1 /* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License. */
17 #include "ofproto/ofproto-dpif-xlate.h"
20 #include <arpa/inet.h>
22 #include <sys/socket.h>
23 #include <netinet/in.h>
25 #include "tnl-arp-cache.h"
30 #include "byte-order.h"
34 #include "dp-packet.h"
36 #include "dynamic-string.h"
42 #include "mac-learning.h"
43 #include "mcast-snooping.h"
44 #include "meta-flow.h"
45 #include "multipath.h"
46 #include "netdev-vport.h"
49 #include "odp-execute.h"
50 #include "ofp-actions.h"
51 #include "ofproto/ofproto-dpif-ipfix.h"
52 #include "ofproto/ofproto-dpif-mirror.h"
53 #include "ofproto/ofproto-dpif-monitor.h"
54 #include "ofproto/ofproto-dpif-sflow.h"
55 #include "ofproto/ofproto-dpif.h"
56 #include "ofproto/ofproto-provider.h"
57 #include "ovs-router.h"
58 #include "tnl-ports.h"
60 #include "openvswitch/vlog.h"
62 COVERAGE_DEFINE(xlate_actions
);
63 COVERAGE_DEFINE(xlate_actions_oversize
);
64 COVERAGE_DEFINE(xlate_actions_too_many_output
);
66 VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate
);
68 /* Maximum depth of flow table recursion (due to resubmit actions) in a
69 * flow translation. */
70 #define MAX_RESUBMIT_RECURSION 64
71 #define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in
74 /* Maximum number of resubmit actions in a flow translation, whether they are
75 * recursive or not. */
76 #define MAX_RESUBMITS (MAX_RESUBMIT_RECURSION * MAX_RESUBMIT_RECURSION)
79 struct hmap_node hmap_node
; /* Node in global 'xbridges' map. */
80 struct ofproto_dpif
*ofproto
; /* Key in global 'xbridges' map. */
82 struct ovs_list xbundles
; /* Owned xbundles. */
83 struct hmap xports
; /* Indexed by ofp_port. */
85 char *name
; /* Name used in log messages. */
86 struct dpif
*dpif
; /* Datapath interface. */
87 struct mac_learning
*ml
; /* Mac learning handle. */
88 struct mcast_snooping
*ms
; /* Multicast Snooping handle. */
89 struct mbridge
*mbridge
; /* Mirroring. */
90 struct dpif_sflow
*sflow
; /* SFlow handle, or null. */
91 struct dpif_ipfix
*ipfix
; /* Ipfix handle, or null. */
92 struct netflow
*netflow
; /* Netflow handle, or null. */
93 struct stp
*stp
; /* STP or null if disabled. */
94 struct rstp
*rstp
; /* RSTP or null if disabled. */
96 bool has_in_band
; /* Bridge has in band control? */
97 bool forward_bpdu
; /* Bridge forwards STP BPDUs? */
99 /* Datapath feature support. */
100 struct dpif_backer_support support
;
104 struct hmap_node hmap_node
; /* In global 'xbundles' map. */
105 struct ofbundle
*ofbundle
; /* Key in global 'xbundles' map. */
107 struct ovs_list list_node
; /* In parent 'xbridges' list. */
108 struct xbridge
*xbridge
; /* Parent xbridge. */
110 struct ovs_list xports
; /* Contains "struct xport"s. */
112 char *name
; /* Name used in log messages. */
113 struct bond
*bond
; /* Nonnull iff more than one port. */
114 struct lacp
*lacp
; /* LACP handle or null. */
116 enum port_vlan_mode vlan_mode
; /* VLAN mode. */
117 int vlan
; /* -1=trunk port, else a 12-bit VLAN ID. */
118 unsigned long *trunks
; /* Bitmap of trunked VLANs, if 'vlan' == -1.
119 * NULL if all VLANs are trunked. */
120 bool use_priority_tags
; /* Use 802.1p tag for frames in VLAN 0? */
121 bool floodable
; /* No port has OFPUTIL_PC_NO_FLOOD set? */
125 struct hmap_node hmap_node
; /* Node in global 'xports' map. */
126 struct ofport_dpif
*ofport
; /* Key in global 'xports map. */
128 struct hmap_node ofp_node
; /* Node in parent xbridge 'xports' map. */
129 ofp_port_t ofp_port
; /* Key in parent xbridge 'xports' map. */
131 odp_port_t odp_port
; /* Datapath port number or ODPP_NONE. */
133 struct ovs_list bundle_node
; /* In parent xbundle (if it exists). */
134 struct xbundle
*xbundle
; /* Parent xbundle or null. */
136 struct netdev
*netdev
; /* 'ofport''s netdev. */
138 struct xbridge
*xbridge
; /* Parent bridge. */
139 struct xport
*peer
; /* Patch port peer or null. */
141 enum ofputil_port_config config
; /* OpenFlow port configuration. */
142 enum ofputil_port_state state
; /* OpenFlow port state. */
143 int stp_port_no
; /* STP port number or -1 if not in use. */
144 struct rstp_port
*rstp_port
; /* RSTP port or null. */
146 struct hmap skb_priorities
; /* Map of 'skb_priority_to_dscp's. */
148 bool may_enable
; /* May be enabled in bonds. */
149 bool is_tunnel
; /* Is a tunnel port. */
151 struct cfm
*cfm
; /* CFM handle or null. */
152 struct bfd
*bfd
; /* BFD handle or null. */
153 struct lldp
*lldp
; /* LLDP handle or null. */
157 struct xlate_in
*xin
;
158 struct xlate_out
*xout
;
160 const struct xbridge
*xbridge
;
162 /* Flow tables version at the beginning of the translation. */
163 cls_version_t tables_version
;
165 /* Flow at the last commit. */
166 struct flow base_flow
;
168 /* Tunnel IP destination address as received. This is stored separately
169 * as the base_flow.tunnel is cleared on init to reflect the datapath
170 * behavior. Used to make sure not to send tunneled output to ourselves,
171 * which might lead to an infinite loop. This could happen easily
172 * if a tunnel is marked as 'ip_remote=flow', and the flow does not
173 * actually set the tun_dst field. */
174 ovs_be32 orig_tunnel_ip_dst
;
176 /* Stack for the push and pop actions. Each stack element is of type
177 * "union mf_subvalue". */
180 /* The rule that we are currently translating, or NULL. */
181 struct rule_dpif
*rule
;
183 /* Flow translation populates this with wildcards relevant in translation.
184 * When 'xin->wc' is nonnull, this is the same pointer. When 'xin->wc' is
185 * null, this is a pointer to uninitialized scratch memory. This allows
186 * code to blindly write to 'ctx->wc' without worrying about whether the
187 * caller really wants wildcards. */
188 struct flow_wildcards
*wc
;
190 /* Output buffer for datapath actions. When 'xin->odp_actions' is nonnull,
191 * this is the same pointer. When 'xin->odp_actions' is null, this points
192 * to a scratch ofpbuf. This allows code to add actions to
193 * 'ctx->odp_actions' without worrying about whether the caller really
195 struct ofpbuf
*odp_actions
;
197 /* Resubmit statistics, via xlate_table_action(). */
198 int recurse
; /* Current resubmit nesting depth. */
199 int resubmits
; /* Total number of resubmits. */
200 bool in_group
; /* Currently translating ofgroup, if true. */
201 bool in_action_set
; /* Currently translating action_set, if true. */
203 uint8_t table_id
; /* OpenFlow table ID where flow was found. */
204 ovs_be64 rule_cookie
; /* Cookie of the rule being translated. */
205 uint32_t orig_skb_priority
; /* Priority when packet arrived. */
206 uint32_t sflow_n_outputs
; /* Number of output ports. */
207 odp_port_t sflow_odp_port
; /* Output port for composing sFlow action. */
208 ofp_port_t nf_output_iface
; /* Output interface index for NetFlow. */
209 bool exit
; /* No further actions should be processed. */
210 mirror_mask_t mirrors
; /* Bitmap of associated mirrors. */
212 /* These are used for non-bond recirculation. The recirculation IDs are
213 * stored in xout and must be associated with a datapath flow (ukey),
214 * otherwise they will be freed when the xout is uninitialized.
217 * Steps in Recirculation Translation
218 * ==================================
220 * At some point during translation, the code recognizes the need for
221 * recirculation. For example, recirculation is necessary when, after
222 * popping the last MPLS label, an action or a match tries to examine or
223 * modify a field that has been newly revealed following the MPLS label.
225 * The simplest part of the work to be done is to commit existing changes to
226 * the packet, which produces datapath actions corresponding to the changes,
227 * and after this, add an OVS_ACTION_ATTR_RECIRC datapath action.
229 * The main problem here is preserving state. When the datapath executes
230 * OVS_ACTION_ATTR_RECIRC, it will upcall to userspace to get a translation
231 * for the post-recirculation actions. At this point userspace has to
232 * resume the translation where it left off, which means that it has to
233 * execute the following:
235 * - The action that prompted recirculation, and any actions following
236 * it within the same flow.
238 * - If the action that prompted recirculation was invoked within a
239 * NXAST_RESUBMIT, then any actions following the resubmit. These
240 * "resubmit"s can be nested, so this has to go all the way up the
243 * - The OpenFlow 1.1+ action set.
245 * State that actions and flow table lookups can depend on, such as the
246 * following, must also be preserved:
248 * - Metadata fields (input port, registers, OF1.1+ metadata, ...).
250 * - Action set, stack
252 * - The table ID and cookie of the flow being translated at each level
253 * of the control stack (since OFPAT_CONTROLLER actions send these to
256 * Translation allows for the control of this state preservation via these
257 * members. When a need for recirculation is identified, the translation
260 * 1. Sets 'recirc_action_offset' to the current size of 'action_set'. The
261 * action set is part of what needs to be preserved, so this allows the
262 * action set and the additional state to share the 'action_set' buffer.
263 * Later steps can tell that setup for recirculation is in progress from
264 * the nonnegative value of 'recirc_action_offset'.
266 * 2. Sets 'exit' to true to tell later steps that we're exiting from the
267 * translation process.
269 * 3. Adds an OFPACT_UNROLL_XLATE action to 'action_set'. This action
270 * holds the current table ID and cookie so that they can be restored
271 * during a post-recirculation upcall translation.
273 * 4. Adds the action that prompted recirculation and any actions following
274 * it within the same flow to 'action_set', so that they can be executed
275 * during a post-recirculation upcall translation.
279 * 6. The action that prompted recirculation might be nested in a stack of
280 * nested "resubmit"s that have actions remaining. Each of these notices
281 * that we're exiting (from 'exit') and that recirculation setup is in
282 * progress (from 'recirc_action_offset') and responds by adding more
283 * OFPACT_UNROLL_XLATE actions to 'action_set', as necessary, and any
284 * actions that were yet unprocessed.
286 * The caller stores all the state produced by this process associated with
287 * the recirculation ID. For post-recirculation upcall translation, the
288 * caller passes it back in for the new translation to execute. The
289 * process yielded a set of ofpacts that can be translated directly, so it
290 * is not much of a special case at that point.
292 int recirc_action_offset
; /* Offset in 'action_set' to actions to be
293 * executed after recirculation, or -1. */
294 int last_unroll_offset
; /* Offset in 'action_set' to the latest unroll
297 /* True if a packet was but is no longer MPLS (due to an MPLS pop action).
298 * This is a trigger for recirculation in cases where translating an action
299 * or looking up a flow requires access to the fields of the packet after
300 * the MPLS label stack that was originally present. */
303 /* OpenFlow 1.1+ action set.
305 * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS.
306 * When translation is otherwise complete, ofpacts_execute_action_set()
307 * converts it to a set of "struct ofpact"s that can be translated into
308 * datapath actions. */
309 bool action_set_has_group
; /* Action set contains OFPACT_GROUP? */
310 struct ofpbuf action_set
; /* Action set. */
313 static void xlate_action_set(struct xlate_ctx
*ctx
);
316 ctx_trigger_recirculation(struct xlate_ctx
*ctx
)
319 ctx
->recirc_action_offset
= ctx
->action_set
.size
;
323 ctx_first_recirculation_action(const struct xlate_ctx
*ctx
)
325 return ctx
->recirc_action_offset
== ctx
->action_set
.size
;
329 exit_recirculates(const struct xlate_ctx
*ctx
)
331 /* When recirculating the 'recirc_action_offset' has a non-negative value.
333 return ctx
->recirc_action_offset
>= 0;
336 static void compose_recirculate_action(struct xlate_ctx
*ctx
);
338 /* A controller may use OFPP_NONE as the ingress port to indicate that
339 * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for
340 * when an input bundle is needed for validation (e.g., mirroring or
341 * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have
342 * any 'port' structs, so care must be taken when dealing with it. */
343 static struct xbundle ofpp_none_bundle
= {
345 .vlan_mode
= PORT_VLAN_TRUNK
348 /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from
349 * 'priority' (the datapath's term for QoS queue) to the dscp bits which all
350 * traffic egressing the 'ofport' with that priority should be marked with. */
351 struct skb_priority_to_dscp
{
352 struct hmap_node hmap_node
; /* Node in 'ofport_dpif''s 'skb_priorities'. */
353 uint32_t skb_priority
; /* Priority of this queue (see struct flow). */
355 uint8_t dscp
; /* DSCP bits to mark outgoing traffic with. */
371 /* xlate_cache entries hold enough information to perform the side effects of
372 * xlate_actions() for a rule, without needing to perform rule translation
373 * from scratch. The primary usage of these is to submit statistics to objects
374 * that a flow relates to, although they may be used for other effects as well
375 * (for instance, refreshing hard timeouts for learned flows). */
379 struct rule_dpif
*rule
;
386 struct netflow
*netflow
;
391 struct mbridge
*mbridge
;
392 mirror_mask_t mirrors
;
400 struct ofproto_dpif
*ofproto
;
401 struct ofputil_flow_mod
*fm
;
402 struct ofpbuf
*ofpacts
;
405 struct ofproto_dpif
*ofproto
;
410 struct rule_dpif
*rule
;
415 struct group_dpif
*group
;
416 struct ofputil_bucket
*bucket
;
419 char br_name
[IFNAMSIZ
];
425 #define XC_ENTRY_FOR_EACH(entry, entries, xcache) \
426 entries = xcache->entries; \
427 for (entry = ofpbuf_try_pull(&entries, sizeof *entry); \
429 entry = ofpbuf_try_pull(&entries, sizeof *entry))
432 struct ofpbuf entries
;
435 /* Xlate config contains hash maps of all bridges, bundles and ports.
