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
->ml
,
622 ofproto
->stp
, ofproto
->rstp
, ofproto
->ms
,
623 ofproto
->mbridge
, ofproto
->sflow
, ofproto
->ipfix
,
625 ofproto
->up
.forward_bpdu
,
626 connmgr_has_in_band(ofproto
->up
.connmgr
),
627 ofproto
->backer
->enable_recirc
,
628 ofproto
->backer
->variable_length_userdata
,
629 ofproto
->backer
->max_mpls_depth
,
630 ofproto
->backer
->masked_set_action
);
632 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
633 xlate_bundle_set(ofproto
, bundle
, bundle
->name
,
634 bundle
->vlan_mode
, bundle
->vlan
,
635 bundle
->trunks
, bundle
->use_priority_tags
,
636 bundle
->bond
, bundle
->lacp
,
640 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
641 int stp_port
= ofport
->stp_port
642 ? stp_port_no(ofport
->stp_port
)
644 xlate_ofport_set(ofproto
, ofport
->bundle
, ofport
,
645 ofport
->up
.ofp_port
, ofport
->odp_port
,
646 ofport
->up
.netdev
, ofport
->cfm
,
647 ofport
->bfd
, ofport
->peer
, stp_port
,
648 ofport
->rstp_port
, ofport
->qdscp
,
649 ofport
->n_qdscp
, ofport
->up
.pp
.config
,
650 ofport
->up
.pp
.state
, ofport
->is_tunnel
,
656 udpif_revalidate(backer
->udpif
);
659 process_dpif_port_changes(backer
);
664 /* Check for and handle port changes in 'backer''s dpif. */
666 process_dpif_port_changes(struct dpif_backer
*backer
)
672 error
= dpif_port_poll(backer
->dpif
, &devname
);
678 process_dpif_all_ports_changed(backer
);
682 process_dpif_port_change(backer
, devname
);
687 process_dpif_port_error(backer
, error
);
694 process_dpif_all_ports_changed(struct dpif_backer
*backer
)
696 struct ofproto_dpif
*ofproto
;
697 struct dpif_port dpif_port
;
698 struct dpif_port_dump dump
;
699 struct sset devnames
;
702 sset_init(&devnames
);
703 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
704 if (ofproto
->backer
== backer
) {
705 struct ofport
*ofport
;
707 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
708 sset_add(&devnames
, netdev_get_name(ofport
->netdev
));
712 DPIF_PORT_FOR_EACH (&dpif_port
, &dump
, backer
->dpif
) {
713 sset_add(&devnames
, dpif_port
.name
);
716 SSET_FOR_EACH (devname
, &devnames
) {
717 process_dpif_port_change(backer
, devname
);
719 sset_destroy(&devnames
);
723 process_dpif_port_change(struct dpif_backer
*backer
, const char *devname
)
725 struct ofproto_dpif
*ofproto
;
726 struct dpif_port port
;
728 /* Don't report on the datapath's device. */
729 if (!strcmp(devname
, dpif_base_name(backer
->dpif
))) {
733 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
,
734 &all_ofproto_dpifs
) {
735 if (simap_contains(&ofproto
->backer
->tnl_backers
, devname
)) {
740 ofproto
= lookup_ofproto_dpif_by_port_name(devname
);
741 if (dpif_port_query_by_name(backer
->dpif
, devname
, &port
)) {
742 /* The port was removed. If we know the datapath,
743 * report it through poll_set(). If we don't, it may be
744 * notifying us of a removal we initiated, so ignore it.
745 * If there's a pending ENOBUFS, let it stand, since
746 * everything will be reevaluated. */
747 if (ofproto
&& ofproto
->port_poll_errno
!= ENOBUFS
) {
748 sset_add(&ofproto
->port_poll_set
, devname
);
749 ofproto
->port_poll_errno
= 0;
751 } else if (!ofproto
) {
752 /* The port was added, but we don't know with which
753 * ofproto we should associate it. Delete it. */
754 dpif_port_del(backer
->dpif
, port
.port_no
);
756 struct ofport_dpif
*ofport
;
758 ofport
= ofport_dpif_cast(shash_find_data(
759 &ofproto
->up
.port_by_name
, devname
));
761 && ofport
->odp_port
!= port
.port_no
762 && !odp_port_to_ofport(backer
, port
.port_no
))
764 /* 'ofport''s datapath port number has changed from
765 * 'ofport->odp_port' to 'port.port_no'. Update our internal data
766 * structures to match. */
767 ovs_rwlock_wrlock(&backer
->odp_to_ofport_lock
);
768 hmap_remove(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
);
769 ofport
->odp_port
= port
.port_no
;
770 hmap_insert(&backer
->odp_to_ofport_map
, &ofport
->odp_port_node
,
771 hash_odp_port(port
.port_no
));
772 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
773 backer
->need_revalidate
= REV_RECONFIGURE
;
776 dpif_port_destroy(&port
);
779 /* Propagate 'error' to all ofprotos based on 'backer'. */
781 process_dpif_port_error(struct dpif_backer
*backer
, int error
)
783 struct ofproto_dpif
*ofproto
;
785 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
786 if (ofproto
->backer
== backer
) {
787 sset_clear(&ofproto
->port_poll_set
);
788 ofproto
->port_poll_errno
= error
;
794 type_wait(const char *type
)
796 struct dpif_backer
*backer
;
798 backer
= shash_find_data(&all_dpif_backers
, type
);
800 /* This is not necessarily a problem, since backers are only
801 * created on demand. */
805 dpif_wait(backer
->dpif
);
808 /* Basic life-cycle. */
810 static int add_internal_flows(struct ofproto_dpif
*);
812 static struct ofproto
*
815 struct ofproto_dpif
*ofproto
= xmalloc(sizeof *ofproto
);
820 dealloc(struct ofproto
*ofproto_
)
822 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
827 close_dpif_backer(struct dpif_backer
*backer
)
829 ovs_assert(backer
->refcount
> 0);
831 if (--backer
->refcount
) {
835 udpif_destroy(backer
->udpif
);
837 simap_destroy(&backer
->tnl_backers
);
838 ovs_rwlock_destroy(&backer
->odp_to_ofport_lock
);
839 hmap_destroy(&backer
->odp_to_ofport_map
);
840 shash_find_and_delete(&all_dpif_backers
, backer
->type
);
841 recirc_id_pool_destroy(backer
->rid_pool
);
843 free(backer
->dp_version_string
);
844 dpif_close(backer
->dpif
);
848 /* Datapath port slated for removal from datapath. */
850 struct list list_node
;
854 static bool check_variable_length_userdata(struct dpif_backer
*backer
);
855 static size_t check_max_mpls_depth(struct dpif_backer
*backer
);
856 static bool check_recirc(struct dpif_backer
*backer
);
857 static bool check_masked_set_action(struct dpif_backer
*backer
);
860 open_dpif_backer(const char *type
, struct dpif_backer
**backerp
)
862 struct dpif_backer
*backer
;
863 struct dpif_port_dump port_dump
;
864 struct dpif_port port
;
865 struct shash_node
*node
;
866 struct list garbage_list
;
867 struct odp_garbage
*garbage
, *next
;
874 backer
= shash_find_data(&all_dpif_backers
, type
);
881 backer_name
= xasprintf("ovs-%s", type
);
883 /* Remove any existing datapaths, since we assume we're the only
884 * userspace controlling the datapath. */
886 dp_enumerate_names(type
, &names
);
887 SSET_FOR_EACH(name
, &names
) {
888 struct dpif
*old_dpif
;
890 /* Don't remove our backer if it exists. */
891 if (!strcmp(name
, backer_name
)) {
895 if (dpif_open(name
, type
, &old_dpif
)) {
896 VLOG_WARN("couldn't open old datapath %s to remove it", name
);
898 dpif_delete(old_dpif
);
899 dpif_close(old_dpif
);
902 sset_destroy(&names
);
904 backer
= xmalloc(sizeof *backer
);
906 error
= dpif_create_and_open(backer_name
, type
, &backer
->dpif
);
909 VLOG_ERR("failed to open datapath of type %s: %s", type
,
910 ovs_strerror(error
));
914 backer
->udpif
= udpif_create(backer
, backer
->dpif
);
916 backer
->type
= xstrdup(type
);
917 backer
->refcount
= 1;
918 hmap_init(&backer
->odp_to_ofport_map
);
919 ovs_rwlock_init(&backer
->odp_to_ofport_lock
);
920 backer
->need_revalidate
= 0;
921 simap_init(&backer
->tnl_backers
);
922 backer
->recv_set_enable
= !ofproto_get_flow_restore_wait();
925 if (backer
->recv_set_enable
) {
926 dpif_flow_flush(backer
->dpif
);
929 /* Loop through the ports already on the datapath and remove any
930 * that we don't need anymore. */
931 list_init(&garbage_list
);
932 dpif_port_dump_start(&port_dump
, backer
->dpif
);
933 while (dpif_port_dump_next(&port_dump
, &port
)) {
934 node
= shash_find(&init_ofp_ports
, port
.name
);
935 if (!node
&& strcmp(port
.name
, dpif_base_name(backer
->dpif
))) {
936 garbage
= xmalloc(sizeof *garbage
);
937 garbage
->odp_port
= port
.port_no
;
938 list_push_front(&garbage_list
, &garbage
->list_node
);
941 dpif_port_dump_done(&port_dump
);
943 LIST_FOR_EACH_SAFE (garbage
, next
, list_node
, &garbage_list
) {
944 dpif_port_del(backer
->dpif
, garbage
->odp_port
);
945 list_remove(&garbage
->list_node
);
949 shash_add(&all_dpif_backers
, type
, backer
);
951 backer
->enable_recirc
= check_recirc(backer
);
952 backer
->max_mpls_depth
= check_max_mpls_depth(backer
);
953 backer
->masked_set_action
= check_masked_set_action(backer
);
954 backer
->rid_pool
= recirc_id_pool_create();
956 error
= dpif_recv_set(backer
->dpif
, backer
->recv_set_enable
);
958 VLOG_ERR("failed to listen on datapath of type %s: %s",
959 type
, ovs_strerror(error
));
960 close_dpif_backer(backer
);
964 if (backer
->recv_set_enable
) {
965 udpif_set_threads(backer
->udpif
, n_handlers
, n_revalidators
);
968 /* This check fails if performed before udpif threads have been set,
969 * as the kernel module checks that the 'pid' in userspace action
971 backer
->variable_length_userdata
= check_variable_length_userdata(backer
);
972 backer
->dp_version_string
= dpif_get_dp_version(backer
->dpif
);
977 /* Tests whether 'backer''s datapath supports recirculation. Only newer
978 * datapaths support OVS_KEY_ATTR_RECIRC_ID in keys. We need to disable some
979 * features on older datapaths that don't support this feature.
981 * Returns false if 'backer' definitely does not support recirculation, true if
982 * it seems to support recirculation or if at least the error we get is
985 check_recirc(struct dpif_backer
*backer
)
988 struct odputil_keybuf keybuf
;
991 bool enable_recirc
= false;
993 memset(&flow
, 0, sizeof flow
);
997 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
998 odp_flow_key_from_flow(&key
, &flow
, NULL
, 0, true);
1000 error
= dpif_flow_put(backer
->dpif
, DPIF_FP_CREATE
| DPIF_FP_PROBE
,
1001 ofpbuf_data(&key
), ofpbuf_size(&key
), NULL
, 0, NULL
,
1003 if (error
&& error
!= EEXIST
) {
1004 if (error
!= EINVAL
) {
1005 VLOG_WARN("%s: Reciculation flow probe failed (%s)",
1006 dpif_name(backer
->dpif
), ovs_strerror(error
));
1011 error
= dpif_flow_del(backer
->dpif
, ofpbuf_data(&key
), ofpbuf_size(&key
),
1014 VLOG_WARN("%s: failed to delete recirculation feature probe flow",
1015 dpif_name(backer
->dpif
));
1018 enable_recirc
= true;
1021 if (enable_recirc
) {
1022 VLOG_INFO("%s: Datapath supports recirculation",
1023 dpif_name(backer
->dpif
));
1025 VLOG_INFO("%s: Datapath does not support recirculation",
1026 dpif_name(backer
->dpif
));
1029 return enable_recirc
;
1032 /* Tests whether 'backer''s datapath supports variable-length
1033 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. We need
1034 * to disable some features on older datapaths that don't support this
1037 * Returns false if 'backer' definitely does not support variable-length
1038 * userdata, true if it seems to support them or if at least the error we get
1041 check_variable_length_userdata(struct dpif_backer
*backer
)
1043 struct eth_header
*eth
;
1044 struct ofpbuf actions
;
1045 struct dpif_execute execute
;
1046 struct ofpbuf packet
;
1050 /* Compose a userspace action that will cause an ERANGE error on older
1051 * datapaths that don't support variable-length userdata.
1053 * We really test for using userdata longer than 8 bytes, but older
1054 * datapaths accepted these, silently truncating the userdata to 8 bytes.
1055 * The same older datapaths rejected userdata shorter than 8 bytes, so we
1056 * test for that instead as a proxy for longer userdata support. */
1057 ofpbuf_init(&actions
, 64);
1058 start
= nl_msg_start_nested(&actions
, OVS_ACTION_ATTR_USERSPACE
);
1059 nl_msg_put_u32(&actions
, OVS_USERSPACE_ATTR_PID
,
1060 dpif_port_get_pid(backer
->dpif
, ODPP_NONE
, 0));
1061 nl_msg_put_unspec_zero(&actions
, OVS_USERSPACE_ATTR_USERDATA
, 4);
1062 nl_msg_end_nested(&actions
, start
);
1064 /* Compose a dummy ethernet packet. */
1065 ofpbuf_init(&packet
, ETH_HEADER_LEN
);
1066 eth
= ofpbuf_put_zeros(&packet
, ETH_HEADER_LEN
);
1067 eth
->eth_type
= htons(0x1234);
1069 /* Execute the actions. On older datapaths this fails with ERANGE, on
1070 * newer datapaths it succeeds. */
1071 execute
.actions
= ofpbuf_data(&actions
);
1072 execute
.actions_len
= ofpbuf_size(&actions
);
1073 execute
.packet
= &packet
;
1074 execute
.md
= PKT_METADATA_INITIALIZER(0);
1075 execute
.needs_help
= false;
1076 execute
.probe
= true;
1078 error
= dpif_execute(backer
->dpif
, &execute
);
1080 ofpbuf_uninit(&packet
);
1081 ofpbuf_uninit(&actions
);
1088 /* Variable-length userdata is not supported. */
1089 VLOG_WARN("%s: datapath does not support variable-length userdata "
1090 "feature (needs Linux 3.10+ or kernel module from OVS "
1091 "1..11+). The NXAST_SAMPLE action will be ignored.",
1092 dpif_name(backer
->dpif
));
1096 /* Something odd happened. We're not sure whether variable-length
1097 * userdata is supported. Default to "yes". */
1098 VLOG_WARN("%s: variable-length userdata feature probe failed (%s)",
1099 dpif_name(backer
->dpif
), ovs_strerror(error
));
1104 /* Tests the MPLS label stack depth supported by 'backer''s datapath.
1106 * Returns the number of elements in a struct flow's mpls_lse field
1107 * if the datapath supports at least that many entries in an
1109 * Otherwise returns the number of MPLS push actions supported by
1112 check_max_mpls_depth(struct dpif_backer
*backer
)
1117 for (n
= 0; n
< FLOW_MAX_MPLS_LABELS
; n
++) {
1118 struct odputil_keybuf keybuf
;
1122 memset(&flow
, 0, sizeof flow
);
1123 flow
.dl_type
= htons(ETH_TYPE_MPLS
);
1124 flow_set_mpls_bos(&flow
, n
, 1);
1126 ofpbuf_use_stack(&key
, &keybuf
, sizeof keybuf
);
1127 odp_flow_key_from_flow(&key
, &flow
, NULL
, 0, false);
1129 error
= dpif_flow_put(backer
->dpif
, DPIF_FP_CREATE
| DPIF_FP_PROBE
,
1130 ofpbuf_data(&key
), ofpbuf_size(&key
), NULL
, 0,
1132 if (error
&& error
!= EEXIST
) {
1133 if (error
!= EINVAL
) {
1134 VLOG_WARN("%s: MPLS stack length feature probe failed (%s)",
1135 dpif_name(backer
->dpif
), ovs_strerror(error
));
1140 error
= dpif_flow_del(backer
->dpif
, ofpbuf_data(&key
),
1141 ofpbuf_size(&key
), NULL
);
1143 VLOG_WARN("%s: failed to delete MPLS feature probe flow",
1144 dpif_name(backer
->dpif
));
1148 VLOG_INFO("%s: MPLS label stack length probed as %d",
1149 dpif_name(backer
->dpif
), n
);
1153 /* Tests whether 'backer''s datapath supports masked data in
1154 * OVS_ACTION_ATTR_SET actions. We need to disable some features on older
1155 * datapaths that don't support this feature. */
1157 check_masked_set_action(struct dpif_backer
*backer
)
1159 struct eth_header
*eth
;
1160 struct ofpbuf actions
;
1161 struct dpif_execute execute
;
1162 struct ofpbuf packet
;
1164 struct ovs_key_ethernet key
, mask
;
1166 /* Compose a set action that will cause an EINVAL error on older
1167 * datapaths that don't support masked set actions.
