if (mac_learning_set_flood_vlans(ofproto
->ml
, flood_vlans
)) {
2941 mac_learning_flush(ofproto
->ml
);
2943 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2948 is_mirror_output_bundle(const struct ofproto
*ofproto_
, void *aux
)
2950 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2951 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
2952 return bundle
&& mirror_bundle_out(ofproto
->mbridge
, bundle
) != 0;
2956 forward_bpdu_changed(struct ofproto
*ofproto_
)
2958 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2959 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2963 set_mac_table_config(struct ofproto
*ofproto_
, unsigned int idle_time
,
2966 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2967 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2968 mac_learning_set_idle_time(ofproto
->ml
, idle_time
);
2969 mac_learning_set_max_entries(ofproto
->ml
, max_entries
);
2970 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2973 /* Configures multicast snooping on 'ofport' using the settings
2974 * defined in 's'. */
2976 set_mcast_snooping(struct ofproto
*ofproto_
,
2977 const struct ofproto_mcast_snooping_settings
*s
)
2979 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2981 /* Only revalidate flows if the configuration changed. */
2982 if (!s
!= !ofproto
->ms
) {
2983 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2988 ofproto
->ms
= mcast_snooping_create();
2991 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
2992 mcast_snooping_set_idle_time(ofproto
->ms
, s
->idle_time
);
2993 mcast_snooping_set_max_entries(ofproto
->ms
, s
->max_entries
);
2994 if (mcast_snooping_set_flood_unreg(ofproto
->ms
, s
->flood_unreg
)) {
2995 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2997 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
2999 mcast_snooping_unref(ofproto
->ms
);
3006 /* Configures multicast snooping port's flood setting on 'ofproto'. */
3008 set_mcast_snooping_port(struct ofproto
*ofproto_
, void *aux
, bool flood
)
3010 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3011 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3014 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3015 mcast_snooping_set_port_flood(ofproto
->ms
, bundle
->vlan
, bundle
,
3017 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3025 static struct ofport_dpif
*
3026 get_ofp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
3028 struct ofport
*ofport
= ofproto_get_port(&ofproto
->up
, ofp_port
);
3029 return ofport
? ofport_dpif_cast(ofport
) : NULL
;
3033 ofproto_port_from_dpif_port(struct ofproto_dpif
*ofproto
,
3034 struct ofproto_port
*ofproto_port
,
3035 struct dpif_port
*dpif_port
)
3037 ofproto_port
->name
= dpif_port
->name
;
3038 ofproto_port
->type
= dpif_port
->type
;
3039 ofproto_port
->ofp_port
= odp_port_to_ofp_port(ofproto
, dpif_port
->port_no
);
3043 ofport_update_peer(struct ofport_dpif
*ofport
)
3045 const struct ofproto_dpif
*ofproto
;
3046 struct dpif_backer
*backer
;
3049 if (!netdev_vport_is_patch(ofport
->up
.netdev
)) {
3053 backer
= ofproto_dpif_cast(ofport
->up
.ofproto
)->backer
;
3054 backer
->need_revalidate
= REV_RECONFIGURE
;
3057 ofport
->peer
->peer
= NULL
;
3058 ofport
->peer
= NULL
;
3061 peer_name
= netdev_vport_patch_peer(ofport
->up
.netdev
);
3066 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
3067 struct ofport
*peer_ofport
;
3068 struct ofport_dpif
*peer
;
3071 if (ofproto
->backer
!= backer
) {
3075 peer_ofport
= shash_find_data(&ofproto
->up
.port_by_name
, peer_name
);
3080 peer
= ofport_dpif_cast(peer_ofport
);
3081 peer_peer
= netdev_vport_patch_peer(peer
->up
.netdev
);
3082 if (peer_peer
&& !strcmp(netdev_get_name(ofport
->up
.netdev
),
3084 ofport
->peer
= peer
;
3085 ofport
->peer
->peer
= ofport
;
3095 port_run(struct ofport_dpif
*ofport
)
3097 long long int carrier_seq
= netdev_get_carrier_resets(ofport
->up
.netdev
);
3098 bool carrier_changed
= carrier_seq
!= ofport
->carrier_seq
;
3099 bool enable
= netdev_get_carrier(ofport
->up
.netdev
);
3100 bool cfm_enable
= false;
3101 bool bfd_enable
= false;
3103 ofport
->carrier_seq
= carrier_seq
;
3106 int cfm_opup
= cfm_get_opup(ofport
->cfm
);
3108 cfm_enable
= !cfm_get_fault(ofport
->cfm
);
3110 if (cfm_opup
>= 0) {
3111 cfm_enable
= cfm_enable
&& cfm_opup
;
3116 bfd_enable
= bfd_forwarding(ofport
->bfd
);
3119 if (ofport
->bfd
|| ofport
->cfm
) {
3120 enable
= enable
&& (cfm_enable
|| bfd_enable
);
3123 if (ofport
->bundle
) {
3124 enable
= enable
&& lacp_slave_may_enable(ofport
->bundle
->lacp
, ofport
);
3125 if (carrier_changed
) {
3126 lacp_slave_carrier_changed(ofport
->bundle
->lacp
, ofport
);
3130 if (ofport
->may_enable
!= enable
) {
3131 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3132 ofproto
->backer
->need_revalidate
= REV_PORT_TOGGLED
;
3135 ofport
->may_enable
= enable
;
3137 if (ofport
->rstp_port
) {
3138 if (rstp_port_get_mac_operational(ofport
->rstp_port
) != enable
) {
3139 rstp_port_set_mac_operational(ofport
->rstp_port
, enable
);
3145 port_query_by_name(const struct ofproto
*ofproto_
, const char *devname
,
3146 struct ofproto_port
*ofproto_port
)
3148 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3149 struct dpif_port dpif_port
;
3152 if (sset_contains(&ofproto
->ghost_ports
, devname
)) {
3153 const char *type
= netdev_get_type_from_name(devname
);
3155 /* We may be called before ofproto->up.port_by_name is populated with
3156 * the appropriate ofport. For this reason, we must get the name and
3157 * type from the netdev layer directly. */
3159 const struct ofport
*ofport
;
3161 ofport
= shash_find_data(&ofproto
->up
.port_by_name
, devname
);
3162 ofproto_port
->ofp_port
= ofport
? ofport
->ofp_port
: OFPP_NONE
;
3163 ofproto_port
->name
= xstrdup(devname
);
3164 ofproto_port
->type
= xstrdup(type
);
3170 if (!sset_contains(&ofproto
->ports
, devname
)) {
3173 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
3174 devname
, &dpif_port
);
3176 ofproto_port_from_dpif_port(ofproto
, ofproto_port
, &dpif_port
);
3182 port_add(struct ofproto
*ofproto_
, struct netdev
*netdev
)
3184 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3185 const char *devname
= netdev_get_name(netdev
);
3186 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
3187 const char *dp_port_name
;
3189 if (netdev_vport_is_patch(netdev
)) {
3190 sset_add(&ofproto
->ghost_ports
, netdev_get_name(netdev
));
3194 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
3195 if (!dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
3196 odp_port_t port_no
= ODPP_NONE
;
3199 error
= dpif_port_add(ofproto
->backer
->dpif
, netdev
, &port_no
);
3203 if (netdev_get_tunnel_config(netdev
)) {
3204 simap_put(&ofproto
->backer
->tnl_backers
,
3205 dp_port_name
, odp_to_u32(port_no
));
3209 if (netdev_get_tunnel_config(netdev
)) {
3210 sset_add(&ofproto
->ghost_ports
, devname
);
3212 sset_add(&ofproto
->ports
, devname
);
3218 port_del(struct ofproto
*ofproto_
, ofp_port_t ofp_port
)
3220 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3221 struct ofport_dpif
*ofport
= get_ofp_port(ofproto
, ofp_port
);
3228 sset_find_and_delete(&ofproto
->ghost_ports
,
3229 netdev_get_name(ofport
->up
.netdev
));
3230 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3231 if (!ofport
->is_tunnel
&& !netdev_vport_is_patch(ofport
->up
.netdev
)) {
3232 error
= dpif_port_del(ofproto
->backer
->dpif
, ofport
->odp_port
);
3234 /* The caller is going to close ofport->up.netdev. If this is a
3235 * bonded port, then the bond is using that netdev, so remove it
3236 * from the bond. The client will need to reconfigure everything
3237 * after deleting ports, so then the slave will get re-added. */
3238 bundle_remove(&ofport
->up
);
3245 port_get_stats(const struct ofport
*ofport_
, struct netdev_stats
*stats
)
3247 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3250 error
= netdev_get_stats(ofport
->up
.netdev
, stats
);
3252 if (!error
&& ofport_
->ofp_port
== OFPP_LOCAL
) {
3253 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3255 ovs_mutex_lock(&ofproto
->stats_mutex
);
3256 /* ofproto->stats.tx_packets represents packets that we created
3257 * internally and sent to some port (e.g. packets sent with
3258 * ofproto_dpif_send_packet()). Account for them as if they had
3259 * come from OFPP_LOCAL and got forwarded. */
3261 if (stats
->rx_packets
!= UINT64_MAX
) {
3262 stats
->rx_packets
+= ofproto
->stats
.tx_packets
;
3265 if (stats
->rx_bytes
!= UINT64_MAX
) {
3266 stats
->rx_bytes
+= ofproto
->stats
.tx_bytes
;
3269 /* ofproto->stats.rx_packets represents packets that were received on
3270 * some port and we processed internally and dropped (e.g. STP).