436 * Xcfgp contains the pointer to the current xlate configuration.
437 * When the main thread needs to change the configuration, it copies xcfgp to
438 * new_xcfg and edits new_xcfg. This enables the use of RCU locking which
439 * does not block handler and revalidator threads. */
441 struct hmap xbridges
;
442 struct hmap xbundles
;
445 static OVSRCU_TYPE(struct xlate_cfg
*) xcfgp
= OVSRCU_INITIALIZER(NULL
);
446 static struct xlate_cfg
*new_xcfg
= NULL
;
448 static bool may_receive(const struct xport
*, struct xlate_ctx
*);
449 static void do_xlate_actions(const struct ofpact
*, size_t ofpacts_len
,
451 static void xlate_normal(struct xlate_ctx
*);
452 static inline void xlate_report(struct xlate_ctx
*, const char *, ...)
453 OVS_PRINTF_FORMAT(2, 3);
454 static void xlate_table_action(struct xlate_ctx
*, ofp_port_t in_port
,
455 uint8_t table_id
, bool may_packet_in
,
456 bool honor_table_miss
);
457 static bool input_vid_is_valid(uint16_t vid
, struct xbundle
*, bool warn
);
458 static uint16_t input_vid_to_vlan(const struct xbundle
*, uint16_t vid
);
459 static void output_normal(struct xlate_ctx
*, const struct xbundle
*,
462 /* Optional bond recirculation parameter to compose_output_action(). */
463 struct xlate_bond_recirc
{
464 uint32_t recirc_id
; /* !0 Use recirculation instead of output. */
465 uint8_t hash_alg
; /* !0 Compute hash for recirc before. */
466 uint32_t hash_basis
; /* Compute hash for recirc before. */
469 static void compose_output_action(struct xlate_ctx
*, ofp_port_t ofp_port
,
470 const struct xlate_bond_recirc
*xr
);
472 static struct xbridge
*xbridge_lookup(struct xlate_cfg
*,
473 const struct ofproto_dpif
*);
474 static struct xbundle
*xbundle_lookup(struct xlate_cfg
*,
475 const struct ofbundle
*);
476 static struct xport
*xport_lookup(struct xlate_cfg
*,
477 const struct ofport_dpif
*);
478 static struct xport
*get_ofp_port(const struct xbridge
*, ofp_port_t ofp_port
);
479 static struct skb_priority_to_dscp
*get_skb_priority(const struct xport
*,
480 uint32_t skb_priority
);
481 static void clear_skb_priorities(struct xport
*);
482 static size_t count_skb_priorities(const struct xport
*);
483 static bool dscp_from_skb_priority(const struct xport
*, uint32_t skb_priority
,
486 static struct xc_entry
*xlate_cache_add_entry(struct xlate_cache
*xc
,
488 static void xlate_xbridge_init(struct xlate_cfg
*, struct xbridge
*);
489 static void xlate_xbundle_init(struct xlate_cfg
*, struct xbundle
*);
490 static void xlate_xport_init(struct xlate_cfg
*, struct xport
*);
491 static void xlate_xbridge_set(struct xbridge
*, struct dpif
*,
492 const struct mac_learning
*, struct stp
*,
493 struct rstp
*, const struct mcast_snooping
*,
494 const struct mbridge
*,
495 const struct dpif_sflow
*,
496 const struct dpif_ipfix
*,
497 const struct netflow
*,
498 bool forward_bpdu
, bool has_in_band
,
499 const struct dpif_backer_support
*);
500 static void xlate_xbundle_set(struct xbundle
*xbundle
,
501 enum port_vlan_mode vlan_mode
, int vlan
,
502 unsigned long *trunks
, bool use_priority_tags
,
503 const struct bond
*bond
, const struct lacp
*lacp
,
505 static void xlate_xport_set(struct xport
*xport
, odp_port_t odp_port
,
506 const struct netdev
*netdev
, const struct cfm
*cfm
,
507 const struct bfd
*bfd
, const struct lldp
*lldp
,
508 int stp_port_no
, const struct rstp_port
*rstp_port
,
509 enum ofputil_port_config config
,
510 enum ofputil_port_state state
, bool is_tunnel
,
512 static void xlate_xbridge_remove(struct xlate_cfg
*, struct xbridge
*);
513 static void xlate_xbundle_remove(struct xlate_cfg
*, struct xbundle
*);
514 static void xlate_xport_remove(struct xlate_cfg
*, struct xport
*);
515 static void xlate_xbridge_copy(struct xbridge
*);
516 static void xlate_xbundle_copy(struct xbridge
*, struct xbundle
*);
517 static void xlate_xport_copy(struct xbridge
*, struct xbundle
*,
519 static void xlate_xcfg_free(struct xlate_cfg
*);
522 xlate_report(struct xlate_ctx
*ctx
, const char *format
, ...)
524 if (OVS_UNLIKELY(ctx
->xin
->report_hook
)) {
527 va_start(args
, format
);
528 ctx
->xin
->report_hook(ctx
->xin
, ctx
->recurse
, format
, args
);
534 xlate_report_actions(struct xlate_ctx
*ctx
, const char *title
,
535 const struct ofpact
*ofpacts
, size_t ofpacts_len
)
537 if (OVS_UNLIKELY(ctx
->xin
->report_hook
)) {
538 struct ds s
= DS_EMPTY_INITIALIZER
;
539 ofpacts_format(ofpacts
, ofpacts_len
, &s
);
540 xlate_report(ctx
, "%s: %s", title
, ds_cstr(&s
));
546 xlate_xbridge_init(struct xlate_cfg
*xcfg
, struct xbridge
*xbridge
)
548 list_init(&xbridge
->xbundles
);
549 hmap_init(&xbridge
->xports
);
550 hmap_insert(&xcfg
->xbridges
, &xbridge
->hmap_node
,
551 hash_pointer(xbridge
->ofproto
, 0));
555 xlate_xbundle_init(struct xlate_cfg
*xcfg
, struct xbundle
*xbundle
)
557 list_init(&xbundle
->xports
);
558 list_insert(&xbundle
->xbridge
->xbundles
, &xbundle
->list_node
);
559 hmap_insert(&xcfg
->xbundles
, &xbundle
->hmap_node
,
560 hash_pointer(xbundle
->ofbundle
, 0));
564 xlate_xport_init(struct xlate_cfg
*xcfg
, struct xport
*xport
)
566 hmap_init(&xport
->skb_priorities
);
567 hmap_insert(&xcfg
->xports
, &xport
->hmap_node
,
568 hash_pointer(xport
->ofport
, 0));
569 hmap_insert(&xport
->xbridge
->xports
, &xport
->ofp_node
,
570 hash_ofp_port(xport
->ofp_port
));
574 xlate_xbridge_set(struct xbridge
*xbridge
,
576 const struct mac_learning
*ml
, struct stp
*stp
,
577 struct rstp
*rstp
, const struct mcast_snooping
*ms
,
578 const struct mbridge
*mbridge
,
579 const struct dpif_sflow
*sflow
,
580 const struct dpif_ipfix
*ipfix
,
581 const struct netflow
*netflow
,
582 bool forward_bpdu
, bool has_in_band
,
583 const struct dpif_backer_support
*support
)
585 if (xbridge
->ml
!= ml
) {
586 mac_learning_unref(xbridge
->ml
);
587 xbridge
->ml
= mac_learning_ref(ml
);
590 if (xbridge
->ms
!= ms
) {
591 mcast_snooping_unref(xbridge
->ms
);
592 xbridge
->ms
= mcast_snooping_ref(ms
);
595 if (xbridge
->mbridge
!= mbridge
) {
596 mbridge_unref(xbridge
->mbridge
);
597 xbridge
->mbridge
= mbridge_ref(mbridge
);
600 if (xbridge
->sflow
!= sflow
) {
601 dpif_sflow_unref(xbridge
->sflow
);
602 xbridge
->sflow
= dpif_sflow_ref(sflow
);
605 if (xbridge
->ipfix
!= ipfix
) {
606 dpif_ipfix_unref(xbridge
->ipfix
);
607 xbridge
->ipfix
= dpif_ipfix_ref(ipfix
);
610 if (xbridge
->stp
!= stp
) {
611 stp_unref(xbridge
->stp
);
612 xbridge
->stp
= stp_ref(stp
);
615 if (xbridge
->rstp
!= rstp
) {
616 rstp_unref(xbridge
->rstp
);
617 xbridge
->rstp
= rstp_ref(rstp
);
620 if (xbridge
->netflow
!= netflow
) {
621 netflow_unref(xbridge
->netflow
);
622 xbridge
->netflow
= netflow_ref(netflow
);
625 xbridge
->dpif
= dpif
;
626 xbridge
->forward_bpdu
= forward_bpdu
;
627 xbridge
->has_in_band
= has_in_band
;
628 xbridge
->support
= *support
;
632 xlate_xbundle_set(struct xbundle
*xbundle
,
633 enum port_vlan_mode vlan_mode
, int vlan
,
634 unsigned long *trunks
, bool use_priority_tags
,
635 const struct bond
*bond
, const struct lacp
*lacp
,
638 ovs_assert(xbundle
->xbridge
);
640 xbundle
->vlan_mode
= vlan_mode
;
641 xbundle
->vlan
= vlan
;
642 xbundle
->trunks
= trunks
;
643 xbundle
->use_priority_tags
= use_priority_tags
;
644 xbundle
->floodable
= floodable
;
646 if (xbundle
->bond
!= bond
) {
647 bond_unref(xbundle
->bond
);
648 xbundle
->bond
= bond_ref(bond
);
651 if (xbundle
->lacp
!= lacp
) {
652 lacp_unref(xbundle
->lacp
);
653 xbundle
->lacp
= lacp_ref(lacp
);
658 xlate_xport_set(struct xport
*xport
, odp_port_t odp_port
,
659 const struct netdev
*netdev
, const struct cfm
*cfm
,
660 const struct bfd
*bfd
, const struct lldp
*lldp
, int stp_port_no
,
661 const struct rstp_port
* rstp_port
,
662 enum ofputil_port_config config
, enum ofputil_port_state state
,
663 bool is_tunnel
, bool may_enable
)
665 xport
->config
= config
;
666 xport
->state
= state
;
667 xport
->stp_port_no
= stp_port_no
;
668 xport
->is_tunnel
= is_tunnel
;
669 xport
->may_enable
= may_enable
;
670 xport
->odp_port
= odp_port
;
672 if (xport
->rstp_port
!= rstp_port
) {
673 rstp_port_unref(xport
->rstp_port
);
674 xport
->rstp_port
= rstp_port_ref(rstp_port
);
677 if (xport
->cfm
!= cfm
) {
678 cfm_unref(xport
->cfm
);
679 xport
->cfm
= cfm_ref(cfm
);
682 if (xport
->bfd
!= bfd
) {
683 bfd_unref(xport
->bfd
);
684 xport
->bfd
= bfd_ref(bfd
);
687 if (xport
->lldp
!= lldp
) {
688 lldp_unref(xport
->lldp
);
689 xport
->lldp
= lldp_ref(lldp
);
692 if (xport
->netdev
!= netdev
) {
693 netdev_close(xport
->netdev
);
694 xport
->netdev
= netdev_ref(netdev
);
699 xlate_xbridge_copy(struct xbridge
*xbridge
)
701 struct xbundle
*xbundle
;
703 struct xbridge
*new_xbridge
= xzalloc(sizeof *xbridge
);
704 new_xbridge
->ofproto
= xbridge
->ofproto
;
705 new_xbridge
->name
= xstrdup(xbridge
->name
);
706 xlate_xbridge_init(new_xcfg
, new_xbridge
);
708 xlate_xbridge_set(new_xbridge
,
709 xbridge
->dpif
, xbridge
->ml
, xbridge
->stp
,
710 xbridge
->rstp
, xbridge
->ms
, xbridge
->mbridge
,
711 xbridge
->sflow
, xbridge
->ipfix
, xbridge
->netflow
,
712 xbridge
->forward_bpdu
, xbridge
->has_in_band
,
714 LIST_FOR_EACH (xbundle
, list_node
, &xbridge
->xbundles
) {
715 xlate_xbundle_copy(new_xbridge
, xbundle
);
718 /* Copy xports which are not part of a xbundle */
719 HMAP_FOR_EACH (xport
, ofp_node
, &xbridge
->xports
) {
720 if (!xport
->xbundle
) {
721 xlate_xport_copy(new_xbridge
, NULL
, xport
);
727 xlate_xbundle_copy(struct xbridge
*xbridge
, struct xbundle
*xbundle
)
730 struct xbundle
*new_xbundle
= xzalloc(sizeof *xbundle
);
731 new_xbundle
->ofbundle
= xbundle
->ofbundle
;
732 new_xbundle
->xbridge
= xbridge
;
733 new_xbundle
->name
= xstrdup(xbundle
->name
);
734 xlate_xbundle_init(new_xcfg
, new_xbundle
);
736 xlate_xbundle_set(new_xbundle
, xbundle
->vlan_mode
,
737 xbundle
->vlan
, xbundle
->trunks
,
738 xbundle
->use_priority_tags
, xbundle
->bond
, xbundle
->lacp
,
740 LIST_FOR_EACH (xport
, bundle_node
, &xbundle
->xports
) {
741 xlate_xport_copy(xbridge
, new_xbundle
, xport
);
746 xlate_xport_copy(struct xbridge
*xbridge
, struct xbundle
*xbundle
,
749 struct skb_priority_to_dscp
*pdscp
, *new_pdscp
;
750 struct xport
*new_xport
= xzalloc(sizeof *xport
);
751 new_xport
->ofport
= xport
->ofport
;
752 new_xport
->ofp_port
= xport
->ofp_port
;
753 new_xport
->xbridge
= xbridge
;
754 xlate_xport_init(new_xcfg
, new_xport
);
756 xlate_xport_set(new_xport
, xport
->odp_port
, xport
->netdev
, xport
->cfm
,
757 xport
->bfd
, xport
->lldp
, xport
->stp_port_no
,
758 xport
->rstp_port
, xport
->config
, xport
->state
,
759 xport
->is_tunnel
, xport
->may_enable
);
762 struct xport
*peer
= xport_lookup(new_xcfg
, xport
->peer
->ofport
);
764 new_xport
->peer
= peer
;
765 new_xport
->peer
->peer
= new_xport
;
770 new_xport
->xbundle
= xbundle
;
771 list_insert(&new_xport
->xbundle
->xports
, &new_xport
->bundle_node
);
774 HMAP_FOR_EACH (pdscp
, hmap_node
, &xport
->skb_priorities
) {
775 new_pdscp
= xmalloc(sizeof *pdscp
);
776 new_pdscp
->skb_priority
= pdscp
->skb_priority
;
777 new_pdscp
->dscp
= pdscp
->dscp
;
778 hmap_insert(&new_xport
->skb_priorities
, &new_pdscp
->hmap_node
,
779 hash_int(new_pdscp
->skb_priority
, 0));
783 /* Sets the current xlate configuration to new_xcfg and frees the old xlate
784 * configuration in xcfgp.