1168 * Avoid using a full mask, as it could be translated to a non-masked
1169 * set action instead. */
1170 ofpbuf_init(&actions
, 64);
1171 memset(&key
, 0x53, sizeof key
);
1172 memset(&mask
, 0x7f, sizeof mask
);
1173 commit_masked_set_action(&actions
, OVS_KEY_ATTR_ETHERNET
, &key
, &mask
,
1176 /* Compose a dummy ethernet packet. */
1177 ofpbuf_init(&packet
, ETH_HEADER_LEN
);
1178 eth
= ofpbuf_put_zeros(&packet
, ETH_HEADER_LEN
);
1179 eth
->eth_type
= htons(0x1234);
1181 /* Execute the actions. On older datapaths this fails with EINVAL, on
1182 * newer datapaths it succeeds. */
1183 execute
.actions
= ofpbuf_data(&actions
);
1184 execute
.actions_len
= ofpbuf_size(&actions
);
1185 execute
.packet
= &packet
;
1186 execute
.md
= PKT_METADATA_INITIALIZER(0);
1187 execute
.needs_help
= false;
1188 execute
.probe
= true;
1190 error
= dpif_execute(backer
->dpif
, &execute
);
1192 ofpbuf_uninit(&packet
);
1193 ofpbuf_uninit(&actions
);
1196 /* Masked set action is not supported. */
1197 VLOG_INFO("%s: datapath does not support masked set action feature.",
1198 dpif_name(backer
->dpif
));
1204 construct(struct ofproto
*ofproto_
)
1206 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1207 struct shash_node
*node
, *next
;
1210 error
= open_dpif_backer(ofproto
->up
.type
, &ofproto
->backer
);
1215 ofproto
->netflow
= NULL
;
1216 ofproto
->sflow
= NULL
;
1217 ofproto
->ipfix
= NULL
;
1218 ofproto
->stp
= NULL
;
1219 ofproto
->rstp
= NULL
;
1220 ofproto
->dump_seq
= 0;
1221 hmap_init(&ofproto
->bundles
);
1222 ofproto
->ml
= mac_learning_create(MAC_ENTRY_DEFAULT_IDLE_TIME
);
1224 ofproto
->mbridge
= mbridge_create();
1225 ofproto
->has_bonded_bundles
= false;
1226 ofproto
->lacp_enabled
= false;
1227 ovs_mutex_init_adaptive(&ofproto
->stats_mutex
);
1228 ovs_mutex_init(&ofproto
->vsp_mutex
);
1230 guarded_list_init(&ofproto
->pins
);
1232 ofproto_dpif_unixctl_init();
1233 ovs_router_unixctl_register();
1235 hmap_init(&ofproto
->vlandev_map
);
1236 hmap_init(&ofproto
->realdev_vid_map
);
1238 sset_init(&ofproto
->ports
);
1239 sset_init(&ofproto
->ghost_ports
);
1240 sset_init(&ofproto
->port_poll_set
);
1241 ofproto
->port_poll_errno
= 0;
1242 ofproto
->change_seq
= 0;
1243 ofproto
->pins_seq
= seq_create();
1244 ofproto
->pins_seqno
= seq_read(ofproto
->pins_seq
);
1247 SHASH_FOR_EACH_SAFE (node
, next
, &init_ofp_ports
) {
1248 struct iface_hint
*iface_hint
= node
->data
;
1250 if (!strcmp(iface_hint
->br_name
, ofproto
->up
.name
)) {
1251 /* Check if the datapath already has this port. */
1252 if (dpif_port_exists(ofproto
->backer
->dpif
, node
->name
)) {
1253 sset_add(&ofproto
->ports
, node
->name
);
1256 free(iface_hint
->br_name
);
1257 free(iface_hint
->br_type
);
1259 shash_delete(&init_ofp_ports
, node
);
1263 hmap_insert(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
,
1264 hash_string(ofproto
->up
.name
, 0));
1265 memset(&ofproto
->stats
, 0, sizeof ofproto
->stats
);
1267 ofproto_init_tables(ofproto_
, N_TABLES
);
1268 error
= add_internal_flows(ofproto
);
1270 ofproto
->up
.tables
[TBL_INTERNAL
].flags
= OFTABLE_HIDDEN
| OFTABLE_READONLY
;
1276 add_internal_miss_flow(struct ofproto_dpif
*ofproto
, int id
,
1277 const struct ofpbuf
*ofpacts
, struct rule_dpif
**rulep
)
1283 match_init_catchall(&match
);
1284 match_set_reg(&match
, 0, id
);
1286 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 0, 0, ofpacts
,
1288 *rulep
= error
? NULL
: rule_dpif_cast(rule
);
1294 add_internal_flows(struct ofproto_dpif
*ofproto
)
1296 struct ofpact_controller
*controller
;
1297 uint64_t ofpacts_stub
[128 / 8];
1298 struct ofpbuf ofpacts
;
1299 struct rule
*unused_rulep OVS_UNUSED
;
1300 struct ofpact_resubmit
*resubmit
;
1305 ofpbuf_use_stack(&ofpacts
, ofpacts_stub
, sizeof ofpacts_stub
);
1308 controller
= ofpact_put_CONTROLLER(&ofpacts
);
1309 controller
->max_len
= UINT16_MAX
;
1310 controller
->controller_id
= 0;
1311 controller
->reason
= OFPR_NO_MATCH
;
1312 ofpact_pad(&ofpacts
);
1314 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1315 &ofproto
->miss_rule
);
1320 ofpbuf_clear(&ofpacts
);
1321 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1322 &ofproto
->no_packet_in_rule
);
1327 error
= add_internal_miss_flow(ofproto
, id
++, &ofpacts
,
1328 &ofproto
->drop_frags_rule
);
1333 /* Continue non-recirculation rule lookups from table 0.
1335 * (priority=2), recirc=0, actions=resubmit(, 0)
1337 resubmit
= ofpact_put_RESUBMIT(&ofpacts
);
1338 resubmit
->in_port
= OFPP_IN_PORT
;
1339 resubmit
->table_id
= 0;
1341 match_init_catchall(&match
);
1342 match_set_recirc_id(&match
, 0);
1344 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 2, 0, &ofpacts
,
1350 /* Drop any run away recirc rule lookups. Recirc_id has to be
1351 * non-zero when reaching this rule.
1353 * (priority=1), *, actions=drop
1355 ofpbuf_clear(&ofpacts
);
1356 match_init_catchall(&match
);
1357 error
= ofproto_dpif_add_internal_flow(ofproto
, &match
, 1, 0, &ofpacts
,
1364 destruct(struct ofproto
*ofproto_
)
1366 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1367 struct ofproto_packet_in
*pin
, *next_pin
;
1368 struct rule_dpif
*rule
;
1369 struct oftable
*table
;
1372 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1374 xlate_remove_ofproto(ofproto
);
1377 /* Ensure that the upcall processing threads have no remaining references
1378 * to the ofproto or anything in it. */
1379 udpif_synchronize(ofproto
->backer
->udpif
);
1381 hmap_remove(&all_ofproto_dpifs
, &ofproto
->all_ofproto_dpifs_node
);
1383 OFPROTO_FOR_EACH_TABLE (table
, &ofproto
->up
) {
1384 CLS_FOR_EACH_SAFE (rule
, up
.cr
, &table
->cls
) {
1385 ofproto_rule_delete(&ofproto
->up
, &rule
->up
);
1389 guarded_list_pop_all(&ofproto
->pins
, &pins
);
1390 LIST_FOR_EACH_SAFE (pin
, next_pin
, list_node
, &pins
) {
1391 list_remove(&pin
->list_node
);
1392 free(CONST_CAST(void *, pin
->up
.packet
));
1395 guarded_list_destroy(&ofproto
->pins
);
1397 mbridge_unref(ofproto
->mbridge
);
1399 netflow_unref(ofproto
->netflow
);
1400 dpif_sflow_unref(ofproto
->sflow
);
1401 dpif_ipfix_unref(ofproto
->ipfix
);
1402 hmap_destroy(&ofproto
->bundles
);
1403 mac_learning_unref(ofproto
->ml
);
1404 mcast_snooping_unref(ofproto
->ms
);
1406 hmap_destroy(&ofproto
->vlandev_map
);
1407 hmap_destroy(&ofproto
->realdev_vid_map
);
1409 sset_destroy(&ofproto
->ports
);
1410 sset_destroy(&ofproto
->ghost_ports
);
1411 sset_destroy(&ofproto
->port_poll_set
);
1413 ovs_mutex_destroy(&ofproto
->stats_mutex
);
1414 ovs_mutex_destroy(&ofproto
->vsp_mutex
);
1416 seq_destroy(ofproto
->pins_seq
);
1418 close_dpif_backer(ofproto
->backer
);
1422 run(struct ofproto
*ofproto_
)
1424 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1425 uint64_t new_seq
, new_dump_seq
;
1427 if (mbridge_need_revalidate(ofproto
->mbridge
)) {
1428 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1429 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1430 mac_learning_flush(ofproto
->ml
);
1431 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1432 mcast_snooping_mdb_flush(ofproto
->ms
);
1435 /* Always updates the ofproto->pins_seqno to avoid frequent wakeup during
1436 * flow restore. Even though nothing is processed during flow restore,
1437 * all queued 'pins' will be handled immediately when flow restore
1439 ofproto
->pins_seqno
= seq_read(ofproto
->pins_seq
);
1441 /* Do not perform any periodic activity required by 'ofproto' while
1442 * waiting for flow restore to complete. */
1443 if (!ofproto_get_flow_restore_wait()) {
1444 struct ofproto_packet_in
*pin
, *next_pin
;
1447 guarded_list_pop_all(&ofproto
->pins
, &pins
);
1448 LIST_FOR_EACH_SAFE (pin
, next_pin
, list_node
, &pins
) {
1449 connmgr_send_packet_in(ofproto
->up
.connmgr
, pin
);
1450 list_remove(&pin
->list_node
);
1451 free(CONST_CAST(void *, pin
->up
.packet
));
1456 if (ofproto
->netflow
) {
1457 netflow_run(ofproto
->netflow
);
1459 if (ofproto
->sflow
) {
1460 dpif_sflow_run(ofproto
->sflow
);
1462 if (ofproto
->ipfix
) {
1463 dpif_ipfix_run(ofproto
->ipfix
);
1466 new_seq
= seq_read(connectivity_seq_get());
1467 if (ofproto
->change_seq
!= new_seq
) {
1468 struct ofport_dpif
*ofport
;
1470 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1474 ofproto
->change_seq
= new_seq
;
1476 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1477 struct ofbundle
*bundle
;
1479 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1486 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
1487 if (mac_learning_run(ofproto
->ml
)) {
1488 ofproto
->backer
->need_revalidate
= REV_MAC_LEARNING
;
1490 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1492 if (mcast_snooping_run(ofproto
->ms
)) {
1493 ofproto
->backer
->need_revalidate
= REV_MCAST_SNOOPING
;
1496 new_dump_seq
= seq_read(udpif_dump_seq(ofproto
->backer
->udpif
));
1497 if (ofproto
->dump_seq
!= new_dump_seq
) {
1498 struct rule
*rule
, *next_rule
;
1500 /* We know stats are relatively fresh, so now is a good time to do some
1502 ofproto
->dump_seq
= new_dump_seq
;
1504 /* Expire OpenFlow flows whose idle_timeout or hard_timeout
1506 ovs_mutex_lock(&ofproto_mutex
);
1507 LIST_FOR_EACH_SAFE (rule
, next_rule
, expirable
,
1508 &ofproto
->up
.expirable
) {
1509 rule_expire(rule_dpif_cast(rule
));
1511 ovs_mutex_unlock(&ofproto_mutex
);
1513 /* All outstanding data in existing flows has been accounted, so it's a
1514 * good time to do bond rebalancing. */
1515 if (ofproto
->has_bonded_bundles
) {
1516 struct ofbundle
*bundle
;
1518 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1520 bond_rebalance(bundle
->bond
);
1530 wait(struct ofproto
*ofproto_
)
1532 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1534 if (ofproto_get_flow_restore_wait()) {
1538 if (ofproto
->sflow
) {
1539 dpif_sflow_wait(ofproto
->sflow
);
1541 if (ofproto
->ipfix
) {
1542 dpif_ipfix_wait(ofproto
->ipfix
);
1544 if (ofproto
->lacp_enabled
|| ofproto
->has_bonded_bundles
) {
1545 struct ofbundle
*bundle
;
1547 HMAP_FOR_EACH (bundle
, hmap_node
, &ofproto
->bundles
) {
1548 bundle_wait(bundle
);
1551 if (ofproto
->netflow
) {
1552 netflow_wait(ofproto
->netflow
);
1554 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
1555 mac_learning_wait(ofproto
->ml
);
1556 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
1557 mcast_snooping_wait(ofproto
->ms
);
1559 if (ofproto
->backer
->need_revalidate
) {
1560 /* Shouldn't happen, but if it does just go around again. */
1561 VLOG_DBG_RL(&rl
, "need revalidate in ofproto_wait_cb()");
1562 poll_immediate_wake();
1565 seq_wait(udpif_dump_seq(ofproto
->backer
->udpif
), ofproto
->dump_seq
);
1566 seq_wait(ofproto
->pins_seq
, ofproto
->pins_seqno
);
1570 type_get_memory_usage(const char *type
, struct simap
*usage
)
1572 struct dpif_backer
*backer
;
1574 backer
= shash_find_data(&all_dpif_backers
, type
);
1576 udpif_get_memory_usage(backer
->udpif
, usage
);
1581 flush(struct ofproto
*ofproto_
)
1583 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1584 struct dpif_backer
*backer
= ofproto
->backer
;
1587 udpif_flush(backer
->udpif
);
1592 query_tables(struct ofproto
*ofproto
,
1593 struct ofputil_table_features
*features
,
1594 struct ofputil_table_stats
*stats
)
1596 strcpy(features
->name
, "classifier");
1601 for (i
= 0; i
< ofproto
->n_tables
; i
++) {
1602 unsigned long missed
, matched
;
1604 atomic_read_relaxed(&ofproto
->tables
[i
].n_matched
, &matched
);
1605 atomic_read_relaxed(&ofproto
->tables
[i
].n_missed
, &missed
);
1607 stats
[i
].matched_count
= matched
;
1608 stats
[i
].lookup_count
= matched
+ missed
;
1613 static struct ofport
*
1616 struct ofport_dpif
*port
= xmalloc(sizeof *port
);
1621 port_dealloc(struct ofport
*port_
)
1623 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1628 port_construct(struct ofport
*port_
)
1630 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1631 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1632 const struct netdev
*netdev
= port
->up
.netdev
;
1633 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1634 struct dpif_port dpif_port
;
1637 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1638 port
->bundle
= NULL
;
1641 port
->may_enable
= false;
1642 port
->stp_port
= NULL
;
1643 port
->stp_state
= STP_DISABLED
;
1644 port
->rstp_port
= NULL
;
1645 port
->rstp_state
= RSTP_DISABLED
;
1646 port
->is_tunnel
= false;
1650 port
->realdev_ofp_port
= 0;
1651 port
->vlandev_vid
= 0;
1652 port
->carrier_seq
= netdev_get_carrier_resets(netdev
);
1653 port
->is_layer3
= netdev_vport_is_layer3(netdev
);
1655 if (netdev_vport_is_patch(netdev
)) {
1656 /* By bailing out here, we don't submit the port to the sFlow module
1657 * to be considered for counter polling export. This is correct
1658 * because the patch port represents an interface that sFlow considers
1659 * to be "internal" to the switch as a whole, and therefore not an
1660 * candidate for counter polling. */
1661 port
->odp_port
= ODPP_NONE
;
1662 ofport_update_peer(port
);
1666 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
1667 netdev_vport_get_dpif_port(netdev
, namebuf
,
1674 port
->odp_port
= dpif_port
.port_no
;
1676 if (netdev_get_tunnel_config(netdev
)) {
1677 tnl_port_add(port
, port
->up
.netdev
, port
->odp_port
);
1678 port
->is_tunnel
= true;
1679 if (ofproto
->ipfix
) {
1680 dpif_ipfix_add_tunnel_port(ofproto
->ipfix
, port_
, port
->odp_port
);
1683 /* Sanity-check that a mapping doesn't already exist. This
1684 * shouldn't happen for non-tunnel ports. */
1685 if (odp_port_to_ofp_port(ofproto
, port
->odp_port
) != OFPP_NONE
) {
1686 VLOG_ERR("port %s already has an OpenFlow port number",
1688 dpif_port_destroy(&dpif_port
);
1692 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1693 hmap_insert(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
,
1694 hash_odp_port(port
->odp_port
));
1695 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1697 dpif_port_destroy(&dpif_port
);
1699 if (ofproto
->sflow
) {
1700 dpif_sflow_add_port(ofproto
->sflow
, port_
, port
->odp_port
);
1707 port_destruct(struct ofport
*port_
)
1709 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1710 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1711 const char *devname
= netdev_get_name(port
->up
.netdev
);
1712 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
1713 const char *dp_port_name
;
1715 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1717 xlate_ofport_remove(port
);
1720 dp_port_name
= netdev_vport_get_dpif_port(port
->up
.netdev
, namebuf
,
1722 if (dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
1723 /* The underlying device is still there, so delete it. This
1724 * happens when the ofproto is being destroyed, since the caller
1725 * assumes that removal of attached ports will happen as part of
1727 if (!port
->is_tunnel
) {
1728 dpif_port_del(ofproto
->backer
->dpif
, port
->odp_port
);
1733 port
->peer
->peer
= NULL
;
1737 if (port
->odp_port
!= ODPP_NONE
&& !port
->is_tunnel
) {
1738 ovs_rwlock_wrlock(&ofproto
->backer
->odp_to_ofport_lock
);
1739 hmap_remove(&ofproto
->backer
->odp_to_ofport_map
, &port
->odp_port_node
);
1740 ovs_rwlock_unlock(&ofproto
->backer
->odp_to_ofport_lock
);
1743 if (port
->is_tunnel
&& ofproto
->ipfix
) {
1744 dpif_ipfix_del_tunnel_port(ofproto
->ipfix
, port
->odp_port
);
1748 sset_find_and_delete(&ofproto
->ports
, devname
);
1749 sset_find_and_delete(&ofproto
->ghost_ports
, devname
);
1750 bundle_remove(port_
);
1751 set_cfm(port_
, NULL
);
1752 set_bfd(port_
, NULL
);
1753 if (port
->stp_port
) {
1754 stp_port_disable(port
->stp_port
);
1756 set_rstp_port(port_
, NULL
);
1757 if (ofproto
->sflow
) {
1758 dpif_sflow_del_port(ofproto
->sflow
, port
->odp_port
);
1765 port_modified(struct ofport
*port_
)
1767 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1769 if (port
->bundle
&& port
->bundle
->bond
) {
1770 bond_slave_set_netdev(port
->bundle
->bond
, port
, port
->up
.netdev
);
1774 cfm_set_netdev(port
->cfm
, port
->up
.netdev
);
1778 bfd_set_netdev(port
->bfd
, port
->up
.netdev
);
1781 ofproto_dpif_monitor_port_update(port
, port
->bfd
, port
->cfm
,
1782 port
->up
.pp
.hw_addr
);
1784 if (port
->is_tunnel
&& tnl_port_reconfigure(port
, port
->up
.netdev
,
1786 ofproto_dpif_cast(port
->up
.ofproto
)->backer
->need_revalidate
=
1790 ofport_update_peer(port
);
1794 port_reconfigured(struct ofport
*port_
, enum ofputil_port_config old_config
)
1796 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
1797 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
1798 enum ofputil_port_config changed
= old_config
^ port
->up
.pp
.config
;
1800 if (changed
& (OFPUTIL_PC_NO_RECV
| OFPUTIL_PC_NO_RECV_STP
|
1801 OFPUTIL_PC_NO_FWD
| OFPUTIL_PC_NO_FLOOD
|
1802 OFPUTIL_PC_NO_PACKET_IN
)) {
1803 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1805 if (changed
& OFPUTIL_PC_NO_FLOOD
&& port
->bundle
) {
1806 bundle_update(port
->bundle
);
