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 "ovs-router.h"
62 #include "poll-loop.h"
68 #include "unaligned.h"
70 #include "vlan-bitmap.h"
73 VLOG_DEFINE_THIS_MODULE(ofproto_dpif
);
75 COVERAGE_DEFINE(ofproto_dpif_expired
);
76 COVERAGE_DEFINE(packet_in_overflow
);
85 * - Do include packets and bytes from datapath flows which have not
86 * recently been processed by a revalidator. */
87 struct ovs_mutex stats_mutex
;
88 struct dpif_flow_stats stats OVS_GUARDED
;
90 /* If non-zero then the recirculation id that has
91 * been allocated for use with this rule.
92 * The recirculation id and associated internal flow should
93 * be freed when the rule is freed */
97 /* RULE_CAST() depends on this. */
98 BUILD_ASSERT_DECL(offsetof(struct rule_dpif
, up
) == 0);
100 static void rule_get_stats(struct rule
*, uint64_t *packets
, uint64_t *bytes
,
101 long long int *used
);
102 static struct rule_dpif
*rule_dpif_cast(const struct rule
*);
103 static void rule_expire(struct rule_dpif
*);
110 * - Do include packets and bytes from datapath flows which have not
111 * recently been processed by a revalidator. */
112 struct ovs_mutex stats_mutex
;
113 uint64_t packet_count OVS_GUARDED
; /* Number of packets received. */
114 uint64_t byte_count OVS_GUARDED
; /* Number of bytes received. */
118 struct hmap_node hmap_node
; /* In struct ofproto's "bundles" hmap. */
119 struct ofproto_dpif
*ofproto
; /* Owning ofproto. */
120 void *aux
; /* Key supplied by ofproto's client. */
121 char *name
; /* Identifier for log messages. */
124 struct list ports
; /* Contains "struct ofport"s. */
125 enum port_vlan_mode vlan_mode
; /* VLAN mode */
126 int vlan
; /* -1=trunk port, else a 12-bit VLAN ID. */
127 unsigned long *trunks
; /* Bitmap of trunked VLANs, if 'vlan' == -1.
128 * NULL if all VLANs are trunked. */
129 struct lacp
*lacp
; /* LACP if LACP is enabled, otherwise NULL. */
130 struct bond
*bond
; /* Nonnull iff more than one port. */
131 bool use_priority_tags
; /* Use 802.1p tag for frames in VLAN 0? */
134 bool floodable
; /* True if no port has OFPUTIL_PC_NO_FLOOD set. */
137 static void bundle_remove(struct ofport
*);
138 static void bundle_update(struct ofbundle
*);
139 static void bundle_destroy(struct ofbundle
*);
140 static void bundle_del_port(struct ofport_dpif
*);
141 static void bundle_run(struct ofbundle
*);
142 static void bundle_wait(struct ofbundle
*);
144 static void stp_run(struct ofproto_dpif
*ofproto
);
145 static void stp_wait(struct ofproto_dpif
*ofproto
);
146 static int set_stp_port(struct ofport
*,
147 const struct ofproto_port_stp_settings
*);
149 static void rstp_run(struct ofproto_dpif
*ofproto
);
150 static void set_rstp_port(struct ofport
*,
151 const struct ofproto_port_rstp_settings
*);
154 struct hmap_node odp_port_node
; /* In dpif_backer's "odp_to_ofport_map". */
158 struct ofbundle
*bundle
; /* Bundle that contains this port, if any. */
159 struct list bundle_node
; /* In struct ofbundle's "ports" list. */
160 struct cfm
*cfm
; /* Connectivity Fault Management, if any. */
161 struct bfd
*bfd
; /* BFD, if any. */
162 bool may_enable
; /* May be enabled in bonds. */
163 bool is_tunnel
; /* This port is a tunnel. */
164 bool is_layer3
; /* This is a layer 3 port. */
165 long long int carrier_seq
; /* Carrier status changes. */
166 struct ofport_dpif
*peer
; /* Peer if patch port. */
169 struct stp_port
*stp_port
; /* Spanning Tree Protocol, if any. */
170 enum stp_state stp_state
; /* Always STP_DISABLED if STP not in use. */
171 long long int stp_state_entered
;
173 /* Rapid Spanning Tree. */
174 struct rstp_port
*rstp_port
; /* Rapid Spanning Tree Protocol, if any. */
175 enum rstp_state rstp_state
; /* Always RSTP_DISABLED if RSTP not in use. */
177 /* Queue to DSCP mapping. */
178 struct ofproto_port_queue
*qdscp
;
181 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
183 * This is deprecated. It is only for compatibility with broken device
184 * drivers in old versions of Linux that do not properly support VLANs when
185 * VLAN devices are not used. When broken device drivers are no longer in
186 * widespread use, we will delete these interfaces. */
187 ofp_port_t realdev_ofp_port
;
191 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
193 * This is deprecated. It is only for compatibility with broken device drivers
194 * in old versions of Linux that do not properly support VLANs when VLAN
195 * devices are not used. When broken device drivers are no longer in
196 * widespread use, we will delete these interfaces. */
197 struct vlan_splinter
{
198 struct hmap_node realdev_vid_node
;
199 struct hmap_node vlandev_node
;
200 ofp_port_t realdev_ofp_port
;
201 ofp_port_t vlandev_ofp_port
;
205 static void vsp_remove(struct ofport_dpif
*);
206 static void vsp_add(struct ofport_dpif
*, ofp_port_t realdev_ofp_port
, int vid
);
208 static odp_port_t
ofp_port_to_odp_port(const struct ofproto_dpif
*,
211 static ofp_port_t
odp_port_to_ofp_port(const struct ofproto_dpif
*,
214 static struct ofport_dpif
*
215 ofport_dpif_cast(const struct ofport
*ofport
)
217 return ofport
? CONTAINER_OF(ofport
, struct ofport_dpif
, up
) : NULL
;
220 static void port_run(struct ofport_dpif
*);
221 static int set_bfd(struct ofport
*, const struct smap
*);
222 static int set_cfm(struct ofport
*, const struct cfm_settings
*);
223 static void ofport_update_peer(struct ofport_dpif
*);
225 /* Reasons that we might need to revalidate every datapath flow, and
226 * corresponding coverage counters.
228 * A value of 0 means that there is no need to revalidate.
230 * It would be nice to have some cleaner way to integrate with coverage
231 * counters, but with only a few reasons I guess this is good enough for
233 enum revalidate_reason
{
234 REV_RECONFIGURE
= 1, /* Switch configuration changed. */
235 REV_STP
, /* Spanning tree protocol port status change. */
236 REV_RSTP
, /* RSTP port status change. */
237 REV_BOND
, /* Bonding changed. */
238 REV_PORT_TOGGLED
, /* Port enabled or disabled by CFM, LACP, ...*/
239 REV_FLOW_TABLE
, /* Flow table changed. */
240 REV_MAC_LEARNING
, /* Mac learning changed. */
241 REV_MCAST_SNOOPING
, /* Multicast snooping changed. */
243 COVERAGE_DEFINE(rev_reconfigure
);
244 COVERAGE_DEFINE(rev_stp
);
245 COVERAGE_DEFINE(rev_rstp
);
246 COVERAGE_DEFINE(rev_bond
);
247 COVERAGE_DEFINE(rev_port_toggled
);
248 COVERAGE_DEFINE(rev_flow_table
);
249 COVERAGE_DEFINE(rev_mac_learning
);
250 COVERAGE_DEFINE(rev_mcast_snooping
);
252 /* All datapaths of a given type share a single dpif backer instance. */
259 struct ovs_rwlock odp_to_ofport_lock
;
260 struct hmap odp_to_ofport_map OVS_GUARDED
; /* Contains "struct ofport"s. */
262 struct simap tnl_backers
; /* Set of dpif ports backing tunnels. */
264 enum revalidate_reason need_revalidate
; /* Revalidate all flows. */
266 bool recv_set_enable
; /* Enables or disables receiving packets. */
269 struct recirc_id_pool
*rid_pool
; /* Recirculation ID pool. */
270 bool enable_recirc
; /* True if the datapath supports recirculation */
272 /* True if the datapath supports variable-length
273 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions.
274 * False if the datapath supports only 8-byte (or shorter) userdata. */
275 bool variable_length_userdata
;
277 /* True if the datapath supports masked data in OVS_ACTION_ATTR_SET
279 bool masked_set_action
;
281 /* Maximum number of MPLS label stack entries that the datapath supports
283 size_t max_mpls_depth
;
285 /* Version string of the datapath stored in OVSDB. */
286 char *dp_version_string
;
289 /* All existing ofproto_backer instances, indexed by ofproto->up.type. */
290 static struct shash all_dpif_backers
= SHASH_INITIALIZER(&all_dpif_backers
);
292 struct ofproto_dpif
{
293 struct hmap_node all_ofproto_dpifs_node
; /* In 'all_ofproto_dpifs'. */
295 struct dpif_backer
*backer
;
297 uint64_t dump_seq
; /* Last read of udpif_dump_seq(). */
299 /* Special OpenFlow rules. */
300 struct rule_dpif
*miss_rule
; /* Sends flow table misses to controller. */
301 struct rule_dpif
*no_packet_in_rule
; /* Drops flow table misses. */
302 struct rule_dpif
*drop_frags_rule
; /* Used in OFPC_FRAG_DROP mode. */
305 struct netflow
*netflow
;
306 struct dpif_sflow
*sflow
;
307 struct dpif_ipfix
*ipfix
;
308 struct hmap bundles
; /* Contains "struct ofbundle"s. */
309 struct mac_learning
*ml
;
310 struct mcast_snooping
*ms
;
311 bool has_bonded_bundles
;
313 struct mbridge
*mbridge
;
315 struct ovs_mutex stats_mutex
;
316 struct netdev_stats stats OVS_GUARDED
; /* To account packets generated and
317 * consumed in userspace. */
321 long long int stp_last_tick
;
323 /* Rapid Spanning Tree. */
325 long long int rstp_last_tick
;
327 /* VLAN splinters. */
328 struct ovs_mutex vsp_mutex
;
329 struct hmap realdev_vid_map OVS_GUARDED
; /* (realdev,vid) -> vlandev. */
330 struct hmap vlandev_map OVS_GUARDED
; /* vlandev -> (realdev,vid). */
333 struct sset ports
; /* Set of standard port names. */
334 struct sset ghost_ports
; /* Ports with no datapath port. */
335 struct sset port_poll_set
; /* Queued names for port_poll() reply. */
336 int port_poll_errno
; /* Last errno for port_poll() reply. */
337 uint64_t change_seq
; /* Connectivity status changes. */
340 struct guarded_list pins
; /* Contains "struct ofputil_packet_in"s. */
341 struct seq
*pins_seq
; /* For notifying 'pins' reception. */
345 /* All existing ofproto_dpif instances, indexed by ->up.name. */
346 static struct hmap all_ofproto_dpifs
= HMAP_INITIALIZER(&all_ofproto_dpifs
);
348 static void ofproto_dpif_unixctl_init(void);
350 static inline struct ofproto_dpif
*
351 ofproto_dpif_cast(const struct ofproto
*ofproto
)
353 ovs_assert(ofproto
->ofproto_class
== &ofproto_dpif_class
);
354 return CONTAINER_OF(ofproto
, struct ofproto_dpif
, up
);
358 ofproto_dpif_get_max_mpls_depth(const struct ofproto_dpif
*ofproto
)
360 return ofproto
->backer
->max_mpls_depth
;
364 ofproto_dpif_get_enable_recirc(const struct ofproto_dpif
*ofproto
)
366 return ofproto
->backer
->enable_recirc
;
369 static struct ofport_dpif
*get_ofp_port(const struct ofproto_dpif
*ofproto
,
370 ofp_port_t ofp_port
);
371 static void ofproto_trace(struct ofproto_dpif
*, struct flow
*,
372 const struct ofpbuf
*packet
,
373 const struct ofpact
[], size_t ofpacts_len
,
376 /* Global variables. */
377 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
379 /* Initial mappings of port to bridge mappings. */
380 static struct shash init_ofp_ports
= SHASH_INITIALIZER(&init_ofp_ports
);
382 /* Executes 'fm'. The caller retains ownership of 'fm' and everything in
385 ofproto_dpif_flow_mod(struct ofproto_dpif
*ofproto
,
386 struct ofputil_flow_mod
*fm
)
388 ofproto_flow_mod(&ofproto
->up
, fm
);
391 /* Appends 'pin' to the queue of "packet ins" to be sent to the controller.
392 * Takes ownership of 'pin' and pin->packet. */
394 ofproto_dpif_send_packet_in(struct ofproto_dpif
*ofproto
,
395 struct ofproto_packet_in
*pin
)
397 if (!guarded_list_push_back(&ofproto
->pins
, &pin
->list_node
, 1024)) {
398 COVERAGE_INC(packet_in_overflow
);
399 free(CONST_CAST(void *, pin
->up
.packet
));
403 /* Wakes up main thread for packet-in I/O. */
404 seq_change(ofproto
->pins_seq
);
407 /* The default "table-miss" behaviour for OpenFlow1.3+ is to drop the
408 * packet rather than to send the packet to the controller.
410 * This function returns false to indicate that a packet_in message
411 * for a "table-miss" should be sent to at least one controller.
412 * False otherwise. */
414 ofproto_dpif_wants_packet_in_on_miss(struct ofproto_dpif
*ofproto
)
416 return connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
);
419 /* Factory functions. */
422 init(const struct shash
*iface_hints
)
424 struct shash_node
*node
;
426 /* Make a local copy, since we don't own 'iface_hints' elements. */
427 SHASH_FOR_EACH(node
, iface_hints
) {
428 const struct iface_hint
*orig_hint
= node
->data
;
429 struct iface_hint
*new_hint
= xmalloc(sizeof *new_hint
);
431 new_hint
->br_name
= xstrdup(orig_hint
->br_name
);
432 new_hint
->br_type
= xstrdup(orig_hint
->br_type
);
433 new_hint
->ofp_port
= orig_hint
->ofp_port
;
435 shash_add(&init_ofp_ports
, node
->name
, new_hint
);
440 enumerate_types(struct sset
*types
)
442 dp_enumerate_types(types
);
446 enumerate_names(const char *type
, struct sset
*names
)
448 struct ofproto_dpif
*ofproto
;
451 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
452 if (strcmp(type
, ofproto
->up
.type
)) {
455 sset_add(names
, ofproto
->up
.name
);
462 del(const char *type
, const char *name
)
467 error
= dpif_open(name
, type
, &dpif
);
469 error
= dpif_delete(dpif
);
476 port_open_type(const char *datapath_type
, const char *port_type
)
478 return dpif_port_open_type(datapath_type
, port_type
);
481 /* Type functions. */
483 static void process_dpif_port_changes(struct dpif_backer
*);
484 static void process_dpif_all_ports_changed(struct dpif_backer
*);
485 static void process_dpif_port_change(struct dpif_backer
*,
486 const char *devname
);
487 static void process_dpif_port_error(struct dpif_backer
*, int error
);
489 static struct ofproto_dpif
*
490 lookup_ofproto_dpif_by_port_name(const char *name
)
492 struct ofproto_dpif
*ofproto
;
494 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
495 if (sset_contains(&ofproto
->ports
, name
)) {
504 type_run(const char *type
)
506 struct dpif_backer
*backer
;
508 backer
= shash_find_data(&all_dpif_backers
, type
);
510 /* This is not necessarily a problem, since backers are only
511 * created on demand. */
515 dpif_run(backer
->dpif
);
516 udpif_run(backer
->udpif
);
518 /* If vswitchd started with other_config:flow_restore_wait set as "true",
519 * and the configuration has now changed to "false", enable receiving
520 * packets from the datapath. */
521 if (!backer
->recv_set_enable
&& !ofproto_get_flow_restore_wait()) {
524 backer
->recv_set_enable
= true;
526 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
528 VLOG_ERR("Failed to enable receiving packets in dpif.");
531 dpif_flow_flush(backer
->dpif
);
532 backer
->need_revalidate
= REV_RECONFIGURE
;
535 if (backer
->recv_set_enable
) {
536 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
539 dpif_poll_threads_set(backer
->dpif
, n_dpdk_rxqs
, pmd_cpu_mask
);
541 if (backer
->need_revalidate
) {
542 struct ofproto_dpif
*ofproto
;
543 struct simap_node
*node
;
544 struct simap tmp_backers
;
546 /* Handle tunnel garbage collection. */
547 simap_init(&tmp_backers
);
548 simap_swap(&backer
->tnl_backers
, &tmp_backers
);
550 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
551 struct ofport_dpif
*iter
;
553 if (backer
!= ofproto
->backer
) {
557 HMAP_FOR_EACH (iter
, up
.hmap_node
, &ofproto
->up
.ports
) {
558 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
561 if (!iter
->is_tunnel
) {
565 dp_port
= netdev_vport_get_dpif_port(iter
->up
.netdev
,
566 namebuf
, sizeof namebuf
);
567 node
= simap_find(&tmp_backers
, dp_port
);
569 simap_put(&backer
->tnl_backers
, dp_port
, node
->data
);
570 simap_delete(&tmp_backers
, node
);
571 node
= simap_find(&backer
->tnl_backers
, dp_port
);
573 node
= simap_find(&backer
->tnl_backers
, dp_port
);
575 odp_port_t odp_port
= ODPP_NONE
;
577 if (!dpif_port_add(backer
->dpif
, iter
->up
.netdev
,
579 simap_put(&backer
->tnl_backers
, dp_port
,
580 odp_to_u32(odp_port
));
581 node
= simap_find(&backer
->tnl_backers
, dp_port
);
586 iter
->odp_port
= node
? u32_to_odp(node
->data
) : ODPP_NONE
;
587 if (tnl_port_reconfigure(iter
, iter
->up
.netdev
,
589 backer
->need_revalidate
= REV_RECONFIGURE
;
594 SIMAP_FOR_EACH (node
, &tmp_backers
) {
595 dpif_port_del(backer
->dpif
, u32_to_odp(node
->data
));
597 simap_destroy(&tmp_backers
);
599 switch (backer
->need_revalidate
) {
600 case REV_RECONFIGURE
: COVERAGE_INC(rev_reconfigure
); break;
601 case REV_STP
: COVERAGE_INC(rev_stp
); break;
602 case REV_RSTP
: COVERAGE_INC(rev_rstp
); break;
603 case REV_BOND
: COVERAGE_INC(rev_bond
); break;
604 case REV_PORT_TOGGLED
: COVERAGE_INC(rev_port_toggled
); break;
605 case REV_FLOW_TABLE
: COVERAGE_INC(rev_flow_table
); break;
606 case REV_MAC_LEARNING
: COVERAGE_INC(rev_mac_learning
); break;
607 case REV_MCAST_SNOOPING
: COVERAGE_INC(rev_mcast_snooping
); break;
609 backer
->need_revalidate
= 0;
611 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
612 struct ofport_dpif
*ofport
;
613 struct ofbundle
*bundle
;
615 if (ofproto
->backer
!= backer
) {
620 xlate_ofproto_set(ofproto
, ofproto
->up
.name
,
621 ofproto
->backer
->dpif
, ofproto
->miss_rule
,
622 ofproto
->no_packet_in_rule
,
623 ofproto
->drop_frags_rule
, ofproto
->ml
,
624 ofproto
->stp
, ofproto
->rstp
, ofproto
->ms
,
625 ofproto
->mbridge
, ofproto
->sflow
, ofproto
->ipfix
,
626 ofproto
->netflow
, ofproto
->up
.frag_handling
,
627 ofproto
->up
.forward_bpdu
,
628 connmgr_has_in_band(ofproto
->up
.connmgr
),
629 ofproto
->backer
->enable_recirc
,
630 ofproto
->backer
->variable_length_userdata
,
631 ofproto
->backer
->max_mpls_depth
,
632 ofproto
->backer
->masked_set_action
);
634 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
635 xlate_bundle_set(ofproto
, bundle
, bundle
->name
,
636 bundle
->vlan_mode
, bundle
->vlan
,
637 bundle
->trunks
, bundle
->use_priority_tags
,
638 bundle
->bond
, bundle
->lacp
,
642 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
643 int stp_port
= ofport
->stp_port
644 ? stp_port_no(ofport
->stp_port
)
646 xlate_ofport_set(ofproto
, ofport
->bundle
, ofport
,
647 ofport
->up
.ofp_port
, ofport
->odp_port
,
648 ofport
->up
.netdev
, ofport
->cfm
,
649 ofport
->bfd
, ofport
->peer
, stp_port
,
650 ofport
->rstp_port
, ofport
->qdscp
,
651 ofport
->n_qdscp
, ofport
->up
.pp
.config
,
652 ofport
->up
.pp
.state
, ofport
->is_tunnel
,
658 udpif_revalidate(backer
->udpif
);
661 process_dpif_port_changes(backer
);
666 /* Check for and handle port changes in 'backer''s dpif. */
668 process_dpif_port_changes(struct dpif_backer
*backer
)
674 error
= dpif_port_poll(backer
->dpif
, &devname
);
680 process_dpif_all_ports_changed(backer
);
684 process_dpif_port_change(backer
, devname
);
689 process_dpif_port_error(backer
, error
);
696 process_dpif_all_ports_changed(struct dpif_backer
*backer
)
698 struct ofproto_dpif
*ofproto
;
699 struct dpif_port dpif_port
;
700 struct dpif_port_dump dump
;
701 struct sset devnames
;
704 sset_init(&devnames
);
705 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
706 if (ofproto
->backer
== backer
) {
707 struct ofport
*ofport
;
709 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
710 sset_add(&devnames
, netdev_get_name(ofport
->netdev
));
714 DPIF_PORT_FOR_EACH (&dpif_port
, &dump
, backer
->dpif
) {
715 sset_add(&devnames
, dpif_port
.name
);
718 SSET_FOR_EACH (devname
, &devnames
) {
719 process_dpif_port_change(backer
, devname
);
721 sset_destroy(&devnames
);
725 process_dpif_port_change(struct dpif_backer
*backer
, const char *devname
)
727 struct ofproto_dpif
*ofproto
;
728 struct dpif_port port
;
730 /* Don't report on the datapath's device. */
731 if (!strcmp(devname
, dpif_base_name(backer
->dpif
))) {
735 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
,
736 &all_ofproto_dpifs
) {
737 if (simap_contains(&ofproto
->backer
->tnl_backers
, devname
)) {
742 ofproto
= lookup_ofproto_dpif_by_port_name(devname
);
743 if (dpif_port_query_by_name(backer
->dpif
, devname
, &port
)) {
744 /* The port was removed. If we know the datapath,
745 * report it through poll_set(). If we don't, it may be
746 * notifying us of a removal we initiated, so ignore it.