786 * This needs to be called after editing the xlate configuration.
788 * Functions that edit the new xlate configuration are
789 * xlate_<ofport/bundle/ofport>_set and xlate_<ofport/bundle/ofport>_remove.
795 * edit_xlate_configuration();
797 * xlate_txn_commit(); */
799 xlate_txn_commit(void)
801 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
803 ovsrcu_set(&xcfgp
, new_xcfg
);
804 ovsrcu_synchronize();
805 xlate_xcfg_free(xcfg
);
809 /* Copies the current xlate configuration in xcfgp to new_xcfg.
811 * This needs to be called prior to editing the xlate configuration. */
813 xlate_txn_start(void)
815 struct xbridge
*xbridge
;
816 struct xlate_cfg
*xcfg
;
818 ovs_assert(!new_xcfg
);
820 new_xcfg
= xmalloc(sizeof *new_xcfg
);
821 hmap_init(&new_xcfg
->xbridges
);
822 hmap_init(&new_xcfg
->xbundles
);
823 hmap_init(&new_xcfg
->xports
);
825 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
830 HMAP_FOR_EACH (xbridge
, hmap_node
, &xcfg
->xbridges
) {
831 xlate_xbridge_copy(xbridge
);
837 xlate_xcfg_free(struct xlate_cfg
*xcfg
)
839 struct xbridge
*xbridge
, *next_xbridge
;
845 HMAP_FOR_EACH_SAFE (xbridge
, next_xbridge
, hmap_node
, &xcfg
->xbridges
) {
846 xlate_xbridge_remove(xcfg
, xbridge
);
849 hmap_destroy(&xcfg
->xbridges
);
850 hmap_destroy(&xcfg
->xbundles
);
851 hmap_destroy(&xcfg
->xports
);
856 xlate_ofproto_set(struct ofproto_dpif
*ofproto
, const char *name
,
858 const struct mac_learning
*ml
, struct stp
*stp
,
859 struct rstp
*rstp
, const struct mcast_snooping
*ms
,
860 const struct mbridge
*mbridge
,
861 const struct dpif_sflow
*sflow
,
862 const struct dpif_ipfix
*ipfix
,
863 const struct netflow
*netflow
,
864 bool forward_bpdu
, bool has_in_band
,
865 const struct dpif_backer_support
*support
)
867 struct xbridge
*xbridge
;
869 ovs_assert(new_xcfg
);
871 xbridge
= xbridge_lookup(new_xcfg
, ofproto
);
873 xbridge
= xzalloc(sizeof *xbridge
);
874 xbridge
->ofproto
= ofproto
;
876 xlate_xbridge_init(new_xcfg
, xbridge
);
880 xbridge
->name
= xstrdup(name
);
882 xlate_xbridge_set(xbridge
, dpif
, ml
, stp
, rstp
, ms
, mbridge
, sflow
, ipfix
,
883 netflow
, forward_bpdu
, has_in_band
, support
);
887 xlate_xbridge_remove(struct xlate_cfg
*xcfg
, struct xbridge
*xbridge
)
889 struct xbundle
*xbundle
, *next_xbundle
;
890 struct xport
*xport
, *next_xport
;
896 HMAP_FOR_EACH_SAFE (xport
, next_xport
, ofp_node
, &xbridge
->xports
) {
897 xlate_xport_remove(xcfg
, xport
);
900 LIST_FOR_EACH_SAFE (xbundle
, next_xbundle
, list_node
, &xbridge
->xbundles
) {
901 xlate_xbundle_remove(xcfg
, xbundle
);
904 hmap_remove(&xcfg
->xbridges
, &xbridge
->hmap_node
);
905 mac_learning_unref(xbridge
->ml
);
906 mcast_snooping_unref(xbridge
->ms
);
907 mbridge_unref(xbridge
->mbridge
);
908 dpif_sflow_unref(xbridge
->sflow
);
909 dpif_ipfix_unref(xbridge
->ipfix
);
910 stp_unref(xbridge
->stp
);
911 rstp_unref(xbridge
->rstp
);
912 hmap_destroy(&xbridge
->xports
);
918 xlate_remove_ofproto(struct ofproto_dpif
*ofproto
)
920 struct xbridge
*xbridge
;
922 ovs_assert(new_xcfg
);
924 xbridge
= xbridge_lookup(new_xcfg
, ofproto
);
925 xlate_xbridge_remove(new_xcfg
, xbridge
);
929 xlate_bundle_set(struct ofproto_dpif
*ofproto
, struct ofbundle
*ofbundle
,
930 const char *name
, enum port_vlan_mode vlan_mode
, int vlan
,
931 unsigned long *trunks
, bool use_priority_tags
,
932 const struct bond
*bond
, const struct lacp
*lacp
,
935 struct xbundle
*xbundle
;
937 ovs_assert(new_xcfg
);
939 xbundle
= xbundle_lookup(new_xcfg
, ofbundle
);
941 xbundle
= xzalloc(sizeof *xbundle
);
942 xbundle
->ofbundle
= ofbundle
;
943 xbundle
->xbridge
= xbridge_lookup(new_xcfg
, ofproto
);
945 xlate_xbundle_init(new_xcfg
, xbundle
);
949 xbundle
->name
= xstrdup(name
);
951 xlate_xbundle_set(xbundle
, vlan_mode
, vlan
, trunks
,
952 use_priority_tags
, bond
, lacp
, floodable
);
956 xlate_xbundle_remove(struct xlate_cfg
*xcfg
, struct xbundle
*xbundle
)
964 LIST_FOR_EACH_POP (xport
, bundle_node
, &xbundle
->xports
) {
965 xport
->xbundle
= NULL
;
968 hmap_remove(&xcfg
->xbundles
, &xbundle
->hmap_node
);
969 list_remove(&xbundle
->list_node
);
970 bond_unref(xbundle
->bond
);
971 lacp_unref(xbundle
->lacp
);
977 xlate_bundle_remove(struct ofbundle
*ofbundle
)
979 struct xbundle
*xbundle
;
981 ovs_assert(new_xcfg
);
983 xbundle
= xbundle_lookup(new_xcfg
, ofbundle
);
984 xlate_xbundle_remove(new_xcfg
, xbundle
);
988 xlate_ofport_set(struct ofproto_dpif
*ofproto
, struct ofbundle
*ofbundle
,
989 struct ofport_dpif
*ofport
, ofp_port_t ofp_port
,
990 odp_port_t odp_port
, const struct netdev
*netdev
,
991 const struct cfm
*cfm
, const struct bfd
*bfd
,
992 const struct lldp
*lldp
, struct ofport_dpif
*peer
,
993 int stp_port_no
, const struct rstp_port
*rstp_port
,
994 const struct ofproto_port_queue
*qdscp_list
, size_t n_qdscp
,
995 enum ofputil_port_config config
,
996 enum ofputil_port_state state
, bool is_tunnel
,
1000 struct xport
*xport
;
1002 ovs_assert(new_xcfg
);
1004 xport
= xport_lookup(new_xcfg
, ofport
);
1006 xport
= xzalloc(sizeof *xport
);
1007 xport
->ofport
= ofport
;
1008 xport
->xbridge
= xbridge_lookup(new_xcfg
, ofproto
);
1009 xport
->ofp_port
= ofp_port
;
1011 xlate_xport_init(new_xcfg
, xport
);
1014 ovs_assert(xport
->ofp_port
== ofp_port
);
1016 xlate_xport_set(xport
, odp_port
, netdev
, cfm
, bfd
, lldp
,
1017 stp_port_no
, rstp_port
, config
, state
, is_tunnel
,
1021 xport
->peer
->peer
= NULL
;
1023 xport
->peer
= xport_lookup(new_xcfg
, peer
);
1025 xport
->peer
->peer
= xport
;
1028 if (xport
->xbundle
) {
1029 list_remove(&xport
->bundle_node
);
1031 xport
->xbundle
= xbundle_lookup(new_xcfg
, ofbundle
);
1032 if (xport
->xbundle
) {
1033 list_insert(&xport
->xbundle
->xports
, &xport
->bundle_node
);
1036 clear_skb_priorities(xport
);
1037 for (i
= 0; i
< n_qdscp
; i
++) {
1038 struct skb_priority_to_dscp
*pdscp
;
1039 uint32_t skb_priority
;
1041 if (dpif_queue_to_priority(xport
->xbridge
->dpif
, qdscp_list
[i
].queue
,
1046 pdscp
= xmalloc(sizeof *pdscp
);
1047 pdscp
->skb_priority
= skb_priority
;
1048 pdscp
->dscp
= (qdscp_list
[i
].dscp
<< 2) & IP_DSCP_MASK
;
1049 hmap_insert(&xport
->skb_priorities
, &pdscp
->hmap_node
,
1050 hash_int(pdscp
->skb_priority
, 0));
1055 xlate_xport_remove(struct xlate_cfg
*xcfg
, struct xport
*xport
)
1062 xport
->peer
->peer
= NULL
;
1066 if (xport
->xbundle
) {
1067 list_remove(&xport
->bundle_node
);
1070 clear_skb_priorities(xport
);
1071 hmap_destroy(&xport
->skb_priorities
);
1073 hmap_remove(&xcfg
->xports
, &xport
->hmap_node
);
1074 hmap_remove(&xport
->xbridge
->xports
, &xport
->ofp_node
);
1076 netdev_close(xport
->netdev
);
1077 rstp_port_unref(xport
->rstp_port
);
1078 cfm_unref(xport
->cfm
);
1079 bfd_unref(xport
->bfd
);
1080 lldp_unref(xport
->lldp
);
1085 xlate_ofport_remove(struct ofport_dpif
*ofport
)
1087 struct xport
*xport
;
1089 ovs_assert(new_xcfg
);
1091 xport
= xport_lookup(new_xcfg
, ofport
);
1092 xlate_xport_remove(new_xcfg
, xport
);
1095 static struct ofproto_dpif
*
1096 xlate_lookup_ofproto_(const struct dpif_backer
*backer
, const struct flow
*flow
,
1097 ofp_port_t
*ofp_in_port
, const struct xport
**xportp
)
1099 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1100 const struct xport
*xport
;
1102 xport
= xport_lookup(xcfg
, tnl_port_should_receive(flow
)
1103 ? tnl_port_receive(flow
)
1104 : odp_port_to_ofport(backer
, flow
->in_port
.odp_port
));
1105 if (OVS_UNLIKELY(!xport
)) {
1110 *ofp_in_port
= xport
->ofp_port
;
1112 return xport
->xbridge
->ofproto
;
1115 /* Given a datapath and flow metadata ('backer', and 'flow' respectively)
1116 * returns the corresponding struct ofproto_dpif and OpenFlow port number. */
1117 struct ofproto_dpif
*
1118 xlate_lookup_ofproto(const struct dpif_backer
*backer
, const struct flow
*flow
,
1119 ofp_port_t
*ofp_in_port
)
1121 const struct xport
*xport
;
1123 return xlate_lookup_ofproto_(backer
, flow
, ofp_in_port
, &xport
);
1126 /* Given a datapath and flow metadata ('backer', and 'flow' respectively),
1127 * optionally populates 'ofproto' with the ofproto_dpif, 'ofp_in_port' with the
1128 * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate
1129 * handles for those protocols if they're enabled. Caller may use the returned
1130 * pointers until quiescing, for longer term use additional references must
1133 * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto.