1812 set_sflow(struct ofproto
*ofproto_
,
1813 const struct ofproto_sflow_options
*sflow_options
)
1815 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1816 struct dpif_sflow
*ds
= ofproto
->sflow
;
1818 if (sflow_options
) {
1820 struct ofport_dpif
*ofport
;
1822 ds
= ofproto
->sflow
= dpif_sflow_create();
1823 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1824 dpif_sflow_add_port(ds
, &ofport
->up
, ofport
->odp_port
);
1826 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1828 dpif_sflow_set_options(ds
, sflow_options
);
1831 dpif_sflow_unref(ds
);
1832 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1833 ofproto
->sflow
= NULL
;
1841 struct ofproto
*ofproto_
,
1842 const struct ofproto_ipfix_bridge_exporter_options
*bridge_exporter_options
,
1843 const struct ofproto_ipfix_flow_exporter_options
*flow_exporters_options
,
1844 size_t n_flow_exporters_options
)
1846 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
1847 struct dpif_ipfix
*di
= ofproto
->ipfix
;
1848 bool has_options
= bridge_exporter_options
|| flow_exporters_options
;
1849 bool new_di
= false;
1851 if (has_options
&& !di
) {
1852 di
= ofproto
->ipfix
= dpif_ipfix_create();
1857 /* Call set_options in any case to cleanly flush the flow
1858 * caches in the last exporters that are to be destroyed. */
1859 dpif_ipfix_set_options(
1860 di
, bridge_exporter_options
, flow_exporters_options
,
1861 n_flow_exporters_options
);
1863 /* Add tunnel ports only when a new ipfix created */
1864 if (new_di
== true) {
1865 struct ofport_dpif
*ofport
;
1866 HMAP_FOR_EACH (ofport
, up
.hmap_node
, &ofproto
->up
.ports
) {
1867 if (ofport
->is_tunnel
== true) {
1868 dpif_ipfix_add_tunnel_port(di
, &ofport
->up
, ofport
->odp_port
);
1874 dpif_ipfix_unref(di
);
1875 ofproto
->ipfix
= NULL
;
1883 set_cfm(struct ofport
*ofport_
, const struct cfm_settings
*s
)
1885 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1890 struct ofproto_dpif
*ofproto
;
1892 ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
1893 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1894 ofport
->cfm
= cfm_create(ofport
->up
.netdev
);
1897 if (cfm_configure(ofport
->cfm
, s
)) {
1904 cfm_unref(ofport
->cfm
);
1907 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
1908 ofport
->up
.pp
.hw_addr
);
1913 cfm_status_changed(struct ofport
*ofport_
)
1915 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1917 return ofport
->cfm
? cfm_check_status_change(ofport
->cfm
) : true;
1921 get_cfm_status(const struct ofport
*ofport_
,
1922 struct cfm_status
*status
)
1924 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1928 cfm_get_status(ofport
->cfm
, status
);
1937 set_bfd(struct ofport
*ofport_
, const struct smap
*cfg
)
1939 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
1940 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1944 ofport
->bfd
= bfd_configure(old
, netdev_get_name(ofport
->up
.netdev
),
1945 cfg
, ofport
->up
.netdev
);
1946 if (ofport
->bfd
!= old
) {
1947 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
1949 ofproto_dpif_monitor_port_update(ofport
, ofport
->bfd
, ofport
->cfm
,
1950 ofport
->up
.pp
.hw_addr
);
1955 bfd_status_changed(struct ofport
*ofport_
)
1957 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1959 return ofport
->bfd
? bfd_check_status_change(ofport
->bfd
) : true;
1963 get_bfd_status(struct ofport
*ofport_
, struct smap
*smap
)
1965 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
1969 bfd_get_status(ofport
->bfd
, smap
);
1977 /* Spanning Tree. */
1979 /* Called while rstp_mutex is held. */
1981 rstp_send_bpdu_cb(struct ofpbuf
*pkt
, void *ofport_
, void *ofproto_
)
1983 struct ofproto_dpif
*ofproto
= ofproto_
;
1984 struct ofport_dpif
*ofport
= ofport_
;
1985 struct eth_header
*eth
= ofpbuf_l2(pkt
);
1987 netdev_get_etheraddr(ofport
->up
.netdev
, eth
->eth_src
);
1988 if (eth_addr_is_zero(eth
->eth_src
)) {
1989 VLOG_WARN_RL(&rl
, "%s port %d: cannot send RSTP BPDU on a port which "
1990 "does not have a configured source MAC address.",
1991 ofproto
->up
.name
, ofp_to_u16(ofport
->up
.ofp_port
));
1993 ofproto_dpif_send_packet(ofport
, pkt
);
1999 send_bpdu_cb(struct ofpbuf
*pkt
, int port_num
, void *ofproto_
)
2001 struct ofproto_dpif
*ofproto
= ofproto_
;
2002 struct stp_port
*sp
= stp_get_port(ofproto
->stp
, port_num
);
2003 struct ofport_dpif
*ofport
;
2005 ofport
= stp_port_get_aux(sp
);
2007 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on unknown port %d",
2008 ofproto
->up
.name
, port_num
);
2010 struct eth_header
*eth
= ofpbuf_l2(pkt
);
2012 netdev_get_etheraddr(ofport
->up
.netdev
, eth
->eth_src
);
2013 if (eth_addr_is_zero(eth
->eth_src
)) {
2014 VLOG_WARN_RL(&rl
, "%s: cannot send BPDU on port %d "
2015 "with unknown MAC", ofproto
->up
.name
, port_num
);
2017 ofproto_dpif_send_packet(ofport
, pkt
);
2023 /* Configure RSTP on 'ofproto_' using the settings defined in 's'. */
2025 set_rstp(struct ofproto
*ofproto_
, const struct ofproto_rstp_settings
*s
)
2027 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2029 /* Only revalidate flows if the configuration changed. */
2030 if (!s
!= !ofproto
->rstp
) {
2031 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2035 if (!ofproto
->rstp
) {
2036 ofproto
->rstp
= rstp_create(ofproto_
->name
, s
->address
,
2037 rstp_send_bpdu_cb
, ofproto
);
2038 ofproto
->rstp_last_tick
= time_msec();
2040 rstp_set_bridge_address(ofproto
->rstp
, s
->address
);
2041 rstp_set_bridge_priority(ofproto
->rstp
, s
->priority
);
2042 rstp_set_bridge_ageing_time(ofproto
->rstp
, s
->ageing_time
);
2043 rstp_set_bridge_force_protocol_version(ofproto
->rstp
,
2044 s
->force_protocol_version
);
2045 rstp_set_bridge_max_age(ofproto
->rstp
, s
->bridge_max_age
);
2046 rstp_set_bridge_forward_delay(ofproto
->rstp
, s
->bridge_forward_delay
);
2047 rstp_set_bridge_transmit_hold_count(ofproto
->rstp
,
2048 s
->transmit_hold_count
);
2050 struct ofport
*ofport
;
2051 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2052 set_rstp_port(ofport
, NULL
);
2054 rstp_unref(ofproto
->rstp
);
2055 ofproto
->rstp
= NULL
;
2060 get_rstp_status(struct ofproto
*ofproto_
, struct ofproto_rstp_status
*s
)
2062 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2064 if (ofproto
->rstp
) {
2066 s
->root_id
= rstp_get_root_id(ofproto
->rstp
);
2067 s
->bridge_id
= rstp_get_bridge_id(ofproto
->rstp
);
2068 s
->designated_id
= rstp_get_designated_id(ofproto
->rstp
);
2069 s
->root_path_cost
= rstp_get_root_path_cost(ofproto
->rstp
);
2070 s
->designated_port_id
= rstp_get_designated_port_id(ofproto
->rstp
);
2071 s
->bridge_port_id
= rstp_get_bridge_port_id(ofproto
->rstp
);
2078 update_rstp_port_state(struct ofport_dpif
*ofport
)
2080 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2081 enum rstp_state state
;
2083 /* Figure out new state. */
2084 state
= ofport
->rstp_port
? rstp_port_get_state(ofport
->rstp_port
)
2088 if (ofport
->rstp_state
!= state
) {
2089 enum ofputil_port_state of_state
;
2092 VLOG_DBG("port %s: RSTP state changed from %s to %s",
2093 netdev_get_name(ofport
->up
.netdev
),
2094 rstp_state_name(ofport
->rstp_state
),
2095 rstp_state_name(state
));
2096 if (rstp_learn_in_state(ofport
->rstp_state
)
2097 != rstp_learn_in_state(state
)) {
2098 /* xxx Learning action flows should also be flushed. */
2099 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2100 mac_learning_flush(ofproto
->ml
);
2101 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2103 fwd_change
= rstp_forward_in_state(ofport
->rstp_state
)
2104 != rstp_forward_in_state(state
);
2106 ofproto
->backer
->need_revalidate
= REV_RSTP
;
2107 ofport
->rstp_state
= state
;
2109 if (fwd_change
&& ofport
->bundle
) {
2110 bundle_update(ofport
->bundle
);
2113 /* Update the RSTP state bits in the OpenFlow port description. */
2114 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2115 of_state
|= (state
== RSTP_LEARNING
? OFPUTIL_PS_STP_LEARN
2116 : state
== RSTP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2117 : state
== RSTP_DISCARDING
? OFPUTIL_PS_STP_LISTEN
2119 ofproto_port_set_state(&ofport
->up
, of_state
);
2124 rstp_run(struct ofproto_dpif
*ofproto
)
2126 if (ofproto
->rstp
) {
2127 long long int now
= time_msec();
2128 long long int elapsed
= now
- ofproto
->rstp_last_tick
;
2129 struct rstp_port
*rp
;
2130 struct ofport_dpif
*ofport
;
2132 /* Every second, decrease the values of the timers. */
2133 if (elapsed
>= 1000) {
2134 rstp_tick_timers(ofproto
->rstp
);
2135 ofproto
->rstp_last_tick
= now
;
2138 while ((ofport
= rstp_get_next_changed_port_aux(ofproto
->rstp
, &rp
))) {
2139 update_rstp_port_state(ofport
);
2141 /* FIXME: This check should be done on-event (i.e., when setting
2142 * p->fdb_flush) and not periodically.
2144 if (rstp_check_and_reset_fdb_flush(ofproto
->rstp
)) {
2145 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2146 /* FIXME: RSTP should be able to flush the entries pertaining to a
2147 * single port, not the whole table.
2149 mac_learning_flush(ofproto
->ml
);
2150 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2155 /* Configures STP on 'ofproto_' using the settings defined in 's'. */
2157 set_stp(struct ofproto
*ofproto_
, const struct ofproto_stp_settings
*s
)
2159 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2161 /* Only revalidate flows if the configuration changed. */
2162 if (!s
!= !ofproto
->stp
) {
2163 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2167 if (!ofproto
->stp
) {
2168 ofproto
->stp
= stp_create(ofproto_
->name
, s
->system_id
,
2169 send_bpdu_cb
, ofproto
);
2170 ofproto
->stp_last_tick
= time_msec();
2173 stp_set_bridge_id(ofproto
->stp
, s
->system_id
);
2174 stp_set_bridge_priority(ofproto
->stp
, s
->priority
);
2175 stp_set_hello_time(ofproto
->stp
, s
->hello_time
);
2176 stp_set_max_age(ofproto
->stp
, s
->max_age
);
2177 stp_set_forward_delay(ofproto
->stp
, s
->fwd_delay
);
2179 struct ofport
*ofport
;
2181 HMAP_FOR_EACH (ofport
, hmap_node
, &ofproto
->up
.ports
) {
2182 set_stp_port(ofport
, NULL
);
2185 stp_unref(ofproto
->stp
);
2186 ofproto
->stp
= NULL
;
2193 get_stp_status(struct ofproto
*ofproto_
, struct ofproto_stp_status
*s
)
2195 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2199 s
->bridge_id
= stp_get_bridge_id(ofproto
->stp
);
2200 s
->designated_root
= stp_get_designated_root(ofproto
->stp
);
2201 s
->root_path_cost
= stp_get_root_path_cost(ofproto
->stp
);
2210 update_stp_port_state(struct ofport_dpif
*ofport
)
2212 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2213 enum stp_state state
;
2215 /* Figure out new state. */
2216 state
= ofport
->stp_port
? stp_port_get_state(ofport
->stp_port
)
2220 if (ofport
->stp_state
!= state
) {
2221 enum ofputil_port_state of_state
;
2224 VLOG_DBG("port %s: STP state changed from %s to %s",
2225 netdev_get_name(ofport
->up
.netdev
),
2226 stp_state_name(ofport
->stp_state
),
2227 stp_state_name(state
));
2228 if (stp_learn_in_state(ofport
->stp_state
)
2229 != stp_learn_in_state(state
)) {
2230 /* xxx Learning action flows should also be flushed. */
2231 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2232 mac_learning_flush(ofproto
->ml
);
2233 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2234 mcast_snooping_mdb_flush(ofproto
->ms
);
2236 fwd_change
= stp_forward_in_state(ofport
->stp_state
)
2237 != stp_forward_in_state(state
);
2239 ofproto
->backer
->need_revalidate
= REV_STP
;
2240 ofport
->stp_state
= state
;
2241 ofport
->stp_state_entered
= time_msec();
2243 if (fwd_change
&& ofport
->bundle
) {
2244 bundle_update(ofport
->bundle
);
2247 /* Update the STP state bits in the OpenFlow port description. */
2248 of_state
= ofport
->up
.pp
.state
& ~OFPUTIL_PS_STP_MASK
;
2249 of_state
|= (state
== STP_LISTENING
? OFPUTIL_PS_STP_LISTEN
2250 : state
== STP_LEARNING
? OFPUTIL_PS_STP_LEARN
2251 : state
== STP_FORWARDING
? OFPUTIL_PS_STP_FORWARD
2252 : state
== STP_BLOCKING
? OFPUTIL_PS_STP_BLOCK
2254 ofproto_port_set_state(&ofport
->up
, of_state
);
2258 /* Configures STP on 'ofport_' using the settings defined in 's'. The
2259 * caller is responsible for assigning STP port numbers and ensuring
2260 * there are no duplicates. */
2262 set_stp_port(struct ofport
*ofport_
,
2263 const struct ofproto_port_stp_settings
*s
)
2265 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2266 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2267 struct stp_port
*sp
= ofport
->stp_port
;
2269 if (!s
|| !s
->enable
) {
2271 ofport
->stp_port
= NULL
;
2272 stp_port_disable(sp
);
2273 update_stp_port_state(ofport
);
2276 } else if (sp
&& stp_port_no(sp
) != s
->port_num
2277 && ofport
== stp_port_get_aux(sp
)) {
2278 /* The port-id changed, so disable the old one if it's not
2279 * already in use by another port. */
2280 stp_port_disable(sp
);
2283 sp
= ofport
->stp_port
= stp_get_port(ofproto
->stp
, s
->port_num
);
2285 /* Set name before enabling the port so that debugging messages can print
2287 stp_port_set_name(sp
, netdev_get_name(ofport
->up
.netdev
));
2288 stp_port_enable(sp
);
2290 stp_port_set_aux(sp
, ofport
);
2291 stp_port_set_priority(sp
, s
->priority
);
2292 stp_port_set_path_cost(sp
, s
->path_cost
);
2294 update_stp_port_state(ofport
);
2300 get_stp_port_status(struct ofport
*ofport_
,
2301 struct ofproto_port_stp_status
*s
)
2303 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2304 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2305 struct stp_port
*sp
= ofport
->stp_port
;
2307 if (!ofproto
->stp
|| !sp
) {
2313 s
->port_id
= stp_port_get_id(sp
);
2314 s
->state
= stp_port_get_state(sp
);
2315 s
->sec_in_state
= (time_msec() - ofport
->stp_state_entered
) / 1000;
2316 s
->role
= stp_port_get_role(sp
);
2322 get_stp_port_stats(struct ofport
*ofport_
,
2323 struct ofproto_port_stp_stats
*s
)
2325 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2326 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2327 struct stp_port
*sp
= ofport
->stp_port
;
2329 if (!ofproto
->stp
|| !sp
) {
2335 stp_port_get_counts(sp
, &s
->tx_count
, &s
->rx_count
, &s
->error_count
);
2341 stp_run(struct ofproto_dpif
*ofproto
)
2344 long long int now
= time_msec();
2345 long long int elapsed
= now
- ofproto
->stp_last_tick
;
2346 struct stp_port
*sp
;
2349 stp_tick(ofproto
->stp
, MIN(INT_MAX
, elapsed
));
2350 ofproto
->stp_last_tick
= now
;
2352 while (stp_get_changed_port(ofproto
->stp
, &sp
)) {
2353 struct ofport_dpif
*ofport
= stp_port_get_aux(sp
);
2356 update_stp_port_state(ofport
);
2360 if (stp_check_and_reset_fdb_flush(ofproto
->stp
)) {
2361 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2362 mac_learning_flush(ofproto
->ml
);
2363 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2364 mcast_snooping_mdb_flush(ofproto
->ms
);
2370 stp_wait(struct ofproto_dpif
*ofproto
)
2373 poll_timer_wait(1000);
2377 /* Configures RSTP on 'ofport_' using the settings defined in 's'. The
2378 * caller is responsible for assigning RSTP port numbers and ensuring
2379 * there are no duplicates. */
2381 set_rstp_port(struct ofport
*ofport_
,
2382 const struct ofproto_port_rstp_settings
*s
)
2384 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2385 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2386 struct rstp_port
*rp
= ofport
->rstp_port
;
2388 if (!s
|| !s
->enable
) {
2390 rstp_port_unref(rp
);
2391 ofport
->rstp_port
= NULL
;
2392 update_rstp_port_state(ofport
);
2397 /* Check if need to add a new port. */
2399 rp
= ofport
->rstp_port
= rstp_add_port(ofproto
->rstp
);
2402 rstp_port_set(rp
, s
->port_num
, s
->priority
, s
->path_cost
,
2403 s
->admin_edge_port
, s
->auto_edge
, s
->mcheck
, ofport
);
2404 update_rstp_port_state(ofport
);
2405 /* Synchronize operational status. */
2406 rstp_port_set_mac_operational(rp
, ofport
->may_enable
);
2410 get_rstp_port_status(struct ofport
*ofport_
,
2411 struct ofproto_port_rstp_status
*s
)
2413 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2414 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2415 struct rstp_port
*rp
= ofport
->rstp_port
;
2417 if (!ofproto
->rstp
|| !rp
) {
2423 rstp_port_get_status(rp
, &s
->port_id
, &s
->state
, &s
->role
, &s
->tx_count
,
2424 &s
->rx_count
, &s
->error_count
, &s
->uptime
);
2429 set_queues(struct ofport
*ofport_
, const struct ofproto_port_queue
*qdscp
,
2432 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
2433 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
2435 if (ofport
->n_qdscp
!= n_qdscp
2436 || (n_qdscp
&& memcmp(ofport
->qdscp
, qdscp
,
2437 n_qdscp
* sizeof *qdscp
))) {
2438 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2439 free(ofport
->qdscp
);
2440 ofport
->qdscp
= n_qdscp
2441 ? xmemdup(qdscp
, n_qdscp
* sizeof *qdscp
)
2443 ofport
->n_qdscp
= n_qdscp
;
2451 /* Expires all MAC learning entries associated with 'bundle' and forces its
2452 * ofproto to revalidate every flow.