747 * If there's a pending ENOBUFS, let it stand, since
748 * everything will be reevaluated. */
749 if (ofproto
&& ofproto
->port_poll_errno
!= ENOBUFS
) {
750 sset_add(&ofproto
->port_poll_set
, devname
);
751 ofproto
->port_poll_errno
= 0;
753 } else if (!ofproto
) {
754 /* The port was added, but we don't know with which
755 * ofproto we should associate it. Delete it. */
756 dpif_port_del(backer
->dpif
, port
.port_no
);
758 struct ofport_dpif
*ofport
;
760 ofport
= ofport_dpif_cast(shash_find_data(
761 &ofproto
->up
.port_by_name
, devname
));
763 && ofport
->odp_port
!= port
.port_no
764 && !odp_port_to_ofport(backer
, port
.port_no
))
766 /* 'ofport''s datapath port number has changed from
767 * 'ofport->odp_port' to 'port.port_no'. Update our internal data
768 * structures to match. */
769 ovs_rwlock_wrlock(&backer
->odp_to_ofport_lock
);
770 hmap_remove(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
);
771 ofport
->odp_port
= port
.port_no
;
772 hmap_insert(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
,
773 hash_odp_port(port
.port_no
));
774 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
775 backer
->need_revalidate
= REV_RECONFIGURE
;
778 dpif_port_destroy(&port
);
781 /* Propagate 'error' to all ofprotos based on 'backer'. */
783 process_dpif_port_error(struct dpif_backer
*backer
, int error
)
785 struct ofproto_dpif
*ofproto
;
787 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
788 if (ofproto
->backer
== backer
) {
789 sset_clear(&ofproto
->port_poll_set
);
790 ofproto
->port_poll_errno
= error
;
796 type_wait(const char *type
)
798 struct dpif_backer
*backer
;
800 backer
= shash_find_data(&all_dpif_backers
, type
);
802 /* This is not necessarily a problem, since backers are only
803 * created on demand. */
807 dpif_wait(backer
->dpif
);
810 /* Basic life-cycle. */
812 static int add_internal_flows(struct ofproto_dpif
*);
814 static struct ofproto
*
817 struct ofproto_dpif
*ofproto
= xmalloc(sizeof *ofproto
);
822 dealloc(struct ofproto
*ofproto_
)
824 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
829 close_dpif_backer(struct dpif_backer
*backer
)
831 ovs_assert(backer
->refcount
> 0);
833 if (--backer
->refcount
) {
837 udpif_destroy(backer
->udpif
);
839 simap_destroy(&backer
->tnl_backers
);
840 ovs_rwlock_destroy(&backer
->odp_to_ofport_lock
);
841 hmap_destroy(&backer
->odp_to_ofport_map
);
842 shash_find_and_delete(&all_dpif_backers
, backer
->type
);
843 recirc_id_pool_destroy(backer
->rid_pool
);
845 free(backer
->dp_version_string
);
846 dpif_close(backer
->dpif
);
850 /* Datapath port slated for removal from datapath. */
852 struct list list_node
;
856 static bool check_variable_length_userdata(struct dpif_backer
*backer
);
857 static size_t check_max_mpls_depth(struct dpif_backer
*backer
);
858 static bool check_recirc(struct dpif_backer
*backer
);
859 static bool check_masked_set_action(struct dpif_backer
*backer
);
862 open_dpif_backer(const char *type
, struct dpif_backer
**backerp
)
864 struct dpif_backer
*backer
;
865 struct dpif_port_dump port_dump
;
866 struct dpif_port port
;
867 struct shash_node
*node
;
868 struct list garbage_list
;
869 struct odp_garbage
*garbage
, *next
;
876 backer
= shash_find_data(&all_dpif_backers
, type
);
883 backer_name
= xasprintf("ovs-%s", type
);
885 /* Remove any existing datapaths, since we assume we're the only
886 * userspace controlling the datapath. */
888 dp_enumerate_names(type
, &names
);
889 SSET_FOR_EACH(name
, &names
) {
890 struct dpif
*old_dpif
;
892 /* Don't remove our backer if it exists. */
893 if (!strcmp(name
, backer_name
)) {
897 if (dpif_open(name
, type
, &old_dpif
)) {
898 VLOG_WARN("couldn't open old datapath %s to remove it", name
);
900 dpif_delete(old_dpif
);
901 dpif_close(old_dpif
);
904 sset_destroy(&names
);
906 backer
= xmalloc(sizeof *backer
);
908 error
= dpif_create_and_open(backer_name
, type
, &backer
->dpif
);
911 VLOG_ERR("failed to open datapath of type %s: %s", type
,
912 ovs_strerror(error
));
916 backer
->udpif
= udpif_create(backer
, backer
->dpif
);
918 backer
->type
= xstrdup(type
);
919 backer
->refcount
= 1;
920 hmap_init(&backer
->odp_to_ofport_map
);
921 ovs_rwlock_init(&backer
->odp_to_ofport_lock
);
922 backer
->need_revalidate
= 0;
923 simap_init(&backer
->tnl_backers
);
924 backer
->recv_set_enable
= !ofproto_get_flow_restore_wait();
927 if (backer
->recv_set_enable
) {
928 dpif_flow_flush(backer
->dpif
);
931 /* Loop through the ports already on the datapath and remove any
932 * that we don't need anymore. */
933 list_init(&garbage_list
);
934 dpif_port_dump_start(&port_dump
, backer
->dpif
);
935 while (dpif_port_dump_next(&port_dump
, &port
)) {
936 node
= shash_find(&init_ofp_ports
, port
.name
);
937 if (!node
&& strcmp(port
.name
, dpif_base_name(backer
->dpif
))) {
938 garbage
= xmalloc(sizeof *garbage
);
939 garbage
->odp_port
= port
.port_no
;
940 list_push_front(&garbage_list
, &garbage
->list_node
);
943 dpif_port_dump_done(&port_dump
);
945 LIST_FOR_EACH_SAFE (garbage
, next
, list_node
, &garbage_list
) {
946 dpif_port_del(backer
->dpif
, garbage
->odp_port
);
947 list_remove(&garbage
->list_node
);
951 shash_add(&all_dpif_backers
, type
, backer
);
953 backer
->enable_recirc
= check_recirc(backer
);
954 backer
->max_mpls_depth
= check_max_mpls_depth(backer
);
955 backer
->masked_set_action
= check_masked_set_action(backer
);
956 backer
->rid_pool
= recirc_id_pool_create();
958 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
960 VLOG_ERR("failed to listen on datapath of type %s: %s",
961 type
, ovs_strerror(error
));
962 close_dpif_backer(backer
);
966 if (backer
->recv_set_enable
) {
967 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
970 /* This check fails if performed before udpif threads have been set,
971 * as the kernel module checks that the 'pid' in userspace action
973 backer
->variable_length_userdata
= check_variable_length_userdata(backer
);
974 backer
->dp_version_string
= dpif_get_dp_version(backer
->dpif
);
979 /* Tests whether 'backer''s datapath supports recirculation. Only newer
980 * datapaths support OVS_KEY_ATTR_RECIRC_ID in keys. We need to disable some
981 * features on older datapaths that don't support this feature.
983 * Returns false if 'backer' definitely does not support recirculation, true if
984 * it seems to support recirculation or if at least the error we get is
987 check_recirc(struct dpif_backer
*backer
)
990 struct odputil_keybuf keybuf
;
993 bool enable_recirc
= false;
995 memset(&flow
, 0, sizeof flow
);
999 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1000 odp_flow_key_from_flow(&key
, &flow
, NULL
, 0, true);
1002 error
= dpif_flow_put(backer
->dpif
, DPIF_FP_CREATE
| DPIF_FP_PROBE
,
1003 ofpbuf_data(&key
), ofpbuf_size(&key
), NULL
, 0, NULL
,
1005 if (error
&& error
!= EEXIST
) {
1006 if (error
!= EINVAL
) {
1007 VLOG_WARN("%s: Reciculation flow probe failed (%s)",
1008 dpif_name(backer
->dpif
), ovs_strerror(error
));
1013 error
= dpif_flow_del(backer
->dpif
, ofpbuf_data(&key
), ofpbuf_size(&key
),
1016 VLOG_WARN("%s: failed to delete recirculation feature probe flow",
1017 dpif_name(backer
->dpif
));
1020 enable_recirc
= true;
1023 if (enable_recirc
) {
1024 VLOG_INFO("%s: Datapath supports recirculation",
1025 dpif_name(backer
->dpif
));
1027 VLOG_INFO("%s: Datapath does not support recirculation",
1028 dpif_name(backer
->dpif
));
1031 return enable_recirc
;
1034 /* Tests whether 'backer''s datapath supports variable-length
1035 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. We need
1036 * to disable some features on older datapaths that don't support this
1039 * Returns false if 'backer' definitely does not support variable-length
1040 * userdata, true if it seems to support them or if at least the error we get
1043 check_variable_length_userdata(struct dpif_backer
*backer
)
1045 struct eth_header
*eth
;
1046 struct ofpbuf actions
;
1047 struct dpif_execute execute
;
1048 struct ofpbuf packet
;
1052 /* Compose a userspace action that will cause an ERANGE error on older
1053 * datapaths that don't support variable-length userdata.
1055 * We really test for using userdata longer than 8 bytes, but older
1056 * datapaths accepted these, silently truncating the userdata to 8 bytes.
1057 * The same older datapaths rejected userdata shorter than 8 bytes, so we
1058 * test for that instead as a proxy for longer userdata support. */
1059 ofpbuf_init(&actions
, 64);
1060 start
= nl_msg_start_nested(&actions
, OVS_ACTION_ATTR_USERSPACE
);
1061 nl_msg_put_u32(&actions
, OVS_USERSPACE_ATTR_PID
,
1062 dpif_port_get_pid(backer
->dpif
, ODPP_NONE
, 0));
1063 nl_msg_put_unspec_zero(&actions
, OVS_USERSPACE_ATTR_USERDATA
, 4);
1064 nl_msg_end_nested(&actions
, start
);
1066 /* Compose a dummy ethernet packet. */
1067 ofpbuf_init(&packet
, ETH_HEADER_LEN
);
1068 eth
= ofpbuf_put_zeros(&packet
, ETH_HEADER_LEN
);
1069 eth
->eth_type
= htons(0x1234);
1071 /* Execute the actions. On older datapaths this fails with ERANGE, on
1072 * newer datapaths it succeeds. */
1073 execute
.actions
= ofpbuf_data(&actions
);
1074 execute
.actions_len
= ofpbuf_size(&actions
);
1075 execute
.packet
= &packet
;
1076 execute
.md
= PKT_METADATA_INITIALIZER(0);
1077 execute
.needs_help
= false;
1078 execute
.probe
= true;
1080 error
= dpif_execute(backer
->dpif
, &execute
);
1082 ofpbuf_uninit(&packet
);
1083 ofpbuf_uninit(&actions
);
1090 /* Variable-length userdata is not supported. */
1091 VLOG_WARN("%s: datapath does not support variable-length userdata "
1092 "feature (needs Linux 3.10+ or kernel module from OVS "
1093 "1..11+). The NXAST_SAMPLE action will be ignored.",
1094 dpif_name(backer
->dpif
));
1098 /* Something odd happened. We're not sure whether variable-length
1099 * userdata is supported. Default to "yes". */
1100 VLOG_WARN("%s: variable-length userdata feature probe failed (%s)",
1101 dpif_name(backer
->dpif
), ovs_strerror(error
));
1106 /* Tests the MPLS label stack depth supported by 'backer''s datapath.
1108 * Returns the number of elements in a struct flow's mpls_lse field
1109 * if the datapath supports at least that many entries in an
1111 * Otherwise returns the number of MPLS push actions supported by
1114 check_max_mpls_depth(struct dpif_backer
*backer
)
1119 for (n
= 0; n
< FLOW_MAX_MPLS_LABELS
; n
++) {
1120 struct odputil_keybuf keybuf
;
1124 memset(&flow
, 0, sizeof flow
);
1125 flow
.dl_type
= htons(ETH_TYPE_MPLS
);
1126 flow_set_mpls_bos(&flow
, n
, 1);
1128 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1129 odp_flow_key_from_flow(&key
, &flow
, NULL
, 0, false);
1131 error
= dpif_flow_put(backer
->dpif
, DPIF_FP_CREATE
| DPIF_FP_PROBE
,
1132 ofpbuf_data(&key
), ofpbuf_size(&key
), NULL
, 0,
1134 if (error
&& error
!= EEXIST
) {
1135 if (error
!= EINVAL
) {
1136 VLOG_WARN("%s: MPLS stack length feature probe failed (%s)",
1137 dpif_name(backer
->dpif
), ovs_strerror(error
));
1142 error
= dpif_flow_del(backer
->dpif
, ofpbuf_data(&key
),
1143 ofpbuf_size(&key
), NULL
);
1145 VLOG_WARN("%s: failed to delete MPLS feature probe flow",
1146 dpif_name(backer
->dpif
));
1150 VLOG_INFO("%s: MPLS label stack length probed as %d",
1151 dpif_name(backer
->dpif
), n
);
1155 /* Tests whether 'backer''s datapath supports masked data in
1156 * OVS_ACTION_ATTR_SET actions. We need to disable some features on older
1157 * datapaths that don't support this feature. */
1159 check_masked_set_action(struct dpif_backer
*backer
)
1161 struct eth_header
*eth
;
1162 struct ofpbuf actions
;
1163 struct dpif_execute execute
;
1164 struct ofpbuf packet
;
1166 struct ovs_key_ethernet key
, mask
;
1168 /* Compose a set action that will cause an EINVAL error on older
1169 * datapaths that don't support masked set actions.
1170 * Avoid using a full mask, as it could be translated to a non-masked
1171 * set action instead. */
1172 ofpbuf_init(&actions
, 64);
1173 memset(&key
, 0x53, sizeof key
);
1174 memset(&mask
, 0x7f, sizeof mask
);
1175 commit_masked_set_action(&actions
, OVS_KEY_ATTR_ETHERNET
, &key
, &mask
,
1178 /* Compose a dummy ethernet packet. */
1179 ofpbuf_init(&packet
, ETH_HEADER_LEN
);
1180 eth
= ofpbuf_put_zeros(&packet
, ETH_HEADER_LEN
);
1181 eth
->eth_type
= htons(0x1234);
1183 /* Execute the actions. On older datapaths this fails with EINVAL, on
1184 * newer datapaths it succeeds. */
1185 execute
.actions
= ofpbuf_data(&actions
);
1186 execute
.actions_len
= ofpbuf_size(&actions
);
1187 execute
.packet
= &packet
;
1188 execute
.md
= PKT_METADATA_INITIALIZER(0);
1189 execute
.needs_help
= false;
1190 execute
.probe
= true;
1192 error
= dpif_execute(backer
->dpif
, &execute
);
1194 ofpbuf_uninit(&packet
);
1195 ofpbuf_uninit(&actions
);
1198 /* Masked set action is not supported. */
1199 VLOG_INFO("%s: datapath does not support masked set action feature.",
1200 dpif_name(backer
->dpif
));
1206 construct(struct ofproto
*ofproto_
)
1208 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1209 struct shash_node
*node
, *next
;
1212 error
= open_dpif_backer(ofproto
->up
.type
, &ofproto
->backer
);
1217 ofproto
->netflow
= NULL
;
1218 ofproto
->sflow
= NULL
;
1219 ofproto
->ipfix
= NULL
;
1220 ofproto
->stp
= NULL
;
1221 ofproto
->rstp
= NULL
;
1222 ofproto
->dump_seq
= 0;
1223 hmap_init(&ofproto
->bundles
);
1224 ofproto
->ml
= mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME
);
1226 ofproto
->mbridge
= mbridge_create();
1227 ofproto
->has_bonded_bundles
= false;
1228 ofproto
->lacp_enabled
= false;
1229 ovs_mutex_init_adaptive(&ofproto
->stats_mutex
);
1230 ovs_mutex_init(&ofproto
->vsp_mutex
);
1232 guarded_list_init(&ofproto
->pins
);
1234 ofproto_dpif_unixctl_init();
1235 ovs_router_unixctl_register();
1237 hmap_init(&ofproto
->vlandev_map
);
1238 hmap_init(&ofproto
->realdev_vid_map
);
1240 sset_init(&ofproto
->ports
);
1241 sset_init(&ofproto
->ghost_ports
);
1242 sset_init(&ofproto
->port_poll_set
);
1243 ofproto
->port_poll_errno
= 0;
1244 ofproto
->change_seq
= 0;
1245 ofproto
->pins_seq
= seq_create();
1246 ofproto
->pins_seqno
= seq_read(ofproto
->pins_seq
);
1249 SHASH_FOR_EACH_SAFE (node
, next
, &init_ofp_ports
) {
1250 struct iface_hint
*iface_hint
= node
->data
;
1252 if (!strcmp(iface_hint
->br_name
, ofproto
->up
.name
)) {
1253 /* Check if the datapath already has this port. */
1254 if (dpif_port_exists(ofproto
->backer
->dpif
, node
->name
)) {
1255 sset_add(&ofproto
->ports
, node
->name
);
1258 free(iface_hint
->br_name
);
1259 free(iface_hint
->br_type
);
1261 shash_delete(&init_ofp_ports
, node
);
1265 hmap_insert(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
,
1266 hash_string(ofproto
->up
.name
, 0));
1267 memset(&ofproto
->stats
, 0, sizeof ofproto
->stats
);
1269 ofproto_init_tables(ofproto_
, N_TABLES
);
1270 error
= add_internal_flows(ofproto
);
1272 ofproto
->up
.tables
[TBL_INTERNAL
].flags
= OFTABLE_HIDDEN
| OFTABLE_READONLY
;
1278 add_internal_miss_flow(struct ofproto_dpif
*ofproto
, int id
,
1279 const struct ofpbuf
*ofpacts
, struct rule_dpif
**rulep
)
1285 match_init_catchall(&match
);
1286 match_set_reg(&match
, 0, id
);
1288 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 0, 0, ofpacts
,
1290 *rulep
= error
? NULL
: rule_dpif_cast(rule
);
1296 add_internal_flows(struct ofproto_dpif
*ofproto
)
1298 struct ofpact_controller
*controller
;
1299 uint64_t ofpacts_stub
[128 / 8];
1300 struct ofpbuf ofpacts
;
1301 struct rule
*unused_rulep OVS_UNUSED
;
1302 struct ofpact_resubmit
*resubmit
;
1307 ofpbuf_use_stack(&ofpacts
, ofpacts_stub
, sizeof ofpacts_stub
);
1310 controller
= ofpact_put_CONTROLLER(&ofpacts
);
1311 controller
->max_len
= UINT16_MAX
;
1312 controller
->controller_id
= 0;
1313 controller
->reason
= OFPR_NO_MATCH
;
1314 ofpact_pad(&ofpacts
);
1316 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1317 &ofproto
->miss_rule
);
1322 ofpbuf_clear(&ofpacts
);
1323 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1324 &ofproto
->no_packet_in_rule
);
1329 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1330 &ofproto
->drop_frags_rule
);
1335 /* Continue non-recirculation rule lookups from table 0.