1136 xlate_lookup(const struct dpif_backer
*backer
, const struct flow
*flow
,
1137 struct ofproto_dpif
**ofprotop
, struct dpif_ipfix
**ipfix
,
1138 struct dpif_sflow
**sflow
, struct netflow
**netflow
,
1139 ofp_port_t
*ofp_in_port
)
1141 struct ofproto_dpif
*ofproto
;
1142 const struct xport
*xport
;
1144 ofproto
= xlate_lookup_ofproto_(backer
, flow
, ofp_in_port
, &xport
);
1151 *ofprotop
= ofproto
;
1155 *ipfix
= xport
? xport
->xbridge
->ipfix
: NULL
;
1159 *sflow
= xport
? xport
->xbridge
->sflow
: NULL
;
1163 *netflow
= xport
? xport
->xbridge
->netflow
: NULL
;
1169 static struct xbridge
*
1170 xbridge_lookup(struct xlate_cfg
*xcfg
, const struct ofproto_dpif
*ofproto
)
1172 struct hmap
*xbridges
;
1173 struct xbridge
*xbridge
;
1175 if (!ofproto
|| !xcfg
) {
1179 xbridges
= &xcfg
->xbridges
;
1181 HMAP_FOR_EACH_IN_BUCKET (xbridge
, hmap_node
, hash_pointer(ofproto
, 0),
1183 if (xbridge
->ofproto
== ofproto
) {
1190 static struct xbundle
*
1191 xbundle_lookup(struct xlate_cfg
*xcfg
, const struct ofbundle
*ofbundle
)
1193 struct hmap
*xbundles
;
1194 struct xbundle
*xbundle
;
1196 if (!ofbundle
|| !xcfg
) {
1200 xbundles
= &xcfg
->xbundles
;
1202 HMAP_FOR_EACH_IN_BUCKET (xbundle
, hmap_node
, hash_pointer(ofbundle
, 0),
1204 if (xbundle
->ofbundle
== ofbundle
) {
1211 static struct xport
*
1212 xport_lookup(struct xlate_cfg
*xcfg
, const struct ofport_dpif
*ofport
)
1214 struct hmap
*xports
;
1215 struct xport
*xport
;
1217 if (!ofport
|| !xcfg
) {
1221 xports
= &xcfg
->xports
;
1223 HMAP_FOR_EACH_IN_BUCKET (xport
, hmap_node
, hash_pointer(ofport
, 0),
1225 if (xport
->ofport
== ofport
) {
1232 static struct stp_port
*
1233 xport_get_stp_port(const struct xport
*xport
)
1235 return xport
->xbridge
->stp
&& xport
->stp_port_no
!= -1
1236 ? stp_get_port(xport
->xbridge
->stp
, xport
->stp_port_no
)
1241 xport_stp_learn_state(const struct xport
*xport
)
1243 struct stp_port
*sp
= xport_get_stp_port(xport
);
1245 ? stp_learn_in_state(stp_port_get_state(sp
))
1250 xport_stp_forward_state(const struct xport
*xport
)
1252 struct stp_port
*sp
= xport_get_stp_port(xport
);
1254 ? stp_forward_in_state(stp_port_get_state(sp
))
1259 xport_stp_should_forward_bpdu(const struct xport
*xport
)
1261 struct stp_port
*sp
= xport_get_stp_port(xport
);
1262 return stp_should_forward_bpdu(sp
? stp_port_get_state(sp
) : STP_DISABLED
);
1265 /* Returns true if STP should process 'flow'. Sets fields in 'wc' that
1266 * were used to make the determination.*/
1268 stp_should_process_flow(const struct flow
*flow
, struct flow_wildcards
*wc
)
1270 /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */
1271 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
1272 return is_stp(flow
);
1276 stp_process_packet(const struct xport
*xport
, const struct dp_packet
*packet
)
1278 struct stp_port
*sp
= xport_get_stp_port(xport
);
1279 struct dp_packet payload
= *packet
;
1280 struct eth_header
*eth
= dp_packet_data(&payload
);
1282 /* Sink packets on ports that have STP disabled when the bridge has
1284 if (!sp
|| stp_port_get_state(sp
) == STP_DISABLED
) {
1288 /* Trim off padding on payload. */
1289 if (dp_packet_size(&payload
) > ntohs(eth
->eth_type
) + ETH_HEADER_LEN
) {
1290 dp_packet_set_size(&payload
, ntohs(eth
->eth_type
) + ETH_HEADER_LEN
);
1293 if (dp_packet_try_pull(&payload
, ETH_HEADER_LEN
+ LLC_HEADER_LEN
)) {
1294 stp_received_bpdu(sp
, dp_packet_data(&payload
), dp_packet_size(&payload
));
1298 static enum rstp_state
1299 xport_get_rstp_port_state(const struct xport
*xport
)
1301 return xport
->rstp_port
1302 ? rstp_port_get_state(xport
->rstp_port
)
1307 xport_rstp_learn_state(const struct xport
*xport
)
1309 return xport
->xbridge
->rstp
&& xport
->rstp_port
1310 ? rstp_learn_in_state(xport_get_rstp_port_state(xport
))
1315 xport_rstp_forward_state(const struct xport
*xport
)
1317 return xport
->xbridge
->rstp
&& xport
->rstp_port
1318 ? rstp_forward_in_state(xport_get_rstp_port_state(xport
))
1323 xport_rstp_should_manage_bpdu(const struct xport
*xport
)
1325 return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport
));
1329 rstp_process_packet(const struct xport
*xport
, const struct dp_packet
*packet
)
1331 struct dp_packet payload
= *packet
;
1332 struct eth_header
*eth
= dp_packet_data(&payload
);
1334 /* Sink packets on ports that have no RSTP. */
1335 if (!xport
->rstp_port
) {
1339 /* Trim off padding on payload. */
1340 if (dp_packet_size(&payload
) > ntohs(eth
->eth_type
) + ETH_HEADER_LEN
) {
1341 dp_packet_set_size(&payload
, ntohs(eth
->eth_type
) + ETH_HEADER_LEN
);
1344 if (dp_packet_try_pull(&payload
, ETH_HEADER_LEN
+ LLC_HEADER_LEN
)) {
1345 rstp_port_received_bpdu(xport
->rstp_port
, dp_packet_data(&payload
),
1346 dp_packet_size(&payload
));
1350 static struct xport
*
1351 get_ofp_port(const struct xbridge
*xbridge
, ofp_port_t ofp_port
)
1353 struct xport
*xport
;
1355 HMAP_FOR_EACH_IN_BUCKET (xport
, ofp_node
, hash_ofp_port(ofp_port
),
1357 if (xport
->ofp_port
== ofp_port
) {
1365 ofp_port_to_odp_port(const struct xbridge
*xbridge
, ofp_port_t ofp_port
)
1367 const struct xport
*xport
= get_ofp_port(xbridge
, ofp_port
);
1368 return xport
? xport
->odp_port
: ODPP_NONE
;
1372 odp_port_is_alive(const struct xlate_ctx
*ctx
, ofp_port_t ofp_port
)
1374 struct xport
*xport
= get_ofp_port(ctx
->xbridge
, ofp_port
);
1375 return xport
&& xport
->may_enable
;
1378 static struct ofputil_bucket
*
1379 group_first_live_bucket(const struct xlate_ctx
*, const struct group_dpif
*,
1383 group_is_alive(const struct xlate_ctx
*ctx
, uint32_t group_id
, int depth
)
1385 struct group_dpif
*group
;
1387 if (group_dpif_lookup(ctx
->xbridge
->ofproto
, group_id
, &group
)) {
1388 struct ofputil_bucket
*bucket
;
1390 bucket
= group_first_live_bucket(ctx
, group
, depth
);
1391 group_dpif_unref(group
);
1392 return bucket
== NULL
;
1398 #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */
1401 bucket_is_alive(const struct xlate_ctx
*ctx
,
1402 struct ofputil_bucket
*bucket
, int depth
)
1404 if (depth
>= MAX_LIVENESS_RECURSION
) {
1405 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
1407 VLOG_WARN_RL(&rl
, "bucket chaining exceeded %d links",
1408 MAX_LIVENESS_RECURSION
);
1412 return (!ofputil_bucket_has_liveness(bucket
)
1413 || (bucket
->watch_port
!= OFPP_ANY
1414 && odp_port_is_alive(ctx
, bucket
->watch_port
))
1415 || (bucket
->watch_group
!= OFPG_ANY
1416 && group_is_alive(ctx
, bucket
->watch_group
, depth
+ 1)));
1419 static struct ofputil_bucket
*
1420 group_first_live_bucket(const struct xlate_ctx
*ctx
,
1421 const struct group_dpif
*group
, int depth
)
1423 struct ofputil_bucket
*bucket
;
1424 const struct ovs_list
*buckets
;
1426 group_dpif_get_buckets(group
, &buckets
);
1427 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
1428 if (bucket_is_alive(ctx
, bucket
, depth
)) {
1436 static struct ofputil_bucket
*
1437 group_best_live_bucket(const struct xlate_ctx
*ctx
,
1438 const struct group_dpif
*group
,
1441 struct ofputil_bucket
*best_bucket
= NULL
;
1442 uint32_t best_score
= 0;
1445 struct ofputil_bucket
*bucket
;
1446 const struct ovs_list
*buckets
;
1448 group_dpif_get_buckets(group
, &buckets
);
1449 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
1450 if (bucket_is_alive(ctx
, bucket
, 0)) {
1451 uint32_t score
= (hash_int(i
, basis
) & 0xffff) * bucket
->weight
;
1452 if (score
>= best_score
) {
1453 best_bucket
= bucket
;
1464 xbundle_trunks_vlan(const struct xbundle
*bundle
, uint16_t vlan
)
1466 return (bundle
->vlan_mode
!= PORT_VLAN_ACCESS
1467 && (!bundle
->trunks
|| bitmap_is_set(bundle
->trunks
, vlan
)));
1471 xbundle_includes_vlan(const struct xbundle
*xbundle
, uint16_t vlan
)
1473 return vlan
== xbundle
->vlan
|| xbundle_trunks_vlan(xbundle
, vlan
);
1476 static mirror_mask_t
1477 xbundle_mirror_out(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
1479 return xbundle
!= &ofpp_none_bundle
1480 ? mirror_bundle_out(xbridge
->mbridge
, xbundle
->ofbundle
)
1484 static mirror_mask_t
1485 xbundle_mirror_src(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
1487 return xbundle
!= &ofpp_none_bundle
1488 ? mirror_bundle_src(xbridge
->mbridge
, xbundle
->ofbundle
)
1492 static mirror_mask_t
1493 xbundle_mirror_dst(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
1495 return xbundle
!= &ofpp_none_bundle
1496 ? mirror_bundle_dst(xbridge
->mbridge
, xbundle
->ofbundle
)
1500 static struct xbundle
*
1501 lookup_input_bundle(const struct xbridge
*xbridge
, ofp_port_t in_port
,
1502 bool warn
, struct xport
**in_xportp
)
1504 struct xport
*xport
;
1506 /* Find the port and bundle for the received packet. */
1507 xport
= get_ofp_port(xbridge
, in_port
);
1511 if (xport
&& xport
->xbundle
) {
1512 return xport
->xbundle
;
1515 /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+),
1516 * which a controller may use as the ingress port for traffic that
1517 * it is sourcing. */
1518 if (in_port
== OFPP_CONTROLLER
|| in_port
== OFPP_NONE
) {
1519 return &ofpp_none_bundle
;
1522 /* Odd. A few possible reasons here:
1524 * - We deleted a port but there are still a few packets queued up
1527 * - Someone externally added a port (e.g. "ovs-dpctl add-if") that
1528 * we don't know about.
1530 * - The ofproto client didn't configure the port as part of a bundle.
1531 * This is particularly likely to happen if a packet was received on the
1532 * port after it was created, but before the client had a chance to
1533 * configure its bundle.
1536 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1538 VLOG_WARN_RL(&rl
, "bridge %s: received packet on unknown "
1539 "port %"PRIu16
, xbridge
->name
, in_port
);
1545 mirror_packet(struct xlate_ctx
*ctx
, struct xbundle
*xbundle
,
1546 mirror_mask_t mirrors
)
1548 bool warn
= ctx
->xin
->packet
!= NULL
;
1549 uint16_t vid
= vlan_tci_to_vid(ctx
->xin
->flow
.vlan_tci
);
1550 if (!input_vid_is_valid(vid
, xbundle
, warn
)) {
1553 uint16_t vlan
= input_vid_to_vlan(xbundle
, vid
);
1555 const struct xbridge
*xbridge
= ctx
->xbridge
;
1557 /* Don't mirror to destinations that we've already mirrored to. */
1558 mirrors
&= ~ctx
->mirrors
;
1563 /* Record these mirrors so that we don't mirror to them again. */
1564 ctx
->mirrors
|= mirrors
;
1566 if (ctx
->xin
->resubmit_stats
) {
1567 mirror_update_stats(xbridge
->mbridge
, mirrors
,
1568 ctx
->xin
->resubmit_stats
->n_packets
,
1569 ctx
->xin
->resubmit_stats
->n_bytes
);
1571 if (ctx
->xin
->xcache
) {
1572 struct xc_entry
*entry
;
1574 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_MIRROR
);
1575 entry
->u
.mirror
.mbridge
= mbridge_ref(xbridge
->mbridge
);
1576 entry
->u
.mirror
.mirrors
= mirrors
;
1580 const unsigned long *vlans
;
1581 mirror_mask_t dup_mirrors
;
1582 struct ofbundle
*out
;
1585 bool has_mirror
= mirror_get(xbridge
->mbridge
, raw_ctz(mirrors
),
1586 &vlans
, &dup_mirrors
, &out
, &out_vlan
);
1587 ovs_assert(has_mirror
);
1590 ctx
->wc
->masks
.vlan_tci
|= htons(VLAN_CFI
| VLAN_VID_MASK
);
1593 if (vlans
&& !bitmap_is_set(vlans
, vlan
)) {
1594 mirrors
= zero_rightmost_1bit(mirrors
);
1598 mirrors
&= ~dup_mirrors
;
1599 ctx
->mirrors
|= dup_mirrors
;
1601 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1602 struct xbundle
*out_xbundle
= xbundle_lookup(xcfg
, out
);
1604 output_normal(ctx
, out_xbundle
, vlan
);
1606 } else if (vlan
!= out_vlan
1607 && !eth_addr_is_reserved(ctx
->xin
->flow
.dl_dst
)) {
1608 struct xbundle
*xbundle
;
1610 LIST_FOR_EACH (xbundle
, list_node
, &xbridge
->xbundles
) {
1611 if (xbundle_includes_vlan(xbundle
, out_vlan
)
1612 && !xbundle_mirror_out(xbridge
, xbundle
)) {
1613 output_normal(ctx
, xbundle
, out_vlan
);
1621 mirror_ingress_packet(struct xlate_ctx
*ctx
)
1623 if (mbridge_has_mirrors(ctx
->xbridge
->mbridge
)) {
1624 bool warn
= ctx
->xin
->packet
!= NULL
;
1625 struct xbundle
*xbundle
= lookup_input_bundle(
1626 ctx
->xbridge
, ctx
->xin
->flow
.in_port
.ofp_port
, warn
, NULL
);
1628 mirror_packet(ctx
, xbundle
,
1629 xbundle_mirror_src(ctx
->xbridge
, xbundle
));
1634 /* Given 'vid', the VID obtained from the 802.1Q header that was received as
1635 * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle',
1636 * the bundle on which the packet was received, returns the VLAN to which the
1639 * Both 'vid' and the return value are in the range 0...4095. */
1641 input_vid_to_vlan(const struct xbundle
*in_xbundle
, uint16_t vid
)
1643 switch (in_xbundle
->vlan_mode
) {
1644 case PORT_VLAN_ACCESS
:
1645 return in_xbundle
->vlan
;
1648 case PORT_VLAN_TRUNK
:
1651 case PORT_VLAN_NATIVE_UNTAGGED
:
1652 case PORT_VLAN_NATIVE_TAGGED
:
1653 return vid
? vid
: in_xbundle
->vlan
;
1660 /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'.