2454 * Normally MAC learning entries are removed only from the ofproto associated
2455 * with 'bundle', but if 'all_ofprotos' is true, then the MAC learning entries
2456 * are removed from every ofproto. When patch ports and SLB bonds are in use
2457 * and a VM migration happens and the gratuitous ARPs are somehow lost, this
2458 * avoids a MAC_ENTRY_IDLE_TIME delay before the migrated VM can communicate
2459 * with the host from which it migrated. */
2461 bundle_flush_macs(struct ofbundle
*bundle
, bool all_ofprotos
)
2463 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2464 struct mac_learning
*ml
= ofproto
->ml
;
2465 struct mac_entry
*mac
, *next_mac
;
2467 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2468 ovs_rwlock_wrlock(&ml
->rwlock
);
2469 LIST_FOR_EACH_SAFE (mac
, next_mac
, lru_node
, &ml
->lrus
) {
2470 if (mac
->port
.p
== bundle
) {
2472 struct ofproto_dpif
*o
;
2474 HMAP_FOR_EACH (o
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
2476 struct mac_entry
*e
;
2478 ovs_rwlock_wrlock(&o
->ml
->rwlock
);
2479 e
= mac_learning_lookup(o
->ml
, mac
->mac
, mac
->vlan
);
2481 mac_learning_expire(o
->ml
, e
);
2483 ovs_rwlock_unlock(&o
->ml
->rwlock
);
2488 mac_learning_expire(ml
, mac
);
2491 ovs_rwlock_unlock(&ml
->rwlock
);
2494 static struct ofbundle
*
2495 bundle_lookup(const struct ofproto_dpif
*ofproto
, void *aux
)
2497 struct ofbundle
*bundle
;
2499 HMAP_FOR_EACH_IN_BUCKET (bundle
, hmap_node
, hash_pointer(aux
, 0),
2500 &ofproto
->bundles
) {
2501 if (bundle
->aux
== aux
) {
2509 bundle_update(struct ofbundle
*bundle
)
2511 struct ofport_dpif
*port
;
2513 bundle
->floodable
= true;
2514 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2515 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2517 || !stp_forward_in_state(port
->stp_state
)) {
2518 bundle
->floodable
= false;
2525 bundle_del_port(struct ofport_dpif
*port
)
2527 struct ofbundle
*bundle
= port
->bundle
;
2529 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2531 list_remove(&port
->bundle_node
);
2532 port
->bundle
= NULL
;
2535 lacp_slave_unregister(bundle
->lacp
, port
);
2538 bond_slave_unregister(bundle
->bond
, port
);
2541 bundle_update(bundle
);
2545 bundle_add_port(struct ofbundle
*bundle
, ofp_port_t ofp_port
,
2546 struct lacp_slave_settings
*lacp
)
2548 struct ofport_dpif
*port
;
2550 port
= get_ofp_port(bundle
->ofproto
, ofp_port
);
2555 if (port
->bundle
!= bundle
) {
2556 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2558 bundle_remove(&port
->up
);
2561 port
->bundle
= bundle
;
2562 list_push_back(&bundle
->ports
, &port
->bundle_node
);
2563 if (port
->up
.pp
.config
& OFPUTIL_PC_NO_FLOOD
2565 || !stp_forward_in_state(port
->stp_state
)) {
2566 bundle
->floodable
= false;
2570 bundle
->ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2571 lacp_slave_register(bundle
->lacp
, port
, lacp
);
2578 bundle_destroy(struct ofbundle
*bundle
)
2580 struct ofproto_dpif
*ofproto
;
2581 struct ofport_dpif
*port
, *next_port
;
2587 ofproto
= bundle
->ofproto
;
2588 mbridge_unregister_bundle(ofproto
->mbridge
, bundle
->aux
);
2591 xlate_bundle_remove(bundle
);
2594 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2595 bundle_del_port(port
);
2598 bundle_flush_macs(bundle
, true);
2599 hmap_remove(&ofproto
->bundles
, &bundle
->hmap_node
);
2601 free(bundle
->trunks
);
2602 lacp_unref(bundle
->lacp
);
2603 bond_unref(bundle
->bond
);
2608 bundle_set(struct ofproto
*ofproto_
, void *aux
,
2609 const struct ofproto_bundle_settings
*s
)
2611 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2612 bool need_flush
= false;
2613 struct ofport_dpif
*port
;
2614 struct ofbundle
*bundle
;
2615 unsigned long *trunks
;
2621 bundle_destroy(bundle_lookup(ofproto
, aux
));
2625 ovs_assert(s
->n_slaves
== 1 || s
->bond
!= NULL
);
2626 ovs_assert((s
->lacp
!= NULL
) == (s
->lacp_slaves
!= NULL
));
2628 bundle
= bundle_lookup(ofproto
, aux
);
2630 bundle
= xmalloc(sizeof *bundle
);
2632 bundle
->ofproto
= ofproto
;
2633 hmap_insert(&ofproto
->bundles
, &bundle
->hmap_node
,
2634 hash_pointer(aux
, 0));
2636 bundle
->name
= NULL
;
2638 list_init(&bundle
->ports
);
2639 bundle
->vlan_mode
= PORT_VLAN_TRUNK
;
2641 bundle
->trunks
= NULL
;
2642 bundle
->use_priority_tags
= s
->use_priority_tags
;
2643 bundle
->lacp
= NULL
;
2644 bundle
->bond
= NULL
;
2646 bundle
->floodable
= true;
2647 mbridge_register_bundle(ofproto
->mbridge
, bundle
);
2650 if (!bundle
->name
|| strcmp(s
->name
, bundle
->name
)) {
2652 bundle
->name
= xstrdup(s
->name
);
2657 ofproto
->lacp_enabled
= true;
2658 if (!bundle
->lacp
) {
2659 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2660 bundle
->lacp
= lacp_create();
2662 lacp_configure(bundle
->lacp
, s
->lacp
);
2664 lacp_unref(bundle
->lacp
);
2665 bundle
->lacp
= NULL
;
2668 /* Update set of ports. */
2670 for (i
= 0; i
< s
->n_slaves
; i
++) {
2671 if (!bundle_add_port(bundle
, s
->slaves
[i
],
2672 s
->lacp
? &s
->lacp_slaves
[i
] : NULL
)) {
2676 if (!ok
|| list_size(&bundle
->ports
) != s
->n_slaves
) {
2677 struct ofport_dpif
*next_port
;
2679 LIST_FOR_EACH_SAFE (port
, next_port
, bundle_node
, &bundle
->ports
) {
2680 for (i
= 0; i
< s
->n_slaves
; i
++) {
2681 if (s
->slaves
[i
] == port
->up
.ofp_port
) {
2686 bundle_del_port(port
);
2690 ovs_assert(list_size(&bundle
->ports
) <= s
->n_slaves
);
2692 if (list_is_empty(&bundle
->ports
)) {
2693 bundle_destroy(bundle
);
2697 /* Set VLAN tagging mode */
2698 if (s
->vlan_mode
!= bundle
->vlan_mode
2699 || s
->use_priority_tags
!= bundle
->use_priority_tags
) {
2700 bundle
->vlan_mode
= s
->vlan_mode
;
2701 bundle
->use_priority_tags
= s
->use_priority_tags
;
2706 vlan
= (s
->vlan_mode
== PORT_VLAN_TRUNK
? -1
2707 : s
->vlan
>= 0 && s
->vlan
<= 4095 ? s
->vlan
2709 if (vlan
!= bundle
->vlan
) {
2710 bundle
->vlan
= vlan
;
2714 /* Get trunked VLANs. */
2715 switch (s
->vlan_mode
) {
2716 case PORT_VLAN_ACCESS
:
2720 case PORT_VLAN_TRUNK
:
2721 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2724 case PORT_VLAN_NATIVE_UNTAGGED
:
2725 case PORT_VLAN_NATIVE_TAGGED
:
2726 if (vlan
!= 0 && (!s
->trunks
2727 || !bitmap_is_set(s
->trunks
, vlan
)
2728 || bitmap_is_set(s
->trunks
, 0))) {
2729 /* Force trunking the native VLAN and prohibit trunking VLAN 0. */
2731 trunks
= bitmap_clone(s
->trunks
, 4096);
2733 trunks
= bitmap_allocate1(4096);
2735 bitmap_set1(trunks
, vlan
);
2736 bitmap_set0(trunks
, 0);
2738 trunks
= CONST_CAST(unsigned long *, s
->trunks
);
2745 if (!vlan_bitmap_equal(trunks
, bundle
->trunks
)) {
2746 free(bundle
->trunks
);
2747 if (trunks
== s
->trunks
) {
2748 bundle
->trunks
= vlan_bitmap_clone(trunks
);
2750 bundle
->trunks
= trunks
;
2755 if (trunks
!= s
->trunks
) {
2760 if (!list_is_short(&bundle
->ports
)) {
2761 bundle
->ofproto
->has_bonded_bundles
= true;
2763 if (bond_reconfigure(bundle
->bond
, s
->bond
)) {
2764 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2767 bundle
->bond
= bond_create(s
->bond
, ofproto
);
2768 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2771 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2772 bond_slave_register(bundle
->bond
, port
,
2773 port
->up
.ofp_port
, port
->up
.netdev
);
2776 bond_unref(bundle
->bond
);
2777 bundle
->bond
= NULL
;
2780 /* If we changed something that would affect MAC learning, un-learn
2781 * everything on this port and force flow revalidation. */
2783 bundle_flush_macs(bundle
, false);
2790 bundle_remove(struct ofport
*port_
)
2792 struct ofport_dpif
*port
= ofport_dpif_cast(port_
);
2793 struct ofbundle
*bundle
= port
->bundle
;
2796 bundle_del_port(port
);
2797 if (list_is_empty(&bundle
->ports
)) {
2798 bundle_destroy(bundle
);
2799 } else if (list_is_short(&bundle
->ports
)) {
2800 bond_unref(bundle
->bond
);
2801 bundle
->bond
= NULL
;
2807 send_pdu_cb(void *port_
, const void *pdu
, size_t pdu_size
)
2809 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 10);
2810 struct ofport_dpif
*port
= port_
;
2811 uint8_t ea
[ETH_ADDR_LEN
];
2814 error
= netdev_get_etheraddr(port
->up
.netdev
, ea
);
2816 struct ofpbuf packet
;
2819 ofpbuf_init(&packet
, 0);
2820 packet_pdu
= eth_compose(&packet
, eth_addr_lacp
, ea
, ETH_TYPE_LACP
,
2822 memcpy(packet_pdu
, pdu
, pdu_size
);
2824 ofproto_dpif_send_packet(port
, &packet
);
2825 ofpbuf_uninit(&packet
);
2827 VLOG_ERR_RL(&rl
, "port %s: cannot obtain Ethernet address of iface "
2828 "%s (%s)", port
->bundle
->name
,
2829 netdev_get_name(port
->up
.netdev
), ovs_strerror(error
));
2834 bundle_send_learning_packets(struct ofbundle
*bundle
)
2836 struct ofproto_dpif
*ofproto
= bundle
->ofproto
;
2837 struct ofpbuf
*learning_packet
;
2838 int error
, n_packets
, n_errors
;
2839 struct mac_entry
*e
;
2840 struct list packets
;
2842 list_init(&packets
);
2843 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
2844 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
2845 if (e
->port
.p
!= bundle
) {
2848 learning_packet
= bond_compose_learning_packet(bundle
->bond
,
2851 /* Temporarily use 'frame' as a private pointer (see below). */
2852 ovs_assert(learning_packet
->frame
== ofpbuf_data(learning_packet
));
2853 learning_packet
->frame
= port_void
;
2854 list_push_back(&packets
, &learning_packet
->list_node
);
2857 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2859 error
= n_packets
= n_errors
= 0;
2860 LIST_FOR_EACH (learning_packet
, list_node
, &packets
) {
2862 void *port_void
= learning_packet
->frame
;
2864 /* Restore 'frame'. */
2865 learning_packet
->frame
= ofpbuf_data(learning_packet
);
2866 ret
= ofproto_dpif_send_packet(port_void
, learning_packet
);
2873 ofpbuf_list_delete(&packets
);
2876 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2877 VLOG_WARN_RL(&rl
, "bond %s: %d errors sending %d gratuitous learning "
2878 "packets, last error was: %s",
2879 bundle
->name
, n_errors
, n_packets
, ovs_strerror(error
));
2881 VLOG_DBG("bond %s: sent %d gratuitous learning packets",
2882 bundle
->name
, n_packets
);
2887 bundle_run(struct ofbundle
*bundle
)
2890 lacp_run(bundle
->lacp
, send_pdu_cb
);
2893 struct ofport_dpif
*port
;
2895 LIST_FOR_EACH (port
, bundle_node
, &bundle
->ports
) {
2896 bond_slave_set_may_enable(bundle
->bond
, port
, port
->may_enable
);
2899 if (bond_run(bundle
->bond
, lacp_status(bundle
->lacp
))) {
2900 bundle
->ofproto
->backer
->need_revalidate
= REV_BOND
;
2903 if (bond_should_send_learning_packets(bundle
->bond
)) {
2904 bundle_send_learning_packets(bundle
);
2910 bundle_wait(struct ofbundle
*bundle
)
2913 lacp_wait(bundle
->lacp
);
2916 bond_wait(bundle
->bond
);
2923 mirror_set__(struct ofproto
*ofproto_
, void *aux
,
2924 const struct ofproto_mirror_settings
*s
)
2926 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2927 struct ofbundle
**srcs
, **dsts
;
2932 mirror_destroy(ofproto
->mbridge
, aux
);
2936 srcs
= xmalloc(s
->n_srcs
* sizeof *srcs
);
2937 dsts
= xmalloc(s
->n_dsts
* sizeof *dsts
);
2939 for (i
= 0; i
< s
->n_srcs
; i
++) {
2940 srcs
[i
] = bundle_lookup(ofproto
, s
->srcs
[i
]);
2943 for (i
= 0; i
< s
->n_dsts
; i
++) {
2944 dsts
[i
] = bundle_lookup(ofproto
, s
->dsts
[i
]);
2947 error
= mirror_set(ofproto
->mbridge
, aux
, s
->name
, srcs
, s
->n_srcs
, dsts
,
2948 s
->n_dsts
, s
->src_vlans
,
2949 bundle_lookup(ofproto
, s
->out_bundle
), s
->out_vlan
);
2956 mirror_get_stats__(struct ofproto
*ofproto
, void *aux
,
2957 uint64_t *packets
, uint64_t *bytes
)
2959 return mirror_get_stats(ofproto_dpif_cast(ofproto
)->mbridge
, aux
, packets
,
2964 set_flood_vlans(struct ofproto
*ofproto_
, unsigned long *flood_vlans
)
2966 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2967 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2968 if (mac_learning_set_flood_vlans(ofproto
->ml
, flood_vlans
)) {
2969 mac_learning_flush(ofproto
->ml
);
2971 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
2976 is_mirror_output_bundle(const struct ofproto
*ofproto_
, void *aux
)
2978 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2979 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
2980 return bundle
&& mirror_bundle_out(ofproto
->mbridge
, bundle
) != 0;
2984 forward_bpdu_changed(struct ofproto
*ofproto_
)
2986 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2987 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
2991 set_mac_table_config(struct ofproto
*ofproto_
, unsigned int idle_time
,
2994 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
2995 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
2996 mac_learning_set_idle_time(ofproto
->ml
, idle_time
);
2997 mac_learning_set_max_entries(ofproto
->ml
, max_entries
);
2998 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
3001 /* Configures multicast snooping on 'ofport' using the settings
3002 * defined in 's'. */
3004 set_mcast_snooping(struct ofproto
*ofproto_
,
3005 const struct ofproto_mcast_snooping_settings
*s
)
3007 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3009 /* Only revalidate flows if the configuration changed. */
3010 if (!s
!= !ofproto
->ms
) {
3011 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3016 ofproto
->ms
= mcast_snooping_create();
3019 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3020 mcast_snooping_set_idle_time(ofproto
->ms
, s
->idle_time
);
3021 mcast_snooping_set_max_entries(ofproto
->ms
, s
->max_entries
);
3022 if (mcast_snooping_set_flood_unreg(ofproto
->ms
, s
->flood_unreg
)) {
3023 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3025 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3027 mcast_snooping_unref(ofproto
->ms
);
3034 /* Configures multicast snooping port's flood setting on 'ofproto'. */
3036 set_mcast_snooping_port(struct ofproto
*ofproto_
, void *aux
, bool flood
)
3038 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3039 struct ofbundle
*bundle
= bundle_lookup(ofproto
, aux
);
3042 ovs_rwlock_wrlock(&ofproto
->ms
->rwlock
);
3043 mcast_snooping_set_port_flood(ofproto
->ms
, bundle
->vlan
, bundle
,
3045 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
3053 static struct ofport_dpif
*
3054 get_ofp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
3056 struct ofport
*ofport
= ofproto_get_port(&ofproto
->up
, ofp_port
);
3057 return ofport
? ofport_dpif_cast(ofport
) : NULL
;
3061 ofproto_port_from_dpif_port(struct ofproto_dpif
*ofproto
,
3062 struct ofproto_port
*ofproto_port
,
3063 struct dpif_port
*dpif_port
)
3065 ofproto_port
->name
= dpif_port
->name
;
3066 ofproto_port
->type
= dpif_port
->type
;
3067 ofproto_port
->ofp_port
= odp_port_to_ofp_port(ofproto
, dpif_port
->port_no
);
3071 ofport_update_peer(struct ofport_dpif
*ofport
)
3073 const struct ofproto_dpif
*ofproto
;
3074 struct dpif_backer
*backer
;
3077 if (!netdev_vport_is_patch(ofport
->up
.netdev
)) {
3081 backer
= ofproto_dpif_cast(ofport
->up
.ofproto
)->backer
;
3082 backer
->need_revalidate
= REV_RECONFIGURE
;
3085 ofport
->peer
->peer
= NULL
;
3086 ofport
->peer
= NULL
;
3089 peer_name
= netdev_vport_patch_peer(ofport
->up
.netdev
);
3094 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
3095 struct ofport
*peer_ofport
;
3096 struct ofport_dpif
*peer
;
3099 if (ofproto
->backer
!= backer
) {
3103 peer_ofport
= shash_find_data(&ofproto
->up
.port_by_name
, peer_name
);
3108 peer
= ofport_dpif_cast(peer_ofport
);
3109 peer_peer
= netdev_vport_patch_peer(peer
->up
.netdev
);
3110 if (peer_peer
&& !strcmp(netdev_get_name(ofport
->up
.netdev
),
3112 ofport
->peer
= peer
;
3113 ofport
->peer
->peer
= ofport
;
3123 port_run(struct ofport_dpif
*ofport
)
3125 long long int carrier_seq
= netdev_get_carrier_resets(ofport
->up
.netdev
);
3126 bool carrier_changed
= carrier_seq
!= ofport
->carrier_seq
;
3127 bool enable
= netdev_get_carrier(ofport
->up
.netdev
);
3128 bool cfm_enable
= false;
3129 bool bfd_enable
= false;
3131 ofport
->carrier_seq
= carrier_seq
;
3134 int cfm_opup
= cfm_get_opup(ofport
->cfm
);
3136 cfm_enable
= !cfm_get_fault(ofport
->cfm
);
3138 if (cfm_opup
>= 0) {
3139 cfm_enable
= cfm_enable
&& cfm_opup
;
3144 bfd_enable
= bfd_forwarding(ofport
->bfd
);
3147 if (ofport
->bfd
|| ofport
->cfm
) {
3148 enable
= enable
&& (cfm_enable
|| bfd_enable
);
3151 if (ofport
->bundle
) {
3152 enable
= enable
&& lacp_slave_may_enable(ofport
->bundle
->lacp
, ofport
);
3153 if (carrier_changed
) {
3154 lacp_slave_carrier_changed(ofport
->bundle
->lacp
, ofport
);
3158 if (ofport
->may_enable
!= enable
) {
3159 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3161 ofproto
->backer
->need_revalidate
= REV_PORT_TOGGLED
;
3163 if (ofport
->rstp_port
) {
3164 rstp_port_set_mac_operational(ofport
->rstp_port
, enable
);
3168 ofport
->may_enable
= enable
;
3172 port_query_by_name(const struct ofproto
*ofproto_
, const char *devname
,
3173 struct ofproto_port
*ofproto_port
)
3175 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3176 struct dpif_port dpif_port
;
3179 if (sset_contains(&ofproto
->ghost_ports
, devname
)) {
3180 const char *type
= netdev_get_type_from_name(devname
);
3182 /* We may be called before ofproto->up.port_by_name is populated with
3183 * the appropriate ofport. For this reason, we must get the name and
3184 * type from the netdev layer directly. */
3186 const struct ofport
*ofport
;
3188 ofport
= shash_find_data(&ofproto
->up
.port_by_name
, devname
);
3189 ofproto_port
->ofp_port
= ofport
? ofport
->ofp_port
: OFPP_NONE
;
3190 ofproto_port
->name
= xstrdup(devname
);
3191 ofproto_port
->type
= xstrdup(type
);
3197 if (!sset_contains(&ofproto
->ports
, devname
)) {
3200 error
= dpif_port_query_by_name(ofproto
->backer
->dpif
,
3201 devname
, &dpif_port
);
3203 ofproto_port_from_dpif_port(ofproto
, ofproto_port
, &dpif_port
);
3209 port_add(struct ofproto
*ofproto_
, struct netdev
*netdev
)
3211 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3212 const char *devname
= netdev_get_name(netdev
);
3213 char namebuf
[NETDEV_VPORT_NAME_BUFSIZE
];
3214 const char *dp_port_name
;
3216 if (netdev_vport_is_patch(netdev
)) {
3217 sset_add(&ofproto
->ghost_ports
, netdev_get_name(netdev
));
3221 dp_port_name
= netdev_vport_get_dpif_port(netdev
, namebuf
, sizeof namebuf
);
3222 if (!dpif_port_exists(ofproto
->backer
->dpif
, dp_port_name
)) {
3223 odp_port_t port_no
= ODPP_NONE
;
3226 error
= dpif_port_add(ofproto
->backer
->dpif
, netdev
, &port_no
);
3230 if (netdev_get_tunnel_config(netdev
)) {
3231 simap_put(&ofproto
->backer
->tnl_backers
,
3232 dp_port_name
, odp_to_u32(port_no
));
3236 if (netdev_get_tunnel_config(netdev
)) {
3237 sset_add(&ofproto
->ghost_ports
, devname
);
3239 sset_add(&ofproto
->ports
, devname
);
3245 port_del(struct ofproto
*ofproto_
, ofp_port_t ofp_port
)
3247 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3248 struct ofport_dpif
*ofport
= get_ofp_port(ofproto
, ofp_port
);
3255 sset_find_and_delete(&ofproto
->ghost_ports
,
3256 netdev_get_name(ofport
->up
.netdev
));
3257 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
3258 if (!ofport
->is_tunnel
&& !netdev_vport_is_patch(ofport
->up
.netdev
)) {
3259 error
= dpif_port_del(ofproto
->backer
->dpif
, ofport
->odp_port
);
3261 /* The caller is going to close ofport->up.netdev. If this is a
3262 * bonded port, then the bond is using that netdev, so remove it
3263 * from the bond. The client will need to reconfigure everything
3264 * after deleting ports, so then the slave will get re-added. */
3265 bundle_remove(&ofport
->up
);
3272 port_get_stats(const struct ofport
*ofport_
, struct netdev_stats
*stats
)
3274 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3277 error
= netdev_get_stats(ofport
->up
.netdev
, stats
);
3279 if (!error
&& ofport_
->ofp_port
== OFPP_LOCAL
) {
3280 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
3282 ovs_mutex_lock(&ofproto
->stats_mutex
);
3283 /* ofproto->stats.tx_packets represents packets that we created
3284 * internally and sent to some port (e.g. packets sent with
3285 * ofproto_dpif_send_packet()). Account for them as if they had
3286 * come from OFPP_LOCAL and got forwarded. */
3288 if (stats
->rx_packets
!= UINT64_MAX
) {
3289 stats
->rx_packets
+= ofproto
->stats
.tx_packets
;
3292 if (stats
->rx_bytes
!= UINT64_MAX
) {
3293 stats
->rx_bytes
+= ofproto
->stats
.tx_bytes
;
3296 /* ofproto->stats.rx_packets represents packets that were received on
3297 * some port and we processed internally and dropped (e.g. STP).