3271 * Account for them as if they had been forwarded to OFPP_LOCAL. */
3273 if (stats
->tx_packets
!= UINT64_MAX
) {
3274 stats
->tx_packets
+= ofproto
->stats
.rx_packets
;
3277 if (stats
->tx_bytes
!= UINT64_MAX
) {
3278 stats
->tx_bytes
+= ofproto
->stats
.rx_bytes
;
3280 ovs_mutex_unlock(&ofproto
->stats_mutex
);
3286 struct port_dump_state
{
3291 struct ofproto_port port
;
3296 port_dump_start(const struct ofproto
*ofproto_ OVS_UNUSED
, void **statep
)
3298 *statep
= xzalloc(sizeof(struct port_dump_state
));
3303 port_dump_next(const struct ofproto
*ofproto_
, void *state_
,
3304 struct ofproto_port
*port
)
3306 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3307 struct port_dump_state
*state
= state_
;
3308 const struct sset
*sset
;
3309 struct sset_node
*node
;
3311 if (state
->has_port
) {
3312 ofproto_port_destroy(&state
->port
);
3313 state
->has_port
= false;
3315 sset
= state
->ghost
? &ofproto
->ghost_ports
: &ofproto
->ports
;
3316 while ((node
= sset_at_position(sset
, &state
->bucket
, &state
->offset
))) {
3319 error
= port_query_by_name(ofproto_
, node
->name
, &state
->port
);
3321 *port
= state
->port
;
3322 state
->has_port
= true;
3324 } else if (error
!= ENODEV
) {
3329 if (!state
->ghost
) {
3330 state
->ghost
= true;
3333 return port_dump_next(ofproto_
, state_
, port
);
3340 port_dump_done(const struct ofproto
*ofproto_ OVS_UNUSED
, void *state_
)
3342 struct port_dump_state
*state
= state_
;
3344 if (state
->has_port
) {
3345 ofproto_port_destroy(&state
->port
);
3352 port_poll(const struct ofproto
*ofproto_
, char **devnamep
)
3354 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3356 if (ofproto
->port_poll_errno
) {
3357 int error
= ofproto
->port_poll_errno
;
3358 ofproto
->port_poll_errno
= 0;
3362 if (sset_is_empty(&ofproto
->port_poll_set
)) {
3366 *devnamep
= sset_pop(&ofproto
->port_poll_set
);
3371 port_poll_wait(const struct ofproto
*ofproto_
)
3373 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3374 dpif_port_poll_wait(ofproto
->backer
->dpif
);
3378 port_is_lacp_current(const struct ofport
*ofport_
)
3380 const struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3381 return (ofport
->bundle
&& ofport
->bundle
->lacp
3382 ? lacp_slave_is_current(ofport
->bundle
->lacp
, ofport
)
3386 /* If 'rule' is an OpenFlow rule, that has expired according to OpenFlow rules,
3387 * then delete it entirely. */
3389 rule_expire(struct rule_dpif
*rule
)
3390 OVS_REQUIRES(ofproto_mutex
)
3392 uint16_t hard_timeout
, idle_timeout
;
3393 long long int now
= time_msec();
3396 hard_timeout
= rule
->up
.hard_timeout
;
3397 idle_timeout
= rule
->up
.idle_timeout
;
3399 /* Has 'rule' expired? */
3401 long long int modified
;
3403 ovs_mutex_lock(&rule
->up
.mutex
);
3404 modified
= rule
->up
.modified
;
3405 ovs_mutex_unlock(&rule
->up
.mutex
);
3407 if (now
> modified
+ hard_timeout
* 1000) {
3408 reason
= OFPRR_HARD_TIMEOUT
;
3412 if (reason
< 0 && idle_timeout
) {
3415 ovs_mutex_lock(&rule
->stats_mutex
);
3416 used
= rule
->stats
.used
;
3417 ovs_mutex_unlock(&rule
->stats_mutex
);
3419 if (now
> used
+ idle_timeout
* 1000) {
3420 reason
= OFPRR_IDLE_TIMEOUT
;
3425 COVERAGE_INC(ofproto_dpif_expired
);
3426 ofproto_rule_expire(&rule
->up
, reason
);
3430 /* Executes, within 'ofproto', the actions in 'rule' or 'ofpacts' on 'packet'.
3431 * 'flow' must reflect the data in 'packet'. */
3433 ofproto_dpif_execute_actions(struct ofproto_dpif
*ofproto
,
3434 const struct flow
*flow
,
3435 struct rule_dpif
*rule
,
3436 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3437 struct ofpbuf
*packet
)
3439 struct dpif_flow_stats stats
;
3440 struct xlate_out xout
;
3441 struct xlate_in xin
;
3443 struct dpif_execute execute
;
3446 ovs_assert((rule
!= NULL
) != (ofpacts
!= NULL
));
3448 dpif_flow_stats_extract(flow
, packet
, time_msec(), &stats
);
3451 rule_dpif_credit_stats(rule
, &stats
);
3454 xlate_in_init(&xin
, ofproto
, flow
, flow
->in_port
.ofp_port
, rule
,
3455 stats
.tcp_flags
, packet
);
3456 xin
.ofpacts
= ofpacts
;
3457 xin
.ofpacts_len
= ofpacts_len
;
3458 xin
.resubmit_stats
= &stats
;
3459 xlate_actions(&xin
, &xout
);
3461 execute
.actions
= ofpbuf_data(xout
.odp_actions
);
3462 execute
.actions_len
= ofpbuf_size(xout
.odp_actions
);
3463 execute
.packet
= packet
;
3464 execute
.md
= pkt_metadata_from_flow(flow
);
3465 execute
.needs_help
= (xout
.slow
& SLOW_ACTION
) != 0;
3467 /* Fix up in_port. */
3468 in_port
= flow
->in_port
.ofp_port
;
3469 if (in_port
== OFPP_NONE
) {
3470 in_port
= OFPP_LOCAL
;
3472 execute
.md
.in_port
.odp_port
= ofp_port_to_odp_port(ofproto
, in_port
);
3474 error
= dpif_execute(ofproto
->backer
->dpif
, &execute
);
3476 xlate_out_uninit(&xout
);
3482 rule_dpif_credit_stats(struct rule_dpif
*rule
,
3483 const struct dpif_flow_stats
*stats
)
3485 ovs_mutex_lock(&rule
->stats_mutex
);
3486 rule
->stats
.n_packets
+= stats
->n_packets
;
3487 rule
->stats
.n_bytes
+= stats
->n_bytes
;
3488 rule
->stats
.used
= MAX(rule
->stats
.used
, stats
->used
);
3489 ovs_mutex_unlock(&rule
->stats_mutex
);
3493 rule_dpif_get_flow_cookie(const struct rule_dpif
*rule
)
3494 OVS_REQUIRES(rule
->up
.mutex
)
3496 return rule
->up
.flow_cookie
;
3500 rule_dpif_reduce_timeouts(struct rule_dpif
*rule
, uint16_t idle_timeout
,
3501 uint16_t hard_timeout
)
3503 ofproto_rule_reduce_timeouts(&rule
->up
, idle_timeout
, hard_timeout
);
3506 /* Returns 'rule''s actions. The caller owns a reference on the returned
3507 * actions and must eventually release it (with rule_actions_unref()) to avoid
3509 const struct rule_actions
*
3510 rule_dpif_get_actions(const struct rule_dpif
*rule
)
3512 return rule_get_actions(&rule
->up
);
3515 /* Sets 'rule''s recirculation id. */
3517 rule_dpif_set_recirc_id(struct rule_dpif
*rule
, uint32_t id
)
3518 OVS_REQUIRES(rule
->up
.mutex
)
3520 ovs_assert(!rule
->recirc_id
);
3521 rule
->recirc_id
= id
;
3524 /* Returns 'rule''s recirculation id. */
3526 rule_dpif_get_recirc_id(struct rule_dpif
*rule
)
3527 OVS_REQUIRES(rule
->up
.mutex
)
3529 if (!rule
->recirc_id
) {
3530 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3532 rule_dpif_set_recirc_id(rule
, ofproto_dpif_alloc_recirc_id(ofproto
));
3534 return rule
->recirc_id
;
3537 /* Sets 'rule''s recirculation id. */
3539 rule_set_recirc_id(struct rule
*rule_
, uint32_t id
)
3541 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3543 ovs_mutex_lock(&rule
->up
.mutex
);
3544 rule_dpif_set_recirc_id(rule
, id
);
3545 ovs_mutex_unlock(&rule
->up
.mutex
);
3548 /* Lookup 'flow' in table 0 of 'ofproto''s classifier.
3549 * If 'wc' is non-null, sets the fields that were relevant as part of
3550 * the lookup. Returns the table_id where a match or miss occurred.
3552 * The return value will be zero unless there was a miss and
3553 * OFPTC11_TABLE_MISS_CONTINUE is in effect for the sequence of tables
3554 * where misses occur.
3556 * The rule is returned in '*rule', which is valid at least until the next
3557 * RCU quiescent period. If the '*rule' needs to stay around longer,
3558 * a non-zero 'take_ref' must be passed in to cause a reference to be taken
3559 * on it before this returns. */
3561 rule_dpif_lookup(struct ofproto_dpif
*ofproto
, struct flow
*flow
,
3562 struct flow_wildcards
*wc
, struct rule_dpif
**rule
,
3563 bool take_ref
, const struct dpif_flow_stats
*stats
)
3565 enum rule_dpif_lookup_verdict verdict
;
3566 enum ofputil_port_config config
= 0;
3569 if (ofproto_dpif_get_enable_recirc(ofproto
)) {
3570 /* Always exactly match recirc_id since datapath supports
3573 wc
->masks
.recirc_id
= UINT32_MAX
;
3576 /* Start looking up from internal table for post recirculation flows
3577 * or packets. We can also simply send all, including normal flows
3578 * or packets to the internal table. They will not match any post
3579 * recirculation rules except the 'catch all' rule that resubmit
3582 * As an optimization, we send normal flows and packets to table 0
3583 * directly, saving one table lookup. */
3584 table_id
= flow
->recirc_id
? TBL_INTERNAL
: 0;
3589 verdict
= rule_dpif_lookup_from_table(ofproto
, flow
, wc
, true,
3590 &table_id
, rule
, take_ref
, stats
);
3593 case RULE_DPIF_LOOKUP_VERDICT_MATCH
:
3595 case RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
: {
3596 struct ofport_dpif
*port
;
3598 port
= get_ofp_port(ofproto
, flow
->in_port
.ofp_port
);
3600 VLOG_WARN_RL(&rl
, "packet-in on unknown OpenFlow port %"PRIu16
,
3601 flow
->in_port
.ofp_port
);
3603 config
= port
? port
->up
.pp
.config
: 0;
3606 case RULE_DPIF_LOOKUP_VERDICT_DROP
:
3607 config
= OFPUTIL_PC_NO_PACKET_IN
;
3609 case RULE_DPIF_LOOKUP_VERDICT_DEFAULT
:
3610 if (!connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
)) {
3611 config
= OFPUTIL_PC_NO_PACKET_IN
;
3618 choose_miss_rule(config
, ofproto
->miss_rule
,
3619 ofproto
->no_packet_in_rule
, rule
, take_ref
);
3623 /* The returned rule is valid at least until the next RCU quiescent period.