1661 * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs
1664 * 'vid' should be the VID obtained from the 802.1Q header that was received as
1665 * part of a packet (specify 0 if there was no 802.1Q header), in the range
1668 input_vid_is_valid(uint16_t vid
, struct xbundle
*in_xbundle
, bool warn
)
1670 /* Allow any VID on the OFPP_NONE port. */
1671 if (in_xbundle
== &ofpp_none_bundle
) {
1675 switch (in_xbundle
->vlan_mode
) {
1676 case PORT_VLAN_ACCESS
:
1679 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1680 VLOG_WARN_RL(&rl
, "dropping VLAN %"PRIu16
" tagged "
1681 "packet received on port %s configured as VLAN "
1682 "%"PRIu16
" access port", vid
, in_xbundle
->name
,
1689 case PORT_VLAN_NATIVE_UNTAGGED
:
1690 case PORT_VLAN_NATIVE_TAGGED
:
1692 /* Port must always carry its native VLAN. */
1696 case PORT_VLAN_TRUNK
:
1697 if (!xbundle_includes_vlan(in_xbundle
, vid
)) {
1699 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1700 VLOG_WARN_RL(&rl
, "dropping VLAN %"PRIu16
" packet "
1701 "received on port %s not configured for trunking "
1702 "VLAN %"PRIu16
, vid
, in_xbundle
->name
, vid
);
1714 /* Given 'vlan', the VLAN that a packet belongs to, and
1715 * 'out_xbundle', a bundle on which the packet is to be output, returns the VID
1716 * that should be included in the 802.1Q header. (If the return value is 0,
1717 * then the 802.1Q header should only be included in the packet if there is a
1720 * Both 'vlan' and the return value are in the range 0...4095. */
1722 output_vlan_to_vid(const struct xbundle
*out_xbundle
, uint16_t vlan
)
1724 switch (out_xbundle
->vlan_mode
) {
1725 case PORT_VLAN_ACCESS
:
1728 case PORT_VLAN_TRUNK
:
1729 case PORT_VLAN_NATIVE_TAGGED
:
1732 case PORT_VLAN_NATIVE_UNTAGGED
:
1733 return vlan
== out_xbundle
->vlan
? 0 : vlan
;
1741 output_normal(struct xlate_ctx
*ctx
, const struct xbundle
*out_xbundle
,
1744 ovs_be16
*flow_tci
= &ctx
->xin
->flow
.vlan_tci
;
1746 ovs_be16 tci
, old_tci
;
1747 struct xport
*xport
;
1748 struct xlate_bond_recirc xr
;
1749 bool use_recirc
= false;
1751 vid
= output_vlan_to_vid(out_xbundle
, vlan
);
1752 if (list_is_empty(&out_xbundle
->xports
)) {
1753 /* Partially configured bundle with no slaves. Drop the packet. */
1755 } else if (!out_xbundle
->bond
) {
1756 xport
= CONTAINER_OF(list_front(&out_xbundle
->xports
), struct xport
,
1759 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1760 struct flow_wildcards
*wc
= ctx
->wc
;
1761 struct ofport_dpif
*ofport
;
1763 if (ctx
->xbridge
->support
.odp
.recirc
) {
1764 use_recirc
= bond_may_recirc(
1765 out_xbundle
->bond
, &xr
.recirc_id
, &xr
.hash_basis
);
1768 /* Only TCP mode uses recirculation. */
1769 xr
.hash_alg
= OVS_HASH_ALG_L4
;
1770 bond_update_post_recirc_rules(out_xbundle
->bond
, false);
1772 /* Recirculation does not require unmasking hash fields. */
1777 ofport
= bond_choose_output_slave(out_xbundle
->bond
,
1778 &ctx
->xin
->flow
, wc
, vid
);
1779 xport
= xport_lookup(xcfg
, ofport
);
1782 /* No slaves enabled, so drop packet. */
1786 /* If use_recirc is set, the main thread will handle stats
1787 * accounting for this bond. */
1789 if (ctx
->xin
->resubmit_stats
) {
1790 bond_account(out_xbundle
->bond
, &ctx
->xin
->flow
, vid
,
1791 ctx
->xin
->resubmit_stats
->n_bytes
);
1793 if (ctx
->xin
->xcache
) {
1794 struct xc_entry
*entry
;
1797 flow
= &ctx
->xin
->flow
;
1798 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_BOND
);
1799 entry
->u
.bond
.bond
= bond_ref(out_xbundle
->bond
);
1800 entry
->u
.bond
.flow
= xmemdup(flow
, sizeof *flow
);
1801 entry
->u
.bond
.vid
= vid
;
1806 old_tci
= *flow_tci
;
1808 if (tci
|| out_xbundle
->use_priority_tags
) {
1809 tci
|= *flow_tci
& htons(VLAN_PCP_MASK
);
1811 tci
|= htons(VLAN_CFI
);
1816 compose_output_action(ctx
, xport
->ofp_port
, use_recirc
? &xr
: NULL
);
1817 *flow_tci
= old_tci
;
1820 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
1821 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
1822 * indicate this; newer upstream kernels use gratuitous ARP requests. */
1824 is_gratuitous_arp(const struct flow
*flow
, struct flow_wildcards
*wc
)
1826 if (flow
->dl_type
!= htons(ETH_TYPE_ARP
)) {
1830 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
1831 if (!eth_addr_is_broadcast(flow
->dl_dst
)) {
1835 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
1836 if (flow
->nw_proto
== ARP_OP_REPLY
) {
1838 } else if (flow
->nw_proto
== ARP_OP_REQUEST
) {
1839 memset(&wc
->masks
.nw_src
, 0xff, sizeof wc
->masks
.nw_src
);
1840 memset(&wc
->masks
.nw_dst
, 0xff, sizeof wc
->masks
.nw_dst
);
1842 return flow
->nw_src
== flow
->nw_dst
;
1848 /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or
1849 * dropped. Returns true if they may be forwarded, false if they should be
1852 * 'in_port' must be the xport that corresponds to flow->in_port.
1853 * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull).
1855 * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as
1856 * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as
1857 * checked by input_vid_is_valid().
1859 * May also add tags to '*tags', although the current implementation only does
1860 * so in one special case.
1863 is_admissible(struct xlate_ctx
*ctx
, struct xport
*in_port
,
1866 struct xbundle
*in_xbundle
= in_port
->xbundle
;
1867 const struct xbridge
*xbridge
= ctx
->xbridge
;
1868 struct flow
*flow
= &ctx
->xin
->flow
;
1870 /* Drop frames for reserved multicast addresses
1871 * only if forward_bpdu option is absent. */
1872 if (!xbridge
->forward_bpdu
&& eth_addr_is_reserved(flow
->dl_dst
)) {
1873 xlate_report(ctx
, "packet has reserved destination MAC, dropping");
1877 if (in_xbundle
->bond
) {
1878 struct mac_entry
*mac
;
1880 switch (bond_check_admissibility(in_xbundle
->bond
, in_port
->ofport
,
1886 xlate_report(ctx
, "bonding refused admissibility, dropping");
1889 case BV_DROP_IF_MOVED
:
1890 ovs_rwlock_rdlock(&xbridge
->ml
->rwlock
);
1891 mac
= mac_learning_lookup(xbridge
->ml
, flow
->dl_src
, vlan
);
1893 && mac_entry_get_port(xbridge
->ml
, mac
) != in_xbundle
->ofbundle
1894 && (!is_gratuitous_arp(flow
, ctx
->wc
)
1895 || mac_entry_is_grat_arp_locked(mac
))) {
1896 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
1897 xlate_report(ctx
, "SLB bond thinks this packet looped back, "
1901 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
1909 /* Checks whether a MAC learning update is necessary for MAC learning table
1910 * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in
1913 * Most packets processed through the MAC learning table do not actually
1914 * change it in any way. This function requires only a read lock on the MAC
1915 * learning table, so it is much cheaper in this common case.
1917 * Keep the code here synchronized with that in update_learning_table__()
1920 is_mac_learning_update_needed(const struct mac_learning
*ml
,
1921 const struct flow
*flow
,
1922 struct flow_wildcards
*wc
,
1923 int vlan
, struct xbundle
*in_xbundle
)
1924 OVS_REQ_RDLOCK(ml
->rwlock
)
1926 struct mac_entry
*mac
;
1928 if (!mac_learning_may_learn(ml
, flow
->dl_src
, vlan
)) {
1932 mac
= mac_learning_lookup(ml
, flow
->dl_src
, vlan
);
1933 if (!mac
|| mac_entry_age(ml
, mac
)) {
1937 if (is_gratuitous_arp(flow
, wc
)) {
1938 /* We don't want to learn from gratuitous ARP packets that are
1939 * reflected back over bond slaves so we lock the learning table. */
1940 if (!in_xbundle
->bond
) {
1942 } else if (mac_entry_is_grat_arp_locked(mac
)) {
1947 return mac_entry_get_port(ml
, mac
) != in_xbundle
->ofbundle
;
1951 /* Updates MAC learning table 'ml' given that a packet matching 'flow' was
1952 * received on 'in_xbundle' in 'vlan'.
1954 * This code repeats all the checks in is_mac_learning_update_needed() because
1955 * the lock was released between there and here and thus the MAC learning state
1956 * could have changed.
1958 * Keep the code here synchronized with that in is_mac_learning_update_needed()
1961 update_learning_table__(const struct xbridge
*xbridge
,
1962 const struct flow
*flow
, struct flow_wildcards
*wc
,
1963 int vlan
, struct xbundle
*in_xbundle
)
1964 OVS_REQ_WRLOCK(xbridge
->ml
->rwlock
)
1966 struct mac_entry
*mac
;
1968 if (!mac_learning_may_learn(xbridge
->ml
, flow
->dl_src
, vlan
)) {
1972 mac
= mac_learning_insert(xbridge
->ml
, flow
->dl_src
, vlan
);
1973 if (is_gratuitous_arp(flow
, wc
)) {
1974 /* We don't want to learn from gratuitous ARP packets that are
1975 * reflected back over bond slaves so we lock the learning table. */
1976 if (!in_xbundle
->bond
) {
1977 mac_entry_set_grat_arp_lock(mac
);
1978 } else if (mac_entry_is_grat_arp_locked(mac
)) {
1983 if (mac_entry_get_port(xbridge
->ml
, mac
) != in_xbundle
->ofbundle
) {
1984 /* The log messages here could actually be useful in debugging,
1985 * so keep the rate limit relatively high. */
1986 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(30, 300);
1988 VLOG_DBG_RL(&rl
, "bridge %s: learned that "ETH_ADDR_FMT
" is "
1989 "on port %s in VLAN %d",
1990 xbridge
->name
, ETH_ADDR_ARGS(flow
->dl_src
),
1991 in_xbundle
->name
, vlan
);
1993 mac_entry_set_port(xbridge
->ml
, mac
, in_xbundle
->ofbundle
);
1998 update_learning_table(const struct xbridge
*xbridge
,
1999 const struct flow
*flow
, struct flow_wildcards
*wc
,
2000 int vlan
, struct xbundle
*in_xbundle
)
2004 /* Don't learn the OFPP_NONE port. */
2005 if (in_xbundle
== &ofpp_none_bundle
) {
2009 /* First try the common case: no change to MAC learning table. */
2010 ovs_rwlock_rdlock(&xbridge
->ml
->rwlock
);
2011 need_update
= is_mac_learning_update_needed(xbridge
->ml
, flow
, wc
, vlan
,
2013 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
2016 /* Slow path: MAC learning table might need an update. */
2017 ovs_rwlock_wrlock(&xbridge
->ml
->rwlock
);
2018 update_learning_table__(xbridge
, flow
, wc
, vlan
, in_xbundle
);
2019 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
2023 /* Updates multicast snooping table 'ms' given that a packet matching 'flow'
2024 * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */
2026 update_mcast_snooping_table4__(const struct xbridge
*xbridge
,
2027 const struct flow
*flow
,
2028 struct mcast_snooping
*ms
, int vlan
,
2029 struct xbundle
*in_xbundle
,
2030 const struct dp_packet
*packet
)
2031 OVS_REQ_WRLOCK(ms
->rwlock
)
2033 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(60, 30);
2035 ovs_be32 ip4
= flow
->igmp_group_ip4
;
2037 switch (ntohs(flow
->tp_src
)) {
2038 case IGMP_HOST_MEMBERSHIP_REPORT
:
2039 case IGMPV2_HOST_MEMBERSHIP_REPORT
:
2040 if (mcast_snooping_add_group4(ms
, ip4
, vlan
, in_xbundle
->ofbundle
)) {
2041 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping learned that "
2042 IP_FMT
" is on port %s in VLAN %d",
2043 xbridge
->name
, IP_ARGS(ip4
), in_xbundle
->name
, vlan
);
2046 case IGMP_HOST_LEAVE_MESSAGE
:
2047 if (mcast_snooping_leave_group4(ms
, ip4
, vlan
, in_xbundle
->ofbundle
)) {
2048 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping leaving "
2049 IP_FMT
" is on port %s in VLAN %d",
2050 xbridge
->name
, IP_ARGS(ip4
), in_xbundle
->name
, vlan
);
2053 case IGMP_HOST_MEMBERSHIP_QUERY
:
2054 if (flow
->nw_src
&& mcast_snooping_add_mrouter(ms
, vlan
,
2055 in_xbundle
->ofbundle
)) {
2056 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping query from "
2057 IP_FMT
" is on port %s in VLAN %d",
2058 xbridge
->name
, IP_ARGS(flow
->nw_src
),
2059 in_xbundle
->name
, vlan
);
2062 case IGMPV3_HOST_MEMBERSHIP_REPORT
:
2063 if ((count
= mcast_snooping_add_report(ms
, packet
, vlan
,
2064 in_xbundle
->ofbundle
))) {
2065 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping processed %d "
2066 "addresses on port %s in VLAN %d",
2067 xbridge
->name
, count
, in_xbundle
->name
, vlan
);
2074 update_mcast_snooping_table6__(const struct xbridge
*xbridge
,
2075 const struct flow
*flow
,
2076 struct mcast_snooping
*ms
, int vlan
,
2077 struct xbundle
*in_xbundle
,
2078 const struct dp_packet
*packet
)
2079 OVS_REQ_WRLOCK(ms
->rwlock
)
2081 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(60, 30);
2084 switch (ntohs(flow
->tp_src
)) {
2086 if (!ipv6_addr_equals(&flow
->ipv6_src
, &in6addr_any
)
2087 && mcast_snooping_add_mrouter(ms
, vlan
, in_xbundle
->ofbundle
)) {
2088 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping query on port %s"
2090 xbridge
->name
, in_xbundle
->name
, vlan
);
2096 count
= mcast_snooping_add_mld(ms
, packet
, vlan
, in_xbundle
->ofbundle
);