3298 * Account for them as if they had been forwarded to OFPP_LOCAL. */
3300 if (stats
->tx_packets
!= UINT64_MAX
) {
3301 stats
->tx_packets
+= ofproto
->stats
.rx_packets
;
3304 if (stats
->tx_bytes
!= UINT64_MAX
) {
3305 stats
->tx_bytes
+= ofproto
->stats
.rx_bytes
;
3307 ovs_mutex_unlock(&ofproto
->stats_mutex
);
3313 struct port_dump_state
{
3318 struct ofproto_port port
;
3323 port_dump_start(const struct ofproto
*ofproto_ OVS_UNUSED
, void **statep
)
3325 *statep
= xzalloc(sizeof(struct port_dump_state
));
3330 port_dump_next(const struct ofproto
*ofproto_
, void *state_
,
3331 struct ofproto_port
*port
)
3333 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3334 struct port_dump_state
*state
= state_
;
3335 const struct sset
*sset
;
3336 struct sset_node
*node
;
3338 if (state
->has_port
) {
3339 ofproto_port_destroy(&state
->port
);
3340 state
->has_port
= false;
3342 sset
= state
->ghost
? &ofproto
->ghost_ports
: &ofproto
->ports
;
3343 while ((node
= sset_at_position(sset
, &state
->bucket
, &state
->offset
))) {
3346 error
= port_query_by_name(ofproto_
, node
->name
, &state
->port
);
3348 *port
= state
->port
;
3349 state
->has_port
= true;
3351 } else if (error
!= ENODEV
) {
3356 if (!state
->ghost
) {
3357 state
->ghost
= true;
3360 return port_dump_next(ofproto_
, state_
, port
);
3367 port_dump_done(const struct ofproto
*ofproto_ OVS_UNUSED
, void *state_
)
3369 struct port_dump_state
*state
= state_
;
3371 if (state
->has_port
) {
3372 ofproto_port_destroy(&state
->port
);
3379 port_poll(const struct ofproto
*ofproto_
, char **devnamep
)
3381 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3383 if (ofproto
->port_poll_errno
) {
3384 int error
= ofproto
->port_poll_errno
;
3385 ofproto
->port_poll_errno
= 0;
3389 if (sset_is_empty(&ofproto
->port_poll_set
)) {
3393 *devnamep
= sset_pop(&ofproto
->port_poll_set
);
3398 port_poll_wait(const struct ofproto
*ofproto_
)
3400 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
3401 dpif_port_poll_wait(ofproto
->backer
->dpif
);
3405 port_is_lacp_current(const struct ofport
*ofport_
)
3407 const struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
3408 return (ofport
->bundle
&& ofport
->bundle
->lacp
3409 ? lacp_slave_is_current(ofport
->bundle
->lacp
, ofport
)
3413 /* If 'rule' is an OpenFlow rule, that has expired according to OpenFlow rules,
3414 * then delete it entirely. */
3416 rule_expire(struct rule_dpif
*rule
)
3417 OVS_REQUIRES(ofproto_mutex
)
3419 uint16_t hard_timeout
, idle_timeout
;
3420 long long int now
= time_msec();
3423 hard_timeout
= rule
->up
.hard_timeout
;
3424 idle_timeout
= rule
->up
.idle_timeout
;
3426 /* Has 'rule' expired? */
3428 long long int modified
;
3430 ovs_mutex_lock(&rule
->up
.mutex
);
3431 modified
= rule
->up
.modified
;
3432 ovs_mutex_unlock(&rule
->up
.mutex
);
3434 if (now
> modified
+ hard_timeout
* 1000) {
3435 reason
= OFPRR_HARD_TIMEOUT
;
3439 if (reason
< 0 && idle_timeout
) {
3442 ovs_mutex_lock(&rule
->stats_mutex
);
3443 used
= rule
->stats
.used
;
3444 ovs_mutex_unlock(&rule
->stats_mutex
);
3446 if (now
> used
+ idle_timeout
* 1000) {
3447 reason
= OFPRR_IDLE_TIMEOUT
;
3452 COVERAGE_INC(ofproto_dpif_expired
);
3453 ofproto_rule_expire(&rule
->up
, reason
);
3457 /* Executes, within 'ofproto', the actions in 'rule' or 'ofpacts' on 'packet'.
3458 * 'flow' must reflect the data in 'packet'. */
3460 ofproto_dpif_execute_actions(struct ofproto_dpif
*ofproto
,
3461 const struct flow
*flow
,
3462 struct rule_dpif
*rule
,
3463 const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3464 struct ofpbuf
*packet
)
3466 struct dpif_flow_stats stats
;
3467 struct xlate_out xout
;
3468 struct xlate_in xin
;
3470 struct dpif_execute execute
;
3473 ovs_assert((rule
!= NULL
) != (ofpacts
!= NULL
));
3475 dpif_flow_stats_extract(flow
, packet
, time_msec(), &stats
);
3478 rule_dpif_credit_stats(rule
, &stats
);
3481 xlate_in_init(&xin
, ofproto
, flow
, flow
->in_port
.ofp_port
, rule
,
3482 stats
.tcp_flags
, packet
);
3483 xin
.ofpacts
= ofpacts
;
3484 xin
.ofpacts_len
= ofpacts_len
;
3485 xin
.resubmit_stats
= &stats
;
3486 xlate_actions(&xin
, &xout
);
3488 execute
.actions
= ofpbuf_data(xout
.odp_actions
);
3489 execute
.actions_len
= ofpbuf_size(xout
.odp_actions
);
3490 execute
.packet
= packet
;
3491 execute
.md
= pkt_metadata_from_flow(flow
);
3492 execute
.needs_help
= (xout
.slow
& SLOW_ACTION
) != 0;
3493 execute
.probe
= false;
3495 /* Fix up in_port. */
3496 in_port
= flow
->in_port
.ofp_port
;
3497 if (in_port
== OFPP_NONE
) {
3498 in_port
= OFPP_LOCAL
;
3500 execute
.md
.in_port
.odp_port
= ofp_port_to_odp_port(ofproto
, in_port
);
3502 error
= dpif_execute(ofproto
->backer
->dpif
, &execute
);
3504 xlate_out_uninit(&xout
);
3510 rule_dpif_credit_stats(struct rule_dpif
*rule
,
3511 const struct dpif_flow_stats
*stats
)
3513 ovs_mutex_lock(&rule
->stats_mutex
);
3514 rule
->stats
.n_packets
+= stats
->n_packets
;
3515 rule
->stats
.n_bytes
+= stats
->n_bytes
;
3516 rule
->stats
.used
= MAX(rule
->stats
.used
, stats
->used
);
3517 ovs_mutex_unlock(&rule
->stats_mutex
);
3521 rule_dpif_get_flow_cookie(const struct rule_dpif
*rule
)
3522 OVS_REQUIRES(rule
->up
.mutex
)
3524 return rule
->up
.flow_cookie
;
3528 rule_dpif_reduce_timeouts(struct rule_dpif
*rule
, uint16_t idle_timeout
,
3529 uint16_t hard_timeout
)
3531 ofproto_rule_reduce_timeouts(&rule
->up
, idle_timeout
, hard_timeout
);
3534 /* Returns 'rule''s actions. The returned actions are RCU-protected, and can
3535 * be read until the calling thread quiesces. */
3536 const struct rule_actions
*
3537 rule_dpif_get_actions(const struct rule_dpif
*rule
)
3539 return rule_get_actions(&rule
->up
);
3542 /* Sets 'rule''s recirculation id. */
3544 rule_dpif_set_recirc_id(struct rule_dpif
*rule
, uint32_t id
)
3545 OVS_REQUIRES(rule
->up
.mutex
)
3547 ovs_assert(!rule
->recirc_id
);
3548 rule
->recirc_id
= id
;
3551 /* Returns 'rule''s recirculation id. */
3553 rule_dpif_get_recirc_id(struct rule_dpif
*rule
)
3554 OVS_REQUIRES(rule
->up
.mutex
)
3556 if (!rule
->recirc_id
) {
3557 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3559 rule_dpif_set_recirc_id(rule
, ofproto_dpif_alloc_recirc_id(ofproto
));
3561 return rule
->recirc_id
;
3564 /* Sets 'rule''s recirculation id. */
3566 rule_set_recirc_id(struct rule
*rule_
, uint32_t id
)
3568 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3570 ovs_mutex_lock(&rule
->up
.mutex
);
3571 rule_dpif_set_recirc_id(rule
, id
);
3572 ovs_mutex_unlock(&rule
->up
.mutex
);
3575 /* Lookup 'flow' in table 0 of 'ofproto''s classifier.
3576 * If 'wc' is non-null, sets the fields that were relevant as part of
3577 * the lookup. Returns the table id where a match or miss occurred via
3578 * 'table_id'. This will be zero unless there was a miss and
3579 * OFPTC11_TABLE_MISS_CONTINUE is in effect for the sequence of tables
3580 * where misses occur, or TBL_INTERNAL if the rule has a non-zero
3581 * recirculation ID, and a match was found in the internal table, or if
3582 * there was no match and one of the special rules (drop_frags_rule,
3583 * miss_rule, or no_packet_in_rule) was returned.
3585 * The return value is the found rule, which is valid at least until the next
3586 * RCU quiescent period. If the rule needs to stay around longer,
3587 * a non-zero 'take_ref' must be passed in to cause a reference to be taken
3588 * on it before this returns. */
3590 rule_dpif_lookup(struct ofproto_dpif
*ofproto
, struct flow
*flow
,
3591 struct flow_wildcards
*wc
, bool take_ref
,
3592 const struct dpif_flow_stats
*stats
, uint8_t *table_id
)
3596 if (ofproto_dpif_get_enable_recirc(ofproto
)) {
3597 /* Always exactly match recirc_id since datapath supports
3600 wc
->masks
.recirc_id
= UINT32_MAX
;
3602 if (flow
->recirc_id
) {
3603 /* Start looking up from internal table for post recirculation
3604 * flows or packets. */
3605 *table_id
= TBL_INTERNAL
;
3609 return rule_dpif_lookup_from_table(ofproto
, flow
, wc
, take_ref
, stats
,
3610 table_id
, flow
->in_port
.ofp_port
, true,
3614 /* The returned rule (if any) is valid at least until the next RCU quiescent
3615 * period. If the rule needs to stay around longer, a non-zero 'take_ref'
3616 * must be passed in to cause a reference to be taken on it. */
3617 static struct rule_dpif
*
3618 rule_dpif_lookup_in_table(struct ofproto_dpif
*ofproto
, uint8_t table_id
,
3619 const struct flow
*flow
, struct flow_wildcards
*wc
,
3622 struct classifier
*cls
= &ofproto
->up
.tables
[table_id
].cls
;
3623 const struct cls_rule
*cls_rule
;
3624 struct rule_dpif
*rule
;
3627 cls_rule
= classifier_lookup(cls
, flow
, wc
);
3629 rule
= rule_dpif_cast(rule_from_cls_rule(cls_rule
));
3631 /* Try again if the rule was released before we get the reference. */
3632 } while (rule
&& take_ref
&& !rule_dpif_try_ref(rule
));
3637 /* Look up 'flow' in 'ofproto''s classifier starting from table '*table_id'.
3638 * Returns the rule that was found, which may be one of the special rules
3639 * according to packet miss hadling. If 'may_packet_in' is false, returning of
3640 * the miss_rule (which issues packet ins for the controller) is avoided.
3641 * Updates 'wc', if nonnull, to reflect the fields that were used during the
3644 * If 'honor_table_miss' is true, the first lookup occurs in '*table_id', but
3645 * if none is found then the table miss configuration for that table is
3646 * honored, which can result in additional lookups in other OpenFlow tables.
3647 * In this case the function updates '*table_id' to reflect the final OpenFlow
3648 * table that was searched.
3650 * If 'honor_table_miss' is false, then only one table lookup occurs, in
3653 * The rule is returned in '*rule', which is valid at least until the next
3654 * RCU quiescent period. If the '*rule' needs to stay around longer,
3655 * a non-zero 'take_ref' must be passed in to cause a reference to be taken
3656 * on it before this returns.
3658 * 'in_port' allows the lookup to take place as if the in port had the value
3659 * 'in_port'. This is needed for resubmit action support. */
3661 rule_dpif_lookup_from_table(struct ofproto_dpif
*ofproto
, struct flow
*flow
,
3662 struct flow_wildcards
*wc
, bool take_ref
,
3663 const struct dpif_flow_stats
*stats
,
3664 uint8_t *table_id
, ofp_port_t in_port
,
3665 bool may_packet_in
, bool honor_table_miss
)
3667 ovs_be16 old_tp_src
= flow
->tp_src
, old_tp_dst
= flow
->tp_dst
;
3668 ofp_port_t old_in_port
= flow
->in_port
.ofp_port
;
3669 enum ofputil_table_miss miss_config
;
3670 struct rule_dpif
*rule
;
3673 /* We always unwildcard nw_frag (for IP), so they
3674 * need not be unwildcarded here. */
3675 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
3676 && ofproto
->up
.frag_handling
!= OFPC_FRAG_NX_MATCH
) {
3677 if (ofproto
->up
.frag_handling
== OFPC_FRAG_NORMAL
) {
3678 /* We must pretend that transport ports are unavailable. */
3679 flow
->tp_src
= htons(0);
3680 flow
->tp_dst
= htons(0);
3682 /* Must be OFPC_FRAG_DROP (we don't have OFPC_FRAG_REASM).