3624 * If the '*rule' needs to stay around longer, a non-zero 'take_ref' must be
3625 * passed in to cause a reference to be taken on it before this returns. */
3626 static struct rule_dpif
*
3627 rule_dpif_lookup_in_table(struct ofproto_dpif
*ofproto
, uint8_t table_id
,
3628 const struct flow
*flow
, struct flow_wildcards
*wc
,
3631 struct classifier
*cls
= &ofproto
->up
.tables
[table_id
].cls
;
3632 const struct cls_rule
*cls_rule
;
3633 struct rule_dpif
*rule
;
3634 struct flow ofpc_normal_flow
;
3636 if (ofproto
->up
.frag_handling
!= OFPC_FRAG_NX_MATCH
) {
3637 /* We always unwildcard dl_type and nw_frag (for IP), so they
3638 * need not be unwildcarded here. */
3640 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
) {
3641 if (ofproto
->up
.frag_handling
== OFPC_FRAG_NORMAL
) {
3642 /* We must pretend that transport ports are unavailable. */
3643 ofpc_normal_flow
= *flow
;
3644 ofpc_normal_flow
.tp_src
= htons(0);
3645 ofpc_normal_flow
.tp_dst
= htons(0);
3646 flow
= &ofpc_normal_flow
;
3648 /* Must be OFPC_FRAG_DROP (we don't have OFPC_FRAG_REASM).
3649 * Use the drop_frags_rule (which cannot disappear). */
3650 cls_rule
= &ofproto
->drop_frags_rule
->up
.cr
;
3651 rule
= rule_dpif_cast(rule_from_cls_rule(cls_rule
));
3653 rule_dpif_ref(rule
);
3661 cls_rule
= classifier_lookup(cls
, flow
, wc
);
3663 rule
= rule_dpif_cast(rule_from_cls_rule(cls_rule
));
3665 /* Try again if the rule was released before we get the reference. */
3666 } while (rule
&& take_ref
&& !rule_dpif_try_ref(rule
));
3671 /* Look up 'flow' in 'ofproto''s classifier starting from table '*table_id'.
3672 * Stores the rule that was found in '*rule', or NULL if none was found.
3673 * Updates 'wc', if nonnull, to reflect the fields that were used during the
3676 * If 'honor_table_miss' is true, the first lookup occurs in '*table_id', but
3677 * if none is found then the table miss configuration for that table is
3678 * honored, which can result in additional lookups in other OpenFlow tables.
3679 * In this case the function updates '*table_id' to reflect the final OpenFlow
3680 * table that was searched.
3682 * If 'honor_table_miss' is false, then only one table lookup occurs, in
3687 * - RULE_DPIF_LOOKUP_VERDICT_MATCH if a rule (in '*rule') was found.
3689 * - RULE_OFPTC_TABLE_MISS_CONTROLLER if no rule was found and either:
3690 * + 'honor_table_miss' is false
3691 * + a table miss configuration specified that the packet should be
3692 * sent to the controller in this case.
3694 * - RULE_DPIF_LOOKUP_VERDICT_DROP if no rule was found, 'honor_table_miss'
3695 * is true and a table miss configuration specified that the packet
3696 * should be dropped in this case.
3698 * - RULE_DPIF_LOOKUP_VERDICT_DEFAULT if no rule was found,
3699 * 'honor_table_miss' is true and a table miss configuration has
3700 * not been specified in this case.
3702 * The rule is returned in '*rule', which is valid at least until the next
3703 * RCU quiescent period. If the '*rule' needs to stay around longer,
3704 * a non-zero 'take_ref' must be passed in to cause a reference to be taken
3705 * on it before this returns. */
3706 enum rule_dpif_lookup_verdict
3707 rule_dpif_lookup_from_table(struct ofproto_dpif
*ofproto
,
3708 const struct flow
*flow
,
3709 struct flow_wildcards
*wc
,
3710 bool honor_table_miss
,
3711 uint8_t *table_id
, struct rule_dpif
**rule
,
3712 bool take_ref
, const struct dpif_flow_stats
*stats
)
3716 for (next_id
= *table_id
;
3717 next_id
< ofproto
->up
.n_tables
;
3718 next_id
++, next_id
+= (next_id
== TBL_INTERNAL
))
3720 *table_id
= next_id
;
3721 *rule
= rule_dpif_lookup_in_table(ofproto
, *table_id
, flow
, wc
,
3724 struct oftable
*tbl
= &ofproto
->up
.tables
[next_id
];
3727 atomic_add_relaxed(*rule
? &tbl
->n_matched
: &tbl
->n_missed
,
3728 stats
->n_packets
, &orig
);
3731 return RULE_DPIF_LOOKUP_VERDICT_MATCH
;
3732 } else if (!honor_table_miss
) {
3733 return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
;
3735 switch (ofproto_table_get_miss_config(&ofproto
->up
, *table_id
)) {
3736 case OFPUTIL_TABLE_MISS_CONTINUE
:
3739 case OFPUTIL_TABLE_MISS_CONTROLLER
:
3740 return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
;
3742 case OFPUTIL_TABLE_MISS_DROP
:
3743 return RULE_DPIF_LOOKUP_VERDICT_DROP
;
3745 case OFPUTIL_TABLE_MISS_DEFAULT
:
3746 return RULE_DPIF_LOOKUP_VERDICT_DEFAULT
;
3751 return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
;
3754 /* Given a port configuration (specified as zero if there's no port), chooses
3755 * which of 'miss_rule' and 'no_packet_in_rule' should be used in case of a
3758 * The rule is returned in '*rule', which is valid at least until the next
3759 * RCU quiescent period. If the '*rule' needs to stay around longer,
3760 * a reference must be taken on it (rule_dpif_ref()).
3763 choose_miss_rule(enum ofputil_port_config config
, struct rule_dpif
*miss_rule
,
3764 struct rule_dpif
*no_packet_in_rule
, struct rule_dpif
**rule
,
3767 *rule
= config
& OFPUTIL_PC_NO_PACKET_IN
? no_packet_in_rule
: miss_rule
;
3769 rule_dpif_ref(*rule
);
3774 complete_operation(struct rule_dpif
*rule
)
3775 OVS_REQUIRES(ofproto_mutex
)
3777 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3779 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
3782 static struct rule_dpif
*rule_dpif_cast(const struct rule
*rule
)
3784 return rule
? CONTAINER_OF(rule
, struct rule_dpif
, up
) : NULL
;
3787 static struct rule
*
3790 struct rule_dpif
*rule
= xmalloc(sizeof *rule
);
3795 rule_dealloc(struct rule
*rule_
)
3797 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3802 rule_construct(struct rule
*rule_
)
3803 OVS_NO_THREAD_SAFETY_ANALYSIS
3805 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3806 ovs_mutex_init_adaptive(&rule
->stats_mutex
);
3807 rule
->stats
.n_packets
= 0;
3808 rule
->stats
.n_bytes
= 0;
3809 rule
->stats
.used
= rule
->up
.modified
;
3810 rule
->recirc_id
= 0;
3816 rule_insert(struct rule
*rule_
)
3817 OVS_REQUIRES(ofproto_mutex
)
3819 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3820 complete_operation(rule
);
3825 rule_delete(struct rule
*rule_
)
3826 OVS_REQUIRES(ofproto_mutex
)
3828 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3829 complete_operation(rule
);
3833 rule_destruct(struct rule
*rule_
)
3835 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3837 ovs_mutex_destroy(&rule
->stats_mutex
);
3838 if (rule
->recirc_id
) {
3839 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3841 ofproto_dpif_free_recirc_id(ofproto
, rule
->recirc_id
);
3846 rule_get_stats(struct rule
*rule_
, uint64_t *packets
, uint64_t *bytes
,
3847 long long int *used
)
3849 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3851 ovs_mutex_lock(&rule
->stats_mutex
);
3852 *packets
= rule
->stats
.n_packets
;
3853 *bytes
= rule
->stats
.n_bytes
;
3854 *used
= rule
->stats
.used
;
3855 ovs_mutex_unlock(&rule
->stats_mutex
);
3859 rule_dpif_execute(struct rule_dpif
*rule
, const struct flow
*flow
,
3860 struct ofpbuf
*packet
)
3862 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3864 ofproto_dpif_execute_actions(ofproto
, flow
, rule
, NULL
, 0, packet
);
3868 rule_execute(struct rule
*rule
, const struct flow
*flow
,
3869 struct ofpbuf
*packet
)
3871 rule_dpif_execute(rule_dpif_cast(rule
), flow
, packet
);
3872 ofpbuf_delete(packet
);
3877 rule_modify_actions(struct rule
*rule_
, bool reset_counters
)
3878 OVS_REQUIRES(ofproto_mutex
)
3880 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3882 if (reset_counters
) {
3883 ovs_mutex_lock(&rule