2098 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping processed %d "
2099 "addresses on port %s in VLAN %d",
2100 xbridge
->name
, count
, in_xbundle
->name
, vlan
);
2106 /* Updates multicast snooping table 'ms' given that a packet matching 'flow'
2107 * was received on 'in_xbundle' in 'vlan'. */
2109 update_mcast_snooping_table(const struct xbridge
*xbridge
,
2110 const struct flow
*flow
, int vlan
,
2111 struct xbundle
*in_xbundle
,
2112 const struct dp_packet
*packet
)
2114 struct mcast_snooping
*ms
= xbridge
->ms
;
2115 struct xlate_cfg
*xcfg
;
2116 struct xbundle
*mcast_xbundle
;
2117 struct mcast_port_bundle
*fport
;
2119 /* Don't learn the OFPP_NONE port. */
2120 if (in_xbundle
== &ofpp_none_bundle
) {
2124 /* Don't learn from flood ports */
2125 mcast_xbundle
= NULL
;
2126 ovs_rwlock_wrlock(&ms
->rwlock
);
2127 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
2128 LIST_FOR_EACH(fport
, node
, &ms
->fport_list
) {
2129 mcast_xbundle
= xbundle_lookup(xcfg
, fport
->port
);
2130 if (mcast_xbundle
== in_xbundle
) {
2135 if (!mcast_xbundle
|| mcast_xbundle
!= in_xbundle
) {
2136 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
2137 update_mcast_snooping_table4__(xbridge
, flow
, ms
, vlan
,
2138 in_xbundle
, packet
);
2140 update_mcast_snooping_table6__(xbridge
, flow
, ms
, vlan
,
2141 in_xbundle
, packet
);
2144 ovs_rwlock_unlock(&ms
->rwlock
);
2147 /* send the packet to ports having the multicast group learned */
2149 xlate_normal_mcast_send_group(struct xlate_ctx
*ctx
,
2150 struct mcast_snooping
*ms OVS_UNUSED
,
2151 struct mcast_group
*grp
,
2152 struct xbundle
*in_xbundle
, uint16_t vlan
)
2153 OVS_REQ_RDLOCK(ms
->rwlock
)
2155 struct xlate_cfg
*xcfg
;
2156 struct mcast_group_bundle
*b
;
2157 struct xbundle
*mcast_xbundle
;
2159 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
2160 LIST_FOR_EACH(b
, bundle_node
, &grp
->bundle_lru
) {
2161 mcast_xbundle
= xbundle_lookup(xcfg
, b
->port
);
2162 if (mcast_xbundle
&& mcast_xbundle
!= in_xbundle
) {
2163 xlate_report(ctx
, "forwarding to mcast group port");
2164 output_normal(ctx
, mcast_xbundle
, vlan
);
2165 } else if (!mcast_xbundle
) {
2166 xlate_report(ctx
, "mcast group port is unknown, dropping");
2168 xlate_report(ctx
, "mcast group port is input port, dropping");
2173 /* send the packet to ports connected to multicast routers */
2175 xlate_normal_mcast_send_mrouters(struct xlate_ctx
*ctx
,
2176 struct mcast_snooping
*ms
,
2177 struct xbundle
*in_xbundle
, uint16_t vlan
)
2178 OVS_REQ_RDLOCK(ms
->rwlock
)
2180 struct xlate_cfg
*xcfg
;
2181 struct mcast_mrouter_bundle
*mrouter
;
2182 struct xbundle
*mcast_xbundle
;
2184 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
2185 LIST_FOR_EACH(mrouter
, mrouter_node
, &ms
->mrouter_lru
) {
2186 mcast_xbundle
= xbundle_lookup(xcfg
, mrouter
->port
);
2187 if (mcast_xbundle
&& mcast_xbundle
!= in_xbundle
) {
2188 xlate_report(ctx
, "forwarding to mcast router port");
2189 output_normal(ctx
, mcast_xbundle
, vlan
);
2190 } else if (!mcast_xbundle
) {
2191 xlate_report(ctx
, "mcast router port is unknown, dropping");
2193 xlate_report(ctx
, "mcast router port is input port, dropping");
2198 /* send the packet to ports flagged to be flooded */
2200 xlate_normal_mcast_send_fports(struct xlate_ctx
*ctx
,
2201 struct mcast_snooping
*ms
,
2202 struct xbundle
*in_xbundle
, uint16_t vlan
)
2203 OVS_REQ_RDLOCK(ms
->rwlock
)
2205 struct xlate_cfg
*xcfg
;
2206 struct mcast_port_bundle
*fport
;
2207 struct xbundle
*mcast_xbundle
;
2209 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
2210 LIST_FOR_EACH(fport
, node
, &ms
->fport_list
) {
2211 mcast_xbundle
= xbundle_lookup(xcfg
, fport
->port
);
2212 if (mcast_xbundle
&& mcast_xbundle
!= in_xbundle
) {
2213 xlate_report(ctx
, "forwarding to mcast flood port");
2214 output_normal(ctx
, mcast_xbundle
, vlan
);
2215 } else if (!mcast_xbundle
) {
2216 xlate_report(ctx
, "mcast flood port is unknown, dropping");
2218 xlate_report(ctx
, "mcast flood port is input port, dropping");
2223 /* forward the Reports to configured ports */
2225 xlate_normal_mcast_send_rports(struct xlate_ctx
*ctx
,
2226 struct mcast_snooping
*ms
,
2227 struct xbundle
*in_xbundle
, uint16_t vlan
)
2228 OVS_REQ_RDLOCK(ms
->rwlock
)
2230 struct xlate_cfg
*xcfg
;
2231 struct mcast_port_bundle
*rport
;
2232 struct xbundle
*mcast_xbundle
;
2234 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
2235 LIST_FOR_EACH(rport
, node
, &ms
->rport_list
) {
2236 mcast_xbundle
= xbundle_lookup(xcfg
, rport
->port
);
2237 if (mcast_xbundle
&& mcast_xbundle
!= in_xbundle
) {
2238 xlate_report(ctx
, "forwarding Report to mcast flagged port");
2239 output_normal(ctx
, mcast_xbundle
, vlan
);
2240 } else if (!mcast_xbundle
) {
2241 xlate_report(ctx
, "mcast port is unknown, dropping the Report");
2243 xlate_report(ctx
, "mcast port is input port, dropping the Report");
2249 xlate_normal_flood(struct xlate_ctx
*ctx
, struct xbundle
*in_xbundle
,
2252 struct xbundle
*xbundle
;
2254 LIST_FOR_EACH (xbundle
, list_node
, &ctx
->xbridge
->xbundles
) {
2255 if (xbundle
!= in_xbundle
2256 && xbundle_includes_vlan(xbundle
, vlan
)
2257 && xbundle
->floodable
2258 && !xbundle_mirror_out(ctx
->xbridge
, xbundle
)) {
2259 output_normal(ctx
, xbundle
, vlan
);
2262 ctx
->nf_output_iface
= NF_OUT_FLOOD
;
2266 xlate_normal(struct xlate_ctx
*ctx
)
2268 struct flow_wildcards
*wc
= ctx
->wc
;
2269 struct flow
*flow
= &ctx
->xin
->flow
;
2270 struct xbundle
*in_xbundle
;
2271 struct xport
*in_port
;
2272 struct mac_entry
*mac
;
2277 memset(&wc
->masks
.dl_src
, 0xff, sizeof wc
->masks
.dl_src
);
2278 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
2279 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
2281 in_xbundle
= lookup_input_bundle(ctx
->xbridge
, flow
->in_port
.ofp_port
,
2282 ctx
->xin
->packet
!= NULL
, &in_port
);
2284 xlate_report(ctx
, "no input bundle, dropping");
2288 /* Drop malformed frames. */
2289 if (flow
->dl_type
== htons(ETH_TYPE_VLAN
) &&
2290 !(flow
->vlan_tci
& htons(VLAN_CFI
))) {
2291 if (ctx
->xin
->packet
!= NULL
) {
2292 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2293 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet with partial "
2294 "VLAN tag received on port %s",
2295 ctx
->xbridge
->name
, in_xbundle
->name
);
2297 xlate_report(ctx
, "partial VLAN tag, dropping");
2301 /* Drop frames on bundles reserved for mirroring. */
2302 if (xbundle_mirror_out(ctx
->xbridge
, in_xbundle
)) {
2303 if (ctx
->xin
->packet
!= NULL
) {
2304 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2305 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet received on port "
2306 "%s, which is reserved exclusively for mirroring",
2307 ctx
->xbridge
->name
, in_xbundle
->name
);
2309 xlate_report(ctx
, "input port is mirror output port, dropping");
2314 vid
= vlan_tci_to_vid(flow
->vlan_tci
);
2315 if (!input_vid_is_valid(vid
, in_xbundle
, ctx
->xin
->packet
!= NULL
)) {
2316 xlate_report(ctx
, "disallowed VLAN VID for this input port, dropping");
2319 vlan
= input_vid_to_vlan(in_xbundle
, vid
);
2321 /* Check other admissibility requirements. */
2322 if (in_port
&& !is_admissible(ctx
, in_port
, vlan
)) {
2326 /* Learn source MAC. */
2327 if (ctx
->xin
->may_learn
) {
2328 update_learning_table(ctx
->xbridge
, flow
, wc
, vlan
, in_xbundle
);
2330 if (ctx
->xin
->xcache
) {
2331 struct xc_entry
*entry
;
2333 /* Save enough info to update mac learning table later. */
2334 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_NORMAL
);
2335 entry
->u
.normal
.ofproto
= ctx
->xbridge
->ofproto
;
2336 entry
->u
.normal
.flow
= xmemdup(flow
, sizeof *flow
);
2337 entry
->u
.normal
.vlan
= vlan
;
2340 /* Determine output bundle. */
2341 if (mcast_snooping_enabled(ctx
->xbridge
->ms
)
2342 && !eth_addr_is_broadcast(flow
->dl_dst
)
2343 && eth_addr_is_multicast(flow
->dl_dst
)
2344 && is_ip_any(flow
)) {
2345 struct mcast_snooping
*ms
= ctx
->xbridge
->ms
;
2346 struct mcast_group
*grp
= NULL
;
2348 if (is_igmp(flow
)) {
2349 if (mcast_snooping_is_membership(flow
->tp_src
) ||
2350 mcast_snooping_is_query(flow
->tp_src
)) {
2351 if (ctx
->xin
->may_learn
) {
2352 update_mcast_snooping_table(ctx
->xbridge
, flow
, vlan
,
2353 in_xbundle
, ctx
->xin
->packet
);
2356 * IGMP packets need to take the slow path, in order to be
2357 * processed for mdb updates. That will prevent expires
2358 * firing off even after hosts have sent reports.
2360 ctx
->xout
->slow
|= SLOW_ACTION
;
2363 if (mcast_snooping_is_membership(flow
->tp_src
)) {
2364 ovs_rwlock_rdlock(&ms
->rwlock
);
2365 xlate_normal_mcast_send_mrouters(ctx
, ms
, in_xbundle
, vlan
);
2366 /* RFC4541: section 2.1.1, item 1: A snooping switch should
2367 * forward IGMP Membership Reports only to those ports where
2368 * multicast routers are attached. Alternatively stated: a
2369 * snooping switch should not forward IGMP Membership Reports
2370 * to ports on which only hosts are attached.
2371 * An administrative control may be provided to override this
2372 * restriction, allowing the report messages to be flooded to
2374 xlate_normal_mcast_send_rports(ctx
, ms
, in_xbundle
, vlan
);
2375 ovs_rwlock_unlock(&ms
->rwlock
);
2377 xlate_report(ctx
, "multicast traffic, flooding");
2378 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2381 } else if (is_mld(flow
)) {
2382 ctx
->xout
->slow
|= SLOW_ACTION
;
2383 if (ctx
->xin
->may_learn
) {
2384 update_mcast_snooping_table(ctx
->xbridge
, flow
, vlan
,
2385 in_xbundle
, ctx
->xin
->packet
);
2387 if (is_mld_report(flow
)) {
2388 ovs_rwlock_rdlock(&ms
->rwlock
);
2389 xlate_normal_mcast_send_mrouters(ctx
, ms
, in_xbundle
, vlan
);
2390 xlate_normal_mcast_send_rports(ctx
, ms
, in_xbundle
, vlan
);
2391 ovs_rwlock_unlock(&ms
->rwlock
);
2393 xlate_report(ctx
, "MLD query, flooding");
2394 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2397 if ((flow
->dl_type
== htons(ETH_TYPE_IP
)
2398 && ip_is_local_multicast(flow
->nw_dst
))
2399 || (flow
->dl_type
== htons(ETH_TYPE_IPV6
)
2400 && ipv6_is_all_hosts(&flow
->ipv6_dst
))) {
2401 /* RFC4541: section 2.1.2, item 2: Packets with a dst IP
2402 * address in the 224.0.0.x range which are not IGMP must
2403 * be forwarded on all ports */
2404 xlate_report(ctx
, "RFC4541: section 2.1.2, item 2, flooding");
2405 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2410 /* forwarding to group base ports */
2411 ovs_rwlock_rdlock(&ms
->rwlock
);
2412 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
2413 grp
= mcast_snooping_lookup4(ms
, flow
->nw_dst
, vlan
);
2414 } else if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
2415 grp
= mcast_snooping_lookup(ms
, &flow
->ipv6_dst
, vlan
);
2418 xlate_normal_mcast_send_group(ctx
, ms
, grp
, in_xbundle
, vlan
);
2419 xlate_normal_mcast_send_fports(ctx
, ms
, in_xbundle
, vlan
);
2420 xlate_normal_mcast_send_mrouters(ctx
, ms
, in_xbundle
, vlan
);
2422 if (mcast_snooping_flood_unreg(ms
)) {
2423 xlate_report(ctx
, "unregistered multicast, flooding");
2424 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2426 xlate_normal_mcast_send_mrouters(ctx
, ms
, in_xbundle
, vlan
);
2427 xlate_normal_mcast_send_fports(ctx
, ms
, in_xbundle
, vlan
);
2430 ovs_rwlock_unlock(&ms
->rwlock
);
2432 ovs_rwlock_rdlock(&ctx
->xbridge
->ml
->rwlock
);
2433 mac
= mac_learning_lookup(ctx
->xbridge
->ml
, flow
->dl_dst
, vlan
);
2434 mac_port
= mac
? mac_entry_get_port(ctx
->xbridge
->ml
, mac
) : NULL
;
2435 ovs_rwlock_unlock(&ctx
->xbridge
->ml
->rwlock
);
2438 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
2439 struct xbundle
*mac_xbundle
= xbundle_lookup(xcfg
, mac_port
);
2440 if (mac_xbundle
&& mac_xbundle
!= in_xbundle
) {
2441 xlate_report(ctx
, "forwarding to learned port");
2442 output_normal(ctx
, mac_xbundle
, vlan
);
2443 } else if (!mac_xbundle
) {
2444 xlate_report(ctx
, "learned port is unknown, dropping");
2446 xlate_report(ctx
, "learned port is input port, dropping");
2449 xlate_report(ctx
, "no learned MAC for destination, flooding");
2450 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2455 /* Appends a "sample" action for sFlow or IPFIX to 'ctx->odp_actions'. The
2456 * 'probability' is the number of packets out of UINT32_MAX to sample. The
2457 * 'cookie' (of length 'cookie_size' bytes) is passed back in the callback for
2458 * each sampled packet. 'tunnel_out_port', if not ODPP_NONE, is added as the
2459 * OVS_USERSPACE_ATTR_EGRESS_TUN_PORT attribute. If 'include_actions', an
2460 * OVS_USERSPACE_ATTR_ACTIONS attribute is added.