1337 * (priority=2), recirc=0, actions=resubmit(, 0)
1339 resubmit
= ofpact_put_RESUBMIT(&ofpacts
);
1340 resubmit
->in_port
= OFPP_IN_PORT
;
1341 resubmit
->table_id
= 0;
1343 match_init_catchall(&match
);
1344 match_set_recirc_id(&match
, 0);
1346 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 2, 0, &ofpacts
,
1352 /* Drop any run away recirc rule lookups. Recirc_id has to be
1353 * non-zero when reaching this rule.
1355 * (priority=1), *, actions=drop
1357 ofpbuf_clear(&ofpacts
);
1358 match_init_catchall(&match
);
1359 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 1, 0, &ofpacts
,
1366 destruct(struct ofproto
*ofproto_
)
1368 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1369 struct ofproto_packet_in
*pin
, *next_pin
;
1370 struct rule_dpif
*rule
;
1371 struct oftable
*table
;
1374 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1376 xlate_remove_ofproto(ofproto
);
1379 /* Ensure that the upcall processing threads have no remaining references
1380 * to the ofproto or anything in it. */
1381 udpif_synchronize(ofproto
->backer
->udpif
);
1383 hmap_remove(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
);
1385 OFPROTO_FOR_EACH_TABLE (table
, &ofproto
->up
) {
1386 CLS_FOR_EACH_SAFE (rule
, up
.cr
, &table
->cls
) {
1387 ofproto_rule_delete(&ofproto
->up
, &rule
->up
);
1391 guarded_list_pop_all(&ofproto
->pins
, &pins
);
1392 LIST_FOR_EACH_SAFE (pin
, next_pin
, list_node
, &pins
) {
1393 list_remove(&pin
->list_node
);
1394 free(CONST_CAST(void *, pin
->up
.packet
));
1397 guarded_list_destroy(&ofproto
->pins
);
1399 mbridge_unref(ofproto
->mbridge
);
1401 netflow_unref(ofproto
->netflow
);
1402 dpif_sflow_unref(ofproto
->sflow
);
1403 dpif_ipfix_unref(ofproto
->ipfix
);
1404 hmap_destroy(&ofproto
->bundles
);
1405 mac_learning_unref(ofproto
->ml
);
1406 mcast_snooping_unref(ofproto
->ms
);
1408 hmap_destroy(&ofproto
->vlandev_map
);
1409 hmap_destroy(&ofproto
->realdev_vid_map
);
1411 sset_destroy(&ofproto
->ports
);
1412 sset_destroy(&ofproto
->ghost_ports
);
1413 sset_destroy(&ofproto
->port_poll_set
);
1415 ovs_mutex_destroy(&ofproto
->stats_mutex
);
1416 ovs_mutex_destroy(&ofproto
->vsp_mutex
);
1418 seq_destroy(ofproto
->pins_seq
);
1420 close_dpif_backer(ofproto
->backer
);
1424 run(struct ofproto
*ofproto_
)
1426 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1427 uint64_t new_seq
, new_dump_seq
;
1429 if (mbridge_need_revalidate(ofproto
->mbridge
)) {
1430 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1431 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1432 mac_learning_flush(ofproto
->ml
);
1433 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1434 mcast_snooping_mdb_flush(ofproto
->ms
);
1437 /* Always updates the ofproto->pins_seqno to avoid frequent wakeup during
1438 * flow restore. Even though nothing is processed during flow restore,
1439 * all queued 'pins' will be handled immediately when flow restore
1441 ofproto
->pins_seqno
= seq_read(ofproto
->pins_seq
);
1443 /* Do not perform any periodic activity required by 'ofproto' while
1444 * waiting for flow restore to complete. */
1445 if (!ofproto_get_flow_restore_wait()) {
1446 struct ofproto_packet_in
*pin
, *next_pin
;
1449 guarded_list_pop_all(&ofproto
->pins
, &pins
);
1450 LIST_FOR_EACH_SAFE (pin
, next_pin
, list_node
, &pins
) {
1451 connmgr_send_packet_in(ofproto
->up
.connmgr
, pin
);
1452 list_remove(&pin
->list_node
);
1453 free(CONST_CAST(void *, pin
->up
.packet
));
1458 if (ofproto
->netflow
) {
1459 netflow_run(ofproto
->netflow
);
1461 if (ofproto
->sflow
) {
1462 dpif_sflow_run(ofproto
->sflow
);
1464 if (ofproto
->ipfix
) {
1465 dpif_ipfix_run(ofproto
->ipfix
);
1468 new_seq
= seq_read(connectivity_seq_get());
1469 if (ofproto
->change_seq
!= new_seq
) {
1470 struct ofport_dpif
*ofport
;
1472 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1476 ofproto
->change_seq
= new_seq
;
1478 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1479 struct ofbundle
*bundle
;
1481 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1488 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1489 if (mac_learning_run(ofproto
->ml
)) {
1490 ofproto
->backer
->need_revalidate
= REV_MAC_LEARNING
;
1492 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1494 if (mcast_snooping_run(ofproto
->ms
)) {
1495 ofproto
->backer
->need_revalidate
= REV_MCAST_SNOOPING
;
1498 new_dump_seq
= seq_read(udpif_dump_seq(ofproto
->backer
->udpif
));
1499 if (ofproto
->dump_seq
!= new_dump_seq
) {
1500 struct rule
*rule
, *next_rule
;
1502 /* We know stats are relatively fresh, so now is a good time to do some
1504 ofproto
->dump_seq
= new_dump_seq
;
1506 /* Expire OpenFlow flows whose idle_timeout or hard_timeout
1508 ovs_mutex_lock(&ofproto_mutex
);
1509 LIST_FOR_EACH_SAFE (rule
, next_rule
, expirable
,
1510 &ofproto
->up
.expirable
) {
1511 rule_expire(rule_dpif_cast(rule
));
1513 ovs_mutex_unlock(&ofproto_mutex
);
1515 /* All outstanding data in existing flows has been accounted, so it's a
1516 * good time to do bond rebalancing. */
1517 if (ofproto
->has_bonded_bundles
) {
1518 struct ofbundle
*bundle
;
1520 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1522 bond_rebalance(bundle
->bond
);
1532 wait(struct ofproto
*ofproto_
)
1534 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1536 if (ofproto_get_flow_restore_wait()) {
1540 if (ofproto
->sflow
) {
1541 dpif_sflow_wait(ofproto
->sflow
);
1543 if (ofproto
->ipfix
) {
1544 dpif_ipfix_wait(ofproto
->ipfix
);
1546 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1547 struct ofbundle
*bundle
;
1549 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1550 bundle_wait(bundle
);
1553 if (ofproto
->netflow
) {
1554 netflow_wait(ofproto
->netflow
);
1556 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
1557 mac_learning_wait(ofproto
->ml
);
1558 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1559 mcast_snooping_wait(ofproto
->ms
);
1561 if (ofproto
->backer
->need_revalidate
) {
1562 /* Shouldn't happen, but if it does just go around again. */
1563 VLOG_DBG_RL(&rl
, "need revalidate in ofproto_wait_cb()");
1564 poll_immediate_wake();
1567 seq_wait(udpif_dump_seq(ofproto
->backer
->udpif
), ofproto
->dump_seq
);
1568 seq_wait(ofproto
->pins_seq
, ofproto
->pins_seqno
);
1572 type_get_memory_usage(const char *type
, struct simap
*usage
)
1574 struct dpif_backer
*backer
;
1576 backer
= shash_find_data(&all_dpif_backers
, type
);
1578 udpif_get_memory_usage(backer
->udpif
, usage
);
1583 flush(struct ofproto
*ofproto_
)
1585 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1586 struct dpif_backer
*backer
= ofproto
->backer
;
1589 udpif_flush(backer
->udpif
);
1594 query_tables(struct ofproto
*ofproto
,
1595 struct ofputil_table_features
*features
,
1596 struct ofputil_table_stats
*stats
)
1598 strcpy(features
->name
, "classifier");
1603 for (i
= 0; i
< ofproto
->n_tables
; i
++) {
1604 unsigned long missed
, matched
;
1606 atomic_read_relaxed(&ofproto
->tables
[i
].n_matched
, &matched
);
1607 atomic_read_relaxed(&ofproto
->tables
[i
].n_missed
, &missed
);
1609 stats
[i
].matched_count
= matched
;
1610 stats
[i
].lookup_count
= matched
+ missed
;
1615 static struct ofport
*
1618 struct ofport_dpif
*port
= xmalloc(sizeof *port
);
1623 port_dealloc(struct ofport
*port_
)
1625 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1630 port_construct(struct ofport
*port_
)
1632 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1633 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1634 const struct netdev
*netdev
= port
->up
.netdev
;
1635 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1636 struct dpif_port dpif_port
;
1639 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1640 port
->bundle
= NULL
;
1643 port
->may_enable
= false;
1644 port
->stp_port
= NULL
;
1645 port
->stp_state
= STP_DISABLED
;
1646 port
->rstp_port
= NULL
;
1647 port
->rstp_state
= RSTP_DISABLED
;
1648 port
->is_tunnel
= false;
1652 port
->realdev_ofp_port
= 0;
1653 port
->vlandev_vid
= 0;
1654 port
->carrier_seq
= netdev_get_carrier_resets(netdev
);
1655 port
->is_layer3
= netdev_vport_is_layer3(netdev
);
1657 if (netdev_vport_is_patch(netdev
)) {
1658 /* By bailing out here, we don't submit the port to the sFlow module
1659 * to be considered for counter polling export. This is correct
1660 * because the patch port represents an interface that sFlow considers
1661 * to be "internal" to the switch as a whole, and therefore not an
1662 * candidate for counter polling. */
1663 port
->odp_port
= ODPP_NONE
;
1664 ofport_update_peer(port
);
1668 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
1669 netdev_vport_get_dpif_port(netdev
, namebuf
,
1676 port
->odp_port
= dpif_port
.port_no
;
1678 if (netdev_get_tunnel_config(netdev
)) {
1679 tnl_port_add(port
, port
->up
.netdev
, port
->odp_port
);
1680 port
->is_tunnel
= true;
1681 if (ofproto
->ipfix
) {
1682 dpif_ipfix_add_tunnel_port(ofproto
->ipfix
, port_
, port
->odp_port
);
1685 /* Sanity-check that a mapping doesn't already exist. This
1686 * shouldn't happen for non-tunnel ports. */
1687 if (odp_port_to_ofp_port(ofproto
, port
->odp_port
) != OFPP_NONE
) {
1688 VLOG_ERR("port %s already has an OpenFlow port number",
1690 dpif_port_destroy(&dpif_port
);
1694 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1695 hmap_insert(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
,
1696 hash_odp_port(port
->odp_port
));
1697 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1699 dpif_port_destroy(&dpif_port
);
1701 if (ofproto
->sflow
) {
1702 dpif_sflow_add_port(ofproto
->sflow
, port_
, port
->odp_port
);
1709 port_destruct(struct ofport
*port_
)
1711 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1712 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1713 const char *devname
= netdev_get_name(port
->up
.netdev
);
1714 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1715 const char *dp_port_name
;
1717 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1719 xlate_ofport_remove(port
);
1722 dp_port_name
= netdev_vport_get_dpif_port(port
->up
.netdev
, namebuf
,
1724 if (dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
1725 /* The underlying device is still there, so delete it. This
1726 * happens when the ofproto is being destroyed, since the caller
1727 * assumes that removal of attached ports will happen as part of
1729 if (!port
->is_tunnel
) {
1730 dpif_port_del(ofproto
->backer
->dpif
, port
->odp_port
);
1735 port
->peer
->peer
= NULL
;
1739 if (port
->odp_port
!= ODPP_NONE
&& !port
->is_tunnel
) {
1740 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1741 hmap_remove(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
);
1742 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1745 if (port
->is_tunnel
&& ofproto
->ipfix
) {
1746 dpif_ipfix_del_tunnel_port(ofproto
->ipfix
, port
->odp_port
);
1750 sset_find_and_delete(&ofproto
->ports
, devname
);
1751 sset_find_and_delete(&ofproto
->ghost_ports
, devname
);
1752 bundle_remove(port_
);
1753 set_cfm(port_
, NULL
);
1754 set_bfd(port_
, NULL
);
1755 if (port
->stp_port
) {
1756 stp_port_disable(port
->stp_port
);
1758 set_rstp_port(port_
, NULL
);
1759 if (ofproto
->sflow
) {
1760 dpif_sflow_del_port(ofproto
->sflow
, port
->odp_port
);
1767 port_modified(struct ofport
*port_
)
1769 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1771 if (port
->bundle
&& port
->bundle
->bond
) {
1772 bond_slave_set_netdev(port
->bundle
->bond
, port
, port
->up
.netdev
);
1776 cfm_set_netdev(port
->cfm
, port
->up
.netdev
);
1780 bfd_set_netdev(port
->bfd
, port
->up
.netdev
);
1783 ofproto_dpif_monitor_port_update(port
, port
->bfd
, port
->cfm
,
1784 port
->up
.pp
.hw_addr
);
1786 if (port
->is_tunnel
&& tnl_port_reconfigure(port
, port
->up
.netdev
,
1788 ofproto_dpif_cast(port
->up
.ofproto
)->backer
->need_revalidate
=
1792 ofport_update_peer(port
);
1796 port_reconfigured(struct ofport
*port_
, enum ofputil_port_config old_config
)
1798 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1799 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1800 enum ofputil_port_config changed
= old_config
^ port
->up
.pp
.config
;
1802 if (changed
& (OFPUTIL_PC_NO_RECV
| OFPUTIL_PC_NO_RECV_STP
|
1803 OFPUTIL_PC_NO_FWD
| OFPUTIL_PC_NO_FLOOD
|
1804 OFPUTIL_PC_NO_PACKET_IN
)) {
1805 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1807 if (changed
& OFPUTIL_PC_NO_FLOOD
&& port
->bundle
) {
1808 bundle_update(port
->bundle
);
1814 set_sflow(struct ofproto
*ofproto_
,
1815 const struct ofproto_sflow_options
*sflow_options
)
1817 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1818 struct dpif_sflow
*ds
= ofproto
->sflow
;
1820 if (sflow_options
) {
1822 struct ofport_dpif
*ofport
;
1824 ds
= ofproto
->sflow
= dpif_sflow_create();
1825 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1826 dpif_sflow_add_port(ds
, &ofport
->up
, ofport
->odp_port
);
1828 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1830 dpif_sflow_set_options(ds
, sflow_options
);
1833 dpif_sflow_unref(ds
);
1834 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1835 ofproto
->sflow
= NULL
;
1843 struct ofproto
*ofproto_
,
1844 const struct ofproto_ipfix_bridge_exporter_options
*bridge_exporter_options
,
1845 const struct ofproto_ipfix_flow_exporter_options
*flow_exporters_options
,
1846 size_t n_flow_exporters_options
)
1848 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1849 struct dpif_ipfix
*di
= ofproto
->ipfix
;
1850 bool has_options
= bridge_exporter_options
|| flow_exporters_options
;
1851 bool new_di
= false;
1853 if (has_options
&& !di
) {
1854 di
= ofproto
->ipfix
= dpif_ipfix_create();
1859 /* Call set_options in any case to cleanly flush the flow
1860 * caches in the last exporters that are to be destroyed. */
1861 dpif_ipfix_set_options(
1862 di
, bridge_exporter_options
, flow_exporters_options
,
1863 n_flow_exporters_options
);
1865 /* Add tunnel ports only when a new ipfix created */
1866 if (new_di
== true) {
1867 struct ofport_dpif
*ofport
;
1868 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1869 if (ofport
->is_tunnel
== true) {
1870 dpif_ipfix_add_tunnel_port(di
, &ofport
->up
, ofport
->odp_port
);
1876 dpif_ipfix_unref(di
);
1877 ofproto
->ipfix
= NULL
;
1885 set_cfm(struct ofport
*ofport_
, const struct cfm_settings
*s
)
1887 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1892 struct ofproto_dpif
*ofproto
;
1894 ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
1895 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1896 ofport
->cfm
= cfm_create(ofport
->up
.netdev
);
1899 if (cfm_configure(ofport
->cfm
, s
)) {
1906 cfm_unref(ofport
->cfm
);
1909 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
1910 ofport
->up
.pp
.hw_addr
);
1915 cfm_status_changed(struct ofport
*ofport_
)
1917 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1919 return ofport
->cfm
? cfm_check_status_change(ofport
->cfm
) : true;
1923 get_cfm_status(const struct ofport
*ofport_
,
1924 struct cfm_status
*status
)
1926 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1930 cfm_get_status(ofport
->cfm
, status
);
1939 set_bfd(struct ofport
*ofport_
, const struct smap
*cfg
)
1941 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
1942 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1946 ofport
->bfd
= bfd_configure(old
, netdev_get_name(ofport
->up
.netdev
),
1947 cfg
, ofport
->up
.netdev
);
1948 if (ofport
->bfd
!= old
) {
1949 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1951 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
1952 ofport
->up
.pp
.hw_addr
);
1957 bfd_status_changed(struct ofport
*ofport_
)
1959 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1961 return ofport
->bfd
? bfd_check_status_change(ofport
->bfd
) : true;
1965 get_bfd_status(struct ofport
*ofport_
, struct smap
*smap
)
1967 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1971 bfd_get_status(ofport
->bfd
, smap
);
1979 /* Spanning Tree. */
1981 /* Called while rstp_mutex is held. */
1983 rstp_send_bpdu_cb(struct ofpbuf
*pkt
, void *ofport_
, void *ofproto_
)
1985 struct ofproto_dpif
*ofproto
= ofproto_
;
1986 struct ofport_dpif
*ofport
= ofport_
;
1987 struct eth_header
*eth
= ofpbuf_l2(pkt
);
1989 netdev_get_etheraddr(ofport
->up
.netdev
, eth
->eth_src
);
1990 if (eth_addr_is_zero(eth
->eth_src
)) {
1991 VLOG_WARN_RL(&rl
, "%s port %d: cannot send RSTP BPDU on a port which "
1992 "does not have a configured source MAC address.",
1993 ofproto
->up
.name
, ofp_to_u16(ofport
->up
.ofp_port
));
1995 ofproto_dpif_send_packet(ofport
, pkt
);
2001 send_bpdu_cb(struct ofpbuf
*pkt
, int port_num
, void *ofproto_
)
2003 struct ofproto_dpif
*ofproto
= ofproto_
;
2004 struct stp_port
*sp
= stp_get_port(ofproto
->stp
, port_num
);
2005 struct ofport_dpif
*ofport
;
2007 ofport
= stp_port_get_aux(sp
);
2009 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on unknown port %d",
2010 ofproto
->up
.name
, port_num
);
2012 struct eth_header
*eth
= ofpbuf_l2(pkt
);
2014 netdev_get_etheraddr(ofport
->up
.netdev
, eth
->eth_src
);
2015 if (eth_addr_is_zero(eth
->eth_src
)) {
2016 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on port %d "
2017 "with unknown MAC", ofproto
->up
.name
, port_num
);
2019 ofproto_dpif_send_packet(ofport
, pkt
);
2025 /* Configure RSTP on 'ofproto_' using the settings defined in 's'. */
2027 set_rstp(struct ofproto
*ofproto_
, const struct ofproto_rstp_settings
*s
)
2029 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2031 /* Only revalidate flows if the configuration changed. */
2032 if (!s
!= !ofproto
->rstp
) {
2033 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2037 if (!ofproto
->rstp
) {
2038 ofproto
->rstp
= rstp_create(ofproto_
->name
, s
->address
,
2039 rstp_send_bpdu_cb
, ofproto
);
2040 ofproto
->rstp_last_tick
= time_msec();
2042 rstp_set_bridge_address(ofproto
->rstp
, s
->address
);
2043 rstp_set_bridge_priority(ofproto
->rstp
, s
->priority
);
2044 rstp_set_bridge_ageing_time(ofproto
->rstp
, s
->ageing_time
);
2045 rstp_set_bridge_force_protocol_version(ofproto
->rstp
,
2046 s
->force_protocol_version
);
2047 rstp_set_bridge_max_age(ofproto
->rstp
, s
->bridge_max_age
);
2048 rstp_set_bridge_forward_delay(ofproto
->rstp
, s
->bridge_forward_delay
);
2049 rstp_set_bridge_transmit_hold_count(ofproto
->rstp
,
2050 s
->transmit_hold_count
);
2052 struct ofport
*ofport
;
2053 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2054 set_rstp_port(ofport
, NULL
);
2056 rstp_unref(ofproto
->rstp
);
2057 ofproto
->rstp
= NULL
;
2062 get_rstp_status(struct ofproto
*ofproto_
, struct ofproto_rstp_status
*s
)
2064 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2066 if (ofproto
->rstp
) {
2068 s
->root_id
= rstp_get_root_id(ofproto
->rstp
);
2069 s
->bridge_id
= rstp_get_bridge_id(ofproto
->rstp
);
2070 s
->designated_id
= rstp_get_designated_id(ofproto
->rstp
);
2071 s
->root_path_cost
= rstp_get_root_path_cost(ofproto
->rstp
);
2072 s
->designated_port_id
= rstp_get_designated_port_id(ofproto
->rstp
);
2073 s
->bridge_port_id
= rstp_get_bridge_port_id(ofproto
->rstp
);
2080 update_rstp_port_state(struct ofport_dpif
*ofport
)
2082 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2083 enum rstp_state state
;
2085 /* Figure out new state. */
2086 state
= ofport
->rstp_port
? rstp_port_get_state(ofport
->rstp_port
)
2090 if (ofport
->rstp_state
!= state
) {
2091 enum ofputil_port_state of_state
;
2094 VLOG_DBG("port %s: RSTP state changed from %s to %s",
2095 netdev_get_name(ofport
->up
.netdev
),
2096 rstp_state_name(ofport
->rstp_state
),
2097 rstp_state_name(state
));
2098 if (rstp_learn_in_state(ofport
->rstp_state
)
2099 != rstp_learn_in_state(state
)) {
2100 /* xxx Learning action flows should also be flushed. */
2101 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2102 mac_learning_flush(ofproto
->ml
);
2103 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2105 fwd_change
= rstp_forward_in_state(ofport
->rstp_state
)
2106 != rstp_forward_in_state(state
);
2108 ofproto
->backer
->need_revalidate
= REV_RSTP
;
2109 ofport
->rstp_state
= state
;
2111 if (fwd_change
&& ofport
->bundle
) {
2112 bundle_update(ofport
->bundle
);
2115 /* Update the RSTP state bits in the OpenFlow port description. */
2116 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2117 of_state
|= (state
== RSTP_LEARNING
? OFPUTIL_PS_STP_LEARN
2118 : state
== RSTP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2119 : state
== RSTP_DISCARDING
? OFPUTIL_PS_STP_LISTEN
2121 ofproto_port_set_state(&ofport
->up
, of_state
);
2126 rstp_run(struct ofproto_dpif
*ofproto
)
2128 if (ofproto
->rstp
) {
2129 long long int now
= time_msec();
2130 long long int elapsed
= now
- ofproto
->rstp_last_tick
;
2131 struct rstp_port
*rp
;
2132 struct ofport_dpif
*ofport
;
2134 /* Every second, decrease the values of the timers. */
2135 if (elapsed
>= 1000) {
2136 rstp_tick_timers(ofproto
->rstp
);
2137 ofproto
->rstp_last_tick
= now
;
2140 while ((ofport
= rstp_get_next_changed_port_aux(ofproto
->rstp
, &rp
))) {
2141 update_rstp_port_state(ofport
);
2143 /* FIXME: This check should be done on-event (i.e., when setting
2144 * p->fdb_flush) and not periodically.