3683 * Use the drop_frags_rule (which cannot disappear). */
3684 rule
= ofproto
->drop_frags_rule
;
3686 rule_dpif_ref(rule
);
3689 struct oftable
*tbl
= &ofproto
->up
.tables
[*table_id
];
3692 atomic_add_relaxed(&tbl
->n_matched
, stats
->n_packets
, &orig
);
3698 /* Look up a flow with 'in_port' as the input port. Then restore the
3699 * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
3700 * have surprising behavior). */
3701 flow
->in_port
.ofp_port
= in_port
;
3703 /* Our current implementation depends on n_tables == N_TABLES, and
3704 * TBL_INTERNAL being the last table. */
3705 BUILD_ASSERT_DECL(N_TABLES
== TBL_INTERNAL
+ 1);
3707 miss_config
= OFPUTIL_TABLE_MISS_CONTINUE
;
3709 for (next_id
= *table_id
;
3710 next_id
< ofproto
->up
.n_tables
;
3711 next_id
++, next_id
+= (next_id
== TBL_INTERNAL
))
3713 *table_id
= next_id
;
3714 rule
= rule_dpif_lookup_in_table(ofproto
, next_id
, flow
, wc
, take_ref
);
3716 struct oftable
*tbl
= &ofproto
->up
.tables
[next_id
];
3719 atomic_add_relaxed(rule
? &tbl
->n_matched
: &tbl
->n_missed
,
3720 stats
->n_packets
, &orig
);
3723 goto out
; /* Match. */
3725 if (honor_table_miss
) {
3726 miss_config
= ofproto_table_get_miss_config(&ofproto
->up
,
3728 if (miss_config
== OFPUTIL_TABLE_MISS_CONTINUE
) {
3735 rule
= ofproto
->no_packet_in_rule
;
3736 if (may_packet_in
) {
3737 if (miss_config
== OFPUTIL_TABLE_MISS_CONTINUE
3738 || miss_config
== OFPUTIL_TABLE_MISS_CONTROLLER
) {
3739 struct ofport_dpif
*port
;
3741 port
= get_ofp_port(ofproto
, old_in_port
);
3743 VLOG_WARN_RL(&rl
, "packet-in on unknown OpenFlow port %"PRIu16
,
3745 } else if (!(port
->up
.pp
.config
& OFPUTIL_PC_NO_PACKET_IN
)) {
3746 rule
= ofproto
->miss_rule
;
3748 } else if (miss_config
== OFPUTIL_TABLE_MISS_DEFAULT
&&
3749 connmgr_wants_packet_in_on_miss(ofproto
->up
.connmgr
)) {
3750 rule
= ofproto
->miss_rule
;
3754 rule_dpif_ref(rule
);
3757 /* Restore port numbers, as they may have been modified above. */
3758 flow
->tp_src
= old_tp_src
;
3759 flow
->tp_dst
= old_tp_dst
;
3760 /* Restore the old in port. */
3761 flow
->in_port
.ofp_port
= old_in_port
;
3767 complete_operation(struct rule_dpif
*rule
)
3768 OVS_REQUIRES(ofproto_mutex
)
3770 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3772 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
3775 static struct rule_dpif
*rule_dpif_cast(const struct rule
*rule
)
3777 return rule
? CONTAINER_OF(rule
, struct rule_dpif
, up
) : NULL
;
3780 static struct rule
*
3783 struct rule_dpif
*rule
= xmalloc(sizeof *rule
);
3788 rule_dealloc(struct rule
*rule_
)
3790 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3795 rule_construct(struct rule
*rule_
)
3796 OVS_NO_THREAD_SAFETY_ANALYSIS
3798 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3799 ovs_mutex_init_adaptive(&rule
->stats_mutex
);
3800 rule
->stats
.n_packets
= 0;
3801 rule
->stats
.n_bytes
= 0;
3802 rule
->stats
.used
= rule
->up
.modified
;
3803 rule
->recirc_id
= 0;
3809 rule_insert(struct rule
*rule_
)
3810 OVS_REQUIRES(ofproto_mutex
)
3812 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3813 complete_operation(rule
);
3818 rule_delete(struct rule
*rule_
)
3819 OVS_REQUIRES(ofproto_mutex
)
3821 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3822 complete_operation(rule
);
3826 rule_destruct(struct rule
*rule_
)
3828 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3830 ovs_mutex_destroy(&rule
->stats_mutex
);
3831 if (rule
->recirc_id
) {
3832 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3834 ofproto_dpif_free_recirc_id(ofproto
, rule
->recirc_id
);
3839 rule_get_stats(struct rule
*rule_
, uint64_t *packets
, uint64_t *bytes
,
3840 long long int *used
)
3842 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3844 ovs_mutex_lock(&rule
->stats_mutex
);
3845 *packets
= rule
->stats
.n_packets
;
3846 *bytes
= rule
->stats
.n_bytes
;
3847 *used
= rule
->stats
.used
;
3848 ovs_mutex_unlock(&rule
->stats_mutex
);
3852 rule_dpif_execute(struct rule_dpif
*rule
, const struct flow
*flow
,
3853 struct ofpbuf
*packet
)
3855 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(rule
->up
.ofproto
);
3857 ofproto_dpif_execute_actions(ofproto
, flow
, rule
, NULL
, 0, packet
);
3861 rule_execute(struct rule
*rule
, const struct flow
*flow
,
3862 struct ofpbuf
*packet
)
3864 rule_dpif_execute(rule_dpif_cast(rule
), flow
, packet
);
3865 ofpbuf_delete(packet
);
3870 rule_modify_actions(struct rule
*rule_
, bool reset_counters
)
3871 OVS_REQUIRES(ofproto_mutex
)
3873 struct rule_dpif
*rule
= rule_dpif_cast(rule_
);
3875 if (reset_counters
) {
3876 ovs_mutex_lock(&rule
->stats_mutex
);
3877 rule
->stats
.n_packets
= 0;
3878 rule
->stats
.n_bytes
= 0;
3879 ovs_mutex_unlock(&rule
->stats_mutex
);
3882 complete_operation(rule
);
3885 static struct group_dpif
*group_dpif_cast(const struct ofgroup
*group
)
3887 return group
? CONTAINER_OF(group
, struct group_dpif
, up
) : NULL
;
3890 static struct ofgroup
*
3893 struct group_dpif
*group
= xzalloc(sizeof *group
);
3898 group_dealloc(struct ofgroup
*group_
)
3900 struct group_dpif
*group
= group_dpif_cast(group_
);
3905 group_construct_stats(struct group_dpif
*group
)
3906 OVS_REQUIRES(group
->stats_mutex
)
3908 struct ofputil_bucket
*bucket
;
3909 const struct list
*buckets
;
3911 group
->packet_count
= 0;
3912 group
->byte_count
= 0;
3914 group_dpif_get_buckets(group
, &buckets
);
3915 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
3916 bucket
->stats
.packet_count
= 0;
3917 bucket
->stats
.byte_count
= 0;
3922 group_dpif_credit_stats(struct group_dpif
*group
,
3923 struct ofputil_bucket
*bucket
,
3924 const struct dpif_flow_stats
*stats
)
3926 ovs_mutex_lock(&group
->stats_mutex
);
3927 group
->packet_count
+= stats
->n_packets
;
3928 group
->byte_count
+= stats
->n_bytes
;
3930 bucket
->stats
.packet_count
+= stats
->n_packets
;
3931 bucket
->stats
.byte_count
+= stats
->n_bytes
;
3932 } else { /* Credit to all buckets */
3933 const struct list
*buckets
;
3935 group_dpif_get_buckets(group
, &buckets
);
3936 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
3937 bucket
->stats
.packet_count
+= stats
->n_packets
;
3938 bucket
->stats
.byte_count
+= stats
->n_bytes
;
3941 ovs_mutex_unlock(&group
->stats_mutex
);
3945 group_construct(struct ofgroup
*group_
)
3947 struct group_dpif
*group
= group_dpif_cast(group_
);
3948 const struct ofputil_bucket
*bucket
;
3950 /* Prevent group chaining because our locking structure makes it hard to
3951 * implement deadlock-free. (See xlate_group_resource_check().) */
3952 LIST_FOR_EACH (bucket
, list_node
, &group
->up
.buckets
) {
3953 const struct ofpact
*a
;
3955 OFPACT_FOR_EACH (a
, bucket
->ofpacts
, bucket
->ofpacts_len
) {
3956 if (a
->type
== OFPACT_GROUP
) {
3957 return OFPERR_OFPGMFC_CHAINING_UNSUPPORTED
;
3962 ovs_mutex_init_adaptive(&group
->stats_mutex
);
3963 ovs_mutex_lock(&group
->stats_mutex
);
3964 group_construct_stats(group
);
3965 ovs_mutex_unlock(&group
->stats_mutex
);
3970 group_destruct(struct ofgroup
*group_
)
3972 struct group_dpif
*group
= group_dpif_cast(group_
);
3973 ovs_mutex_destroy(&group
->stats_mutex
);
3977 group_modify(struct ofgroup
*group_
)
3979 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(group_
->ofproto
);
3981 ofproto
->backer
->need_revalidate
= REV_FLOW_TABLE
;
3987 group_get_stats(const struct ofgroup
*group_
, struct ofputil_group_stats
*ogs
)
3989 struct group_dpif
*group
= group_dpif_cast(group_
);
3990 struct ofputil_bucket
*bucket
;
3991 const struct list
*buckets
;
3992 struct bucket_counter
*bucket_stats
;
3994 ovs_mutex_lock(&group
->stats_mutex
);
3995 ogs
->packet_count
= group
->packet_count
;
3996 ogs
->byte_count
= group
->byte_count
;
3998 group_dpif_get_buckets(group
, &buckets
);
3999 bucket_stats
= ogs
->bucket_stats
;
4000 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
4001 bucket_stats
->packet_count
= bucket
->stats
.packet_count
;
4002 bucket_stats
->byte_count
= bucket
->stats
.byte_count
;
4005 ovs_mutex_unlock(&group
->stats_mutex
);
4010 /* If the group exists, this function increments the groups's reference count.
4012 * Make sure to call group_dpif_unref() after no longer needing to maintain
4013 * a reference to the group. */
4015 group_dpif_lookup(struct ofproto_dpif
*ofproto
, uint32_t group_id
,
4016 struct group_dpif
**group
)
4018 struct ofgroup
*ofgroup
;
4021 found
= ofproto_group_lookup(&ofproto
->up
, group_id
, &ofgroup
);
4022 *group
= found
? group_dpif_cast(ofgroup
) : NULL
;
4028 group_dpif_get_buckets(const struct group_dpif
*group
,
4029 const struct list
**buckets
)
4031 *buckets
= &group
->up
.buckets
;
4034 enum ofp11_group_type
4035 group_dpif_get_type(const struct group_dpif
*group
)
4037 return group
->up
.type
;
4040 /* Sends 'packet' out 'ofport'.
4041 * May modify 'packet'.
4042 * Returns 0 if successful, otherwise a positive errno value. */
4044 ofproto_dpif_send_packet(const struct ofport_dpif
*ofport
, struct ofpbuf
*packet
)
4046 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport
->up
.ofproto
);
4049 error
= xlate_send_packet(ofport
, packet
);
4051 ovs_mutex_lock(&ofproto
->stats_mutex
);
4052 ofproto
->stats
.tx_packets
++;
4053 ofproto
->stats
.tx_bytes
+= ofpbuf_size(packet
);
4054 ovs_mutex_unlock(&ofproto
->stats_mutex
);
4058 /* Return the version string of the datapath that backs up
4062 get_datapath_version(const struct ofproto
*ofproto_
)
4064 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4066 return ofproto
->backer
->dp_version_string
;
4070 set_frag_handling(struct ofproto
*ofproto_
,
4071 enum ofp_config_flags frag_handling
)
4073 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4074 if (frag_handling
!= OFPC_FRAG_REASM
) {
4075 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4083 packet_out(struct ofproto
*ofproto_
, struct ofpbuf
*packet
,
4084 const struct flow
*flow
,
4085 const struct ofpact
*ofpacts
, size_t ofpacts_len
)
4087 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4089 ofproto_dpif_execute_actions(ofproto
, flow
, NULL
, ofpacts
,
4090 ofpacts_len
, packet
);
4097 set_netflow(struct ofproto
*ofproto_
,
4098 const struct netflow_options
*netflow_options
)
4100 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4102 if (netflow_options
) {
4103 if (!ofproto
->netflow
) {
4104 ofproto
->netflow
= netflow_create();
4105 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4107 return netflow_set_options(ofproto
->netflow
, netflow_options
);
4108 } else if (ofproto
->netflow
) {
4109 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
4110 netflow_unref(ofproto
->netflow
);
4111 ofproto
->netflow
= NULL
;
4118 get_netflow_ids(const struct ofproto
*ofproto_
,
4119 uint8_t *engine_type
, uint8_t *engine_id
)
4121 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofproto_
);
4123 dpif_get_netflow_ids(ofproto
->backer
->dpif
, engine_type
, engine_id
);
4126 static struct ofproto_dpif
*
4127 ofproto_dpif_lookup(const char *name
)
4129 struct ofproto_dpif
*ofproto
;
4131 HMAP_FOR_EACH_WITH_HASH (ofproto
, all_ofproto_dpifs_node
,
4132 hash_string(name
, 0), &all_ofproto_dpifs
) {
4133 if (!strcmp(ofproto
->up
.name
, name
)) {
4141 ofproto_unixctl_fdb_flush(struct unixctl_conn
*conn
, int argc
,
4142 const char *argv
[], void *aux OVS_UNUSED
)
4144 struct ofproto_dpif
*ofproto
;
4147 ofproto
= ofproto_dpif_lookup(argv
[1]);
4149 unixctl_command_reply_error(conn
, "no such bridge");
4152 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4153 mac_learning_flush(ofproto
->ml
);
4154 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4156 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4157 ovs_rwlock_wrlock(&ofproto
->ml
->rwlock
);
4158 mac_learning_flush(ofproto
->ml
);
4159 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4163 unixctl_command_reply(conn
, "table successfully flushed");
4167 ofproto_unixctl_mcast_snooping_flush(struct unixctl_conn
*conn
, int argc
,
4168 const char *argv
[], void *aux OVS_UNUSED
)
4170 struct ofproto_dpif
*ofproto
;
4173 ofproto
= ofproto_dpif_lookup(argv
[1]);
4175 unixctl_command_reply_error(conn
, "no such bridge");
4179 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4180 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4183 mcast_snooping_mdb_flush(ofproto
->ms
);
4185 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4186 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4189 mcast_snooping_mdb_flush(ofproto
->ms
);
4193 unixctl_command_reply(conn
, "table successfully flushed");
4196 static struct ofport_dpif
*
4197 ofbundle_get_a_port(const struct ofbundle
*bundle
)
4199 return CONTAINER_OF(list_front(&bundle
->ports
), struct ofport_dpif
,
4204 ofproto_unixctl_fdb_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4205 const char *argv
[], void *aux OVS_UNUSED
)
4207 struct ds ds
= DS_EMPTY_INITIALIZER
;
4208 const struct ofproto_dpif
*ofproto
;
4209 const struct mac_entry
*e
;
4211 ofproto
= ofproto_dpif_lookup(argv
[1]);
4213 unixctl_command_reply_error(conn
, "no such bridge");
4217 ds_put_cstr(&ds
, " port VLAN MAC Age\n");
4218 ovs_rwlock_rdlock(&ofproto
->ml
->rwlock
);
4219 LIST_FOR_EACH (e
, lru_node
, &ofproto
->ml
->lrus
) {
4220 struct ofbundle
*bundle
= e
->port
.p
;
4221 char name
[OFP_MAX_PORT_NAME_LEN
];
4223 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4225 ds_put_format(&ds
, "%5s %4d "ETH_ADDR_FMT
" %3d\n",
4226 name
, e
->vlan
, ETH_ADDR_ARGS(e
->mac
),
4227 mac_entry_age(ofproto
->ml
, e
));
4229 ovs_rwlock_unlock(&ofproto
->ml
->rwlock
);
4230 unixctl_command_reply(conn
, ds_cstr(&ds
));
4235 ofproto_unixctl_mcast_snooping_show(struct unixctl_conn
*conn
,
4236 int argc OVS_UNUSED
,
4238 void *aux OVS_UNUSED
)
4240 struct ds ds
= DS_EMPTY_INITIALIZER
;
4241 const struct ofproto_dpif
*ofproto
;
4242 const struct ofbundle
*bundle
;
4243 const struct mcast_group
*grp
;
4244 struct mcast_group_bundle
*b
;
4245 struct mcast_mrouter_bundle
*mrouter
;
4247 ofproto
= ofproto_dpif_lookup(argv
[1]);
4249 unixctl_command_reply_error(conn
, "no such bridge");
4253 if (!mcast_snooping_enabled(ofproto
->ms
)) {
4254 unixctl_command_reply_error(conn
, "multicast snooping is disabled");
4258 ds_put_cstr(&ds
, " port VLAN GROUP Age\n");
4259 ovs_rwlock_rdlock(&ofproto
->ms
->rwlock
);
4260 LIST_FOR_EACH (grp
, group_node
, &ofproto
->ms
->group_lru
) {
4261 LIST_FOR_EACH(b
, bundle_node
, &grp
->bundle_lru
) {
4262 char name
[OFP_MAX_PORT_NAME_LEN
];
4265 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4267 ds_put_format(&ds
, "%5s %4d "IP_FMT
" %3d\n",
4268 name
, grp
->vlan
, IP_ARGS(grp
->ip4
),
4269 mcast_bundle_age(ofproto
->ms
, b
));
4273 /* ports connected to multicast routers */
4274 LIST_FOR_EACH(mrouter
, mrouter_node
, &ofproto
->ms
->mrouter_lru
) {
4275 char name
[OFP_MAX_PORT_NAME_LEN
];
4277 bundle
= mrouter
->port
;
4278 ofputil_port_to_string(ofbundle_get_a_port(bundle
)->up
.ofp_port
,
4280 ds_put_format(&ds
, "%5s %4d querier %3d\n",
4281 name
, mrouter
->vlan
,
4282 mcast_mrouter_age(ofproto
->ms
, mrouter
));
4284 ovs_rwlock_unlock(&ofproto
->ms