->stats_mutex
);
3884 rule
->stats
.n_packets
= 0;
3885 rule
->stats
.n_bytes
= 0;
3886 ovs_mutex_unlock(&rule
->stats_mutex
);
3889 complete_operation(rule
);
3892 static struct group_dpif
*group_dpif_cast(const struct ofgroup
*group
)
3894 return group
? CONTAINER_OF(group
, struct group_dpif
, up
) : NULL
;
3897 static struct ofgroup
*
3900 struct group_dpif
*group
= xzalloc(sizeof *group
);
3905 group_dealloc(struct ofgroup
*group_
)
3907 struct group_dpif
*group
= group_dpif_cast(group_
);
3912 group_construct_stats(struct group_dpif
*group
)
3913 OVS_REQUIRES(group
->stats_mutex
)
3915 struct ofputil_bucket
*bucket
;
3916 const struct list
*buckets
;
3918 group
->packet_count
= 0;
3919 group
->byte_count
= 0;
3921 group_dpif_get_buckets(group
, &buckets
);
3922 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
3923 bucket
->stats
.packet_count
= 0;
3924 bucket
->stats
.byte_count
= 0;
3929 group_dpif_credit_stats(struct group_dpif
*group
,
3930 struct ofputil_bucket
*bucket
,
3931 const struct dpif_flow_stats
*stats
)
3933 ovs_mutex_lock(&group
->stats_mutex
);
3934 group
->packet_count
+= stats
->n_packets
;
3935 group
->byte_count
+= stats
->n_bytes
;
3937 bucket
->stats
.packet_count
+= stats
->n_packets
;
3938 bucket
->stats
.byte_count
+= stats
->n_bytes
;
3939 } else { /* Credit to all buckets */
3940 const struct list
*buckets
;
3942 group_dpif_get_buckets(group
, &buckets
);
3943 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
3944 bucket
->stats
.packet_count
+= stats
->n_packets
;
3945 bucket
->stats
.byte_count
+= stats
->n_bytes
;
3948 ovs_mutex_unlock(&group
->stats_mutex
);
3952 group_construct(struct ofgroup
*group_
)
3954 struct group_dpif
*group
= group_dpif_cast(group_
);
3955 const struct ofputil_bucket
*bucket
;
3957 /* Prevent group chaining because our locking structure makes it hard to
3958 * implement deadlock-free. (See xlate_group_resource_check().) */
3959 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
3960 const struct ofpact
*a
;
3962 OFPACT_FOR_EACH (a
, bucket
->ofpacts
, bucket
->ofpacts_len
) {
3963 if (a
->type
== OFPACT_GROUP
) {
3964 return OFPERR_OFPGMFC_CHAINING_UNSUPPORTED
;
3969 ovs_mutex_init_adaptive(&group
->stats_mutex
);
3970 ovs_mutex_lock(&group
->stats_mutex
);
3971 group_construct_stats(group
);
3972 ovs_mutex_unlock(&group
->stats_mutex
);
3977 group_destruct(struct ofgroup
*group_
)
3979 struct group_dpif
*group
= group_dpif_cast(group_
);
3980 ovs_mutex_destroy(&group
->stats_mutex
);
3984 group_modify(struct ofgroup
*group_
)
3986 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(group_
->ofproto
);
3988 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
3994 group_get_stats(const struct ofgroup
*group_
, struct ofputil_group_stats
*ogs
)
3996 struct group_dpif
*group
= group_dpif_cast(group_
);
3997 struct ofputil_bucket
*bucket
;
3998 const struct list
*buckets
;
3999 struct bucket_counter
*bucket_stats
;
4001 ovs_mutex_lock(&group
->stats_mutex
);
4002 ogs
->packet_count
= group
->packet_count
;
4003 ogs
->byte_count
= group
->byte_count
;
4005 group_dpif_get_buckets(group
, &buckets
);
4006 bucket_stats
= ogs
->bucket_stats
;
4007 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
4008 bucket_stats
->packet_count
= bucket
->stats
.packet_count
;
4009 bucket_stats
->byte_count
= bucket
->stats
.byte_count
;
4012 ovs_mutex_unlock(&group
->stats_mutex
);
4017 /* If the group exists, this function increments the groups's reference count.
4019 * Make sure to call group_dpif_unref() after no longer needing to maintain
4020 * a reference to the group. */
4022 group_dpif_lookup(struct ofproto_dpif
*ofproto
, uint32_t group_id
,
4023 struct group_dpif
**group
)
4025 struct ofgroup
*ofgroup
;
4028 found
= ofproto_group_lookup(&ofproto
->up
, group_id
, &ofgroup
);
4029 *group
= found
? group_dpif_cast(ofgroup
) : NULL
;
4035 group_dpif_get_buckets(const struct group_dpif
*group
,
4036 const struct list
**buckets
)
4038 *buckets
= &group
->up
.buckets
;
4041 enum ofp11_group_type
4042 group_dpif_get_type(const struct group_dpif
*group
)
4044 return group
->up
.type
;
4047 /* Sends 'packet' out 'ofport'.
4048 * May modify 'packet'.
4049 * Returns 0 if successful, otherwise a positive errno value. */
4051 ofproto_dpif_send_packet(const struct ofport_dpif
*ofport
, struct ofpbuf
*packet
)
4053 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
4056 error
= xlate_send_packet(ofport
, packet
);
4058 ovs_mutex_lock(&ofproto
->stats_mutex
);
4059 ofproto
->stats
.tx_packets
++;
4060 ofproto
->stats
.tx_bytes
+= ofpbuf_size(packet
);
4061 ovs_mutex_unlock(&ofproto
->stats_mutex
);
4066 set_frag_handling(struct ofproto
*ofproto_
,
4067 enum ofp_config_flags frag_handling
)
4069 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4070 if (frag_handling
!= OFPC_FRAG_REASM
) {
4071 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4079 packet_out(struct ofproto
*ofproto_
, struct ofpbuf
*packet
,
4080 const struct flow
*flow
,
4081 const struct ofpact
*ofpacts
, size_t ofpacts_len
)
4083 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4085 ofproto_dpif_execute_actions(ofproto
, flow
, NULL
, ofpacts
,
4086 ofpacts_len
, packet
);
4093 set_netflow(struct ofproto
*ofproto_
,
4094 const struct netflow_options
*netflow_options
)
4096 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4098 if (netflow_options
) {
4099 if (!ofproto
->netflow
) {
4100 ofproto
->netflow
= netflow_create();
4101 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4103 return netflow_set_options(ofproto
->netflow
, netflow_options
);
4104 } else if (ofproto
->netflow
) {
4105 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4106 netflow_unref(ofproto
->netflow
);
4107 ofproto
->netflow
= NULL
;
4114 get_netflow_ids(const struct ofproto
*ofproto_
,
4115 uint8_t *engine_type
, uint8_t *engine_id
)
4117 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4119 dpif_get_netflow_ids(ofproto
->backer
->dpif
, engine_type
, engine_id
);
4122 static struct ofproto_dpif
*
4123 ofproto_dpif_lookup(const char *name
)
4125 struct ofproto_dpif
*ofproto
;
4127 HMAP_FOR_EACH_WITH_HASH (ofproto
, all_ofproto_dpifs_node
,
4128 hash_string(name
, 0), &all_ofproto_dpifs
) {
4129 if (!strcmp(ofproto
->up
.name
, name
)) {
4137 ofproto_unixctl_fdb_flush(struct unixctl_conn
*conn
, int argc
,
4138 const char *argv
[], void *aux OVS_UNUSED
)
4140 struct ofproto_dpif
*ofproto
;
4143 ofproto
= ofproto_dpif_lookup(argv
[1]);
4145 unixctl_command_reply_error(conn
, "no such bridge");
4148 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4149 mac_learning_flush(ofproto
->ml
);
4150 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4152 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4153 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4154 mac_learning_flush(ofproto
->ml
);
4155 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4159 unixctl_command_reply(conn
, "table successfully flushed");
4163 ofproto_unixctl_mcast_snooping_flush(struct unixctl_conn
*conn
, int argc
,
4164 const char *argv
[], void *aux OVS_UNUSED
)
4166 struct ofproto_dpif
*ofproto
;
4169 ofproto
= ofproto_dpif_lookup(argv
[1]);
4171 unixctl_command_reply_error(conn
, "no such bridge");
4175 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4176 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4179 mcast_snooping_mdb_flush(ofproto
->ms
);
4181 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4182 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4185 mcast_snooping_mdb_flush(ofproto