2146 if (rstp_check_and_reset_fdb_flush(ofproto
->rstp
)) {
2147 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2148 /* FIXME: RSTP should be able to flush the entries pertaining to a
2149 * single port, not the whole table.
2151 mac_learning_flush(ofproto
->ml
);
2152 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2157 /* Configures STP on 'ofproto_' using the settings defined in 's'. */
2159 set_stp(struct ofproto
*ofproto_
, const struct ofproto_stp_settings
*s
)
2161 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2163 /* Only revalidate flows if the configuration changed. */
2164 if (!s
!= !ofproto
->stp
) {
2165 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2169 if (!ofproto
->stp
) {
2170 ofproto
->stp
= stp_create(ofproto_
->name
, s
->system_id
,
2171 send_bpdu_cb
, ofproto
);
2172 ofproto
->stp_last_tick
= time_msec();
2175 stp_set_bridge_id(ofproto
->stp
, s
->system_id
);
2176 stp_set_bridge_priority(ofproto
->stp
, s
->priority
);
2177 stp_set_hello_time(ofproto
->stp
, s
->hello_time
);
2178 stp_set_max_age(ofproto
->stp
, s
->max_age
);
2179 stp_set_forward_delay(ofproto
->stp
, s
->fwd_delay
);
2181 struct ofport
*ofport
;
2183 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2184 set_stp_port(ofport
, NULL
);
2187 stp_unref(ofproto
->stp
);
2188 ofproto
->stp
= NULL
;
2195 get_stp_status(struct ofproto
*ofproto_
, struct ofproto_stp_status
*s
)
2197 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2201 s
->bridge_id
= stp_get_bridge_id(ofproto
->stp
);
2202 s
->designated_root
= stp_get_designated_root(ofproto
->stp
);
2203 s
->root_path_cost
= stp_get_root_path_cost(ofproto
->stp
);
2212 update_stp_port_state(struct ofport_dpif
*ofport
)
2214 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2215 enum stp_state state
;
2217 /* Figure out new state. */
2218 state
= ofport
->stp_port
? stp_port_get_state(ofport
->stp_port
)
2222 if (ofport
->stp_state
!= state
) {
2223 enum ofputil_port_state of_state
;
2226 VLOG_DBG("port %s: STP state changed from %s to %s",
2227 netdev_get_name(ofport
->up
.netdev
),
2228 stp_state_name(ofport
->stp_state
),
2229 stp_state_name(state
));
2230 if (stp_learn_in_state(ofport
->stp_state
)
2231 != stp_learn_in_state(state
)) {
2232 /* xxx Learning action flows should also be flushed. */
2233 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2234 mac_learning_flush(ofproto
->ml
);
2235 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2236 mcast_snooping_mdb_flush(ofproto
->ms
);
2238 fwd_change
= stp_forward_in_state(ofport
->stp_state
)
2239 != stp_forward_in_state(state
);
2241 ofproto
->backer
->need_revalidate
= REV_STP
;
2242 ofport
->stp_state
= state
;
2243 ofport
->stp_state_entered
= time_msec();
2245 if (fwd_change
&& ofport
->bundle
) {
2246 bundle_update(ofport
->bundle
);
2249 /* Update the STP state bits in the OpenFlow port description. */
2250 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2251 of_state
|= (state
== STP_LISTENING
? OFPUTIL_PS_STP_LISTEN
2252 : state
== STP_LEARNING
? OFPUTIL_PS_STP_LEARN
2253 : state
== STP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2254 : state
== STP_BLOCKING
? OFPUTIL_PS_STP_BLOCK
2256 ofproto_port_set_state(&ofport
->up
, of_state
);
2260 /* Configures STP on 'ofport_' using the settings defined in 's'. The
2261 * caller is responsible for assigning STP port numbers and ensuring
2262 * there are no duplicates. */
2264 set_stp_port(struct ofport
*ofport_
,
2265 const struct ofproto_port_stp_settings
*s
)
2267 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2268 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2269 struct stp_port
*sp
= ofport
->stp_port
;
2271 if (!s
|| !s
->enable
) {
2273 ofport
->stp_port
= NULL
;
2274 stp_port_disable(sp
);
2275 update_stp_port_state(ofport
);
2278 } else if (sp
&& stp_port_no(sp
) != s
->port_num
2279 && ofport
== stp_port_get_aux(sp
)) {
2280 /* The port-id changed, so disable the old one if it's not
2281 * already in use by another port. */
2282 stp_port_disable(sp
);
2285 sp
= ofport
->stp_port
= stp_get_port(ofproto
->stp
, s
->port_num
);
2287 /* Set name before enabling the port so that debugging messages can print
2289 stp_port_set_name(sp
, netdev_get_name(ofport
->up
.netdev
));
2290 stp_port_enable(sp
);
2292 stp_port_set_aux(sp
, ofport
);
2293 stp_port_set_priority(sp
, s
->priority
);
2294 stp_port_set_path_cost(sp
, s
->path_cost
);
2296 update_stp_port_state(ofport
);
2302 get_stp_port_status(struct ofport
*ofport_
,
2303 struct ofproto_port_stp_status
*s
)
2305 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2306 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2307 struct stp_port
*sp
= ofport
->stp_port
;
2309 if (!ofproto
->stp
|| !sp
) {
2315 s
->port_id
= stp_port_get_id(sp
);
2316 s
->state
= stp_port_get_state(sp
);
2317 s
->sec_in_state
= (time_msec() - ofport
->stp_state_entered
) / 1000;
2318 s
->role
= stp_port_get_role(sp
);
2324 get_stp_port_stats(struct ofport
*ofport_
,
2325 struct ofproto_port_stp_stats
*s
)
2327 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2328 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2329 struct stp_port
*sp
= ofport
->stp_port
;
2331 if (!ofproto
->stp
|| !sp
) {
2337 stp_port_get_counts(sp
, &s
->tx_count
, &s
->rx_count
, &s
->error_count
);
2343 stp_run(struct ofproto_dpif
*ofproto
)
2346 long long int now
= time_msec();
2347 long long int elapsed
= now
- ofproto
->stp_last_tick
;
2348 struct stp_port
*sp
;
2351 stp_tick(ofproto
->stp
, MIN(INT_MAX
, elapsed
));
2352 ofproto
->stp_last_tick
= now
;
2354 while (stp_get_changed_port(ofproto
->stp
, &sp
)) {
2355 struct ofport_dpif
*ofport
= stp_port_get_aux(sp
);
2358 update_stp_port_state(ofport
);
2362 if (stp_check_and_reset_fdb_flush(ofproto
->stp
)) {
2363 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2364 mac_learning_flush(ofproto
->ml
);
2365 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2366 mcast_snooping_mdb_flush(ofproto
->ms
);
2372 stp_wait(struct ofproto_dpif
*ofproto
)
2375 poll_timer_wait(1000);
2379 /* Configures RSTP on 'ofport_' using the settings defined in 's'. The
2380 * caller is responsible for assigning RSTP port numbers and ensuring
2381 * there are no duplicates. */
2383 set_rstp_port(struct ofport
*ofport_
,
2384 const struct ofproto_port_rstp_settings
*s
)
2386 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2387 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2388 struct rstp_port
*rp
= ofport
->rstp_port
;
2390 if (!s
|| !s
->enable
) {
2392 rstp_port_unref(rp
);
2393 ofport
->rstp_port
= NULL
;
2394 update_rstp_port_state(ofport
);
2399 /* Check if need to add a new port. */
2401 rp
= ofport
->rstp_port
= rstp_add_port(ofproto
->rstp
);
2404 rstp_port_set(rp
, s
->port_num
, s
->priority
, s
->path_cost
,
2405 s
->admin_edge_port
, s
->auto_edge
, s
->mcheck
, ofport
);
2406 update_rstp_port_state(ofport
);
2407 /* Synchronize operational status. */
2408 rstp_port_set_mac_operational(rp
, ofport
->may_enable
);
2412 get_rstp_port_status(struct ofport
*ofport_
,
2413 struct ofproto_port_rstp_status
*s
)
2415 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2416 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2417 struct rstp_port
*rp
= ofport
->rstp_port
;
2419 if (!ofproto
->rstp
|| !rp
) {
2425 rstp_port_get_status(rp
, &s
->port_id
, &s
->state
, &s
->role
, &s
->tx_count
,
2426 &s
->rx_count
, &s
->error_count
, &s
->uptime
);
2431 set_queues(struct ofport
*ofport_
, const struct ofproto_port_queue
*qdscp
,
2434 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2435 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2437 if (ofport
->n_qdscp
!= n_qdscp
2438 || (n_qdscp
&& memcmp(ofport
->qdscp
, qdscp
,
2439 n_qdscp
* sizeof *qdscp
))) {
2440 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2441 free(ofport
->qdscp
);
2442 ofport
->qdscp
= n_qdscp
2443 ? xmemdup(qdscp
, n_qdscp
* sizeof *qdscp
)
2445 ofport
->n_qdscp
= n_qdscp
;
2453 /* Expires all MAC learning entries associated with 'bundle' and forces its
2454 * ofproto to revalidate every flow.
2456 * Normally MAC learning entries are removed only from the ofproto associated
2457 * with 'bundle', but if 'all_ofprotos' is true, then the MAC learning entries
2458 * are removed from every ofproto. When patch ports and SLB bonds are in use
2459 * and a VM migration happens and the gratuitous ARPs are somehow lost, this
2460 * avoids a MAC_ENTRY_IDLE_TIME delay before the migrated VM can communicate
2461 * with the host from which it migrated. */
2463 bundle_flush_macs(struct ofbundle
*bundle
, bool all_ofprotos
)
2465 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2466 struct mac_learning
*ml
= ofproto
->ml
;
2467 struct mac_entry
*mac
, *next_mac
;
2469 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2470 ovs_rwlock_wrlock(&ml
->rwlock
);
2471 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2472 if (mac
->port
.p
== bundle
) {
2474 struct ofproto_dpif
*o
;
2476 HMAP_FOR_EACH (o
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
2478 struct mac_entry
*e
;
2480 ovs_rwlock_wrlock(&o
->ml
->rwlock
);
2481 e
= mac_learning_lookup(o
->ml
, mac
->mac
, mac
->vlan
);
2483 mac_learning_expire(o
->ml
, e
);
2485 ovs_rwlock_unlock(&o
->ml
->rwlock
);
2490 mac_learning_expire(ml
, mac
);
2493 ovs_rwlock_unlock(&ml
->rwlock
);
2496 static struct ofbundle
*
2497 bundle_lookup(const struct ofproto_dpif
*ofproto
, void *aux
)
2499 struct ofbundle
*bundle
;
2501 HMAP_FOR_EACH_IN_BUCKET (bundle
, hmap_node
, hash_pointer(aux
, 0),
2502 &ofproto
->bundles
) {
2503 if (bundle
->aux
== aux
) {
2511 bundle_update(struct ofbundle
*bundle
)
2513 struct ofport_dpif
*port
;
2515 bundle
->floodable
= true;
2516 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2517 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2519 || !stp_forward_in_state(port
->stp_state
)) {
2520 bundle
->floodable
= false;
2527 bundle_del_port(struct ofport_dpif
*port
)
2529 struct ofbundle
*bundle
= port
->bundle
;
2531 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2533 list_remove(&port
->bundle_node
);
2534 port
->bundle
= NULL
;
2537 lacp_slave_unregister(bundle
->lacp
, port
);
2540 bond_slave_unregister(bundle
->bond
, port
);
2543 bundle_update(bundle
);
2547 bundle_add_port(struct ofbundle
*bundle
, ofp_port_t ofp_port
,
2548 struct lacp_slave_settings
*lacp
)
2550 struct ofport_dpif
*port
;
2552 port
= get_ofp_port(bundle
->ofproto
, ofp_port
);
2557 if (port
->bundle
!= bundle
) {
2558 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2560 bundle_remove(&port
->up
);
2563 port
->bundle
= bundle
;
2564 list_push_back(&bundle
->ports
, &port
->bundle_node
);
2565 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2567 || !stp_forward_in_state(port
->stp_state
)) {
2568 bundle
->floodable
= false;
2572 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2573 lacp_slave_register(bundle
->lacp
, port
, lacp
);
2580 bundle_destroy(struct ofbundle
*bundle
)
2582 struct ofproto_dpif
*ofproto
;
2583 struct ofport_dpif
*port
, *next_port
;
2589 ofproto
= bundle
->ofproto
;
2590 mbridge_unregister_bundle(ofproto
->mbridge
, bundle
->aux
);
2593 xlate_bundle_remove(bundle
);
2596 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2597 bundle_del_port(port
);
2600 bundle_flush_macs(bundle
, true);
2601 hmap_remove(&ofproto
->bundles
, &bundle
->hmap_node
);
2603 free(bundle
->trunks
);
2604 lacp_unref(bundle
->lacp
);
2605 bond_unref(bundle
->bond
);
2610 bundle_set(struct ofproto
*ofproto_
, void *aux
,
2611 const struct ofproto_bundle_settings
*s
)
2613 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2614 bool need_flush
= false;
2615 struct ofport_dpif
*port
;
2616 struct ofbundle
*bundle
;
2617 unsigned long *trunks
;
2623 bundle_destroy(bundle_lookup(ofproto
, aux
));
2627 ovs_assert(s
->n_slaves
== 1 || s
->bond
!= NULL
);
2628 ovs_assert((s
->lacp
!= NULL
) == (s
->lacp_slaves
!= NULL
));
2630 bundle
= bundle_lookup(ofproto
, aux
);
2632 bundle
= xmalloc(sizeof *bundle
);
2634 bundle
->ofproto
= ofproto
;
2635 hmap_insert(&ofproto
->bundles
, &bundle
->hmap_node
,
2636 hash_pointer(aux
, 0));
2638 bundle
->name
= NULL
;
2640 list_init(&bundle
->ports
);
2641 bundle
->vlan_mode
= PORT_VLAN_TRUNK
;
2643 bundle
->trunks
= NULL
;
2644 bundle
->use_priority_tags
= s
->use_priority_tags
;
2645 bundle
->lacp
= NULL
;
2646 bundle
->bond
= NULL
;
2648 bundle
->floodable
= true;
2649 mbridge_register_bundle(ofproto
->mbridge
, bundle
);
2652 if (!bundle
->name
|| strcmp(s
->name
, bundle
->name
)) {
2654 bundle
->name
= xstrdup(s
->name
);
2659 ofproto
->lacp_enabled
= true;
2660 if (!bundle
->lacp
) {
2661 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2662 bundle
->lacp
= lacp_create();
2664 lacp_configure(bundle
->lacp
, s
->lacp
);
2666 lacp_unref(bundle
->lacp
);
2667 bundle
->lacp
= NULL
;
2670 /* Update set of ports. */
2672 for (i
= 0; i
< s
->n_slaves
; i
++) {
2673 if (!bundle_add_port(bundle
, s
->slaves
[i
],
2674 s
->lacp
? &s
->lacp_slaves
[i
] : NULL
)) {
2678 if (!ok
|| list_size(&bundle
->ports
) != s
->n_slaves
) {
2679 struct ofport_dpif
*next_port
;
2681 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2682 for (i
= 0; i
< s
->n_slaves
; i
++) {
2683 if (s
->slaves
[i
] == port
->up
.ofp_port
) {
2688 bundle_del_port(port
);
2692 ovs_assert(list_size(&bundle
->ports
) <= s
->n_slaves
);
2694 if (list_is_empty(&bundle
->ports
)) {
2695 bundle_destroy(bundle
);
2699 /* Set VLAN tagging mode */
2700 if (s
->vlan_mode
!= bundle
->vlan_mode
2701 || s
->use_priority_tags
!= bundle
->use_priority_tags
) {
2702 bundle
->vlan_mode
= s
->vlan_mode
;
2703 bundle
->use_priority_tags
= s
->use_priority_tags
;
2708 vlan
= (s
->vlan_mode
== PORT_VLAN_TRUNK
? -1
2709 : s
->vlan
>= 0 && s
->vlan
<= 4095 ? s
->vlan
2711 if (vlan
!= bundle
->vlan
) {
2712 bundle
->vlan
= vlan
;
2716 /* Get trunked VLANs. */
2717 switch (s
->vlan_mode
) {
2718 case PORT_VLAN_ACCESS
:
2722 case PORT_VLAN_TRUNK
:
2723 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2726 case PORT_VLAN_NATIVE_UNTAGGED
:
2727 case PORT_VLAN_NATIVE_TAGGED
:
2728 if (vlan
!= 0 && (!s
->trunks
2729 || !bitmap_is_set(s
->trunks
, vlan
)
2730 || bitmap_is_set(s
->trunks
, 0))) {
2731 /* Force trunking the native VLAN and prohibit trunking VLAN 0. */
2733 trunks
= bitmap_clone(s
->trunks
, 4096);
2735 trunks
= bitmap_allocate1(4096);
2737 bitmap_set1(trunks
, vlan
);
2738 bitmap_set0(trunks
, 0);
2740 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2747 if (!vlan_bitmap_equal(trunks
, bundle
->trunks
)) {
2748 free(bundle
->trunks
);
2749 if (trunks
== s
->trunks
) {
2750 bundle
->trunks
= vlan_bitmap_clone(trunks
);
2752 bundle
->trunks
= trunks
;
2757 if (trunks
!= s
->trunks
) {
2762 if (!list_is_short(&bundle
->ports
)) {
2763 bundle
->ofproto
->has_bonded_bundles
= true;
2765 if (bond_reconfigure(bundle
->bond
, s
->bond
)) {
2766 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2769 bundle
->bond
= bond_create(s
->bond
, ofproto
);
2770 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2773 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2774 bond_slave_register(bundle
->bond
, port
,
2775 port
->up
.ofp_port
, port
->up
.netdev
);
2778 bond_unref(bundle
->bond
);
2779 bundle
->bond
= NULL
;
2782 /* If we changed something that would affect MAC learning, un-learn
2783 * everything on this port and force flow revalidation. */
2785 bundle_flush_macs(bundle
, false);
2792 bundle_remove(struct ofport
*port_
)
2794 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
2795 struct ofbundle
*bundle
= port
->bundle
;
2798 bundle_del_port(port
);
2799 if (list_is_empty(&bundle
->ports
)) {
2800 bundle_destroy(bundle
);
2801 } else if (list_is_short(&bundle
->ports
)) {
2802 bond_unref(bundle
->bond
);
2803 bundle