2463 compose_sample_action(struct xlate_ctx
*ctx
,
2464 const uint32_t probability
,
2465 const union user_action_cookie
*cookie
,
2466 const size_t cookie_size
,
2467 const odp_port_t tunnel_out_port
,
2468 bool include_actions
)
2470 size_t sample_offset
= nl_msg_start_nested(ctx
->odp_actions
,
2471 OVS_ACTION_ATTR_SAMPLE
);
2473 nl_msg_put_u32(ctx
->odp_actions
, OVS_SAMPLE_ATTR_PROBABILITY
, probability
);
2475 size_t actions_offset
= nl_msg_start_nested(ctx
->odp_actions
,
2476 OVS_SAMPLE_ATTR_ACTIONS
);
2478 odp_port_t odp_port
= ofp_port_to_odp_port(
2479 ctx
->xbridge
, ctx
->xin
->flow
.in_port
.ofp_port
);
2480 uint32_t pid
= dpif_port_get_pid(ctx
->xbridge
->dpif
, odp_port
,
2481 flow_hash_5tuple(&ctx
->xin
->flow
, 0));
2482 int cookie_offset
= odp_put_userspace_action(pid
, cookie
, cookie_size
,
2487 nl_msg_end_nested(ctx
->odp_actions
, actions_offset
);
2488 nl_msg_end_nested(ctx
->odp_actions
, sample_offset
);
2490 return cookie_offset
;
2493 /* If sFLow is not enabled, returns 0 without doing anything.
2495 * If sFlow is enabled, appends a template "sample" action to the ODP actions
2496 * in 'ctx'. This action is a template because some of the information needed
2497 * to fill it out is not available until flow translation is complete. In this
2498 * case, this functions returns an offset, which is always nonzero, to pass
2499 * later to fix_sflow_action() to fill in the rest of the template. */
2501 compose_sflow_action(struct xlate_ctx
*ctx
)
2503 struct dpif_sflow
*sflow
= ctx
->xbridge
->sflow
;
2504 if (!sflow
|| ctx
->xin
->flow
.in_port
.ofp_port
== OFPP_NONE
) {
2508 union user_action_cookie cookie
= { .type
= USER_ACTION_COOKIE_SFLOW
};
2509 return compose_sample_action(ctx
, dpif_sflow_get_probability(sflow
),
2510 &cookie
, sizeof cookie
.sflow
, ODPP_NONE
,
2514 /* If IPFIX is enabled, this appends a "sample" action to implement IPFIX to
2515 * 'ctx->odp_actions'. */
2517 compose_ipfix_action(struct xlate_ctx
*ctx
, odp_port_t output_odp_port
)
2519 struct dpif_ipfix
*ipfix
= ctx
->xbridge
->ipfix
;
2520 odp_port_t tunnel_out_port
= ODPP_NONE
;
2522 if (!ipfix
|| ctx
->xin
->flow
.in_port
.ofp_port
== OFPP_NONE
) {
2526 /* For input case, output_odp_port is ODPP_NONE, which is an invalid port
2528 if (output_odp_port
== ODPP_NONE
&&
2529 !dpif_ipfix_get_bridge_exporter_input_sampling(ipfix
)) {
2533 /* For output case, output_odp_port is valid*/
2534 if (output_odp_port
!= ODPP_NONE
) {
2535 if (!dpif_ipfix_get_bridge_exporter_output_sampling(ipfix
)) {
2538 /* If tunnel sampling is enabled, put an additional option attribute:
2539 * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT
2541 if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(ipfix
) &&
2542 dpif_ipfix_get_tunnel_port(ipfix
, output_odp_port
) ) {
2543 tunnel_out_port
= output_odp_port
;
2547 union user_action_cookie cookie
= {
2549 .type
= USER_ACTION_COOKIE_IPFIX
,
2550 .output_odp_port
= output_odp_port
,
2553 compose_sample_action(ctx
,
2554 dpif_ipfix_get_bridge_exporter_probability(ipfix
),
2555 &cookie
, sizeof cookie
.ipfix
, tunnel_out_port
,
2559 /* Fix "sample" action according to data collected while composing ODP actions,
2560 * as described in compose_sflow_action().
2562 * 'user_cookie_offset' must be the offset returned by add_sflow_action(). */
2564 fix_sflow_action(struct xlate_ctx
*ctx
, unsigned int user_cookie_offset
)
2566 const struct flow
*base
= &ctx
->base_flow
;
2567 union user_action_cookie
*cookie
;
2569 cookie
= ofpbuf_at(ctx
->odp_actions
, user_cookie_offset
,
2570 sizeof cookie
->sflow
);
2571 ovs_assert(cookie
->type
== USER_ACTION_COOKIE_SFLOW
);
2573 cookie
->type
= USER_ACTION_COOKIE_SFLOW
;
2574 cookie
->sflow
.vlan_tci
= base
->vlan_tci
;
2576 /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output
2577 * port information") for the interpretation of cookie->output. */
2578 switch (ctx
->sflow_n_outputs
) {
2580 /* 0x40000000 | 256 means "packet dropped for unknown reason". */
2581 cookie
->sflow
.output
= 0x40000000 | 256;
2585 cookie
->sflow
.output
= dpif_sflow_odp_port_to_ifindex(
2586 ctx
->xbridge
->sflow
, ctx
->sflow_odp_port
);
2587 if (cookie
->sflow
.output
) {
2592 /* 0x80000000 means "multiple output ports. */
2593 cookie
->sflow
.output
= 0x80000000 | ctx
->sflow_n_outputs
;
2599 process_special(struct xlate_ctx
*ctx
, const struct xport
*xport
)
2601 const struct flow
*flow
= &ctx
->xin
->flow
;
2602 struct flow_wildcards
*wc
= ctx
->wc
;
2603 const struct xbridge
*xbridge
= ctx
->xbridge
;
2604 const struct dp_packet
*packet
= ctx
->xin
->packet
;
2605 enum slow_path_reason slow
;
2609 } else if (xport
->cfm
&& cfm_should_process_flow(xport
->cfm
, flow
, wc
)) {
2611 cfm_process_heartbeat(xport
->cfm
, packet
);
2614 } else if (xport
->bfd
&& bfd_should_process_flow(xport
->bfd
, flow
, wc
)) {
2616 bfd_process_packet(xport
->bfd
, flow
, packet
);
2617 /* If POLL received, immediately sends FINAL back. */
2618 if (bfd_should_send_packet(xport
->bfd
)) {
2619 ofproto_dpif_monitor_port_send_soon(xport
->ofport
);
2623 } else if (xport
->xbundle
&& xport
->xbundle
->lacp
2624 && flow
->dl_type
== htons(ETH_TYPE_LACP
)) {
2626 lacp_process_packet(xport
->xbundle
->lacp
, xport
->ofport
, packet
);
2629 } else if ((xbridge
->stp
|| xbridge
->rstp
) &&
2630 stp_should_process_flow(flow
, wc
)) {
2633 ? stp_process_packet(xport
, packet
)
2634 : rstp_process_packet(xport
, packet
);
2637 } else if (xport
->lldp
&& lldp_should_process_flow(xport
->lldp
, flow
)) {
2639 lldp_process_packet(xport
->lldp
, packet
);
2647 ctx
->xout
->slow
|= slow
;
2655 tnl_route_lookup_flow(const struct flow
*oflow
,
2656 ovs_be32
*ip
, struct xport
**out_port
)
2658 char out_dev
[IFNAMSIZ
];
2659 struct xbridge
*xbridge
;
2660 struct xlate_cfg
*xcfg
;
2663 if (!ovs_router_lookup(oflow
->tunnel
.ip_dst
, out_dev
, &gw
)) {
2670 *ip
= oflow
->tunnel
.ip_dst
;
2673 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
2676 HMAP_FOR_EACH (xbridge
, hmap_node
, &xcfg
->xbridges
) {
2677 if (!strncmp(xbridge
->name
, out_dev
, IFNAMSIZ
)) {
2680 HMAP_FOR_EACH (port
, ofp_node
, &xbridge
->xports
) {
2681 if (!strncmp(netdev_get_name(port
->netdev
), out_dev
, IFNAMSIZ
)) {
2692 xlate_flood_packet(struct xbridge
*xbridge
, struct dp_packet
*packet
)
2694 struct ofpact_output output
;
2697 ofpact_init(&output
.ofpact
, OFPACT_OUTPUT
, sizeof output
);
2698 /* Use OFPP_NONE as the in_port to avoid special packet processing. */
2699 flow_extract(packet
, &flow
);
2700 flow
.in_port
.ofp_port
= OFPP_NONE
;
2701 output
.port
= OFPP_FLOOD
;
2704 return ofproto_dpif_execute_actions(xbridge
->ofproto
, &flow
, NULL
,
2705 &output
.ofpact
, sizeof output
,
2710 tnl_send_arp_request(const struct xport
*out_dev
,
2711 const struct eth_addr eth_src
,
2712 ovs_be32 ip_src
, ovs_be32 ip_dst
)
2714 struct xbridge
*xbridge
= out_dev
->xbridge
;
2715 struct dp_packet packet
;
2717 dp_packet_init(&packet
, 0);
2718 compose_arp(&packet
, ARP_OP_REQUEST
,
2719 eth_src
, eth_addr_zero
, true, ip_src
, ip_dst
);
2721 xlate_flood_packet(xbridge
, &packet
);
2722 dp_packet_uninit(&packet
);
2726 build_tunnel_send(struct xlate_ctx
*ctx
, const struct xport
*xport
,
2727 const struct flow
*flow
, odp_port_t tunnel_odp_port
)
2729 struct ovs_action_push_tnl tnl_push_data
;
2730 struct xport
*out_dev
= NULL
;
2731 ovs_be32 s_ip
, d_ip
= 0;
2732 struct eth_addr smac
;
2733 struct eth_addr dmac
;
2736 err
= tnl_route_lookup_flow(flow
, &d_ip
, &out_dev
);
2738 xlate_report(ctx
, "native tunnel routing failed");
2741 xlate_report(ctx
, "tunneling to "IP_FMT
" via %s",
2742 IP_ARGS(d_ip
), netdev_get_name(out_dev
->netdev
));
2744 /* Use mac addr of bridge port of the peer. */
2745 err
= netdev_get_etheraddr(out_dev
->netdev
, &smac
);
2747 xlate_report(ctx
, "tunnel output device lacks Ethernet address");
2751 err
= netdev_get_in4(out_dev
->netdev
, (struct in_addr
*) &s_ip
, NULL
);
2753 xlate_report(ctx
, "tunnel output device lacks IPv4 address");
2757 err
= tnl_arp_lookup(out_dev
->xbridge
->name
, d_ip
, &dmac
);
2759 xlate_report(ctx
, "ARP cache miss for "IP_FMT
" on bridge %s, "
2760 "sending ARP request",
2761 IP_ARGS(d_ip
), out_dev
->xbridge
->name
);
2762 tnl_send_arp_request(out_dev
, smac
, s_ip
, d_ip
);
2765 if (ctx
->xin
->xcache
) {
2766 struct xc_entry
*entry
;
2768 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_TNL_ARP
);
2769 ovs_strlcpy(entry
->u
.tnl_arp_cache
.br_name
, out_dev
->xbridge
->name
,
2770 sizeof entry
->u
.tnl_arp_cache
.br_name
);
2771 entry
->u
.tnl_arp_cache
.d_ip
= d_ip
;
2774 xlate_report(ctx
, "tunneling from "ETH_ADDR_FMT
" "IP_FMT
2775 " to "ETH_ADDR_FMT
" "IP_FMT
,
2776 ETH_ADDR_ARGS(smac
), IP_ARGS(s_ip
),
2777 ETH_ADDR_ARGS(dmac
), IP_ARGS(d_ip
));
2778 err
= tnl_port_build_header(xport
->ofport
, flow
,
2779 dmac
, smac
, s_ip
, &tnl_push_data
);
2783 tnl_push_data
.tnl_port
= odp_to_u32(tunnel_odp_port
);
2784 tnl_push_data
.out_port
= odp_to_u32(out_dev
->odp_port
);
2785 odp_put_tnl_push_action(ctx
->odp_actions
, &tnl_push_data
);
2790 compose_output_action__(struct xlate_ctx
*ctx
, ofp_port_t ofp_port
,
2791 const struct xlate_bond_recirc
*xr
, bool check_stp
)
2793 const struct xport
*xport
= get_ofp_port(ctx
->xbridge
, ofp_port
);
2794 struct flow_wildcards
*wc
= ctx
->wc
;
2795 struct flow
*flow
= &ctx
->xin
->flow
;
2796 struct flow_tnl flow_tnl
;
2797 ovs_be16 flow_vlan_tci
;
2798 uint32_t flow_pkt_mark
;
2799 uint8_t flow_nw_tos
;
2800 odp_port_t out_port
, odp_port
;
2801 bool tnl_push_pop_send
= false;
2804 /* If 'struct flow' gets additional metadata, we'll need to zero it out
2805 * before traversing a patch port. */
2806 BUILD_ASSERT_DECL(FLOW_WC_SEQ
== 33);
2807 memset(&flow_tnl
, 0, sizeof flow_tnl
);
2810 xlate_report(ctx
, "Nonexistent output port");
2812 } else if (xport
->config
& OFPUTIL_PC_NO_FWD
) {
2813 xlate_report(ctx
, "OFPPC_NO_FWD set, skipping output");
2815 } else if (check_stp
) {
2816 if (is_stp(&ctx
->base_flow
)) {
2817 if (!xport_stp_should_forward_bpdu(xport
) &&
2818 !xport_rstp_should_manage_bpdu(xport
)) {
2819 if (ctx
->xbridge
->stp
!= NULL
) {
2820 xlate_report(ctx
, "STP not in listening state, "
2821 "skipping bpdu output");
2822 } else if (ctx