->rwlock
);
4285 unixctl_command_reply(conn
, ds_cstr(&ds
));
4290 struct xlate_out xout
;
4291 struct xlate_in xin
;
4292 const struct flow
*key
;
4294 struct flow_wildcards wc
;
4299 trace_format_rule(struct ds
*result
, int level
, const struct rule_dpif
*rule
)
4301 const struct rule_actions
*actions
;
4304 ds_put_char_multiple(result
, '\t', level
);
4306 ds_put_cstr(result
, "No match\n");
4310 ovs_mutex_lock(&rule
->up
.mutex
);
4311 cookie
= rule
->up
.flow_cookie
;
4312 ovs_mutex_unlock(&rule
->up
.mutex
);
4314 ds_put_format(result
, "Rule: table=%"PRIu8
" cookie=%#"PRIx64
" ",
4315 rule
? rule
->up
.table_id
: 0, ntohll(cookie
));
4316 cls_rule_format(&rule
->up
.cr
, result
);
4317 ds_put_char(result
, '\n');
4319 actions
= rule_dpif_get_actions(rule
);
4321 ds_put_char_multiple(result
, '\t', level
);
4322 ds_put_cstr(result
, "OpenFlow actions=");
4323 ofpacts_format(actions
->ofpacts
, actions
->ofpacts_len
, result
);
4324 ds_put_char(result
, '\n');
4328 trace_format_flow(struct ds
*result
, int level
, const char *title
,
4329 struct trace_ctx
*trace
)
4331 ds_put_char_multiple(result
, '\t', level
);
4332 ds_put_format(result
, "%s: ", title
);
4333 /* Do not report unchanged flows for resubmits. */
4334 if ((level
> 0 && flow_equal(&trace
->xin
.flow
, &trace
->flow
))
4335 || (level
== 0 && flow_equal(&trace
->xin
.flow
, trace
->key
))) {
4336 ds_put_cstr(result
, "unchanged");
4338 flow_format(result
, &trace
->xin
.flow
);
4339 trace
->flow
= trace
->xin
.flow
;
4341 ds_put_char(result
, '\n');
4345 trace_format_regs(struct ds
*result
, int level
, const char *title
,
4346 struct trace_ctx
*trace
)
4350 ds_put_char_multiple(result
, '\t', level
);
4351 ds_put_format(result
, "%s:", title
);
4352 for (i
= 0; i
< FLOW_N_REGS
; i
++) {
4353 ds_put_format(result
, " reg%"PRIuSIZE
"=0x%"PRIx32
, i
, trace
->flow
.regs
[i
]);
4355 ds_put_char(result
, '\n');
4359 trace_format_odp(struct ds
*result
, int level
, const char *title
,
4360 struct trace_ctx
*trace
)
4362 struct ofpbuf
*odp_actions
= trace
->xout
.odp_actions
;
4364 ds_put_char_multiple(result
, '\t', level
);
4365 ds_put_format(result
, "%s: ", title
);
4366 format_odp_actions(result
, ofpbuf_data(odp_actions
),
4367 ofpbuf_size(odp_actions
));
4368 ds_put_char(result
, '\n');
4372 trace_format_megaflow(struct ds
*result
, int level
, const char *title
,
4373 struct trace_ctx
*trace
)
4377 ds_put_char_multiple(result
, '\t', level
);
4378 ds_put_format(result
, "%s: ", title
);
4379 flow_wildcards_or(&trace
->wc
, &trace
->xout
.wc
, &trace
->wc
);
4380 match_init(&match
, trace
->key
, &trace
->wc
);
4381 match_format(&match
, result
, OFP_DEFAULT_PRIORITY
);
4382 ds_put_char(result
, '\n');
4385 static void trace_report(struct xlate_in
*xin
, const char *s
, int recurse
);
4388 trace_resubmit(struct xlate_in
*xin
, struct rule_dpif
*rule
, int recurse
)
4390 struct trace_ctx
*trace
= CONTAINER_OF(xin
, struct trace_ctx
, xin
);
4391 struct ds
*result
= trace
->result
;
4394 if (rule
== xin
->ofproto
->miss_rule
) {
4395 trace_report(xin
, "No match, flow generates \"packet in\"s.",
4397 } else if (rule
== xin
->ofproto
->no_packet_in_rule
) {
4398 trace_report(xin
, "No match, packets dropped because "
4399 "OFPPC_NO_PACKET_IN is set on in_port.", recurse
);
4400 } else if (rule
== xin
->ofproto
->drop_frags_rule
) {
4401 trace_report(xin
, "Packets dropped because they are IP "
4402 "fragments and the fragment handling mode is "
4403 "\"drop\".", recurse
);
4407 ds_put_char(result
, '\n');
4409 trace_format_flow(result
, recurse
, "Resubmitted flow", trace
);
4410 trace_format_regs(result
, recurse
, "Resubmitted regs", trace
);
4411 trace_format_odp(result
, recurse
, "Resubmitted odp", trace
);
4412 trace_format_megaflow(result
, recurse
, "Resubmitted megaflow", trace
);
4414 trace_format_rule(result
, recurse
, rule
);
4418 trace_report(struct xlate_in
*xin
, const char *s
, int recurse
)
4420 struct trace_ctx
*trace
= CONTAINER_OF(xin
, struct trace_ctx
, xin
);
4421 struct ds
*result
= trace
->result
;
4423 ds_put_char_multiple(result
, '\t', recurse
);
4424 ds_put_cstr(result
, s
);
4425 ds_put_char(result
, '\n');
4428 /* Parses the 'argc' elements of 'argv', ignoring argv[0]. The following
4429 * forms are supported:
4431 * - [dpname] odp_flow [-generate | packet]
4432 * - bridge br_flow [-generate | packet]
4434 * On success, initializes '*ofprotop' and 'flow' and returns NULL. On failure
4435 * returns a nonnull malloced error message. */
4436 static char * WARN_UNUSED_RESULT
4437 parse_flow_and_packet(int argc
, const char *argv
[],
4438 struct ofproto_dpif
**ofprotop
, struct flow
*flow
,
4439 struct ofpbuf
**packetp
)
4441 const struct dpif_backer
*backer
= NULL
;
4442 const char *error
= NULL
;
4444 struct simap port_names
= SIMAP_INITIALIZER(&port_names
);
4445 struct ofpbuf
*packet
;
4446 struct ofpbuf odp_key
;
4447 struct ofpbuf odp_mask
;
4449 ofpbuf_init(&odp_key
, 0);
4450 ofpbuf_init(&odp_mask
, 0);
4452 /* Handle "-generate" or a hex string as the last argument. */
4453 if (!strcmp(argv
[argc
- 1], "-generate")) {
4454 packet
= ofpbuf_new(0);
4457 error
= eth_from_hex(argv
[argc
- 1], &packet
);
4460 } else if (argc
== 4) {
4461 /* The 3-argument form must end in "-generate' or a hex string. */
4467 /* odp_flow can have its in_port specified as a name instead of port no.
4468 * We do not yet know whether a given flow is a odp_flow or a br_flow.
4469 * But, to know whether a flow is odp_flow through odp_flow_from_string(),
4470 * we need to create a simap of name to port no. */
4472 const char *dp_type
;
4473 if (!strncmp(argv
[1], "ovs-", 4)) {
4474 dp_type
= argv
[1] + 4;
4478 backer
= shash_find_data(&all_dpif_backers
, dp_type
);
4479 } else if (argc
== 2) {
4480 struct shash_node
*node
;
4481 if (shash_count(&all_dpif_backers
) == 1) {
4482 node
= shash_first(&all_dpif_backers
);
4483 backer
= node
->data
;
4486 error
= "Syntax error";
4489 if (backer
&& backer
->dpif
) {
4490 struct dpif_port dpif_port
;
4491 struct dpif_port_dump port_dump
;
4492 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, backer
->dpif
) {
4493 simap_put(&port_names
, dpif_port
.name
,
4494 odp_to_u32(dpif_port
.port_no
));
4498 /* Parse the flow and determine whether a datapath or
4499 * bridge is specified. If function odp_flow_key_from_string()
4500 * returns 0, the flow is a odp_flow. If function
4501 * parse_ofp_exact_flow() returns NULL, the flow is a br_flow. */
4502 if (!odp_flow_from_string(argv
[argc
- 1], &port_names
,
4503 &odp_key
, &odp_mask
)) {
4505 error
= "Cannot find the datapath";
4509 if (odp_flow_key_to_flow(ofpbuf_data(&odp_key
), ofpbuf_size(&odp_key
),
4510 flow
) == ODP_FIT_ERROR
) {
4511 error
= "Failed to parse flow key";
4515 *ofprotop
= xlate_lookup_ofproto(backer
, flow
,
4516 &flow
->in_port
.ofp_port
);
4517 if (*ofprotop
== NULL
) {
4518 error
= "Invalid datapath flow";
4522 vsp_adjust_flow(*ofprotop
, flow
, NULL
);
4525 char *err
= parse_ofp_exact_flow(flow
, NULL
, argv
[argc
- 1], NULL
);
4528 m_err
= xasprintf("Bad flow syntax: %s", err
);
4533 error
= "Must specify bridge name";
4537 *ofprotop
= ofproto_dpif_lookup(argv
[1]);
4539 error
= "Unknown bridge name";
4545 /* Generate a packet, if requested. */
4547 if (!ofpbuf_size(packet
)) {
4548 flow_compose(packet
, flow
);
4550 struct pkt_metadata md
= pkt_metadata_from_flow(flow
);
4552 /* Use the metadata from the flow and the packet argument
4553 * to reconstruct the flow. */
4554 flow_extract(packet
, &md
, flow
);
4559 if (error
&& !m_err
) {
4560 m_err
= xstrdup(error
);
4563 ofpbuf_delete(packet
);
4567 ofpbuf_uninit(&odp_key
);
4568 ofpbuf_uninit(&odp_mask
);
4569 simap_destroy(&port_names
);
4574 ofproto_unixctl_trace(struct unixctl_conn
*conn
, int argc
, const char *argv
[],
4575 void *aux OVS_UNUSED
)
4577 struct ofproto_dpif
*ofproto
;
4578 struct ofpbuf
*packet
;
4582 error
= parse_flow_and_packet(argc
, argv
, &ofproto
, &flow
, &packet
);
4587 ofproto_trace(ofproto
, &flow
, packet
, NULL
, 0, &result
);
4588 unixctl_command_reply(conn
, ds_cstr(&result
));
4589 ds_destroy(&result
);
4590 ofpbuf_delete(packet
);
4592 unixctl_command_reply_error(conn
, error
);
4598 ofproto_unixctl_trace_actions(struct unixctl_conn
*conn
, int argc
,
4599 const char *argv
[], void *aux OVS_UNUSED
)
4601 enum ofputil_protocol usable_protocols
;
4602 struct ofproto_dpif
*ofproto
;
4603 bool enforce_consistency
;
4604 struct ofpbuf ofpacts
;
4605 struct ofpbuf
*packet
;
4610 /* Three kinds of error return values! */
4616 ofpbuf_init(&ofpacts
, 0);
4618 /* Parse actions. */
4619 error
= ofpacts_parse_actions(argv
[--argc
], &ofpacts
, &usable_protocols
);
4621 unixctl_command_reply_error(conn
, error
);
4626 /* OpenFlow 1.1 and later suggest that the switch enforces certain forms of
4627 * consistency between the flow and the actions. With -consistent, we
4628 * enforce consistency even for a flow supported in OpenFlow 1.0. */
4629 if (!strcmp(argv
[1], "-consistent")) {
4630 enforce_consistency
= true;
4634 enforce_consistency
= false;
4637 error
= parse_flow_and_packet(argc
, argv
, &ofproto
, &flow
, &packet
);
4639 unixctl_command_reply_error(conn
, error
);
4644 /* Do the same checks as handle_packet_out() in ofproto.c.
4646 * We pass a 'table_id' of 0 to ofproto_check_ofpacts(), which isn't
4647 * strictly correct because these actions aren't in any table, but it's OK
4648 * because it 'table_id' is used only to check goto_table instructions, but
4649 * packet-outs take a list of actions and therefore it can't include
4652 * We skip the "meter" check here because meter is an instruction, not an
4653 * action, and thus cannot appear in ofpacts. */
4654 in_port
= ofp_to_u16(flow
.in_port
.ofp_port
);
4655 if (in_port
>= ofproto
->up
.max_ports
&& in_port
< ofp_to_u16(OFPP_MAX
)) {
4656 unixctl_command_reply_error(conn
, "invalid in_port");
4659 if (enforce_consistency
) {
4660 retval
= ofpacts_check_consistency(ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
),
4661 &flow
, u16_to_ofp(ofproto
->up
.max_ports
),
4662 0, 0, usable_protocols
);
4664 retval
= ofpacts_check(ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
), &flow
,
4665 u16_to_ofp(ofproto
->up
.max_ports
), 0, 0,
4671 ds_put_format(&result
, "Bad actions: %s", ofperr_to_string(retval
));
4672 unixctl_command_reply_error(conn
, ds_cstr(&result
));
4676 ofproto_trace(ofproto
, &flow
, packet
,
4677 ofpbuf_data(&ofpacts
), ofpbuf_size(&ofpacts
), &result
);
4678 unixctl_command_reply(conn
, ds_cstr(&result
));
4681 ds_destroy(&result
);
4682 ofpbuf_delete(packet
);
4683 ofpbuf_uninit(&ofpacts
);
4686 /* Implements a "trace" through 'ofproto''s flow table, appending a textual
4687 * description of the results to 'ds'.
4689 * The trace follows a packet with the specified 'flow' through the flow
4690 * table. 'packet' may be nonnull to trace an actual packet, with consequent
4691 * side effects (if it is nonnull then its flow must be 'flow').
4693 * If 'ofpacts' is nonnull then its 'ofpacts_len' bytes specify the actions to
4694 * trace, otherwise the actions are determined by a flow table lookup. */
4696 ofproto_trace(struct ofproto_dpif
*ofproto
, struct flow
*flow
,
4697 const struct ofpbuf
*packet
,
4698 const struct ofpact ofpacts
[], size_t ofpacts_len
,
4701 struct trace_ctx trace
;
4703 ds_put_format(ds
, "Bridge: %s\n", ofproto
->up
.name
);
4704 ds_put_cstr(ds
, "Flow: ");
4705 flow_format(ds
, flow
);
4706 ds_put_char(ds
, '\n');
4708 flow_wildcards_init_catchall(&trace
.wc
);
4711 trace
.key
= flow
; /* Original flow key, used for megaflow. */
4712 trace
.flow
= *flow
; /* May be modified by actions. */
4713 xlate_in_init(&trace
.xin
, ofproto
, flow
, flow
->in_port
.ofp_port
, NULL
,
4714 ntohs(flow
->tcp_flags
), packet
);
4715 trace
.xin
.ofpacts
= ofpacts
;
4716 trace
.xin
.ofpacts_len
= ofpacts_len
;
4717 trace
.xin
.resubmit_hook
= trace_resubmit
;
4718 trace
.xin
.report_hook
= trace_report
;
4720 xlate_actions(&trace
.xin
, &trace
.xout
);
4722 ds_put_char(ds
, '\n');
4723 trace_format_flow(ds
, 0, "Final flow", &trace
);
4724 trace_format_megaflow(ds
, 0, "Megaflow", &trace
);
4726 ds_put_cstr(ds
, "Datapath actions: ");
4727 format_odp_actions(ds
, ofpbuf_data(trace
.xout
.odp_actions
),
4728 ofpbuf_size(trace
.xout
.odp_actions
));
4730 if (trace
.xout
.slow
) {
4731 enum slow_path_reason slow
;
4733 ds_put_cstr(ds
, "\nThis flow is handled by the userspace "
4734 "slow path because it:");
4736 slow
= trace
.xout
.slow
;
4738 enum slow_path_reason bit
= rightmost_1bit(slow
);
4740 ds_put_format(ds
, "\n\t- %s.",
4741 slow_path_reason_to_explanation(bit
));
4747 xlate_out_uninit(&trace
.xout
);
4750 /* Store the current ofprotos in 'ofproto_shash'. Returns a sorted list
4751 * of the 'ofproto_shash' nodes. It is the responsibility of the caller
4752 * to destroy 'ofproto_shash' and free the returned value. */
4753 static const struct shash_node
**
4754 get_ofprotos(struct shash
*ofproto_shash
)
4756 const struct ofproto_dpif
*ofproto
;
4758 HMAP_FOR_EACH (ofproto
, all_ofproto_dpifs_node
, &all_ofproto_dpifs
) {
4759 char *name
= xasprintf("%s@%s", ofproto
->up
.type
, ofproto
->up
.name
);
4760 shash_add_nocopy(ofproto_shash
, name
, ofproto
);
4763 return shash_sort(ofproto_shash
);
4767 ofproto_unixctl_dpif_dump_dps(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4768 const char *argv
[] OVS_UNUSED
,
4769 void *aux OVS_UNUSED
)
4771 struct ds ds
= DS_EMPTY_INITIALIZER
;
4772 struct shash ofproto_shash
;
4773 const struct shash_node
**sorted_ofprotos
;
4776 shash_init(&ofproto_shash
);
4777 sorted_ofprotos
= get_ofprotos(&ofproto_shash
);
4778 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4779 const struct shash_node
*node
= sorted_ofprotos
[i
];
4780 ds_put_format(&ds
, "%s\n", node
->name
);
4783 shash_destroy(&ofproto_shash
);
4784 free(sorted_ofprotos
);
4786 unixctl_command_reply(conn
, ds_cstr(&ds
));
4791 dpif_show_backer(const struct dpif_backer
*backer
, struct ds
*ds
)
4793 const struct shash_node
**ofprotos
;
4794 struct dpif_dp_stats dp_stats
;
4795 struct shash ofproto_shash
;
4798 dpif_get_dp_stats(backer
->dpif
, &dp_stats
);
4800 ds_put_format(ds
, "%s: hit:%"PRIu64
" missed:%"PRIu64
"\n",
4801 dpif_name(backer
->dpif
), dp_stats
.n_hit
, dp_stats
.n_missed
);
4803 shash_init(&ofproto_shash
);
4804 ofprotos
= get_ofprotos(&ofproto_shash
);
4805 for (i
= 0; i
< shash_count(&ofproto_shash
); i
++) {
4806 struct ofproto_dpif
*ofproto
= ofprotos
[i
]->data
;
4807 const struct shash_node
**ports
;
4810 if (ofproto
->backer
!= backer
) {
4814 ds_put_format(ds
, "\t%s:\n", ofproto
->up
.name
);
4816 ports
= shash_sort(&ofproto
->up
.port_by_name
);
4817 for (j
= 0; j
< shash_count(&ofproto
->up
.port_by_name
); j
++) {
4818 const struct shash_node
*node
= ports
[j
];
4819 struct ofport
*ofport
= node
->data
;
4821 odp_port_t odp_port
;
4823 ds_put_format(ds
, "\t\t%s %u/", netdev_get_name(ofport
->netdev
),
4826 odp_port
= ofp_port_to_odp_port(ofproto
, ofport
->ofp_port
);
4827 if (odp_port