->bond
= NULL
;
2809 send_pdu_cb(void *port_
, const void *pdu
, size_t pdu_size
)
2811 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 10);
2812 struct ofport_dpif
*port
= port_
;
2813 uint8_t ea
[ETH_ADDR_LEN
];
2816 error
= netdev_get_etheraddr(port
->up
.netdev
, ea
);
2818 struct ofpbuf packet
;
2821 ofpbuf_init(&packet
, 0);
2822 packet_pdu
= eth_compose(&packet
, eth_addr_lacp
, ea
, ETH_TYPE_LACP
,
2824 memcpy(packet_pdu
, pdu
, pdu_size
);
2826 ofproto_dpif_send_packet(port
, &packet
);
2827 ofpbuf_uninit(&packet
);
2829 VLOG_ERR_RL(&rl
, "port %s: cannot obtain Ethernet address of iface "
2830 "%s (%s)", port
->bundle
->name
,
2831 netdev_get_name(port
->up
.netdev
), ovs_strerror(error
));
2836 bundle_send_learning_packets(struct ofbundle
*bundle
)
2838 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2839 struct ofpbuf
*learning_packet
;
2840 int error
, n_packets
, n_errors
;
2841 struct mac_entry
*e
;
2842 struct list packets
;
2844 list_init(&packets
);
2845 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
2846 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
2847 if (e
->port
.p
!= bundle
) {
2850 learning_packet
= bond_compose_learning_packet(bundle
->bond
,
2853 /* Temporarily use 'frame' as a private pointer (see below). */
2854 ovs_assert(learning_packet
->frame
== ofpbuf_data(learning_packet
));
2855 learning_packet
->frame
= port_void
;
2856 list_push_back(&packets
, &learning_packet
->list_node
);
2859 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2861 error
= n_packets
= n_errors
= 0;
2862 LIST_FOR_EACH (learning_packet
, list_node
, &packets
) {
2864 void *port_void
= learning_packet
->frame
;
2866 /* Restore 'frame'. */
2867 learning_packet
->frame
= ofpbuf_data(learning_packet
);
2868 ret
= ofproto_dpif_send_packet(port_void
, learning_packet
);
2875 ofpbuf_list_delete(&packets
);
2878 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2879 VLOG_WARN_RL(&rl
, "bond %s: %d errors sending %d gratuitous learning "
2880 "packets, last error was: %s",
2881 bundle
->name
, n_errors
, n_packets
, ovs_strerror(error
));
2883 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
2884 bundle
->name
, n_packets
);
2889 bundle_run(struct ofbundle
*bundle
)
2892 lacp_run(bundle
->lacp
, send_pdu_cb
);
2895 struct ofport_dpif
*port
;
2897 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2898 bond_slave_set_may_enable(bundle
->bond
, port
, port
->may_enable
);
2901 if (bond_run(bundle
->bond
, lacp_status(bundle
->lacp
))) {
2902 bundle
->ofproto
->backer
->need_revalidate
= REV_BOND
;
2905 if (bond_should_send_learning_packets(bundle
->bond
)) {
2906 bundle_send_learning_packets(bundle
);
2912 bundle_wait(struct ofbundle
*bundle
)
2915 lacp_wait(bundle
->lacp
);
2918 bond_wait(bundle
->bond
);
2925 mirror_set__(struct ofproto
*ofproto_
, void *aux
,
2926 const struct ofproto_mirror_settings
*s
)
2928 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2929 struct ofbundle
**srcs
, **dsts
;
2934 mirror_destroy(ofproto
->mbridge
, aux
);
2938 srcs
= xmalloc(s
->n_srcs
* sizeof *srcs
);
2939 dsts
= xmalloc(s
->n_dsts
* sizeof *dsts
);
2941 for (i
= 0; i
< s
->n_srcs
; i
++) {
2942 srcs
[i
] = bundle_lookup(ofproto
, s
->srcs
[i
]);
2945 for (i
= 0; i
< s
->n_dsts
; i
++) {
2946 dsts
[i
] = bundle_lookup(ofproto
, s
->dsts
[i
]);
2949 error
= mirror_set(ofproto
->mbridge
, aux
, s
->name
, srcs
, s
->n_srcs
, dsts
,
2950 s
->n_dsts
, s
->src_vlans
,
2951 bundle_lookup(ofproto
, s
->out_bundle
), s
->out_vlan
);
2958 mirror_get_stats__(struct ofproto
*ofproto
, void *aux
,
2959 uint64_t *packets
, uint64_t *bytes
)
2961 return mirror_get_stats(ofproto_dpif_cast(ofproto
)->mbridge
, aux
, packets
,
2966 set_flood_vlans(struct ofproto
*ofproto_
, unsigned long *flood_vlans
)
2968 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2969 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2970 if (mac_learning_set_flood_vlans(ofproto
->ml
, flood_vlans
)) {
2971 mac_learning_flush(ofproto
->ml
);
2973 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2978 is_mirror_output_bundle(const struct ofproto
*ofproto_
, void *aux
)
2980 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2981 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
2982 return bundle
&& mirror_bundle_out(ofproto
->mbridge
, bundle
) != 0;
2986 forward_bpdu_changed(struct ofproto
*ofproto_
)
2988 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2989 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2993 set_mac_table_config(struct ofproto
*ofproto_
, unsigned int idle_time
,
2996 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2997 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2998 mac_learning_set_idle_time(ofproto
->ml
, idle_time
);
2999 mac_learning_set_max_entries(ofproto
->ml
, max_entries
);
3000 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3003 /* Configures multicast snooping on 'ofport' using the settings
3004 * defined in 's'. */
3006 set_mcast_snooping(struct ofproto
*ofproto_
,
3007 const struct ofproto_mcast_snooping_settings
*s
)
3009 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3011 /* Only revalidate flows if the configuration changed. */
3012 if (!s
!= !ofproto
->ms
) {
3013 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3018 ofproto
->ms
= mcast_snooping_create();
3021 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3022 mcast_snooping_set_idle_time(ofproto
->ms
, s
->idle_time
);
3023 mcast_snooping_set_max_entries(ofproto
->ms
, s
->max_entries
);
3024 if (mcast_snooping_set_flood_unreg(ofproto
->ms
, s
->flood_unreg
)) {
3025 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3027 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3029 mcast_snooping_unref(ofproto
->ms
);
3036 /* Configures multicast snooping port's flood setting on 'ofproto'. */
3038 set_mcast_snooping_port(struct ofproto
*ofproto_
, void *aux
, bool flood
)
3040 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3041 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3044 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3045 mcast_snooping_set_port_flood(ofproto
->ms
, bundle
->vlan
, bundle
,
3047 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3055 static struct ofport_dpif
*
3056 get_ofp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
3058 struct ofport
*ofport
= ofproto_get_port(&ofproto
->up
, ofp_port
);
3059 return ofport
? ofport_dpif_cast(ofport
) : NULL
;
3063 ofproto_port_from_dpif_port(struct ofproto_dpif
*ofproto
,
3064 struct ofproto_port
*ofproto_port
,
3065 struct dpif_port
*dpif_port
)
3067 ofproto_port
->name
= dpif_port
->name
;
3068 ofproto_port
->type
= dpif_port
->type
;
3069 ofproto_port
->ofp_port
= odp_port_to_ofp_port(ofproto
, dpif_port
->port_no
);
3073 ofport_update_peer(struct ofport_dpif
*ofport
)
3075 const struct ofproto_dpif
*ofproto
;
3076 struct dpif_backer
*backer
;
3079 if (!netdev_vport_is_patch(ofport
->up
.netdev
)) {
3083 backer
= ofproto_dpif_cast(ofport
->up
.ofproto
)->backer
;
3084 backer
->need_revalidate
= REV_RECONFIGURE
;
3087 ofport
->peer
->peer
= NULL
;
3088 ofport
->peer
= NULL
;
3091 peer_name
= netdev_vport_patch_peer(ofport
->up
.netdev
);
3096 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
3097 struct ofport
*peer_ofport
;
3098 struct ofport_dpif
*peer
;
3101 if (ofproto
->backer
!= backer
) {
3105 peer_ofport
= shash_find_data(&ofproto
->up
.port_by_name
, peer_name
);
3110 peer
= ofport_dpif_cast(peer_ofport
);
3111 peer_peer
= netdev_vport_patch_peer(peer
->up
.netdev
);
3112 if (peer_peer
&& !strcmp(netdev_get_name(ofport
->up
.netdev
),
3114 ofport
->peer
= peer
;
3115 ofport
->peer
->peer
= ofport
;
3125 port_run(struct ofport_dpif
*ofport
)
3127 long long int carrier_seq
= netdev_get_carrier_resets(ofport
->up
.netdev
);
3128 bool carrier_changed
= carrier_seq
!= ofport
->carrier_seq
;
3129 bool enable
= netdev_get_carrier(ofport
->up
.netdev
);
3130 bool cfm_enable
= false;
3131 bool bfd_enable
= false;
3133 ofport
->carrier_seq
= carrier_seq
;
3136 int cfm_opup
= cfm_get_opup(ofport
->cfm
);
3138 cfm_enable
= !cfm_get_fault(ofport
->cfm
);
3140 if (cfm_opup
>= 0) {
3141 cfm_enable
= cfm_enable
&& cfm_opup
;
3146 bfd_enable
= bfd_forwarding(ofport
->bfd
);
3149 if (ofport
->bfd
|| ofport
->cfm
) {
3150 enable
= enable
&& (cfm_enable
|| bfd_enable
);
3153 if (ofport
->bundle
) {
3154 enable
= enable
&& lacp_slave_may_enable(ofport
->bundle
->lacp
, ofport
);
3155 if (carrier_changed
) {
3156 lacp_slave_carrier_changed(ofport
->bundle
->lacp
, ofport
);
3160 if (ofport
->may_enable
!= enable
) {
3161 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3163 ofproto
->backer
->need_revalidate
= REV_PORT_TOGGLED
;
3165 if (ofport
->rstp_port
) {
3166 rstp_port_set_mac_operational(ofport
->rstp_port
, enable
);
3170 ofport
->may_enable
= enable
;
3174 port_query_by_name(const struct ofproto
*ofproto_
, const char *devname
,
3175 struct ofproto_port
*ofproto_port
)
3177 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3178 struct dpif_port dpif_port
;
3181 if (sset_contains(&ofproto
->ghost_ports
, devname
)) {
3182 const char *type
= netdev_get_type_from_name(devname
);
3184 /* We may be called before ofproto->up.port_by_name is populated with
3185 * the appropriate ofport. For this reason, we must get the name and
3186 * type from the netdev layer directly. */
3188 const struct ofport
*ofport
;
3190 ofport
= shash_find_data(&ofproto
->up
.port_by_name
, devname
);
3191 ofproto_port
->ofp_port
= ofport
? ofport
->ofp_port
: OFPP_NONE
;
3192 ofproto_port
->name
= xstrdup(devname
);
3193 ofproto_port
->type
= xstrdup(type
);
3199 if (!sset_contains(&ofproto
->ports
, devname
)) {
3202 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
3203 devname
, &dpif_port
);
3205 ofproto_port_from_dpif_port(ofproto
, ofproto_port
, &dpif_port
);
3211 port_add(struct ofproto
*ofproto_
, struct netdev
*netdev
)
3213 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3214 const char *devname
= netdev_get_name(netdev
);
3215 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
3216 const char *dp_port_name
;
3218 if (netdev_vport_is_patch(netdev
)) {
3219 sset_add(&ofproto
->ghost_ports
, netdev_get_name(netdev
));
3223 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
3224 if (!dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
3225 odp_port_t port_no
= ODPP_NONE
;
3228 error
= dpif_port_add(ofproto
->backer
->dpif
, netdev
, &port_no
);
3232 if (netdev_get_tunnel_config(netdev
)) {
3233 simap_put(&ofproto
->backer
->tnl_backers
,
3234 dp_port_name
, odp_to_u32(port_no
));
3238 if (netdev_get_tunnel_config(netdev
)) {
3239 sset_add(&ofproto
->ghost_ports
, devname
);
3241 sset_add(&ofproto
->ports
, devname
);
3247 port_del(struct ofproto
*ofproto_
, ofp_port_t ofp_port
)
3249 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3250 struct ofport_dpif
*ofport
= get_ofp_port(ofproto
, ofp_port
);
3257 sset_find_and_delete(&ofproto
->ghost_ports
,
3258 netdev_get_name(ofport
->up
.netdev
));
3259 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3260 if (!ofport
->is_tunnel
&& !netdev_vport_is_patch(ofport
->up
.netdev
)) {
3261 error
= dpif_port_del(ofproto
->backer
->dpif
, ofport
->odp_port
);
3263 /* The caller is going to close ofport->up.netdev. If this is a
3264 * bonded port, then the bond is using that netdev, so remove it
3265 * from the bond. The client will need to reconfigure everything
3266 * after deleting ports, so then the slave will get re-added. */
3267 bundle_remove(&ofport
->up
);
3274 port_get_stats(const struct ofport
*ofport_
, struct netdev_stats
*stats
)
3276 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3279 error
= netdev_get_stats(ofport
->up
.netdev
, stats
);
3281 if (!error
&& ofport_
->ofp_port
== OFPP_LOCAL
) {
3282 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3284 ovs_mutex_lock(&ofproto
->stats_mutex
);
3285 /* ofproto->stats.tx_packets represents packets that we created
3286 * internally and sent to some port (e.g. packets sent with
3287 * ofproto_dpif_send_packet()). Account for them as if they had
3288 * come from OFPP_LOCAL and got forwarded. */
3290 if (stats
->rx_packets
!= UINT64_MAX
) {
3291 stats
->rx_packets
+= ofproto
->stats
.tx_packets
;
3294 if (stats
->rx_bytes
!= UINT64_MAX
) {
3295 stats
->rx_bytes
+= ofproto
->stats
.tx_bytes
;
3298 /* ofproto->stats.rx_packets represents packets that were received on
3299 * some port and we processed internally and dropped (e.g. STP).
3300 * Account for them as if they had been forwarded to OFPP_LOCAL. */
3302 if (stats
->tx_packets
!= UINT64_MAX
) {
3303 stats
->tx_packets
+= ofproto
->stats
.rx_packets
;
3306 if (stats
->tx_bytes
!= UINT64_MAX
) {
3307 stats
->tx_bytes
+= ofproto
->stats
.rx_bytes
;
3309 ovs_mutex_unlock(&ofproto
->stats_mutex
);
3315 struct port_dump_state
{
3320 struct ofproto_port port
;
3325 port_dump_start(const struct ofproto
*ofproto_ OVS_UNUSED
, void **statep
)
3327 *statep
= xzalloc(sizeof(struct port_dump_state
));
3332 port_dump_next(const struct ofproto
*ofproto_
, void *state_
,
3333 struct ofproto_port
*port
)
3335 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3336 struct port_dump_state
*state
= state_
;
3337 const struct sset
*sset
;
3338 struct sset_node
*node
;
3340 if (state
->has_port
) {
3341 ofproto_port_destroy(&state
->port
);
3342 state
->has_port
= false;
3344 sset
= state
->ghost
? &ofproto
->ghost_ports
: &ofproto
->ports
;
3345 while ((node
= sset_at_position(sset
, &state
->bucket
, &state
->offset
))) {
3348 error
= port_query_by_name(ofproto_
, node
->name
, &state
->port
);
3350 *port
= state
->port
;
3351 state
->has_port
= true;
3353 } else if (error
!= ENODEV
) {
3358 if (!state
->ghost
) {
3359 state
->ghost
= true;
3362 return port_dump_next(ofproto_
, state_
, port
);
3369 port_dump_done(const struct ofproto
*ofproto_ OVS_UNUSED
, void *state_
)
3371 struct port_dump_state
*state
= state_
;
3373 if (state
->has_port
) {
3374 ofproto_port_destroy(&state
->port
);
3381 port_poll(const struct ofproto
*ofproto_
, char **devnamep
)
3383 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3385 if (ofproto
->port_poll_errno
) {
3386 int error
= ofproto
->port_poll_errno
;
3387 ofproto
->port_poll_errno
= 0;
3391 if (sset_is_empty(&ofproto
->port_poll_set
)) {
3395 *devnamep
= sset_pop(&ofproto
->port_poll_set
);
3400 port_poll_wait(const struct ofproto
*ofproto_
)
3402 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3403 dpif_port_poll_wait(ofproto
->backer
->dpif
);
3407 port_is_lacp_current(const struct ofport
*ofport_
)
3409 const struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3410 return (ofport
->bundle
&& ofport
->bundle
->lacp
3411 ? lacp_slave_is_current(ofport
->bundle
->lacp
, ofport
)
3415 /* If 'rule' is an OpenFlow rule, that has expired according to OpenFlow rules,
3416 * then delete it entirely. */
3418 rule_expire(struct rule_dpif
*rule
)
3419 OVS_REQUIRES(ofproto_mutex
)
3421 uint16_t hard_timeout
, idle_timeout
;
3422 long long int now
= time_msec();
3425 hard_timeout
= rule
->up
.hard_timeout
;
3426 idle_timeout
= rule
->up
.idle_timeout
;
3428 /* Has 'rule' expired? */
3430 long long int modified
;
3432 ovs_mutex_lock(&rule
->up
.mutex
);
3433 modified
= rule
->up
.modified
;
3434 ovs_mutex_unlock(&rule
->up
.mutex
);
3436 if (now
> modified
+ hard_timeout
* 1000) {
3437 reason
= OFPRR_HARD_TIMEOUT
;
3441 if (reason
< 0 && idle_timeout
) {
3444 ovs_mutex_lock(&rule
->stats_mutex
);
3445 used
= rule
->stats
.used
;
3446 ovs_mutex_unlock(&rule
->stats_mutex
);
3448 if (now
> used
+ idle_timeout
* 1000) {
3449 reason
= OFPRR_IDLE_TIMEOUT
;
3454 COVERAGE_INC(ofproto_dpif_expired
);
3455 ofproto_rule_expire(&rule
->up
, reason
);
3459 /* Executes, within 'ofproto', the actions in 'rule' or 'ofpacts' on 'packet'.