->ms
);
4189 unixctl_command_reply(conn
, "table successfully flushed");
4192 static struct ofport_dpif
*
4193 ofbundle_get_a_port(const struct ofbundle
*bundle
)
4195 return CONTAINER_OF(list_front(&bundle
->ports
), struct ofport_dpif
,
4200 ofproto_unixctl_fdb_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4201 const char *argv
[], void *aux OVS_UNUSED
)
4203 struct ds ds
= DS_EMPTY_INITIALIZER
;
4204 const struct ofproto_dpif
*ofproto
;
4205 const struct mac_entry
*e
;
4207 ofproto
= ofproto_dpif_lookup(argv
[1]);
4209 unixctl_command_reply_error(conn
, "no such bridge");
4213 ds_put_cstr(&ds
, " port VLAN MAC Age\n");
4214 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
4215 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
4216 struct ofbundle
*bundle
= e
->port
.p
;
4217 char name
[OFP_MAX_PORT_NAME_LEN
];
4219 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4221 ds_put_format(&ds
, "%5s %4d "ETH_ADDR_FMT
" %3d\n",
4222 name
, e
->vlan
, ETH_ADDR_ARGS(e
->mac
),
4223 mac_entry_age(ofproto
->ml
, e
));
4225 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4226 unixctl_command_reply(conn
, ds_cstr(&ds
));
4231 ofproto_unixctl_mcast_snooping_show(struct unixctl_conn
*conn
,
4232 int argc OVS_UNUSED
,
4234 void *aux OVS_UNUSED
)
4236 struct ds ds
= DS_EMPTY_INITIALIZER
;
4237 const struct ofproto_dpif
*ofproto
;
4238 const struct ofbundle
*bundle
;
4239 const struct mcast_group
*grp
;
4240 struct mcast_group_bundle
*b
;
4241 struct mcast_mrouter_bundle
*mrouter
;
4243 ofproto
= ofproto_dpif_lookup(argv
[1]);
4245 unixctl_command_reply_error(conn
, "no such bridge");
4249 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4250 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4254 ds_put_cstr(&ds
, " port VLAN GROUP Age\n");
4255 ovs_rwlock_rdlock(&ofproto
->ms
->rwlock
);
4256 LIST_FOR_EACH (grp
, group_node
, &ofproto
->ms
->group_lru
) {
4257 LIST_FOR_EACH(b
, bundle_node
, &grp
->bundle_lru
) {
4258 char name
[OFP_MAX_PORT_NAME_LEN
];
4261 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4263 ds_put_format(&ds
, "%5s %4d "IP_FMT
" %3d\n",
4264 name
, grp
->vlan
, IP_ARGS(grp
->ip4
),
4265 mcast_bundle_age(ofproto
->ms
, b
));
4269 /* ports connected to multicast routers */
4270 LIST_FOR_EACH(mrouter
, mrouter_node
, &ofproto
->ms
->mrouter_lru
) {
4271 char name
[OFP_MAX_PORT_NAME_LEN
];
4273 bundle
= mrouter
->port
;
4274 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4276 ds_put_format(&ds
, "%5s %4d querier %3d\n",
4277 name
, mrouter
->vlan
,
4278 mcast_mrouter_age(ofproto
->ms
, mrouter
));
4280 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
4281 unixctl_command_reply(conn
, ds_cstr(&ds
));
4286 struct xlate_out xout
;
4287 struct xlate_in xin
;
4288 const struct flow
*key
;
4290 struct flow_wildcards wc
;
4295 trace_format_rule(struct ds
*result
, int level
, const struct rule_dpif
*rule
)
4297 const struct rule_actions
*actions
;
4300 ds_put_char_multiple(result
, '\t', level
);
4302 ds_put_cstr(result
, "No match\n");
4306 ovs_mutex_lock(&rule
->up
.mutex
);
4307 cookie
= rule
->up
.flow_cookie
;
4308 ovs_mutex_unlock(&rule
->up
.mutex
);
4310 ds_put_format(result
, "Rule: table=%"PRIu8
" cookie=%#"PRIx64
" ",
4311 rule
? rule
->up
.table_id
: 0, ntohll(cookie
));
4312 cls_rule_format(&rule
->up
.cr
, result
);
4313 ds_put_char(result
, '\n');
4315 actions
= rule_dpif_get_actions(rule
);
4317 ds_put_char_multiple(result
, '\t', level
);
4318 ds_put_cstr(result
, "OpenFlow actions=");
4319 ofpacts_format(actions
->ofpacts
, actions
->ofpacts_len
, result
);
4320 ds_put_char(result
, '\n');
4324 trace_format_flow(struct ds
*result
, int level
, const char *title
,
4325 struct trace_ctx
*trace
)
4327 ds_put_char_multiple(result
, '\t', level
);
4328 ds_put_format(result
, "%s: ", title
);
4329 /* Do not report unchanged flows for resubmits. */
4330 if ((level
> 0 && flow_equal(&trace
->xin
.flow
, &trace
->flow
))
4331 || (level
== 0 && flow_equal(&trace
->xin
.flow
, trace
->key
))) {
4332 ds_put_cstr(result
, "unchanged");
4334 flow_format(result
, &trace
->xin
.flow
);
4335 trace
->flow
= trace
->xin
.flow
;
4337 ds_put_char(result
, '\n');
4341 trace_format_regs(struct ds
*result
, int level
, const char *title
,
4342 struct trace_ctx
*trace
)
4346 ds_put_char_multiple(result
, '\t', level
);
4347 ds_put_format(result
, "%s:", title
);
4348 for (i
= 0; i
< FLOW_N_REGS
; i
++) {
4349 ds_put_format(result
, " reg%"PRIuSIZE
"=0x%"PRIx32
, i
, trace
->flow
.regs
[i
]);
4351 ds_put_char(result
, '\n');
4355 trace_format_odp(struct ds
*result
, int level
, const char *title
,
4356 struct trace_ctx
*trace
)
4358 struct ofpbuf
*odp_actions
= trace
->xout
.odp_actions
;
4360 ds_put_char_multiple(result
, '\t', level
);
4361 ds_put_format(result
, "%s: ", title
);
4362 format_odp_actions(result
, ofpbuf_data(odp_actions
),
4363 ofpbuf_size(odp_actions
));
4364 ds_put_char(result
, '\n');
4368 trace_format_megaflow(struct ds
*result
, int level
, const char *title
,
4369 struct trace_ctx
*trace
)
4373 ds_put_char_multiple(result
, '\t', level
);
4374 ds_put_format(result
, "%s: ", title
);
4375 flow_wildcards_or(&trace
->wc
, &trace
->xout
.wc
, &trace
->wc
);
4376 match_init(&match
, trace
->key
, &trace
->wc
);
4377 match_format(&match
, result
, OFP_DEFAULT_PRIORITY
);
4378 ds_put_char(result
, '\n');
4382 trace_resubmit(struct xlate_in
*xin
, struct rule_dpif
*rule
, int recurse
)
4384 struct trace_ctx
*trace
= CONTAINER_OF(xin
, struct trace_ctx
, xin
);
4385 struct ds
*result
= trace
->result
;
4387 ds_put_char(result
, '\n');
4388 trace_format_flow(result
, recurse
+ 1, "Resubmitted flow", trace
);
4389 trace_format_regs(result
, recurse
+ 1, "Resubmitted regs", trace
);
4390 trace_format_odp(result
, recurse
+ 1, "Resubmitted odp", trace
);
4391 trace_format_megaflow(result
, recurse
+ 1, "Resubmitted megaflow", trace
);
4392 trace_format_rule(result
, recurse
+ 1, rule
);
4396 trace_report(struct xlate_in
*xin
, const char *s
, int recurse
)
4398 struct trace_ctx
*trace
= CONTAINER_OF(xin
, struct trace_ctx
, xin
);
4399 struct ds
*result
= trace
->result
;
4401 ds_put_char_multiple(result
, '\t', recurse
);
4402 ds_put_cstr(result
, s
);
4403 ds_put_char(result
, '\n');
4406 /* Parses the 'argc' elements of 'argv', ignoring argv[0]. The following
4407 * forms are supported:
4409 * - [dpname] odp_flow [-generate | packet]
4410 * - bridge br_flow [-generate | packet]
4412 * On success, initializes '*ofprotop' and 'flow' and returns NULL. On failure
4413 * returns a nonnull malloced error message. */
4414 static char * WARN_UNUSED_RESULT
4415 parse_flow_and_packet(int argc
, const char *argv
[],
4416 struct ofproto_dpif
**ofprotop
, struct flow
*flow
,
4417 struct ofpbuf
**packetp
)
4419 const struct dpif_backer
*backer
= NULL
;
4420 const char *error
= NULL
;
4422 struct simap port_names
= SIMAP_INITIALIZER(&port_names
);
4423 struct ofpbuf
*packet
;
4424 struct ofpbuf odp_key
;
4425 struct ofpbuf odp_mask
;
4427 ofpbuf_init(&odp_key
, 0);
4428 ofpbuf_init(&odp_mask
, 0);
4430 /* Handle "-generate" or a hex string as the last argument. */
4431 if (!strcmp(argv
[argc
- 1], "-generate")) {
4432 packet
= ofpbuf_new(0);
4435 error
= eth_from_hex(argv
[argc
- 1], &packet
);
4438 } else if (argc
== 4) {
4439 /* The 3-argument form must end in "-generate' or a hex string. */
4445 /* odp_flow can have its in_port specified as a name instead of port no.
4446 * We do not yet know whether a given flow is a odp_flow or a br_flow.