!= ODPP_NONE
) {
4828 ds_put_format(ds
, "%"PRIu32
":", odp_port
);
4830 ds_put_cstr(ds
, "none:");
4833 ds_put_format(ds
, " (%s", netdev_get_type(ofport
->netdev
));
4836 if (!netdev_get_config(ofport
->netdev
, &config
)) {
4837 const struct smap_node
**nodes
;
4840 nodes
= smap_sort(&config
);
4841 for (i
= 0; i
< smap_count(&config
); i
++) {
4842 const struct smap_node
*node
= nodes
[i
];
4843 ds_put_format(ds
, "%c %s=%s", i
? ',' : ':',
4844 node
->key
, node
->value
);
4848 smap_destroy(&config
);
4850 ds_put_char(ds
, ')');
4851 ds_put_char(ds
, '\n');
4855 shash_destroy(&ofproto_shash
);
4860 ofproto_unixctl_dpif_show(struct unixctl_conn
*conn
, int argc OVS_UNUSED
,
4861 const char *argv
[] OVS_UNUSED
, void *aux OVS_UNUSED
)
4863 struct ds ds
= DS_EMPTY_INITIALIZER
;
4864 const struct shash_node
**backers
;
4867 backers
= shash_sort(&all_dpif_backers
);
4868 for (i
= 0; i
< shash_count(&all_dpif_backers
); i
++) {
4869 dpif_show_backer(backers
[i
]->data
, &ds
);
4873 unixctl_command_reply(conn
, ds_cstr(&ds
));
4878 ofproto_unixctl_dpif_dump_flows(struct unixctl_conn
*conn
,
4879 int argc OVS_UNUSED
, const char *argv
[],
4880 void *aux OVS_UNUSED
)
4882 const struct ofproto_dpif
*ofproto
;
4884 struct ds ds
= DS_EMPTY_INITIALIZER
;
4885 bool verbosity
= false;
4887 struct dpif_port dpif_port
;
4888 struct dpif_port_dump port_dump
;
4889 struct hmap portno_names
;
4891 struct dpif_flow_dump
*flow_dump
;
4892 struct dpif_flow_dump_thread
*flow_dump_thread
;
4896 ofproto
= ofproto_dpif_lookup(argv
[argc
- 1]);
4898 unixctl_command_reply_error(conn
, "no such bridge");
4902 if (argc
> 2 && !strcmp(argv
[1], "-m")) {
4906 hmap_init(&portno_names
);
4907 DPIF_PORT_FOR_EACH (&dpif_port
, &port_dump
, ofproto
->backer
->dpif
) {
4908 odp_portno_names_set(&portno_names
, dpif_port
.port_no
, dpif_port
.name
);
4912 flow_dump
= dpif_flow_dump_create(ofproto
->backer
->dpif
);
4913 flow_dump_thread
= dpif_flow_dump_thread_create(flow_dump
);
4914 while (dpif_flow_dump_next(flow_dump_thread
, &f
, 1)) {
4917 if (odp_flow_key_to_flow(f
.key
, f
.key_len
, &flow
) == ODP_FIT_ERROR
4918 || xlate_lookup_ofproto(ofproto
->backer
, &flow
, NULL
) != ofproto
) {
4922 odp_flow_format(f
.key
, f
.key_len
, f
.mask
, f
.mask_len
,
4923 &portno_names
, &ds
, verbosity
);
4924 ds_put_cstr(&ds
, ", ");
4925 dpif_flow_stats_format(&f
.stats
, &ds
);
4926 ds_put_cstr(&ds
, ", actions:");
4927 format_odp_actions(&ds
, f
.actions
, f
.actions_len
);
4928 ds_put_char(&ds
, '\n');
4930 dpif_flow_dump_thread_destroy(flow_dump_thread
);
4931 error
= dpif_flow_dump_destroy(flow_dump
);
4935 ds_put_format(&ds
, "dpif/dump_flows failed: %s", ovs_strerror(errno
));
4936 unixctl_command_reply_error(conn
, ds_cstr(&ds
));
4938 unixctl_command_reply(conn
, ds_cstr(&ds
));
4940 odp_portno_names_destroy(&portno_names
);
4941 hmap_destroy(&portno_names
);
4946 ofproto_dpif_unixctl_init(void)
4948 static bool registered
;
4954 unixctl_command_register(
4956 "{[dp_name] odp_flow | bridge br_flow} [-generate|packet]",
4957 1, 3, ofproto_unixctl_trace
, NULL
);
4958 unixctl_command_register(
4959 "ofproto/trace-packet-out",
4960 "[-consistent] {[dp_name] odp_flow | bridge br_flow} [-generate|packet] actions",
4961 2, 6, ofproto_unixctl_trace_actions
, NULL
);
4962 unixctl_command_register("fdb/flush", "[bridge]", 0, 1,
4963 ofproto_unixctl_fdb_flush
, NULL
);
4964 unixctl_command_register("fdb/show", "bridge", 1, 1,
4965 ofproto_unixctl_fdb_show
, NULL
);
4966 unixctl_command_register("mdb/flush", "[bridge]", 0, 1,
4967 ofproto_unixctl_mcast_snooping_flush
, NULL
);
4968 unixctl_command_register("mdb/show", "bridge", 1, 1,
4969 ofproto_unixctl_mcast_snooping_show
, NULL
);
4970 unixctl_command_register("dpif/dump-dps", "", 0, 0,
4971 ofproto_unixctl_dpif_dump_dps
, NULL
);
4972 unixctl_command_register("dpif/show", "", 0, 0, ofproto_unixctl_dpif_show
,
4974 unixctl_command_register("dpif/dump-flows", "[-m] bridge", 1, 2,
4975 ofproto_unixctl_dpif_dump_flows
, NULL
);
4978 /* Returns true if 'table' is the table used for internal rules,
4979 * false otherwise. */
4981 table_is_internal(uint8_t table_id
)
4983 return table_id
== TBL_INTERNAL
;
4986 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
4988 * This is deprecated. It is only for compatibility with broken device drivers
4989 * in old versions of Linux that do not properly support VLANs when VLAN
4990 * devices are not used. When broken device drivers are no longer in
4991 * widespread use, we will delete these interfaces. */
4994 set_realdev(struct ofport
*ofport_
, ofp_port_t realdev_ofp_port
, int vid
)
4996 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(ofport_
->ofproto
);
4997 struct ofport_dpif
*ofport
= ofport_dpif_cast(ofport_
);
4999 if (realdev_ofp_port
== ofport
->realdev_ofp_port
5000 && vid
== ofport
->vlandev_vid
) {
5004 ofproto
->backer
->need_revalidate
= REV_RECONFIGURE
;
5006 if (ofport
->realdev_ofp_port
) {
5009 if (realdev_ofp_port
&& ofport
->bundle
) {
5010 /* vlandevs are enslaved to their realdevs, so they are not allowed to
5011 * themselves be part of a bundle. */
5012 bundle_set(ofport_
->ofproto
, ofport
->bundle
, NULL
);
5015 ofport
->realdev_ofp_port
= realdev_ofp_port
;
5016 ofport
->vlandev_vid
= vid
;
5018 if (realdev_ofp_port
) {
5019 vsp_add(ofport
, realdev_ofp_port
, vid
);
5026 hash_realdev_vid(ofp_port_t realdev_ofp_port
, int vid
)
5028 return hash_2words(ofp_to_u16(realdev_ofp_port
), vid
);
5032 ofproto_has_vlan_splinters(const struct ofproto_dpif
*ofproto
)
5033 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5035 /* hmap_is_empty is thread safe. */
5036 return !hmap_is_empty(&ofproto
->realdev_vid_map
);
5041 vsp_realdev_to_vlandev__(const struct ofproto_dpif
*ofproto
,
5042 ofp_port_t realdev_ofp_port
, ovs_be16 vlan_tci
)
5043 OVS_REQUIRES(ofproto
->vsp_mutex
)
5045 if (!hmap_is_empty(&ofproto
->realdev_vid_map
)) {
5046 int vid
= vlan_tci_to_vid(vlan_tci
);
5047 const struct vlan_splinter
*vsp
;
5049 HMAP_FOR_EACH_WITH_HASH (vsp
, realdev_vid_node
,
5050 hash_realdev_vid(realdev_ofp_port
, vid
),
5051 &ofproto
->realdev_vid_map
) {
5052 if (vsp
->realdev_ofp_port
== realdev_ofp_port
5053 && vsp
->vid
== vid
) {
5054 return vsp
->vlandev_ofp_port
;
5058 return realdev_ofp_port
;
5061 /* Returns the OFP port number of the Linux VLAN device that corresponds to
5062 * 'vlan_tci' on the network device with port number 'realdev_ofp_port' in
5063 * 'struct ofport_dpif'. For example, given 'realdev_ofp_port' of eth0 and
5064 * 'vlan_tci' 9, it would return the port number of eth0.9.
5066 * Unless VLAN splinters are enabled for port 'realdev_ofp_port', this
5067 * function just returns its 'realdev_ofp_port' argument. */
5069 vsp_realdev_to_vlandev(const struct ofproto_dpif
*ofproto
,
5070 ofp_port_t realdev_ofp_port
, ovs_be16 vlan_tci
)
5071 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5075 /* hmap_is_empty is thread safe, see if we can return immediately. */
5076 if (hmap_is_empty(&ofproto
->realdev_vid_map
)) {
5077 return realdev_ofp_port
;
5079 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5080 ret
= vsp_realdev_to_vlandev__(ofproto
, realdev_ofp_port
, vlan_tci
);
5081 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5085 static struct vlan_splinter
*
5086 vlandev_find(const struct ofproto_dpif
*ofproto
, ofp_port_t vlandev_ofp_port
)
5088 struct vlan_splinter
*vsp
;
5090 HMAP_FOR_EACH_WITH_HASH (vsp
, vlandev_node
,
5091 hash_ofp_port(vlandev_ofp_port
),
5092 &ofproto
->vlandev_map
) {
5093 if (vsp
->vlandev_ofp_port
== vlandev_ofp_port
) {
5101 /* Returns the OpenFlow port number of the "real" device underlying the Linux
5102 * VLAN device with OpenFlow port number 'vlandev_ofp_port' and stores the
5103 * VLAN VID of the Linux VLAN device in '*vid'. For example, given
5104 * 'vlandev_ofp_port' of eth0.9, it would return the OpenFlow port number of
5105 * eth0 and store 9 in '*vid'.
5107 * Returns 0 and does not modify '*vid' if 'vlandev_ofp_port' is not a Linux
5108 * VLAN device. Unless VLAN splinters are enabled, this is what this function
5111 vsp_vlandev_to_realdev(const struct ofproto_dpif
*ofproto
,
5112 ofp_port_t vlandev_ofp_port
, int *vid
)
5113 OVS_REQUIRES(ofproto
->vsp_mutex
)
5115 if (!hmap_is_empty(&ofproto
->vlandev_map
)) {
5116 const struct vlan_splinter
*vsp
;
5118 vsp
= vlandev_find(ofproto
, vlandev_ofp_port
);
5123 return vsp
->realdev_ofp_port
;
5129 /* Given 'flow', a flow representing a packet received on 'ofproto', checks
5130 * whether 'flow->in_port' represents a Linux VLAN device. If so, changes
5131 * 'flow->in_port' to the "real" device backing the VLAN device, sets
5132 * 'flow->vlan_tci' to the VLAN VID, and returns true. Optionally pushes the
5133 * appropriate VLAN on 'packet' if provided. Otherwise (which is always the
5134 * case unless VLAN splinters are enabled), returns false without making any
5137 vsp_adjust_flow(const struct ofproto_dpif
*ofproto
, struct flow
*flow
,
5138 struct ofpbuf
*packet
)
5139 OVS_EXCLUDED(ofproto
->vsp_mutex
)
5144 /* hmap_is_empty is thread safe. */
5145 if (hmap_is_empty(&ofproto
->vlandev_map
)) {
5149 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5150 realdev
= vsp_vlandev_to_realdev(ofproto
, flow
->in_port
.ofp_port
, &vid
);
5151 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5156 /* Cause the flow to be processed as if it came in on the real device with
5157 * the VLAN device's VLAN ID. */
5158 flow
->in_port
.ofp_port
= realdev
;
5159 flow
->vlan_tci
= htons((vid
& VLAN_VID_MASK
) | VLAN_CFI
);
5162 /* Make the packet resemble the flow, so that it gets sent to an
5163 * OpenFlow controller properly, so that it looks correct for sFlow,
5164 * and so that flow_extract() will get the correct vlan_tci if it is
5165 * called on 'packet'. */
5166 eth_push_vlan(packet
, htons(ETH_TYPE_VLAN
), flow
->vlan_tci
);
5173 vsp_remove(struct ofport_dpif
*port
)
5175 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
5176 struct vlan_splinter
*vsp
;
5178 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5179 vsp
= vlandev_find(ofproto
, port
->up
.ofp_port
);
5181 hmap_remove(&ofproto
->vlandev_map
, &vsp
->vlandev_node
);
5182 hmap_remove(&ofproto
->realdev_vid_map
, &vsp
->realdev_vid_node
);
5185 port
->realdev_ofp_port
= 0;
5187 VLOG_ERR("missing vlan device record");
5189 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5193 vsp_add(struct ofport_dpif
*port
, ofp_port_t realdev_ofp_port
, int vid
)
5195 struct ofproto_dpif
*ofproto
= ofproto_dpif_cast(port
->up
.ofproto
);
5197 ovs_mutex_lock(&ofproto
->vsp_mutex
);
5198 if (!vsp_vlandev_to_realdev(ofproto
, port
->up
.ofp_port
, NULL
)
5199 && (vsp_realdev_to_vlandev__(ofproto
, realdev_ofp_port
, htons(vid
))
5200 == realdev_ofp_port
)) {
5201 struct vlan_splinter
*vsp
;
5203 vsp
= xmalloc(sizeof *vsp
);
5204 vsp
->realdev_ofp_port
= realdev_ofp_port
;
5205 vsp
->vlandev_ofp_port
= port
->up
.ofp_port
;
5208 port
->realdev_ofp_port
= realdev_ofp_port
;
5210 hmap_insert(&ofproto
->vlandev_map
, &vsp
->vlandev_node
,
5211 hash_ofp_port(port
->up
.ofp_port
));
5212 hmap_insert(&ofproto
->realdev_vid_map
, &vsp
->realdev_vid_node
,
5213 hash_realdev_vid(realdev_ofp_port
, vid
));
5215 VLOG_ERR("duplicate vlan device record");
5217 ovs_mutex_unlock(&ofproto
->vsp_mutex
);
5221 ofp_port_to_odp_port(const struct ofproto_dpif
*ofproto
, ofp_port_t ofp_port
)
5223 const struct ofport_dpif
*ofport
= get_ofp_port(ofproto
, ofp_port
);
5224 return ofport
? ofport
->odp_port
: ODPP_NONE
;
5227 struct ofport_dpif
*
5228 odp_port_to_ofport(const struct dpif_backer
*backer
, odp_port_t odp_port
)
5230 struct ofport_dpif
*port
;
5232 ovs_rwlock_rdlock(&backer
->odp_to_ofport_lock
);
5233 HMAP_FOR_EACH_IN_BUCKET (port
, odp_port_node
, hash_odp_port(odp_port
),
5234 &backer
->odp_to_ofport_map
) {
5235 if (port
->odp_port
== odp_port
) {
5236 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5241 ovs_rwlock_unlock(&backer
->odp_to_ofport_lock
);
5246 odp_port_to_ofp_port(const struct ofproto_dpif
*ofproto
, odp_port_t odp_port
)
5248 struct ofport_dpif
*port
;
5250 port
= odp_port_to_ofport(ofproto
->backer
, odp_port
);
5251 if (port
&& &ofproto
->up
== port
->up
.ofproto
) {
5252 return port
->up
.ofp_port
;
5259 ofproto_dpif_alloc_recirc_id(struct ofproto_dpif
*ofproto
)
5261 struct dpif_backer
*backer
= ofproto
->backer
;
5263 return recirc_id_alloc(backer
->rid_pool
);
5267 ofproto_dpif_free_recirc_id(struct ofproto_dpif
*ofproto
, uint32_t recirc_id
)
5269 struct dpif_backer
*backer
= ofproto
->backer
;
5271 recirc_id_free(backer
->rid_pool
, recirc_id
);
5275 ofproto_dpif_add_internal_flow(struct ofproto_dpif
*ofproto
,
5276 const struct match
*match
, int priority
,
5277 uint16_t idle_timeout
,
5278 const struct ofpbuf
*ofpacts
,
5279 struct rule
**rulep
)
5281 struct ofputil_flow_mod fm
;
5282 struct rule_dpif
*rule
;
5286 fm
.priority
= priority
;
5287 fm
.new_cookie
= htonll(0);
5288 fm
.cookie
= htonll(0);
5289 fm
.cookie_mask
= htonll(0);
5290 fm
.modify_cookie
= false;
5291 fm
.table_id
= TBL_INTERNAL
;
5292 fm
.command
= OFPFC_ADD
;
5293 fm
.idle_timeout
= idle_timeout
;
5294 fm
.hard_timeout
= 0;
5298 fm
.flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
;
5299 fm
.ofpacts
= ofpbuf_data(ofpacts
);
5300 fm
.ofpacts_len
= ofpbuf_size(ofpacts
);
5302 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5304 VLOG_ERR_RL(&rl
, "failed to add internal flow (%s)",
5305 ofperr_to_string(error
));
5310 rule
= rule_dpif_lookup_in_table(ofproto
, TBL_INTERNAL
, &fm
.match
.flow
,
5311 &fm
.match
.wc
, false);
5321 ofproto_dpif_delete_internal_flow(struct ofproto_dpif
*ofproto
,
5322 struct match
*match
, int priority
)
5324 struct ofputil_flow_mod fm
;
5328 fm
.priority
= priority
;
5329 fm
.new_cookie
= htonll(0);
5330 fm
.cookie
= htonll(0);
5331 fm
.cookie_mask
= htonll(0);
5332 fm
.modify_cookie
= false;
5333 fm
.table_id
= TBL_INTERNAL
;
5334 fm
.flags
= OFPUTIL_FF_HIDDEN_FIELDS
| OFPUTIL_FF_NO_READONLY
;
5335 fm
.command
= OFPFC_DELETE_STRICT
;
5337 error
= ofproto_flow_mod(&ofproto
->up
, &fm
);
5339 VLOG_ERR_RL(&rl
, "failed to delete internal flow (%s)",
5340 ofperr_to_string(error
));
5347 const struct ofproto_class ofproto_dpif_class
= {
5361 NULL
, /* get_memory_usage. */
5362 type_get_memory_usage
,
5380 port_is_lacp_current
,
5381 NULL
, /* rule_choose_table */
5390 NULL
, /* rule_premodify_actions */
5391 rule_modify_actions
,
5407 get_stp_port_status
,
5412 get_rstp_port_status
,
5419 is_mirror_output_bundle
,
5420 forward_bpdu_changed
,
5421 set_mac_table_config
,
5423 set_mcast_snooping_port
,
5425 NULL
, /* meter_get_features */
5426 NULL
, /* meter_set */
5427 NULL
, /* meter_get */
5428 NULL
, /* meter_del */
5429 group_alloc
, /* group_alloc */
5430 group_construct
, /* group_construct */
5431 group_destruct
, /* group_destruct */
5432 group_dealloc
, /* group_dealloc */
5433 group_modify
, /* group_modify */
5434 group_get_stats
, /* group_get_stats */
5435 get_datapath_version
, /* get_datapath_version */