3460 * 'flow' must reflect the data in 'packet'. */
3462 ofproto_dpif_execute_actions(struct ofproto_dpif
*ofproto
,
3463 const struct flow
*flow
,
3464 struct rule_dpif
*rule
,
3465 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3466 struct ofpbuf
*packet
)
3468 struct dpif_flow_stats stats
;
3469 struct xlate_out xout
;
3470 struct xlate_in xin
;
3472 struct dpif_execute execute
;
3475 ovs_assert((rule
!= NULL
) != (ofpacts
!= NULL
));
3477 dpif_flow_stats_extract(flow
, packet
, time_msec(), &stats
);
3480 rule_dpif_credit_stats(rule
, &stats
);
3483 xlate_in_init(&xin
, ofproto
, flow
, flow
->in_port
.ofp_port
, rule
,
3484 stats
.tcp_flags
, packet
);
3485 xin
.ofpacts
= ofpacts
;
3486 xin
.ofpacts_len
= ofpacts_len
;
3487 xin
.resubmit_stats
= &stats
;
3488 xlate_actions(&xin
, &xout
);
3490 execute
.actions
= ofpbuf_data(xout
.odp_actions
);
3491 execute
.actions_len
= ofpbuf_size(xout
.odp_actions
);
3492 execute
.packet
= packet
;
3493 execute
.md
= pkt_metadata_from_flow(flow
);
3494 execute
.needs_help
= (xout
.slow
& SLOW_ACTION
) != 0;
3495 execute
.probe
= false;
3497 /* Fix up in_port. */
3498 in_port
= flow
->in_port
.ofp_port
;
3499 if (in_port
== OFPP_NONE
) {
3500 in_port
= OFPP_LOCAL
;
3502 execute
.md
.in_port
.odp_port
= ofp_port_to_odp_port(ofproto
, in_port
);
3504 error
= dpif_execute(ofproto
->backer
->dpif
, &execute
);
3506 xlate_out_uninit(&xout
);
3512 rule_dpif_credit_stats(struct rule_dpif
*rule
,
3513 const struct dpif_flow_stats
*stats
)
3515 ovs_mutex_lock(&rule
->stats_mutex
);
3516 rule
->stats
.n_packets
+= stats
->n_packets
;
3517 rule
->stats
.n_bytes
+= stats
->n_bytes
;
3518 rule
->stats
.used
= MAX(rule
->stats
.used
, stats
->used
);
3519 ovs_mutex_unlock(&rule
->stats_mutex
);
3523 rule_dpif_get_flow_cookie(const struct rule_dpif
*rule
)
3524 OVS_REQUIRES(rule
->up
.mutex
)
3526 return rule
->up
.flow_cookie
;
3530 rule_dpif_reduce_timeouts(struct rule_dpif
*rule
, uint16_t idle_timeout
,
3531 uint16_t hard_timeout
)
3533 ofproto_rule_reduce_timeouts(&rule
->up
, idle_timeout
, hard_timeout
);
3536 /* Returns 'rule''s actions. The caller owns a reference on the returned
3537 * actions and must eventually release it (with rule_actions_unref()) to avoid
3539 const struct rule_actions
*
3540 rule_dpif_get_actions(const struct rule_dpif
*rule
)
3542 return rule_get_actions(&rule
->up
);
3545 /* Sets 'rule''s recirculation id. */
3547 rule_dpif_set_recirc_id(struct rule_dpif
*rule
, uint32_t id
)
3548 OVS_REQUIRES(rule
->up
.mutex
)
3550 ovs_assert(!rule
->recirc_id
);
3551 rule
->recirc_id
= id
;
3554 /* Returns 'rule''s recirculation id. */
3556 rule_dpif_get_recirc_id(struct rule_dpif
*rule
)
3557 OVS_REQUIRES(rule
->up
.mutex
)
3559 if (!rule
->recirc_id
) {
3560 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3562 rule_dpif_set_recirc_id(rule
, ofproto_dpif_alloc_recirc_id(ofproto
));
3564 return rule
->recirc_id
;
3567 /* Sets 'rule''s recirculation id. */
3569 rule_set_recirc_id(struct rule
*rule_
, uint32_t id
)
3571 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3573 ovs_mutex_lock(&rule
->up
.mutex
);
3574 rule_dpif_set_recirc_id(rule
, id
);
3575 ovs_mutex_unlock(&rule
->up
.mutex
);
3578 /* Lookup 'flow' in table 0 of 'ofproto''s classifier.
3579 * If 'wc' is non-null, sets the fields that were relevant as part of
3580 * the lookup. Returns the table_id where a match or miss occurred.
3582 * The return value will be zero unless there was a miss and
3583 * OFPTC11_TABLE_MISS_CONTINUE is in effect for the sequence of tables
3584 * where misses occur.
3586 * The rule is returned in '*rule', which is valid at least until the next
3587 * RCU quiescent period. If the '*rule' needs to stay around longer,
3588 * a non-zero 'take_ref' must be passed in to cause a reference to be taken
3589 * on it before this returns. */
3591 rule_dpif_lookup(struct ofproto_dpif
*ofproto
, struct flow
*flow
,
3592 struct flow_wildcards
*wc
, struct rule_dpif
**rule
,
3593 bool take_ref
, const struct dpif_flow_stats
*stats
)
3595 enum rule_dpif_lookup_verdict verdict
;
3596 enum ofputil_port_config config
= 0;
3599 if (ofproto_dpif_get_enable_recirc(ofproto
)) {
3600 /* Always exactly match recirc_id since datapath supports
3603 wc
->masks
.recirc_id
= UINT32_MAX
;
3606 /* Start looking up from internal table for post recirculation flows
3607 * or packets. We can also simply send all, including normal flows
3608 * or packets to the internal table. They will not match any post
3609 * recirculation rules except the 'catch all' rule that resubmit
3612 * As an optimization, we send normal flows and packets to table 0
3613 * directly, saving one table lookup. */
3614 table_id
= flow
->recirc_id
? TBL_INTERNAL
: 0;
3619 verdict
= rule_dpif_lookup_from_table(ofproto
, flow
, wc
, true,
3620 &table_id
, rule
, take_ref
, stats
);
3623 case RULE_DPIF_LOOKUP_VERDICT_MATCH
:
3625 case RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
: {
3626 struct ofport_dpif
*port
;
3628 port
= get_ofp_port(ofproto
, flow
->in_port
.ofp_port
);
3630 VLOG_WARN_RL(&rl
, "packet-in on unknown OpenFlow port %"PRIu16
,
3631 flow
->in_port
.ofp_port
);
3633 config
= port
? port
->up
.pp
.config
: 0;
3636 case RULE_DPIF_LOOKUP_VERDICT_DROP
:
3637 config
= OFPUTIL_PC_NO_PACKET_IN
;
3639 case RULE_DPIF_LOOKUP_VERDICT_DEFAULT
:
3640 if (!connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
)) {
3641 config
= OFPUTIL_PC_NO_PACKET_IN
;
3648 choose_miss_rule(config
, ofproto
->miss_rule
,
3649 ofproto
->no_packet_in_rule
, rule
, take_ref
);
3653 /* The returned rule is valid at least until the next RCU quiescent period.
3654 * If the '*rule' needs to stay around longer, a non-zero 'take_ref' must be
3655 * passed in to cause a reference to be taken on it before this returns. */
3656 static struct rule_dpif
*
3657 rule_dpif_lookup_in_table(struct ofproto_dpif
*ofproto
, uint8_t table_id
,
3658 const struct flow
*flow
, struct flow_wildcards
*wc
,
3661 struct classifier
*cls
= &ofproto
->up
.tables
[table_id
].cls
;
3662 const struct cls_rule
*cls_rule
;
3663 struct rule_dpif
*rule
;
3664 struct flow ofpc_normal_flow
;
3666 if (ofproto
->up
.frag_handling
!= OFPC_FRAG_NX_MATCH
) {
3667 /* We always unwildcard dl_type and nw_frag (for IP), so they
3668 * need not be unwildcarded here. */
3670 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
) {
3671 if (ofproto
->up
.frag_handling
== OFPC_FRAG_NORMAL
) {
3672 /* We must pretend that transport ports are unavailable. */
3673 ofpc_normal_flow
= *flow
;
3674 ofpc_normal_flow
.tp_src
= htons(0);
3675 ofpc_normal_flow
.tp_dst
= htons(0);
3676 flow
= &ofpc_normal_flow
;
3678 /* Must be OFPC_FRAG_DROP (we don't have OFPC_FRAG_REASM).
3679 * Use the drop_frags_rule (which cannot disappear). */
3680 cls_rule
= &ofproto
->drop_frags_rule
->up
.cr
;
3681 rule
= rule_dpif_cast(rule_from_cls_rule(cls_rule
));
3683 rule_dpif_ref(rule
);
3691 cls_rule
= classifier_lookup(cls
, flow
, wc
);
3693 rule
= rule_dpif_cast(rule_from_cls_rule(cls_rule
));
3695 /* Try again if the rule was released before we get the reference. */
3696 } while (rule
&& take_ref
&& !rule_dpif_try_ref(rule
));
3701 /* Look up 'flow' in 'ofproto''s classifier starting from table '*table_id'.
3702 * Stores the rule that was found in '*rule', or NULL if none was found.
3703 * Updates 'wc', if nonnull, to reflect the fields that were used during the
3706 * If 'honor_table_miss' is true, the first lookup occurs in '*table_id', but
3707 * if none is found then the table miss configuration for that table is
3708 * honored, which can result in additional lookups in other OpenFlow tables.
3709 * In this case the function updates '*table_id' to reflect the final OpenFlow
3710 * table that was searched.
3712 * If 'honor_table_miss' is false, then only one table lookup occurs, in
3717 * - RULE_DPIF_LOOKUP_VERDICT_MATCH if a rule (in '*rule') was found.
3719 * - RULE_OFPTC_TABLE_MISS_CONTROLLER if no rule was found and either:
3720 * + 'honor_table_miss' is false
3721 * + a table miss configuration specified that the packet should be
3722 * sent to the controller in this case.
3724 * - RULE_DPIF_LOOKUP_VERDICT_DROP if no rule was found, 'honor_table_miss'
3725 * is true and a table miss configuration specified that the packet
3726 * should be dropped in this case.
3728 * - RULE_DPIF_LOOKUP_VERDICT_DEFAULT if no rule was found,
3729 * 'honor_table_miss' is true and a table miss configuration has
3730 * not been specified in this case.
3732 * The rule is returned in '*rule', which is valid at least until the next
3733 * RCU quiescent period. If the '*rule' needs to stay around longer,
3734 * a non-zero 'take_ref' must be passed in to cause a reference to be taken
3735 * on it before this returns. */
3736 enum rule_dpif_lookup_verdict
3737 rule_dpif_lookup_from_table(struct ofproto_dpif
*ofproto
,
3738 const struct flow
*flow
,
3739 struct flow_wildcards
*wc
,
3740 bool honor_table_miss
,
3741 uint8_t *table_id
, struct rule_dpif
**rule
,
3742 bool take_ref
, const struct dpif_flow_stats
*stats
)
3746 for (next_id
= *table_id
;
3747 next_id
< ofproto
->up
.n_tables
;
3748 next_id
++, next_id
+= (next_id
== TBL_INTERNAL
))
3750 *table_id
= next_id
;
3751 *rule
= rule_dpif_lookup_in_table(ofproto
, *table_id
, flow
, wc
,
3754 struct oftable
*tbl
= &ofproto
->up
.tables
[next_id
];
3757 atomic_add_relaxed(*rule
? &tbl
->n_matched
: &tbl
->n_missed
,
3758 stats
->n_packets
, &orig
);
3761 return RULE_DPIF_LOOKUP_VERDICT_MATCH
;
3762 } else if (!honor_table_miss
) {
3763 return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
;
3765 switch (ofproto_table_get_miss_config(&ofproto
->up
, *table_id
)) {
3766 case OFPUTIL_TABLE_MISS_CONTINUE
:
3769 case OFPUTIL_TABLE_MISS_CONTROLLER
:
3770 return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
;
3772 case OFPUTIL_TABLE_MISS_DROP
:
3773 return RULE_DPIF_LOOKUP_VERDICT_DROP
;
3775 case OFPUTIL_TABLE_MISS_DEFAULT
:
3776 return RULE_DPIF_LOOKUP_VERDICT_DEFAULT
;
3781 return RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
;
3784 /* Given a port configuration (specified as zero if there's no port), chooses
3785 * which of 'miss_rule' and 'no_packet_in_rule' should be used in case of a
3788 * The rule is returned in '*rule', which is valid at least until the next
3789 * RCU quiescent period. If the '*rule' needs to stay around longer,
3790 * a reference must be taken on it (rule_dpif_ref()).
3793 choose_miss_rule(enum ofputil_port_config config
, struct rule_dpif
*miss_rule
,
3794 struct rule_dpif
*no_packet_in_rule
, struct rule_dpif
**rule
,
3797 *rule
= config
& OFPUTIL_PC_NO_PACKET_IN
? no_packet_in_rule
: miss_rule
;
3799 rule_dpif_ref(*rule
);
3804 complete_operation(struct rule_dpif
*rule
)
3805 OVS_REQUIRES(ofproto_mutex
)
3807 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3809 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
3812 static struct rule_dpif
*rule_dpif_cast(const struct rule
*rule
)
3814 return rule
? CONTAINER_OF(rule
, struct rule_dpif
, up
) : NULL
;
3817 static struct rule
*
3820 struct rule_dpif
*rule
= xmalloc(sizeof *rule
);
3825 rule_dealloc(struct rule
*rule_
)
3827 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3832 rule_construct(struct rule
*rule_
)
3833 OVS_NO_THREAD_SAFETY_ANALYSIS
3835 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3836 ovs_mutex_init_adaptive(&rule
->stats_mutex
);
3837 rule
->stats
.n_packets
= 0;
3838 rule
->stats
.n_bytes
= 0;
3839 rule
->stats
.used
= rule
->up
.modified
;
3840 rule
->recirc_id
= 0;
3846 rule_insert(struct rule
*rule_
)
3847 OVS_REQUIRES(ofproto_mutex
)
3849 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3850 complete_operation(rule
);
3855 rule_delete(struct rule
*rule_
)
3856 OVS_REQUIRES(ofproto_mutex
)
3858 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3859 complete_operation(rule
);
3863 rule_destruct(struct rule
*rule_
)
3865 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3867 ovs_mutex_destroy(&rule
->stats_mutex
);
3868 if (rule
->recirc_id
) {
3869 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3871 ofproto_dpif_free_recirc_id(ofproto
, rule
->recirc_id
);
3876 rule_get_stats(struct rule
*rule_
, uint64_t *packets
, uint64_t *bytes
,
3877 long long int *used
)
3879 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3881 ovs_mutex_lock(&rule
->stats_mutex
);
3882 *packets
= rule
->stats
.n_packets
;
3883 *bytes
= rule
->stats
.n_bytes
;
3884 *used
= rule
->stats
.used
;
3885 ovs_mutex_unlock(&rule
->stats_mutex
);
3889 rule_dpif_execute(struct rule_dpif
*rule
, const struct flow
*flow
,
3890 struct ofpbuf
*packet
)
3892 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3894 ofproto_dpif_execute_actions(ofproto
, flow
, rule
, NULL
, 0, packet
);
3898 rule_execute(struct rule
*rule
, const struct flow
*flow
,
3899 struct ofpbuf
*packet
)
3901 rule_dpif_execute(rule_dpif_cast(rule
), flow
, packet
);
3902 ofpbuf_delete(packet
);
3907 rule_modify_actions(struct rule
*rule_
, bool reset_counters
)
3908 OVS_REQUIRES(ofproto_mutex
)
3910 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3912 if (reset_counters
) {
3913 ovs_mutex_lock(&rule
->stats_mutex
);
3914 rule
->stats
.n_packets
= 0;
3915 rule
->stats
.n_bytes
= 0;
3916 ovs_mutex_unlock(&rule
->stats_mutex
);
3919 complete_operation(rule
);
3922 static struct group_dpif
*group_dpif_cast(const struct ofgroup
*group
)
3924 return group
? CONTAINER_OF(group
, struct group_dpif
, up
) : NULL
;
3927 static struct ofgroup
*
3930 struct group_dpif
*group
= xzalloc(sizeof *group
);
3935 group_dealloc(struct ofgroup
*group_
)
3937 struct group_dpif
*group
= group_dpif_cast(group_
);
3942 group_construct_stats(struct group_dpif
*group
)
3943 OVS_REQUIRES(group
->stats_mutex
)
3945 struct ofputil_bucket
*bucket
;
3946 const struct list
*buckets
;
3948 group
->packet_count
= 0;
3949 group
->byte_count
= 0;
3951 group_dpif_get_buckets(group
, &buckets
);
3952 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
3953 bucket
->stats
.packet_count
= 0;
3954 bucket
->stats
.byte_count
= 0;
3959 group_dpif_credit_stats(struct group_dpif
*group
,
3960 struct ofputil_bucket
*bucket
,
3961 const struct dpif_flow_stats
*stats
)
3963 ovs_mutex_lock(&group
->stats_mutex
);
3964 group
->packet_count
+= stats
->n_packets
;
3965 group
->byte_count
+= stats
->n_bytes
;
3967 bucket
->stats
.packet_count
+= stats
->n_packets
;
3968 bucket
->stats
.byte_count
+= stats
->n_bytes
;
3969 } else { /* Credit to all buckets */
3970 const struct list
*buckets
;
3972 group_dpif_get_buckets(group
, &buckets
);
3973 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
3974 bucket
->stats
.packet_count
+= stats
->n_packets
;
3975 bucket
->stats
.byte_count
+= stats
->n_bytes
;
3978 ovs_mutex_unlock(&group
->stats_mutex
);
3982 group_construct(struct ofgroup
*group_
)
3984 struct group_dpif
*group
= group_dpif_cast(group_
);
3985 const struct ofputil_bucket
*bucket
;
3987 /* Prevent group chaining because our locking structure makes it hard to
3988 * implement deadlock-free. (See xlate_group_resource_check().) */
3989 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
3990 const struct ofpact
*a
;
3992 OFPACT_FOR_EACH (a
, bucket
->ofpacts
, bucket
->ofpacts_len
) {
3993 if (a
->type
== OFPACT_GROUP
) {
3994 return OFPERR_OFPGMFC_CHAINING_UNSUPPORTED
;
3999 ovs_mutex_init_adaptive(&group
->stats_mutex
);
4000 ovs_mutex_lock(&group
->stats_mutex
);
4001 group_construct_stats(group
);
4002 ovs_mutex_unlock(&group
->stats_mutex
);
4007 group_destruct(struct ofgroup
*group_
)
4009 struct group_dpif
*group
= group_dpif_cast(group_
);
4010 ovs_mutex_destroy(&group
->stats_mutex
);
4014 group_modify(struct ofgroup
*group_
)
4016 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(group_
->ofproto
);
4018 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
4024 group_get_stats(const struct ofgroup
*group_
, struct ofputil_group_stats
*ogs
)
4026 struct group_dpif
*group
= group_dpif_cast(group_
);
4027 struct ofputil_bucket
*bucket
;
4028 const struct list
*buckets
;
4029 struct bucket_counter
*bucket_stats
;
4031 ovs_mutex_lock(&group
->stats_mutex
);
4032 ogs
->packet_count
= group
->packet_count
;
4033 ogs
->byte_count
= group
->byte_count
;
4035 group_dpif_get_buckets(group
, &buckets
);
4036 bucket_stats
= ogs
->bucket_stats
;
4037 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
4038 bucket_stats
->packet_count
= bucket
->stats
.packet_count
;
4039 bucket_stats
->byte_count
= bucket
->stats
.byte_count
;
4042 ovs_mutex_unlock(&group
->stats_mutex
);
4047 /* If the group exists, this function increments the groups's reference count.
4049 * Make sure to call group_dpif_unref() after no longer needing to maintain
4050 * a reference to the group. */
4052 group_dpif_lookup(struct ofproto_dpif
*ofproto
, uint32_t group_id
,
4053 struct group_dpif
**group
)
4055 struct ofgroup
*ofgroup
;
4058 found
= ofproto_group_lookup(&ofproto
->up
, group_id
, &ofgroup
);
4059 *group
= found
? group_dpif_cast(ofgroup
) : NULL
;
4065 group_dpif_get_buckets(const struct group_dpif
*group
,
4066 const struct list
**buckets
)
4068 *buckets
= &group
->up
.buckets
;
4071 enum ofp11_group_type
4072 group_dpif_get_type(const struct group_dpif
*group
)
4074 return group
->up
.type
;
4077 /* Sends 'packet' out 'ofport'.
4078 * May modify 'packet'.