4447 * But, to know whether a flow is odp_flow through odp_flow_from_string(),
4448 * we need to create a simap of name to port no. */
4450 const char *dp_type
;
4451 if (!strncmp(argv
[1], "ovs-", 4)) {
4452 dp_type
= argv
[1] + 4;
4456 backer
= shash_find_data(&all_dpif_backers
, dp_type
);
4457 } else if (argc
== 2) {
4458 struct shash_node
*node
;
4459 if (shash_count(&all_dpif_backers
) == 1) {
4460 node
= shash_first(&all_dpif_backers
);
4461 backer
= node
->data
;
4464 error
= "Syntax error";
4467 if (backer
&& backer
->dpif
) {
4468 struct dpif_port dpif_port
;
4469 struct dpif_port_dump port_dump
;
4470 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, backer
->dpif
) {
4471 simap_put(&port_names
, dpif_port
.name
,
4472 odp_to_u32(dpif_port
.port_no
));
4476 /* Parse the flow and determine whether a datapath or
4477 * bridge is specified. If function odp_flow_key_from_string()
4478 * returns 0, the flow is a odp_flow. If function
4479 * parse_ofp_exact_flow() returns NULL, the flow is a br_flow. */
4480 if (!odp_flow_from_string(argv
[argc
- 1], &port_names
,
4481 &odp_key
, &odp_mask
)) {
4483 error
= "Cannot find the datapath";
4487 if (odp_flow_key_to_flow(ofpbuf_data(&odp_key
), ofpbuf_size(&odp_key
),
4488 flow
) == ODP_FIT_ERROR
) {
4489 error
= "Failed to parse flow key";
4493 *ofprotop
= xlate_lookup_ofproto(backer
, flow
,
4494 &flow
->in_port
.ofp_port
);
4495 if (*ofprotop
== NULL
) {
4496 error
= "Invalid datapath flow";
4500 vsp_adjust_flow(*ofprotop
, flow
, NULL
);
4503 char *err
= parse_ofp_exact_flow(flow
, NULL
, argv
[argc
- 1], NULL
);
4506 m_err
= xasprintf("Bad flow syntax: %s", err
);
4511 error
= "Must specify bridge name";
4515 *ofprotop
= ofproto_dpif_lookup(argv
[1]);
4517 error
= "Unknown bridge name";
4523 /* Generate a packet, if requested. */
4525 if (!ofpbuf_size(packet
)) {
4526 flow_compose(packet
, flow
);
4528 struct pkt_metadata md
= pkt_metadata_from_flow(flow
);
4530 /* Use the metadata from the flow and the packet argument
4531 * to reconstruct the flow. */
4532 flow_extract(packet
, &md
, flow
);
4537 if (error
&& !m_err
) {
4538 m_err
= xstrdup(error
);
4541 ofpbuf_delete(packet
);
4545 ofpbuf_uninit(&odp_key
);
4546 ofpbuf_uninit(&odp_mask
);
4547 simap_destroy(&port_names
);
4552 ofproto_unixctl_trace(struct unixctl_conn
*conn
, int argc
, const char *argv
[],
4553 void *aux OVS_UNUSED
)
4555 struct ofproto_dpif
*ofproto
;
4556 struct ofpbuf
*packet
;
4560 error
= parse_flow_and_packet(argc
, argv
, &ofproto
, &flow
, &packet
);
4565 ofproto_trace(ofproto
, &flow
, packet
, NULL
, 0, &result
);
4566 unixctl_command_reply(conn
, ds_cstr(&result
));
4567 ds_destroy(&result
);
4568 ofpbuf_delete(packet
);
4570 unixctl_command_reply_error(conn
, error
);
4576 ofproto_unixctl_trace_actions(struct unixctl_conn
*conn
, int argc
,
4577 const char *argv
[], void *aux OVS_UNUSED
)
4579 enum ofputil_protocol usable_protocols
;
4580 struct ofproto_dpif
*ofproto
;
4581 bool enforce_consistency
;
4582 struct ofpbuf ofpacts
;
4583 struct ofpbuf
*packet
;
4588 /* Three kinds of error return values! */
4594 ofpbuf_init(&ofpacts
, 0);
4596 /* Parse actions. */
4597 error
= ofpacts_parse_actions(argv
[--argc
], &ofpacts
, &usable_protocols
);
4599 unixctl_command_reply_error(conn
, error
);
4604 /* OpenFlow 1.1 and later suggest that the switch enforces certain forms of
4605 * consistency between the flow and the actions. With -consistent, we
4606 * enforce consistency even for a flow supported in OpenFlow 1.0. */
4607 if (!strcmp(argv
[1], "-consistent")) {
4608 enforce_consistency
= true;
4612 enforce_consistency
= false;
4615 error
= parse_flow_and_packet(argc
, argv
, &ofproto
, &flow
, &packet
);
4617 unixctl_command_reply_error(conn
, error
);
4622 /* Do the same checks as handle_packet_out() in ofproto.c.
4624 * We pass a 'table_id' of 0 to ofproto_check_ofpacts(), which isn't
4625 * strictly correct because these actions aren't in any table, but it's OK
4626 * because it 'table_id' is used only to check goto_table instructions, but
4627 * packet-outs take a list of actions and therefore it can't include
4630 * We skip the "meter" check here because meter is an instruction, not an
4631 * action, and thus cannot appear in ofpacts. */
4632 in_port
= ofp_to_u16(flow
.in_port
.ofp_port
);
4633 if (in_port
>= ofproto
->up
.max_ports
&& in_port
< ofp_to_u16(OFPP_MAX
)) {
4634 unixctl_command_reply_error(conn
, "invalid in_port");
4637 if (enforce_consistency
) {
4638 retval
= ofpacts_check_consistency(ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
),
4639 &flow
, u16_to_ofp(ofproto
->up
.max_ports
),
4640 0, 0, usable_protocols
);
4642 retval
= ofpacts_check(ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
), &flow
,
4643 u16_to_ofp(ofproto
->up
.max_ports
), 0, 0,
4649 ds_put_format(&result
, "Bad actions: %s", ofperr_to_string(retval
));
4650 unixctl_command_reply_error(conn
, ds_cstr(&result
));
4654 ofproto_trace(ofproto
, &flow
, packet
,
4655 ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
), &result
);
4656 unixctl_command_reply(conn
, ds_cstr(&result
));
4659 ds_destroy(&result
);
4660 ofpbuf_delete(packet
);
4661 ofpbuf_uninit(&ofpacts
);
4664 /* Implements a "trace" through 'ofproto''s flow table, appending a textual
4665 * description of the results to 'ds'.
4667 * The trace follows a packet with the specified 'flow' through the flow
4668 * table. 'packet' may be nonnull to trace an actual packet, with consequent
4669 * side effects (if it is nonnull then its flow must be 'flow').
4671 * If 'ofpacts' is nonnull then its 'ofpacts_len' bytes specify the actions to
4672 * trace, otherwise the actions are determined by a flow table lookup. */
4674 ofproto_trace(struct ofproto_dpif
*ofproto
, struct flow
*flow
,
4675 const struct ofpbuf
*packet
,
4676 const struct ofpact ofpacts
[], size_t ofpacts_len
,
4679 struct rule_dpif
*rule
;
4680 struct trace_ctx trace
;
4682 ds_put_format(ds
, "Bridge: %s\n", ofproto
->up
.name
);
4683 ds_put_cstr(ds
, "Flow: ");
4684 flow_format(ds
, flow
);
4685 ds_put_char(ds
, '\n');
4687 flow_wildcards_init_catchall(&trace
.wc
);
4691 rule_dpif_lookup(ofproto
, flow
, &trace
.wc
, &rule
, false, NULL
);
4693 trace_format_rule(ds
, 0, rule
);
4694 if (rule
== ofproto
->miss_rule
) {
4695 ds_put_cstr(ds
, "\nNo match, flow generates \"packet in\"s.\n");
4696 } else if (rule
== ofproto
->no_packet_in_rule
) {
4697 ds_put_cstr(ds
, "\nNo match, packets dropped because "
4698 "OFPPC_NO_PACKET_IN is set on in_port.\n");
4699 } else if (rule
== ofproto
->drop_frags_rule
) {
4700 ds_put_cstr(ds
, "\nPackets dropped because they are IP fragments "
4701 "and the fragment handling mode is \"drop\".\n");
4705 if (rule
|| ofpacts
) {
4707 trace
.key
= flow
; /* Original flow key, used for megaflow. */
4708 trace
.flow
= *flow
; /* May be modified by actions. */
4709 xlate_in_init(&trace
.xin
, ofproto
, flow
, flow
->in_port
.ofp_port
, rule
,
4710 ntohs(flow
->tcp_flags
), packet
);
4712 trace
.xin
.ofpacts
= ofpacts
;
4713 trace
.xin
.ofpacts_len
= ofpacts_len
;
4715 trace
.xin
.resubmit_hook
= trace_resubmit
;
4716 trace
.xin
.report_hook
= trace_report
;
4718 xlate_actions(&trace
.xin
, &trace
.xout
);
4720 ds_put_char(ds
, '\n');
4721 trace_format_flow(ds
, 0, "Final flow", &trace
);
4722 trace_format_megaflow(ds
, 0, "Megaflow", &trace
);
4724 ds_put_cstr(ds
, "Datapath actions: ");
4725 format_odp_actions(ds
, ofpbuf_data(trace
.xout
.odp_actions
),
4726 ofpbuf_size(trace
.xout
.odp_actions
));
4728 if (trace
.xout
.slow
) {
4729 enum slow_path_reason slow
;
4731 ds_put_cstr(ds
, "\nThis flow is handled by the userspace "
4732 "slow path because it:");
4734 slow
= trace
.xout
.slow
;
4736 enum slow_path_reason bit
= rightmost_1bit(slow
);
4738 ds_put_format(ds
, "\n\t- %s.",
4739 slow_path_reason_to_explanation(bit
));
4745 xlate_out_uninit(&trace
.xout
);
4749 /* Store the current ofprotos in 'ofproto_shash'. Returns a sorted list
4750 * of the 'ofproto_shash' nodes. It is the responsibility of the caller
4751 * to destroy 'ofproto_shash' and free the returned value. */
4752 static const struct shash_node
**
4753 get_ofprotos(struct shash
*ofproto_shash
)
4755 const struct ofproto_dpif
*ofproto
;
4757 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4758 char *name
= xasprintf("%s@%s", ofproto
->up
.type
, ofproto
->up
.name
);
4759 shash_add_nocopy(ofproto_shash
, name
, ofproto
);
4762 return shash_sort(ofproto_shash
);
4766 ofproto_unixctl_dpif_dump_dps(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4767 const char *argv
[] OVS_UNUSED
,
4768 void *aux OVS_UNUSED
)
4770 struct ds ds
= DS_EMPTY_INITIALIZER
;
4771 struct shash ofproto_shash
;
4772 const struct shash_node
**sorted_ofprotos
;
4775 shash_init(&ofproto_shash
);
4776 sorted_ofprotos
= get_ofprotos(&ofproto_shash
);
4777 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4778 const struct shash_node
*node
= sorted_ofprotos
[i
];
4779 ds_put_format(&ds
, "%s\n", node
->name
);
4782 shash_destroy(&ofproto_shash
);
4783 free(sorted_ofprotos
);
4785 unixctl_command_reply(conn
, ds_cstr(&ds
));
4790 dpif_show_backer(const struct dpif_backer
*backer
, struct ds
*ds
)
4792 const struct shash_node
**ofprotos
;
4793 struct dpif_dp_stats dp_stats
;
4794 struct shash ofproto_shash
;
4797 dpif_get_dp_stats(backer
->dpif
, &dp_stats
);
4799 ds_put_format(ds
, "%s: hit:%"PRIu64
" missed:%"PRIu64
"\n",
4800 dpif_name(backer
->dpif
), dp_stats
.n_hit
, dp_stats
.n_missed
);
4802 shash_init(&ofproto_shash
);
4803 ofprotos
= get_ofprotos(&ofproto_shash
);
4804 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4805 struct ofproto_dpif
*ofproto
= ofprotos
[i
]->data
;
4806 const struct shash_node
**ports
;
4809 if (ofproto
->backer
!= backer
) {
4813 ds_put_format(ds
, "\t%s:\n", ofproto
->up
.name
);
4815 ports
= shash_sort(&ofproto
->up
.port_by_name
);
4816 for (j
= 0; j
< shash_count(&ofproto
->up
.port_by_name
); j
++) {
4817 const struct shash_node
*node
= ports
[j
];
4818 struct ofport
*ofport
= node
->data
;
4820 odp_port_t odp_port
;
4822 ds_put_format(ds
, "\t\t%s %u/", netdev_get_name(ofport
->netdev
),
4825 odp_port
= ofp_port_to_odp_port(ofproto
, ofport
->ofp_port
);
4826 if (odp_port
!= ODPP_NONE
) {
4827 ds_put_format(ds
, "%"PRIu32
":", odp_port
);
4829 ds_put_cstr(ds
, "none:");
4832 ds_put_format(ds
, " (%s", netdev_get_type(ofport
->netdev
));
4835 if (!netdev_get_config(ofport
->netdev
, &config
)) {
4836 const struct smap_node
**nodes
;
4839 nodes
= smap_sort(&config
);
4840 for (i
= 0; i
< smap_count(&config
); i
++) {
4841 const struct smap_node
*node
= nodes
[i
];
4842 ds_put_format(ds
, "%c %s=%s", i
? ',' : ':',
4843 node
->key
, node
->value
);
4847 smap_destroy(&config
);
4849 ds_put_char(ds
, ')');
4850 ds_put_char(ds
, '\n');
4854 shash_destroy(&ofproto_shash
);
4859 ofproto_unixctl_dpif_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4860 const char *argv
[] OVS_UNUSED
, void *aux OVS_UNUSED
)
4862 struct ds ds
= DS_EMPTY_INITIALIZER
;
4863 const struct shash_node
**backers
;
4866 backers
= shash_sort(&all_dpif_backers
);
4867 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
4868 dpif_show_backer(backers
[i
]->data
, &ds
);
4872 unixctl_command_reply(conn
, ds_cstr(&ds
));
4877 ofproto_unixctl_dpif_dump_flows(struct unixctl_conn
*conn
,
4878 int argc OVS_UNUSED
, const char *argv
[],
4879 void *aux OVS_UNUSED
)
4881 const struct ofproto_dpif
*ofproto
;
4883 struct ds ds
= DS_EMPTY_INITIALIZER
;
4884 bool verbosity
= false;
4886 struct dpif_port dpif_port
;
4887 struct dpif_port_dump port_dump
;
4888 struct hmap portno_names
;
4890 struct dpif_flow_dump
*flow_dump
;
4891 struct dpif_flow_dump_thread
*flow_dump_thread
;
4895 ofproto
= ofproto_dpif_lookup(argv
[argc
- 1]);
4897 unixctl_command_reply_error(conn
, "no such bridge");
4901 if (argc
> 2 && !strcmp(argv
[1], "-m")) {
4905 hmap_init(&portno_names
);
4906 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, ofproto
->backer
->dpif
) {
4907 odp_portno_names_set(&portno_names
, dpif_port
.port_no
, dpif_port
.name
);
4911 flow_dump
= dpif_flow_dump_create(ofproto
->backer
->dpif
);
4912 flow_dump_thread
= dpif_flow_dump_thread_create(flow_dump
);
4913 while (dpif_flow_dump_next(flow_dump_thread
, &f
, 1)) {
4916 if (odp_flow_key_to_flow(f
.key
, f
.key_len
, &flow
) == ODP_FIT_ERROR
4917 || xlate_lookup_ofproto(ofproto
->backer
, &flow
, NULL
) != ofproto
) {
4921 odp_flow_format(f
.key
, f
.key_len
, f
.mask
, f
.mask_len
,
4922 &portno_names
, &ds
, verbosity
);
4923 ds_put_cstr(&ds
, ", ");
4924 dpif_flow_stats_format(&f
.stats
, &ds
);
4925 ds_put_cstr(&ds
, ", actions:");
4926 format_odp_actions(&ds
, f
.actions
, f
.actions_len
);
4927 ds_put_char(&ds
, '\n');
4929 dpif_flow_dump_thread_destroy(flow_dump_thread
);
4930 error
= dpif_flow_dump_destroy(flow_dump
);
4934 ds_put_format(&ds
, "dpif/dump_flows failed: %s", ovs_strerror(errno
));
4935 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
4937 unixctl_command_reply(conn
, ds_cstr(&ds
));
4939 odp_portno_names_destroy(&portno_names
);
4940 hmap_destroy(&portno_names
);
4945 ofproto_dpif_unixctl_init(void)
4947 static bool registered
;
4953 unixctl_command_register(
4955 "{[dp_name] odp_flow | bridge br_flow} [-generate|packet]",
4956 1, 3, ofproto_unixctl_trace
, NULL
);
4957 unixctl_command_register(
4958 "ofproto/trace-packet-out",
4959 "[-consistent] {[dp_name] odp_flow | bridge br_flow} [-generate|packet] actions",
4960 2, 6, ofproto_unixctl_trace_actions
, NULL
);
4961 unixctl_command_register("fdb/flush", "[bridge]", 0, 1,
4962 ofproto_unixctl_fdb_flush
, NULL
);
4963 unixctl_command_register("fdb/show", "bridge", 1, 1,
4964 ofproto_unixctl_fdb_show
, NULL
);
4965 unixctl_command_register("mdb/flush", "[bridge]", 0, 1,
4966 ofproto_unixctl_mcast_snooping_flush
, NULL
);
4967 unixctl_command_register("mdb/show", "bridge", 1, 1,
4968 ofproto_unixctl_mcast_snooping_show
, NULL
);
4969 unixctl_command_register("dpif/dump-dps", "", 0, 0,
4970 ofproto_unixctl_dpif_dump_dps
, NULL
);
4971 unixctl_command_register("dpif/show", "", 0, 0, ofproto_unixctl_dpif_show
,
4973 unixctl_command_register("dpif/dump-flows", "[-m] bridge", 1, 2,
4974 ofproto_unixctl_dpif_dump_flows
, NULL
);
4977 /* Returns true if 'table' is the table used for internal rules,
4978 * false otherwise. */
4980 table_is_internal(uint8_t table_id
)
4982 return table_id
== TBL_INTERNAL
;
4985 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4987 * This is deprecated. It is only for compatibility with broken device drivers
4988 * in old versions of Linux that do not properly support VLANs when VLAN
4989 * devices are not used. When broken device drivers are no longer in
4990 * widespread use, we will delete these interfaces. */
4993 set_realdev(struct ofport
*ofport_
, ofp_port_t realdev_ofp_port
, int vid
)
4995 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
4996 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
4998 if (realdev_ofp_port
== ofport
->realdev_ofp_port
4999 && vid
== ofport
->vlandev_vid
) {
5003 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
5005 if (ofport
->realdev_ofp_port
) {
5008 if (realdev_ofp_port
&& ofport
->bundle
) {
5009 /* vlandevs are enslaved to their realdevs, so they are not allowed to
5010 * themselves be part of a bundle. */
5011 bundle_set(ofport_
->ofproto
, ofport
->bundle
, NULL
);
5014 ofport
->realdev_ofp_port
= realdev_ofp_port
;
5015 ofport
->vlandev_vid
= vid
;
5017 if (realdev_ofp_port
) {
5018 vsp_add(ofport
, realdev_ofp_port
, vid
);
5025 hash_realdev_vid(ofp_port_t realdev_ofp_port
, int vid
)
5027 return hash_2words(ofp_to_u16(realdev_ofp_port
), vid
);
5031 ofproto_has_vlan_splinters(const struct ofproto_dpif
*ofproto
)
5032 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5034 /* hmap_is_empty is thread safe. */
5035 return !hmap_is_empty(&ofproto
->realdev_vid_map
);
5040 vsp_realdev_to_vlandev__(const struct ofproto_dpif
*ofproto
,
5041 ofp_port_t realdev_ofp_port
, ovs_be16 vlan_tci
)
5042 OVS_REQUIRES(ofproto
->vsp_mutex
)
5044 if (!hmap_is_empty(&ofproto
->realdev_vid_map
)) {
5045 int vid
= vlan_tci_to_vid(vlan_tci
);
5046 const struct vlan_splinter
*vsp
;
5048 HMAP_FOR_EACH_WITH_HASH (vsp
, realdev_vid_node
,
5049 hash_realdev_vid(realdev_ofp_port
, vid
),
5050 &ofproto
->realdev_vid_map
) {
5051 if (vsp
->realdev_ofp_port
== realdev_ofp_port
5052 && vsp
->vid
== vid
) {
5053 return vsp
->vlandev_ofp_port
;
5057 return realdev_ofp_port
;
5060 /* Returns the OFP port number of the Linux VLAN device that corresponds to
5061 * 'vlan_tci' on the network device with port number 'realdev_ofp_port' in
5062 * 'struct ofport_dpif'. For example, given 'realdev_ofp_port' of eth0 and
5063 * 'vlan_tci' 9, it would return the port number of eth0.9.