4079 * Returns 0 if successful, otherwise a positive errno value. */
4081 ofproto_dpif_send_packet(const struct ofport_dpif
*ofport
, struct ofpbuf
*packet
)
4083 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
4086 error
= xlate_send_packet(ofport
, packet
);
4088 ovs_mutex_lock(&ofproto
->stats_mutex
);
4089 ofproto
->stats
.tx_packets
++;
4090 ofproto
->stats
.tx_bytes
+= ofpbuf_size(packet
);
4091 ovs_mutex_unlock(&ofproto
->stats_mutex
);
4095 /* Return the version string of the datapath that backs up
4099 get_datapath_version(const struct ofproto
*ofproto_
)
4101 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4103 return ofproto
->backer
->dp_version_string
;
4107 set_frag_handling(struct ofproto
*ofproto_
,
4108 enum ofp_config_flags frag_handling
)
4110 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4111 if (frag_handling
!= OFPC_FRAG_REASM
) {
4112 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4120 packet_out(struct ofproto
*ofproto_
, struct ofpbuf
*packet
,
4121 const struct flow
*flow
,
4122 const struct ofpact
*ofpacts
, size_t ofpacts_len
)
4124 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4126 ofproto_dpif_execute_actions(ofproto
, flow
, NULL
, ofpacts
,
4127 ofpacts_len
, packet
);
4134 set_netflow(struct ofproto
*ofproto_
,
4135 const struct netflow_options
*netflow_options
)
4137 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4139 if (netflow_options
) {
4140 if (!ofproto
->netflow
) {
4141 ofproto
->netflow
= netflow_create();
4142 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4144 return netflow_set_options(ofproto
->netflow
, netflow_options
);
4145 } else if (ofproto
->netflow
) {
4146 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4147 netflow_unref(ofproto
->netflow
);
4148 ofproto
->netflow
= NULL
;
4155 get_netflow_ids(const struct ofproto
*ofproto_
,
4156 uint8_t *engine_type
, uint8_t *engine_id
)
4158 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4160 dpif_get_netflow_ids(ofproto
->backer
->dpif
, engine_type
, engine_id
);
4163 static struct ofproto_dpif
*
4164 ofproto_dpif_lookup(const char *name
)
4166 struct ofproto_dpif
*ofproto
;
4168 HMAP_FOR_EACH_WITH_HASH (ofproto
, all_ofproto_dpifs_node
,
4169 hash_string(name
, 0), &all_ofproto_dpifs
) {
4170 if (!strcmp(ofproto
->up
.name
, name
)) {
4178 ofproto_unixctl_fdb_flush(struct unixctl_conn
*conn
, int argc
,
4179 const char *argv
[], void *aux OVS_UNUSED
)
4181 struct ofproto_dpif
*ofproto
;
4184 ofproto
= ofproto_dpif_lookup(argv
[1]);
4186 unixctl_command_reply_error(conn
, "no such bridge");
4189 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4190 mac_learning_flush(ofproto
->ml
);
4191 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4193 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4194 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4195 mac_learning_flush(ofproto
->ml
);
4196 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4200 unixctl_command_reply(conn
, "table successfully flushed");
4204 ofproto_unixctl_mcast_snooping_flush(struct unixctl_conn
*conn
, int argc
,
4205 const char *argv
[], void *aux OVS_UNUSED
)
4207 struct ofproto_dpif
*ofproto
;
4210 ofproto
= ofproto_dpif_lookup(argv
[1]);
4212 unixctl_command_reply_error(conn
, "no such bridge");
4216 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4217 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4220 mcast_snooping_mdb_flush(ofproto
->ms
);
4222 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4223 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4226 mcast_snooping_mdb_flush(ofproto
->ms
);
4230 unixctl_command_reply(conn
, "table successfully flushed");
4233 static struct ofport_dpif
*
4234 ofbundle_get_a_port(const struct ofbundle
*bundle
)
4236 return CONTAINER_OF(list_front(&bundle
->ports
), struct ofport_dpif
,
4241 ofproto_unixctl_fdb_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4242 const char *argv
[], void *aux OVS_UNUSED
)
4244 struct ds ds
= DS_EMPTY_INITIALIZER
;
4245 const struct ofproto_dpif
*ofproto
;
4246 const struct mac_entry
*e
;
4248 ofproto
= ofproto_dpif_lookup(argv
[1]);
4250 unixctl_command_reply_error(conn
, "no such bridge");
4254 ds_put_cstr(&ds
, " port VLAN MAC Age\n");
4255 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
4256 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
4257 struct ofbundle
*bundle
= e
->port
.p
;
4258 char name
[OFP_MAX_PORT_NAME_LEN
];
4260 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4262 ds_put_format(&ds
, "%5s %4d "ETH_ADDR_FMT
" %3d\n",
4263 name
, e
->vlan
, ETH_ADDR_ARGS(e
->mac
),
4264 mac_entry_age(ofproto
->ml
, e
));
4266 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4267 unixctl_command_reply(conn
, ds_cstr(&ds
));
4272 ofproto_unixctl_mcast_snooping_show(struct unixctl_conn
*conn
,
4273 int argc OVS_UNUSED
,
4275 void *aux OVS_UNUSED
)
4277 struct ds ds
= DS_EMPTY_INITIALIZER
;
4278 const struct ofproto_dpif
*ofproto
;
4279 const struct ofbundle
*bundle
;
4280 const struct mcast_group
*grp
;
4281 struct mcast_group_bundle
*b
;
4282 struct mcast_mrouter_bundle
*mrouter
;
4284 ofproto
= ofproto_dpif_lookup(argv
[1]);
4286 unixctl_command_reply_error(conn
, "no such bridge");
4290 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4291 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4295 ds_put_cstr(&ds
, " port VLAN GROUP Age\n");
4296 ovs_rwlock_rdlock(&ofproto
->ms
->rwlock
);
4297 LIST_FOR_EACH (grp
, group_node
, &ofproto
->ms
->group_lru
) {
4298 LIST_FOR_EACH(b
, bundle_node
, &grp
->bundle_lru
) {
4299 char name
[OFP_MAX_PORT_NAME_LEN
];
4302 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4304 ds_put_format(&ds
, "%5s %4d "IP_FMT
" %3d\n",
4305 name
, grp
->vlan
, IP_ARGS(grp
->ip4
),
4306 mcast_bundle_age(ofproto
->ms
, b
));
4310 /* ports connected to multicast routers */
4311 LIST_FOR_EACH(mrouter
, mrouter_node
, &ofproto
->ms
->mrouter_lru
) {
4312 char name
[OFP_MAX_PORT_NAME_LEN
];
4314 bundle
= mrouter
->port
;
4315 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4317 ds_put_format(&ds
, "%5s %4d querier %3d\n",
4318 name
, mrouter
->vlan
,
4319 mcast_mrouter_age(ofproto
->ms
, mrouter
));
4321 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
4322 unixctl_command_reply(conn
, ds_cstr(&ds
));
4327 struct xlate_out xout
;
4328 struct xlate_in xin
;
4329 const struct flow
*key
;
4331 struct flow_wildcards wc
;
4336 trace_format_rule(struct ds
*result
, int level
, const struct rule_dpif
*rule
)
4338 const struct rule_actions
*actions
;
4341 ds_put_char_multiple(result
, '\t', level
);
4343 ds_put_cstr(result
, "No match\n");
4347 ovs_mutex_lock(&rule
->up
.mutex
);
4348 cookie
= rule
->up
.flow_cookie
;
4349 ovs_mutex_unlock(&rule
->up
.mutex
);
4351 ds_put_format(result
, "Rule: table=%"PRIu8
" cookie=%#"PRIx64
" ",
4352 rule
? rule
->up
.table_id
: 0, ntohll(cookie
));
4353 cls_rule_format(&rule
->up
.cr
, result
);
4354 ds_put_char(result
, '\n');
4356 actions
= rule_dpif_get_actions(rule
);
4358 ds_put_char_multiple(result
, '\t', level
);
4359 ds_put_cstr(result
, "OpenFlow actions=");
4360 ofpacts_format(actions
->ofpacts
, actions
->ofpacts_len
, result
);
4361 ds_put_char(result
, '\n');
4365 trace_format_flow(struct ds
*result
, int level
, const char *title
,
4366 struct trace_ctx
*trace
)
4368 ds_put_char_multiple(result
, '\t', level
);
4369 ds_put_format(result
, "%s: ", title
);
4370 /* Do not report unchanged flows for resubmits. */
4371 if ((level
> 0 && flow_equal(&trace
->xin
.flow
, &trace
->flow
))
4372 || (level
== 0 && flow_equal(&trace
->xin
.flow
, trace
->key
))) {
4373 ds_put_cstr(result
, "unchanged");
4375 flow_format(result
, &trace
->xin
.flow
);
4376 trace
->flow
= trace
->xin
.flow
;
4378 ds_put_char(result
, '\n');
4382 trace_format_regs(struct ds
*result
, int level
, const char *title
,
4383 struct trace_ctx
*trace
)
4387 ds_put_char_multiple(result
, '\t', level
);
4388 ds_put_format(result
, "%s:", title
);
4389 for (i
= 0; i
< FLOW_N_REGS
; i
++) {
4390 ds_put_format(result
, " reg%"PRIuSIZE
"=0x%"PRIx32
, i
, trace
->flow
.regs
[i
]);
4392 ds_put_char(result
, '\n');
4396 trace_format_odp(struct ds
*result
, int level
, const char *title
,
4397 struct trace_ctx
*trace
)
4399 struct ofpbuf
*odp_actions
= trace
->xout
.odp_actions
;
4401 ds_put_char_multiple(result
, '\t', level
);
4402 ds_put_format(result
, "%s: ", title
);
4403 format_odp_actions(result
, ofpbuf_data(odp_actions
),
4404 ofpbuf_size(odp_actions
));
4405 ds_put_char(result
, '\n');
4409 trace_format_megaflow(struct ds
*result
, int level
, const char *title
,
4410 struct trace_ctx
*trace
)
4414 ds_put_char_multiple(result
, '\t', level
);
4415 ds_put_format(result
, "%s: ", title
);
4416 flow_wildcards_or(&trace
->wc
, &trace
->xout
.wc
, &trace
->wc
);
4417 match_init(&match
, trace
->key
, &trace
->wc
);
4418 match_format(&match
, result
, OFP_DEFAULT_PRIORITY
);
4419 ds_put_char(result
, '\n');
4423 trace_resubmit(struct xlate_in
*xin
, struct rule_dpif
*rule
, int recurse
)
4425 struct trace_ctx
*trace
= CONTAINER_OF(xin
, struct trace_ctx
, xin
);
4426 struct ds
*result
= trace
->result
;
4428 ds_put_char(result
, '\n');
4430 trace_format_flow(result
, recurse
, "Resubmitted flow", trace
);
4431 trace_format_regs(result
, recurse
, "Resubmitted regs", trace
);
4432 trace_format_odp(result
, recurse
, "Resubmitted odp", trace
);
4433 trace_format_megaflow(result
, recurse
, "Resubmitted megaflow", trace
);
4435 trace_format_rule(result
, recurse
, rule
);
4439 trace_report(struct xlate_in
*xin
, const char *s
, int recurse
)
4441 struct trace_ctx
*trace
= CONTAINER_OF(xin
, struct trace_ctx
, xin
);
4442 struct ds
*result
= trace
->result
;
4444 ds_put_char_multiple(result
, '\t', recurse
);
4445 ds_put_cstr(result
, s
);
4446 ds_put_char(result
, '\n');
4449 /* Parses the 'argc' elements of 'argv', ignoring argv[0]. The following
4450 * forms are supported:
4452 * - [dpname] odp_flow [-generate | packet]
4453 * - bridge br_flow [-generate | packet]
4455 * On success, initializes '*ofprotop' and 'flow' and returns NULL. On failure
4456 * returns a nonnull malloced error message. */
4457 static char * WARN_UNUSED_RESULT
4458 parse_flow_and_packet(int argc
, const char *argv
[],
4459 struct ofproto_dpif
**ofprotop
, struct flow
*flow
,
4460 struct ofpbuf
**packetp
)
4462 const struct dpif_backer
*backer
= NULL
;
4463 const char *error
= NULL
;
4465 struct simap port_names
= SIMAP_INITIALIZER(&port_names
);
4466 struct ofpbuf
*packet
;
4467 struct ofpbuf odp_key
;
4468 struct ofpbuf odp_mask
;
4470 ofpbuf_init(&odp_key
, 0);
4471 ofpbuf_init(&odp_mask
, 0);
4473 /* Handle "-generate" or a hex string as the last argument. */
4474 if (!strcmp(argv
[argc
- 1], "-generate")) {
4475 packet
= ofpbuf_new(0);
4478 error
= eth_from_hex(argv
[argc
- 1], &packet
);
4481 } else if (argc
== 4) {
4482 /* The 3-argument form must end in "-generate' or a hex string. */
4488 /* odp_flow can have its in_port specified as a name instead of port no.
4489 * We do not yet know whether a given flow is a odp_flow or a br_flow.
4490 * But, to know whether a flow is odp_flow through odp_flow_from_string(),
4491 * we need to create a simap of name to port no. */
4493 const char *dp_type
;
4494 if (!strncmp(argv
[1], "ovs-", 4)) {
4495 dp_type
= argv
[1] + 4;
4499 backer
= shash_find_data(&all_dpif_backers
, dp_type
);
4500 } else if (argc
== 2) {
4501 struct shash_node
*node
;
4502 if (shash_count(&all_dpif_backers
) == 1) {
4503 node
= shash_first(&all_dpif_backers
);
4504 backer
= node
->data
;
4507 error
= "Syntax error";
4510 if (backer
&& backer
->dpif
) {
4511 struct dpif_port dpif_port
;
4512 struct dpif_port_dump port_dump
;
4513 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, backer
->dpif
) {
4514 simap_put(&port_names
, dpif_port
.name
,
4515 odp_to_u32(dpif_port
.port_no
));
4519 /* Parse the flow and determine whether a datapath or
4520 * bridge is specified. If function odp_flow_key_from_string()
4521 * returns 0, the flow is a odp_flow. If function
4522 * parse_ofp_exact_flow() returns NULL, the flow is a br_flow. */
4523 if (!odp_flow_from_string(argv
[argc
- 1], &port_names
,
4524 &odp_key
, &odp_mask
)) {
4526 error
= "Cannot find the datapath";
4530 if (odp_flow_key_to_flow(ofpbuf_data(&odp_key
), ofpbuf_size(&odp_key
),
4531 flow
) == ODP_FIT_ERROR
) {
4532 error
= "Failed to parse flow key";
4536 *ofprotop
= xlate_lookup_ofproto(backer
, flow
,
4537 &flow
->in_port
.ofp_port
);
4538 if (*ofprotop
== NULL
) {
4539 error
= "Invalid datapath flow";
4543 vsp_adjust_flow(*ofprotop
, flow
, NULL
);
4546 char *err
= parse_ofp_exact_flow(flow
, NULL
, argv
[argc
- 1], NULL
);
4549 m_err
= xasprintf("Bad flow syntax: %s", err
);
4554 error
= "Must specify bridge name";
4558 *ofprotop
= ofproto_dpif_lookup(argv
[1]);
4560 error
= "Unknown bridge name";
4566 /* Generate a packet, if requested. */
4568 if (!ofpbuf_size(packet
)) {
4569 flow_compose(packet
, flow
);
4571 struct pkt_metadata md
= pkt_metadata_from_flow(flow
);
4573 /* Use the metadata from the flow and the packet argument
4574 * to reconstruct the flow. */
4575 flow_extract(packet
, &md
, flow
);
4580 if (error
&& !m_err
) {
4581 m_err
= xstrdup(error
);
4584 ofpbuf_delete(packet
);
4588 ofpbuf_uninit(&odp_key
);
4589 ofpbuf_uninit(&odp_mask
);
4590 simap_destroy(&port_names
);
4595 ofproto_unixctl_trace(struct unixctl_conn
*conn
, int argc
, const char *argv
[],
4596 void *aux OVS_UNUSED
)
4598 struct ofproto_dpif
*ofproto
;
4599 struct ofpbuf
*packet
;
4603 error
= parse_flow_and_packet(argc
, argv
, &ofproto
, &flow
, &packet
);
4608 ofproto_trace(ofproto
, &flow
, packet
, NULL
, 0, &result
);
4609 unixctl_command_reply(conn
, ds_cstr(&result
));
4610 ds_destroy(&result
);
4611 ofpbuf_delete(packet
);
4613 unixctl_command_reply_error(conn
, error
);
4619 ofproto_unixctl_trace_actions(struct unixctl_conn
*conn
, int argc
,
4620 const char *argv
[], void *aux OVS_UNUSED
)
4622 enum ofputil_protocol usable_protocols
;
4623 struct ofproto_dpif
*ofproto
;
4624 bool enforce_consistency
;
4625 struct ofpbuf ofpacts
;
4626 struct ofpbuf
*packet
;
4631 /* Three kinds of error return values! */
4637 ofpbuf_init(&ofpacts
, 0);
4639 /* Parse actions. */
4640 error
= ofpacts_parse_actions(argv
[--argc
], &ofpacts
, &usable_protocols
);
4642 unixctl_command_reply_error(conn
, error
);
4647 /* OpenFlow 1.1 and later suggest that the switch enforces certain forms of
4648 * consistency between the flow and the actions. With -consistent, we
4649 * enforce consistency even for a flow supported in OpenFlow 1.0. */
4650 if (!strcmp(argv
[1], "-consistent")) {
4651 enforce_consistency
= true;
4655 enforce_consistency
= false;
4658 error
= parse_flow_and_packet(argc
, argv
, &ofproto
, &flow
, &packet
);
4660 unixctl_command_reply_error(conn
, error
);
4665 /* Do the same checks as handle_packet_out() in ofproto.c.
4667 * We pass a 'table_id' of 0 to ofproto_check_ofpacts(), which isn't
4668 * strictly correct because these actions aren't in any table, but it's OK
4669 * because it 'table_id' is used only to check goto_table instructions, but
4670 * packet-outs take a list of actions and therefore it can't include
4673 * We skip the "meter" check here because meter is an instruction, not an
4674 * action, and thus cannot appear in ofpacts. */
4675 in_port
= ofp_to_u16(flow
.in_port
.ofp_port
);
4676 if (in_port
>= ofproto
->up
.max_ports
&& in_port
< ofp_to_u16(OFPP_MAX
)) {
4677 unixctl_command_reply_error(conn
, "invalid in_port");
4680 if (enforce_consistency
) {
4681 retval
= ofpacts_check_consistency(ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
),
4682 &flow
, u16_to_ofp(ofproto
->up
.max_ports
),
4683 0, 0, usable_protocols
);
4685 retval
= ofpacts_check(ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
), &flow
,
4686 u16_to_ofp(ofproto
->up
.max_ports
), 0, 0,
4692 ds_put_format(&result
, "Bad actions: %s", ofperr_to_string(retval
));
4693 unixctl_command_reply_error(conn
, ds_cstr(&result
));
4697 ofproto_trace(ofproto
, &flow
, packet
,
4698 ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
), &result
);
4699 unixctl_command_reply(conn
, ds_cstr(&result
));
4702 ds_destroy(&result
);
4703 ofpbuf_delete(packet
);
4704 ofpbuf_uninit(&ofpacts
);
4707 /* Implements a "trace" through 'ofproto''s flow table, appending a textual
4708 * description of the results to 'ds'.
4710 * The trace follows a packet with the specified 'flow' through the flow
4711 * table. 'packet' may be nonnull to trace an actual packet, with consequent
4712 * side effects (if it is nonnull then its flow must be 'flow').