5065 * Unless VLAN splinters are enabled for port 'realdev_ofp_port', this
5066 * function just returns its 'realdev_ofp_port' argument. */
5068 vsp_realdev_to_vlandev(const struct ofproto_dpif
*ofproto
,
5069 ofp_port_t realdev_ofp_port
, ovs_be16 vlan_tci
)
5070 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5074 /* hmap_is_empty is thread safe, see if we can return immediately. */
5075 if (hmap_is_empty(&ofproto
->realdev_vid_map
)) {
5076 return realdev_ofp_port
;
5078 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5079 ret
= vsp_realdev_to_vlandev__(ofproto
, realdev_ofp_port
, vlan_tci
);
5080 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5084 static struct vlan_splinter
*
5085 vlandev_find(const struct ofproto_dpif
*ofproto
, ofp_port_t vlandev_ofp_port
)
5087 struct vlan_splinter
*vsp
;
5089 HMAP_FOR_EACH_WITH_HASH (vsp
, vlandev_node
,
5090 hash_ofp_port(vlandev_ofp_port
),
5091 &ofproto
->vlandev_map
) {
5092 if (vsp
->vlandev_ofp_port
== vlandev_ofp_port
) {
5100 /* Returns the OpenFlow port number of the "real" device underlying the Linux
5101 * VLAN device with OpenFlow port number 'vlandev_ofp_port' and stores the
5102 * VLAN VID of the Linux VLAN device in '*vid'. For example, given
5103 * 'vlandev_ofp_port' of eth0.9, it would return the OpenFlow port number of
5104 * eth0 and store 9 in '*vid'.
5106 * Returns 0 and does not modify '*vid' if 'vlandev_ofp_port' is not a Linux
5107 * VLAN device. Unless VLAN splinters are enabled, this is what this function
5110 vsp_vlandev_to_realdev(const struct ofproto_dpif
*ofproto
,
5111 ofp_port_t vlandev_ofp_port
, int *vid
)
5112 OVS_REQUIRES(ofproto
->vsp_mutex
)
5114 if (!hmap_is_empty(&ofproto
->vlandev_map
)) {
5115 const struct vlan_splinter
*vsp
;
5117 vsp
= vlandev_find(ofproto
, vlandev_ofp_port
);
5122 return vsp
->realdev_ofp_port
;
5128 /* Given 'flow', a flow representing a packet received on 'ofproto', checks
5129 * whether 'flow->in_port' represents a Linux VLAN device. If so, changes
5130 * 'flow->in_port' to the "real" device backing the VLAN device, sets
5131 * 'flow->vlan_tci' to the VLAN VID, and returns true. Optionally pushes the
5132 * appropriate VLAN on 'packet' if provided. Otherwise (which is always the
5133 * case unless VLAN splinters are enabled), returns false without making any
5136 vsp_adjust_flow(const struct ofproto_dpif
*ofproto
, struct flow
*flow
,
5137 struct ofpbuf
*packet
)
5138 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5143 /* hmap_is_empty is thread safe. */
5144 if (hmap_is_empty(&ofproto
->vlandev_map
)) {
5148 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5149 realdev
= vsp_vlandev_to_realdev(ofproto
, flow
->in_port
.ofp_port
, &vid
);
5150 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5155 /* Cause the flow to be processed as if it came in on the real device with
5156 * the VLAN device's VLAN ID. */
5157 flow
->in_port
.ofp_port
= realdev
;
5158 flow
->vlan_tci
= htons((vid
& VLAN_VID_MASK
) | VLAN_CFI
);
5161 /* Make the packet resemble the flow, so that it gets sent to an
5162 * OpenFlow controller properly, so that it looks correct for sFlow,
5163 * and so that flow_extract() will get the correct vlan_tci if it is
5164 * called on 'packet'. */
5165 eth_push_vlan(packet
, htons(ETH_TYPE_VLAN
), flow
->vlan_tci
);
5172 vsp_remove(struct ofport_dpif
*port
)
5174 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
5175 struct vlan_splinter
*vsp
;
5177 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5178 vsp
= vlandev_find(ofproto
, port
->up
.ofp_port
);
5180 hmap_remove(&ofproto
->vlandev_map
, &vsp
->vlandev_node
);
5181 hmap_remove(&ofproto
->realdev_vid_map
, &vsp
->realdev_vid_node
);
5184 port
->realdev_ofp_port
= 0;
5186 VLOG_ERR("missing vlan device record");
5188 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5192 vsp_add(struct ofport_dpif
*port
, ofp_port_t realdev_ofp_port
, int vid
)
5194 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
5196 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5197 if (!vsp_vlandev_to_realdev(ofproto
, port
->up
.ofp_port
, NULL
)
5198 && (vsp_realdev_to_vlandev__(ofproto
, realdev_ofp_port
, htons(vid
))
5199 == realdev_ofp_port
)) {
5200 struct vlan_splinter
*vsp
;
5202 vsp
= xmalloc(sizeof *vsp
);
5203 vsp
->realdev_ofp_port
= realdev_ofp_port
;
5204 vsp
->vlandev_ofp_port
= port
->up
.ofp_port
;
5207 port
->realdev_ofp_port
= realdev_ofp_port
;
5209 hmap_insert(&ofproto
->vlandev_map
, &vsp
->vlandev_node
,
5210 hash_ofp_port(port
->up
.ofp_port
));
5211 hmap_insert(&ofproto
->realdev_vid_map
, &vsp
->realdev_vid_node
,
5212 hash_realdev_vid(realdev_ofp_port
, vid
));
5214 VLOG_ERR("duplicate vlan device record");
5216 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5220 ofp_port_to_odp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
5222 const struct ofport_dpif
*ofport
= get_ofp_port(ofproto
, ofp_port
);
5223 return ofport
? ofport
->odp_port
: ODPP_NONE
;
5226 struct ofport_dpif
*
5227 odp_port_to_ofport(const struct dpif_backer
*backer
, odp_port_t odp_port
)
5229 struct ofport_dpif
*port
;
5231 ovs_rwlock_rdlock(&backer
->odp_to_ofport_lock
);
5232 HMAP_FOR_EACH_IN_BUCKET (port
, odp_port_node
, hash_odp_port(odp_port
),
5233 &backer
->odp_to_ofport_map
) {
5234 if (port
->odp_port
== odp_port
) {
5235 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5240 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5245 odp_port_to_ofp_port(const struct ofproto_dpif
*ofproto
, odp_port_t odp_port
)
5247 struct ofport_dpif
*port
;
5249 port
= odp_port_to_ofport(ofproto
->backer
, odp_port
);
5250 if (port
&& &ofproto
->up
== port
->up
.ofproto
) {
5251 return port
->up
.ofp_port
;
5258 ofproto_dpif_alloc_recirc_id(struct ofproto_dpif
*ofproto
)
5260 struct dpif_backer
*backer
= ofproto
->backer
;
5262 return recirc_id_alloc(backer
->rid_pool
);
5266 ofproto_dpif_free_recirc_id(struct ofproto_dpif
*ofproto
, uint32_t recirc_id
)
5268 struct dpif_backer
*backer
= ofproto
->backer
;
5270 recirc_id_free(backer
->rid_pool
, recirc_id
);
5274 ofproto_dpif_add_internal_flow(struct ofproto_dpif
*ofproto
,
5275 const struct match
*match
, int priority
,
5276 uint16_t idle_timeout
,
5277 const struct ofpbuf
*ofpacts
,
5278 struct rule
**rulep
)
5280 struct ofputil_flow_mod fm
;
5281 struct rule_dpif
*rule
;
5285 fm
.priority
= priority
;
5286 fm
.new_cookie
= htonll(0);
5287 fm
.cookie
= htonll(0);
5288 fm
.cookie_mask
= htonll(0);
5289 fm
.modify_cookie
= false;
5290 fm
.table_id
= TBL_INTERNAL
;
5291 fm
.command
= OFPFC_ADD
;
5292 fm
.idle_timeout
= idle_timeout
;
5293 fm
.hard_timeout
= 0;
5296 fm
.flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
;
5297 fm
.ofpacts
= ofpbuf_data(ofpacts
);
5298 fm
.ofpacts_len
= ofpbuf_size(ofpacts
);
5300 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5302 VLOG_ERR_RL(&rl
, "failed to add internal flow (%s)",
5303 ofperr_to_string(error
));
5308 rule
= rule_dpif_lookup_in_table(ofproto
, TBL_INTERNAL
, &fm
.match
.flow
,
5309 &fm
.match
.wc
, false);
5319 ofproto_dpif_delete_internal_flow(struct ofproto_dpif
*ofproto
,
5320 struct match
*match
, int priority
)
5322 struct ofputil_flow_mod fm
;
5326 fm
.priority
= priority
;
5327 fm
.new_cookie
= htonll(0);
5328 fm
.cookie
= htonll(0);
5329 fm
.cookie_mask
= htonll(0);
5330 fm
.modify_cookie
= false;
5331 fm
.table_id
= TBL_INTERNAL
;
5332 fm
.flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
;
5333 fm
.command
= OFPFC_DELETE_STRICT
;
5335 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5337 VLOG_ERR_RL(&rl
, "failed to delete internal flow (%s)",
5338 ofperr_to_string(error
));
5345 const struct ofproto_class ofproto_dpif_class
= {
5359 NULL
, /* get_memory_usage. */
5360 type_get_memory_usage
,
5378 port_is_lacp_current
,
5379 NULL
, /* rule_choose_table */
5388 NULL
, /* rule_premodify_actions */
5389 rule_modify_actions
,
5405 get_stp_port_status
,
5410 get_rstp_port_status
,
5417 is_mirror_output_bundle
,
5418 forward_bpdu_changed
,
5419 set_mac_table_config
,
5421 set_mcast_snooping_port
,
5423 NULL
, /* meter_get_features */
5424 NULL
, /* meter_set */
5425 NULL
, /* meter_get */
5426 NULL
, /* meter_del */
5427 group_alloc
, /* group_alloc */
5428 group_construct
, /* group_construct */
5429 group_destruct
, /* group_destruct */
5430 group_dealloc
, /* group_dealloc */
5431 group_modify
, /* group_modify */
5432 group_get_stats
, /* group_get_stats */