4714 * If 'ofpacts' is nonnull then its 'ofpacts_len' bytes specify the actions to
4715 * trace, otherwise the actions are determined by a flow table lookup. */
4717 ofproto_trace(struct ofproto_dpif
*ofproto
, struct flow
*flow
,
4718 const struct ofpbuf
*packet
,
4719 const struct ofpact ofpacts
[], size_t ofpacts_len
,
4722 struct trace_ctx trace
;
4724 ds_put_format(ds
, "Bridge: %s\n", ofproto
->up
.name
);
4725 ds_put_cstr(ds
, "Flow: ");
4726 flow_format(ds
, flow
);
4727 ds_put_char(ds
, '\n');
4729 flow_wildcards_init_catchall(&trace
.wc
);
4732 trace
.key
= flow
; /* Original flow key, used for megaflow. */
4733 trace
.flow
= *flow
; /* May be modified by actions. */
4734 xlate_in_init(&trace
.xin
, ofproto
, flow
, flow
->in_port
.ofp_port
, NULL
,
4735 ntohs(flow
->tcp_flags
), packet
);
4736 trace
.xin
.ofpacts
= ofpacts
;
4737 trace
.xin
.ofpacts_len
= ofpacts_len
;
4738 trace
.xin
.resubmit_hook
= trace_resubmit
;
4739 trace
.xin
.report_hook
= trace_report
;
4741 xlate_actions(&trace
.xin
, &trace
.xout
);
4743 ds_put_char(ds
, '\n');
4744 trace_format_flow(ds
, 0, "Final flow", &trace
);
4745 trace_format_megaflow(ds
, 0, "Megaflow", &trace
);
4747 ds_put_cstr(ds
, "Datapath actions: ");
4748 format_odp_actions(ds
, ofpbuf_data(trace
.xout
.odp_actions
),
4749 ofpbuf_size(trace
.xout
.odp_actions
));
4751 if (trace
.xout
.slow
) {
4752 enum slow_path_reason slow
;
4754 ds_put_cstr(ds
, "\nThis flow is handled by the userspace "
4755 "slow path because it:");
4757 slow
= trace
.xout
.slow
;
4759 enum slow_path_reason bit
= rightmost_1bit(slow
);
4761 ds_put_format(ds
, "\n\t- %s.",
4762 slow_path_reason_to_explanation(bit
));
4768 xlate_out_uninit(&trace
.xout
);
4771 /* Store the current ofprotos in 'ofproto_shash'. Returns a sorted list
4772 * of the 'ofproto_shash' nodes. It is the responsibility of the caller
4773 * to destroy 'ofproto_shash' and free the returned value. */
4774 static const struct shash_node
**
4775 get_ofprotos(struct shash
*ofproto_shash
)
4777 const struct ofproto_dpif
*ofproto
;
4779 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4780 char *name
= xasprintf("%s@%s", ofproto
->up
.type
, ofproto
->up
.name
);
4781 shash_add_nocopy(ofproto_shash
, name
, ofproto
);
4784 return shash_sort(ofproto_shash
);
4788 ofproto_unixctl_dpif_dump_dps(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4789 const char *argv
[] OVS_UNUSED
,
4790 void *aux OVS_UNUSED
)
4792 struct ds ds
= DS_EMPTY_INITIALIZER
;
4793 struct shash ofproto_shash
;
4794 const struct shash_node
**sorted_ofprotos
;
4797 shash_init(&ofproto_shash
);
4798 sorted_ofprotos
= get_ofprotos(&ofproto_shash
);
4799 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4800 const struct shash_node
*node
= sorted_ofprotos
[i
];
4801 ds_put_format(&ds
, "%s\n", node
->name
);
4804 shash_destroy(&ofproto_shash
);
4805 free(sorted_ofprotos
);
4807 unixctl_command_reply(conn
, ds_cstr(&ds
));
4812 dpif_show_backer(const struct dpif_backer
*backer
, struct ds
*ds
)
4814 const struct shash_node
**ofprotos
;
4815 struct dpif_dp_stats dp_stats
;
4816 struct shash ofproto_shash
;
4819 dpif_get_dp_stats(backer
->dpif
, &dp_stats
);
4821 ds_put_format(ds
, "%s: hit:%"PRIu64
" missed:%"PRIu64
"\n",
4822 dpif_name(backer
->dpif
), dp_stats
.n_hit
, dp_stats
.n_missed
);
4824 shash_init(&ofproto_shash
);
4825 ofprotos
= get_ofprotos(&ofproto_shash
);
4826 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4827 struct ofproto_dpif
*ofproto
= ofprotos
[i
]->data
;
4828 const struct shash_node
**ports
;
4831 if (ofproto
->backer
!= backer
) {
4835 ds_put_format(ds
, "\t%s:\n", ofproto
->up
.name
);
4837 ports
= shash_sort(&ofproto
->up
.port_by_name
);
4838 for (j
= 0; j
< shash_count(&ofproto
->up
.port_by_name
); j
++) {
4839 const struct shash_node
*node
= ports
[j
];
4840 struct ofport
*ofport
= node
->data
;
4842 odp_port_t odp_port
;
4844 ds_put_format(ds
, "\t\t%s %u/", netdev_get_name(ofport
->netdev
),
4847 odp_port
= ofp_port_to_odp_port(ofproto
, ofport
->ofp_port
);
4848 if (odp_port
!= ODPP_NONE
) {
4849 ds_put_format(ds
, "%"PRIu32
":", odp_port
);
4851 ds_put_cstr(ds
, "none:");
4854 ds_put_format(ds
, " (%s", netdev_get_type(ofport
->netdev
));
4857 if (!netdev_get_config(ofport
->netdev
, &config
)) {
4858 const struct smap_node
**nodes
;
4861 nodes
= smap_sort(&config
);
4862 for (i
= 0; i
< smap_count(&config
); i
++) {
4863 const struct smap_node
*node
= nodes
[i
];
4864 ds_put_format(ds
, "%c %s=%s", i
? ',' : ':',
4865 node
->key
, node
->value
);
4869 smap_destroy(&config
);
4871 ds_put_char(ds
, ')');
4872 ds_put_char(ds
, '\n');
4876 shash_destroy(&ofproto_shash
);
4881 ofproto_unixctl_dpif_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4882 const char *argv
[] OVS_UNUSED
, void *aux OVS_UNUSED
)
4884 struct ds ds
= DS_EMPTY_INITIALIZER
;
4885 const struct shash_node
**backers
;
4888 backers
= shash_sort(&all_dpif_backers
);
4889 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
4890 dpif_show_backer(backers
[i
]->data
, &ds
);
4894 unixctl_command_reply(conn
, ds_cstr(&ds
));
4899 ofproto_unixctl_dpif_dump_flows(struct unixctl_conn
*conn
,
4900 int argc OVS_UNUSED
, const char *argv
[],
4901 void *aux OVS_UNUSED
)
4903 const struct ofproto_dpif
*ofproto
;
4905 struct ds ds
= DS_EMPTY_INITIALIZER
;
4906 bool verbosity
= false;
4908 struct dpif_port dpif_port
;
4909 struct dpif_port_dump port_dump
;
4910 struct hmap portno_names
;
4912 struct dpif_flow_dump
*flow_dump
;
4913 struct dpif_flow_dump_thread
*flow_dump_thread
;
4917 ofproto
= ofproto_dpif_lookup(argv
[argc
- 1]);
4919 unixctl_command_reply_error(conn
, "no such bridge");
4923 if (argc
> 2 && !strcmp(argv
[1], "-m")) {
4927 hmap_init(&portno_names
);
4928 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, ofproto
->backer
->dpif
) {
4929 odp_portno_names_set(&portno_names
, dpif_port
.port_no
, dpif_port
.name
);
4933 flow_dump
= dpif_flow_dump_create(ofproto
->backer
->dpif
);
4934 flow_dump_thread
= dpif_flow_dump_thread_create(flow_dump
);
4935 while (dpif_flow_dump_next(flow_dump_thread
, &f
, 1)) {
4938 if (odp_flow_key_to_flow(f
.key
, f
.key_len
, &flow
) == ODP_FIT_ERROR
4939 || xlate_lookup_ofproto(ofproto
->backer
, &flow
, NULL
) != ofproto
) {
4943 odp_flow_format(f
.key
, f
.key_len
, f
.mask
, f
.mask_len
,
4944 &portno_names
, &ds
, verbosity
);
4945 ds_put_cstr(&ds
, ", ");
4946 dpif_flow_stats_format(&f
.stats
, &ds
);
4947 ds_put_cstr(&ds
, ", actions:");
4948 format_odp_actions(&ds
, f
.actions
, f
.actions_len
);
4949 ds_put_char(&ds
, '\n');
4951 dpif_flow_dump_thread_destroy(flow_dump_thread
);
4952 error
= dpif_flow_dump_destroy(flow_dump
);
4956 ds_put_format(&ds
, "dpif/dump_flows failed: %s", ovs_strerror(errno
));
4957 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
4959 unixctl_command_reply(conn
, ds_cstr(&ds
));
4961 odp_portno_names_destroy(&portno_names
);
4962 hmap_destroy(&portno_names
);
4967 ofproto_dpif_unixctl_init(void)
4969 static bool registered
;
4975 unixctl_command_register(
4977 "{[dp_name] odp_flow | bridge br_flow} [-generate|packet]",
4978 1, 3, ofproto_unixctl_trace
, NULL
);
4979 unixctl_command_register(
4980 "ofproto/trace-packet-out",
4981 "[-consistent] {[dp_name] odp_flow | bridge br_flow} [-generate|packet] actions",
4982 2, 6, ofproto_unixctl_trace_actions
, NULL
);
4983 unixctl_command_register("fdb/flush", "[bridge]", 0, 1,
4984 ofproto_unixctl_fdb_flush
, NULL
);
4985 unixctl_command_register("fdb/show", "bridge", 1, 1,
4986 ofproto_unixctl_fdb_show
, NULL
);
4987 unixctl_command_register("mdb/flush", "[bridge]", 0, 1,
4988 ofproto_unixctl_mcast_snooping_flush
, NULL
);
4989 unixctl_command_register("mdb/show", "bridge", 1, 1,
4990 ofproto_unixctl_mcast_snooping_show
, NULL
);
4991 unixctl_command_register("dpif/dump-dps", "", 0, 0,
4992 ofproto_unixctl_dpif_dump_dps
, NULL
);
4993 unixctl_command_register("dpif/show", "", 0, 0, ofproto_unixctl_dpif_show
,
4995 unixctl_command_register("dpif/dump-flows", "[-m] bridge", 1, 2,
4996 ofproto_unixctl_dpif_dump_flows
, NULL
);
4999 /* Returns true if 'table' is the table used for internal rules,
5000 * false otherwise. */
5002 table_is_internal(uint8_t table_id
)
5004 return table_id
== TBL_INTERNAL
;
5007 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
5009 * This is deprecated. It is only for compatibility with broken device drivers
5010 * in old versions of Linux that do not properly support VLANs when VLAN
5011 * devices are not used. When broken device drivers are no longer in
5012 * widespread use, we will delete these interfaces. */
5015 set_realdev(struct ofport
*ofport_
, ofp_port_t realdev_ofp_port
, int vid
)
5017 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
5018 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
5020 if (realdev_ofp_port
== ofport
->realdev_ofp_port
5021 && vid
== ofport
->vlandev_vid
) {
5025 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
5027 if (ofport
->realdev_ofp_port
) {
5030 if (realdev_ofp_port
&& ofport
->bundle
) {
5031 /* vlandevs are enslaved to their realdevs, so they are not allowed to
5032 * themselves be part of a bundle. */
5033 bundle_set(ofport_
->ofproto
, ofport
->bundle
, NULL
);
5036 ofport
->realdev_ofp_port
= realdev_ofp_port
;
5037 ofport
->vlandev_vid
= vid
;
5039 if (realdev_ofp_port
) {
5040 vsp_add(ofport
, realdev_ofp_port
, vid
);
5047 hash_realdev_vid(ofp_port_t realdev_ofp_port
, int vid
)
5049 return hash_2words(ofp_to_u16(realdev_ofp_port
), vid
);
5053 ofproto_has_vlan_splinters(const struct ofproto_dpif
*ofproto
)
5054 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5056 /* hmap_is_empty is thread safe. */
5057 return !hmap_is_empty(&ofproto
->realdev_vid_map
);
5062 vsp_realdev_to_vlandev__(const struct ofproto_dpif
*ofproto
,
5063 ofp_port_t realdev_ofp_port
, ovs_be16 vlan_tci
)
5064 OVS_REQUIRES(ofproto
->vsp_mutex
)
5066 if (!hmap_is_empty(&ofproto
->realdev_vid_map
)) {
5067 int vid
= vlan_tci_to_vid(vlan_tci
);
5068 const struct vlan_splinter
*vsp
;
5070 HMAP_FOR_EACH_WITH_HASH (vsp
, realdev_vid_node
,
5071 hash_realdev_vid(realdev_ofp_port
, vid
),
5072 &ofproto
->realdev_vid_map
) {
5073 if (vsp
->realdev_ofp_port
== realdev_ofp_port
5074 && vsp
->vid
== vid
) {
5075 return vsp
->vlandev_ofp_port
;
5079 return realdev_ofp_port
;
5082 /* Returns the OFP port number of the Linux VLAN device that corresponds to
5083 * 'vlan_tci' on the network device with port number 'realdev_ofp_port' in
5084 * 'struct ofport_dpif'. For example, given 'realdev_ofp_port' of eth0 and
5085 * 'vlan_tci' 9, it would return the port number of eth0.9.
5087 * Unless VLAN splinters are enabled for port 'realdev_ofp_port', this
5088 * function just returns its 'realdev_ofp_port' argument. */
5090 vsp_realdev_to_vlandev(const struct ofproto_dpif
*ofproto
,
5091 ofp_port_t realdev_ofp_port
, ovs_be16 vlan_tci
)
5092 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5096 /* hmap_is_empty is thread safe, see if we can return immediately. */
5097 if (hmap_is_empty(&ofproto
->realdev_vid_map
)) {
5098 return realdev_ofp_port
;
5100 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5101 ret
= vsp_realdev_to_vlandev__(ofproto
, realdev_ofp_port
, vlan_tci
);
5102 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5106 static struct vlan_splinter
*
5107 vlandev_find(const struct ofproto_dpif
*ofproto
, ofp_port_t vlandev_ofp_port
)
5109 struct vlan_splinter
*vsp
;
5111 HMAP_FOR_EACH_WITH_HASH (vsp
, vlandev_node
,
5112 hash_ofp_port(vlandev_ofp_port
),
5113 &ofproto
->vlandev_map
) {
5114 if (vsp
->vlandev_ofp_port
== vlandev_ofp_port
) {
5122 /* Returns the OpenFlow port number of the "real" device underlying the Linux
5123 * VLAN device with OpenFlow port number 'vlandev_ofp_port' and stores the
5124 * VLAN VID of the Linux VLAN device in '*vid'. For example, given
5125 * 'vlandev_ofp_port' of eth0.9, it would return the OpenFlow port number of
5126 * eth0 and store 9 in '*vid'.
5128 * Returns 0 and does not modify '*vid' if 'vlandev_ofp_port' is not a Linux
5129 * VLAN device. Unless VLAN splinters are enabled, this is what this function
5132 vsp_vlandev_to_realdev(const struct ofproto_dpif
*ofproto
,
5133 ofp_port_t vlandev_ofp_port
, int *vid
)
5134 OVS_REQUIRES(ofproto
->vsp_mutex
)
5136 if (!hmap_is_empty(&ofproto
->vlandev_map
)) {
5137 const struct vlan_splinter
*vsp
;
5139 vsp
= vlandev_find(ofproto
, vlandev_ofp_port
);
5144 return vsp
->realdev_ofp_port
;
5150 /* Given 'flow', a flow representing a packet received on 'ofproto', checks
5151 * whether 'flow->in_port' represents a Linux VLAN device. If so, changes
5152 * 'flow->in_port' to the "real" device backing the VLAN device, sets
5153 * 'flow->vlan_tci' to the VLAN VID, and returns true. Optionally pushes the
5154 * appropriate VLAN on 'packet' if provided. Otherwise (which is always the
5155 * case unless VLAN splinters are enabled), returns false without making any
5158 vsp_adjust_flow(const struct ofproto_dpif
*ofproto
, struct flow
*flow
,
5159 struct ofpbuf
*packet
)
5160 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5165 /* hmap_is_empty is thread safe. */
5166 if (hmap_is_empty(&ofproto
->vlandev_map
)) {
5170 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5171 realdev
= vsp_vlandev_to_realdev(ofproto
, flow
->in_port
.ofp_port
, &vid
);
5172 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5177 /* Cause the flow to be processed as if it came in on the real device with
5178 * the VLAN device's VLAN ID. */
5179 flow
->in_port
.ofp_port
= realdev
;
5180 flow
->vlan_tci
= htons((vid
& VLAN_VID_MASK
) | VLAN_CFI
);
5183 /* Make the packet resemble the flow, so that it gets sent to an
5184 * OpenFlow controller properly, so that it looks correct for sFlow,
5185 * and so that flow_extract() will get the correct vlan_tci if it is
5186 * called on 'packet'. */
5187 eth_push_vlan(packet
, htons(ETH_TYPE_VLAN
), flow
->vlan_tci
);
5194 vsp_remove(struct ofport_dpif
*port
)
5196 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
5197 struct vlan_splinter
*vsp
;
5199 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5200 vsp
= vlandev_find(ofproto
, port
->up
.ofp_port
);
5202 hmap_remove(&ofproto
->vlandev_map
, &vsp
->vlandev_node
);
5203 hmap_remove(&ofproto
->realdev_vid_map
, &vsp
->realdev_vid_node
);
5206 port
->realdev_ofp_port
= 0;
5208 VLOG_ERR("missing vlan device record");
5210 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5214 vsp_add(struct ofport_dpif
*port
, ofp_port_t realdev_ofp_port
, int vid
)
5216 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
5218 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5219 if (!vsp_vlandev_to_realdev(ofproto
, port
->up
.ofp_port
, NULL
)
5220 && (vsp_realdev_to_vlandev__(ofproto
, realdev_ofp_port
, htons(vid
))
5221 == realdev_ofp_port
)) {
5222 struct vlan_splinter
*vsp
;
5224 vsp
= xmalloc(sizeof *vsp
);
5225 vsp
->realdev_ofp_port
= realdev_ofp_port
;
5226 vsp
->vlandev_ofp_port
= port
->up
.ofp_port
;
5229 port
->realdev_ofp_port
= realdev_ofp_port
;
5231 hmap_insert(&ofproto
->vlandev_map
, &vsp
->vlandev_node
,
5232 hash_ofp_port(port
->up
.ofp_port
));
5233 hmap_insert(&ofproto
->realdev_vid_map
, &vsp
->realdev_vid_node
,
5234 hash_realdev_vid(realdev_ofp_port
, vid
));
5236 VLOG_ERR("duplicate vlan device record");
5238 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5242 ofp_port_to_odp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
5244 const struct ofport_dpif
*ofport
= get_ofp_port(ofproto
, ofp_port
);
5245 return ofport
? ofport
->odp_port
: ODPP_NONE
;
5248 struct ofport_dpif
*
5249 odp_port_to_ofport(const struct dpif_backer
*backer
, odp_port_t odp_port
)
5251 struct ofport_dpif
*port
;
5253 ovs_rwlock_rdlock(&backer
->odp_to_ofport_lock
);
5254 HMAP_FOR_EACH_IN_BUCKET (port
, odp_port_node
, hash_odp_port(odp_port
),
5255 &backer
->odp_to_ofport_map
) {
5256 if (port
->odp_port
== odp_port
) {
5257 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5262 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5267 odp_port_to_ofp_port(const struct ofproto_dpif
*ofproto
, odp_port_t odp_port
)
5269 struct ofport_dpif
*port
;
5271 port
= odp_port_to_ofport(ofproto
->backer
, odp_port
);
5272 if (port
&& &ofproto
->up
== port
->up
.ofproto
) {
5273 return port
->up
.ofp_port
;
5280 ofproto_dpif_alloc_recirc_id(struct ofproto_dpif
*ofproto
)
5282 struct dpif_backer
*backer
= ofproto
->backer
;
5284 return recirc_id_alloc(backer
->rid_pool
);
5288 ofproto_dpif_free_recirc_id(struct ofproto_dpif
*ofproto
, uint32_t recirc_id
)
5290 struct dpif_backer
*backer
= ofproto
->backer
;
5292 recirc_id_free(backer
->rid_pool
, recirc_id
);
5296 ofproto_dpif_add_internal_flow(struct ofproto_dpif
*ofproto
,
5297 const struct match
*match
, int priority
,
5298 uint16_t idle_timeout
,
5299 const struct ofpbuf
*ofpacts
,
5300 struct rule
**rulep
)
5302 struct ofputil_flow_mod fm
;
5303 struct rule_dpif
*rule
;
5307 fm
.priority
= priority
;
5308 fm
.new_cookie
= htonll(0);
5309 fm
.cookie
= htonll(0);
5310 fm
.cookie_mask
= htonll(0);
5311 fm
.modify_cookie
= false;
5312 fm
.table_id
= TBL_INTERNAL
;
5313 fm
.command
= OFPFC_ADD
;
5314 fm
.idle_timeout
= idle_timeout
;
5315 fm
.hard_timeout
= 0;
5319 fm
.flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
;
5320 fm
.ofpacts
= ofpbuf_data(ofpacts
);
5321 fm
.ofpacts_len
= ofpbuf_size(ofpacts
);
5323 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5325 VLOG_ERR_RL(&rl
, "failed to add internal flow (%s)",
5326 ofperr_to_string(error
));
5331 rule
= rule_dpif_lookup_in_table(ofproto
, TBL_INTERNAL
, &fm
.match
.flow
,
5332 &fm
.match
.wc
, false);
5342 ofproto_dpif_delete_internal_flow(struct ofproto_dpif
*ofproto
,
5343 struct match
*match
, int priority
)
5345 struct ofputil_flow_mod fm
;
5349 fm
.priority
= priority
;
5350 fm
.new_cookie
= htonll(0);
5351 fm
.cookie
= htonll(0);
5352 fm
.cookie_mask
= htonll(0);
5353 fm
.modify_cookie
= false;
5354 fm
.table_id
= TBL_INTERNAL
;
5355 fm
.flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
;
5356 fm
.command
= OFPFC_DELETE_STRICT
;
5358 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5360 VLOG_ERR_RL(&rl
, "failed to delete internal flow (%s)",
5361 ofperr_to_string(error
));
5368 const struct ofproto_class ofproto_dpif_class
= {
5382 NULL
, /* get_memory_usage. */
5383 type_get_memory_usage
,
5401 port_is_lacp_current
,
5402 NULL
, /* rule_choose_table */
5411 NULL
, /* rule_premodify_actions */
5412 rule_modify_actions
,
5428 get_stp_port_status
,
5433 get_rstp_port_status
,
5440 is_mirror_output_bundle
,
5441 forward_bpdu_changed
,
5442 set_mac_table_config
,
5444 set_mcast_snooping_port
,
5446 NULL
, /* meter_get_features */
5447 NULL
, /* meter_set */
5448 NULL
, /* meter_get */
5449 NULL
, /* meter_del */
5450 group_alloc
, /* group_alloc */
5451 group_construct
, /* group_construct */
5452 group_destruct
, /* group_destruct */
5453 group_dealloc
, /* group_dealloc */
5454 group_modify
, /* group_modify */
5455 group_get_stats
, /* group_get_stats */
5456 get_datapath_version
, /* get_datapath_version */