->xbridge
->rstp
!= NULL
) {
2823 xlate_report(ctx
, "RSTP not managing BPDU in this state, "
2824 "skipping bpdu output");
2828 } else if (!xport_stp_forward_state(xport
) ||
2829 !xport_rstp_forward_state(xport
)) {
2830 if (ctx
->xbridge
->stp
!= NULL
) {
2831 xlate_report(ctx
, "STP not in forwarding state, "
2833 } else if (ctx
->xbridge
->rstp
!= NULL
) {
2834 xlate_report(ctx
, "RSTP not in forwarding state, "
2842 const struct xport
*peer
= xport
->peer
;
2843 struct flow old_flow
= ctx
->xin
->flow
;
2844 bool old_was_mpls
= ctx
->was_mpls
;
2845 cls_version_t old_version
= ctx
->tables_version
;
2846 struct ofpbuf old_stack
= ctx
->stack
;
2847 union mf_subvalue new_stack
[1024 / sizeof(union mf_subvalue
)];
2848 struct ofpbuf old_action_set
= ctx
->action_set
;
2849 uint64_t actset_stub
[1024 / 8];
2851 ofpbuf_use_stub(&ctx
->stack
, new_stack
, sizeof new_stack
);
2852 ofpbuf_use_stub(&ctx
->action_set
, actset_stub
, sizeof actset_stub
);
2853 ctx
->xbridge
= peer
->xbridge
;
2854 flow
->in_port
.ofp_port
= peer
->ofp_port
;
2855 flow
->metadata
= htonll(0);
2856 memset(&flow
->tunnel
, 0, sizeof flow
->tunnel
);
2857 memset(flow
->regs
, 0, sizeof flow
->regs
);
2858 flow
->actset_output
= OFPP_UNSET
;
2860 /* The bridge is now known so obtain its table version. */
2862 = ofproto_dpif_get_tables_version(ctx
->xbridge
->ofproto
);
2864 if (!process_special(ctx
, peer
) && may_receive(peer
, ctx
)) {
2865 if (xport_stp_forward_state(peer
) && xport_rstp_forward_state(peer
)) {
2866 xlate_table_action(ctx
, flow
->in_port
.ofp_port
, 0, true, true);
2867 if (ctx
->action_set
.size
) {
2868 /* Translate action set only if not dropping the packet and
2869 * not recirculating. */
2870 if (!exit_recirculates(ctx
)) {
2871 xlate_action_set(ctx
);
2874 /* Check if need to recirculate. */
2875 if (exit_recirculates(ctx
)) {
2876 compose_recirculate_action(ctx
);
2879 /* Forwarding is disabled by STP and RSTP. Let OFPP_NORMAL and
2880 * the learning action look at the packet, then drop it. */
2881 struct flow old_base_flow
= ctx
->base_flow
;
2882 size_t old_size
= ctx
->odp_actions
->size
;
2883 mirror_mask_t old_mirrors
= ctx
->mirrors
;
2885 xlate_table_action(ctx
, flow
->in_port
.ofp_port
, 0, true, true);
2886 ctx
->mirrors
= old_mirrors
;
2887 ctx
->base_flow
= old_base_flow
;
2888 ctx
->odp_actions
->size
= old_size
;
2890 /* Undo changes that may have been done for recirculation. */
2891 if (exit_recirculates(ctx
)) {
2892 ctx
->action_set
.size
= ctx
->recirc_action_offset
;
2893 ctx
->recirc_action_offset
= -1;
2894 ctx
->last_unroll_offset
= -1;
2899 ctx
->xin
->flow
= old_flow
;
2900 ctx
->xbridge
= xport
->xbridge
;
2901 ofpbuf_uninit(&ctx
->action_set
);
2902 ctx
->action_set
= old_action_set
;
2903 ofpbuf_uninit(&ctx
->stack
);
2904 ctx
->stack
= old_stack
;
2906 /* Restore calling bridge's lookup version. */
2907 ctx
->tables_version
= old_version
;
2909 /* The peer bridge popping MPLS should have no effect on the original
2911 ctx
->was_mpls
= old_was_mpls
;
2913 /* The fact that the peer bridge exits (for any reason) does not mean
2914 * that the original bridge should exit. Specifically, if the peer
2915 * bridge recirculates (which typically modifies the packet), the
2916 * original bridge must continue processing with the original, not the
2917 * recirculated packet! */
2920 if (ctx
->xin
->resubmit_stats
) {
2921 netdev_vport_inc_tx(xport
->netdev
, ctx
->xin
->resubmit_stats
);
2922 netdev_vport_inc_rx(peer
->netdev
, ctx
->xin
->resubmit_stats
);
2924 bfd_account_rx(peer
->bfd
, ctx
->xin
->resubmit_stats
);
2927 if (ctx
->xin
->xcache
) {
2928 struct xc_entry
*entry
;
2930 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_NETDEV
);
2931 entry
->u
.dev
.tx
= netdev_ref(xport
->netdev
);
2932 entry
->u
.dev
.rx
= netdev_ref(peer
->netdev
);
2933 entry
->u
.dev
.bfd
= bfd_ref(peer
->bfd
);
2938 flow_vlan_tci
= flow
->vlan_tci
;
2939 flow_pkt_mark
= flow
->pkt_mark
;
2940 flow_nw_tos
= flow
->nw_tos
;
2942 if (count_skb_priorities(xport
)) {
2943 memset(&wc
->masks
.skb_priority
, 0xff, sizeof wc
->masks
.skb_priority
);
2944 if (dscp_from_skb_priority(xport
, flow
->skb_priority
, &dscp
)) {
2945 wc
->masks
.nw_tos
|= IP_DSCP_MASK
;
2946 flow
->nw_tos
&= ~IP_DSCP_MASK
;
2947 flow
->nw_tos
|= dscp
;
2951 if (xport
->is_tunnel
) {
2952 /* Save tunnel metadata so that changes made due to
2953 * the Logical (tunnel) Port are not visible for any further
2954 * matches, while explicit set actions on tunnel metadata are.
2956 flow_tnl
= flow
->tunnel
;
2957 odp_port
= tnl_port_send(xport
->ofport
, flow
, ctx
->wc
);
2958 if (odp_port
== ODPP_NONE
) {
2959 xlate_report(ctx
, "Tunneling decided against output");
2960 goto out
; /* restore flow_nw_tos */
2962 if (flow
->tunnel
.ip_dst
== ctx
->orig_tunnel_ip_dst
) {
2963 xlate_report(ctx
, "Not tunneling to our own address");
2964 goto out
; /* restore flow_nw_tos */
2966 if (ctx
->xin
->resubmit_stats
) {
2967 netdev_vport_inc_tx(xport
->netdev
, ctx
->xin
->resubmit_stats
);
2969 if (ctx
->xin
->xcache
) {
2970 struct xc_entry
*entry
;
2972 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_NETDEV
);
2973 entry
->u
.dev
.tx
= netdev_ref(xport
->netdev
);
2975 out_port
= odp_port
;
2976 if (ovs_native_tunneling_is_on(ctx
->xbridge
->ofproto
)) {
2977 xlate_report(ctx
, "output to native tunnel");
2978 tnl_push_pop_send
= true;
2980 xlate_report(ctx
, "output to kernel tunnel");
2981 commit_odp_tunnel_action(flow
, &ctx
->base_flow
, ctx
->odp_actions
);
2982 flow
->tunnel
= flow_tnl
; /* Restore tunnel metadata */
2985 odp_port
= xport
->odp_port
;
2986 out_port
= odp_port
;
2987 if (ofproto_has_vlan_splinters(ctx
->xbridge
->ofproto
)) {
2988 ofp_port_t vlandev_port
;
2990 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
2991 vlandev_port
= vsp_realdev_to_vlandev(ctx
->xbridge
->ofproto
,
2992 ofp_port
, flow
->vlan_tci
);
2993 if (vlandev_port
!= ofp_port
) {
2994 out_port
= ofp_port_to_odp_port(ctx
->xbridge
, vlandev_port
);
2995 flow
->vlan_tci
= htons(0);
3000 if (out_port
!= ODPP_NONE
) {
3001 bool use_masked
= ctx
->xbridge
->support
.masked_set_action
;
3003 ctx
->xout
->slow
|= commit_odp_actions(flow
, &ctx
->base_flow
,
3008 struct ovs_action_hash
*act_hash
;
3011 act_hash
= nl_msg_put_unspec_uninit(ctx
->odp_actions
,
3012 OVS_ACTION_ATTR_HASH
,
3014 act_hash
->hash_alg
= xr
->hash_alg
;
3015 act_hash
->hash_basis
= xr
->hash_basis
;
3017 /* Recirc action. */
3018 nl_msg_put_u32(ctx
->odp_actions
, OVS_ACTION_ATTR_RECIRC
,
3022 if (tnl_push_pop_send
) {
3023 build_tunnel_send(ctx
, xport
, flow
, odp_port
);
3024 flow
->tunnel
= flow_tnl
; /* Restore tunnel metadata */
3026 odp_port_t odp_tnl_port
= ODPP_NONE
;
3028 /* XXX: Write better Filter for tunnel port. We can use inport
3029 * int tunnel-port flow to avoid these checks completely. */
3030 if (ofp_port
== OFPP_LOCAL
&&
3031 ovs_native_tunneling_is_on(ctx
->xbridge
->ofproto
)) {
3033 odp_tnl_port
= tnl_port_map_lookup(flow
, wc
);
3036 if (odp_tnl_port
!= ODPP_NONE
) {
3037 nl_msg_put_odp_port(ctx
->odp_actions
,
3038 OVS_ACTION_ATTR_TUNNEL_POP
,
3041 /* Tunnel push-pop action is not compatible with
3043 compose_ipfix_action(ctx
, out_port
);
3044 nl_msg_put_odp_port(ctx
->odp_actions
,
3045 OVS_ACTION_ATTR_OUTPUT
,
3051 ctx
->sflow_odp_port
= odp_port
;
3052 ctx
->sflow_n_outputs
++;
3053 ctx
->nf_output_iface
= ofp_port
;
3056 if (mbridge_has_mirrors(ctx
->xbridge
->mbridge
) && xport
->xbundle
) {
3057 mirror_packet(ctx
, xport
->xbundle
,
3058 xbundle_mirror_dst(xport
->xbundle
->xbridge
,
3064 flow
->vlan_tci
= flow_vlan_tci
;
3065 flow
->pkt_mark
= flow_pkt_mark
;
3066 flow
->nw_tos
= flow_nw_tos
;
3070 compose_output_action(struct xlate_ctx
*ctx
, ofp_port_t ofp_port
,
3071 const struct xlate_bond_recirc
*xr
)
3073 compose_output_action__(ctx
, ofp_port
, xr
, true);
3077 xlate_recursively(struct xlate_ctx
*ctx
, struct rule_dpif
*rule
)
3079 struct rule_dpif
*old_rule
= ctx
->rule
;
3080 ovs_be64 old_cookie
= ctx
->rule_cookie
;
3081 const struct rule_actions
*actions
;
3083 if (ctx
->xin
->resubmit_stats
) {
3084 rule_dpif_credit_stats(rule
, ctx
->xin
->resubmit_stats
);
3090 ctx
->rule_cookie
= rule_dpif_get_flow_cookie(rule
);
3091 actions
= rule_dpif_get_actions(rule
);
3092 do_xlate_actions(actions
->ofpacts
, actions
->ofpacts_len
, ctx
);
3093 ctx
->rule_cookie
= old_cookie
;
3094 ctx
->rule
= old_rule
;
3099 xlate_resubmit_resource_check(struct xlate_ctx
*ctx
)
3101 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
3103 if (ctx
->recurse
>= MAX_RESUBMIT_RECURSION
+ MAX_INTERNAL_RESUBMITS
) {
3104 VLOG_ERR_RL(&rl
, "resubmit actions recursed over %d times",
3105 MAX_RESUBMIT_RECURSION
);
3106 } else if (ctx
->resubmits
>= MAX_RESUBMITS
+ MAX_INTERNAL_RESUBMITS
) {
3107 VLOG_ERR_RL(&rl
, "over %d resubmit actions", MAX_RESUBMITS
);
3108 } else if (ctx
->odp_actions
->size
> UINT16_MAX
) {
3109 VLOG_ERR_RL(&rl
, "resubmits yielded over 64 kB of actions");
3110 } else if (ctx
->stack
.size
>= 65536) {
3111 VLOG_ERR_RL(&rl
, "resubmits yielded over 64 kB of stack");
3120 xlate_table_action(struct xlate_ctx
*ctx
, ofp_port_t in_port
, uint8_t table_id
,
3121 bool may_packet_in
, bool honor_table_miss
)
3123 /* Check if we need to recirculate before matching in a table. */
3124 if (ctx
->was_mpls
) {
3125 ctx_trigger_recirculation(ctx
);
3128 if (xlate_resubmit_resource_check(ctx
)) {
3129 uint8_t old_table_id
= ctx
->table_id
;
3130 struct rule_dpif
*rule
;
3132 ctx
->table_id
= table_id
;
3134 rule
= rule_dpif_lookup_from_table(ctx
->xbridge
->ofproto
,
3135 ctx
->tables_version
,
3136 &ctx
->xin
->flow
, ctx
->xin
->wc
,
3137 ctx
->xin
->resubmit_stats
,
3138 &ctx
->table_id
, in_port
,
3139 may_packet_in
, honor_table_miss
);
3141 if (OVS_UNLIKELY(ctx
->xin
->resubmit_hook
)) {
3142 ctx
->xin
->resubmit_hook(ctx
->xin
, rule
, ctx
->recurse
+ 1);
3146 /* Fill in the cache entry here instead of xlate_recursively
3147 * to make the reference counting more explicit. We take a
3148 * reference in the lookups above if we are going to cache the
3150 if (ctx
->xin
->xcache
) {
3151 struct xc_entry
*entry
;
3153 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_RULE
);
3154 entry
->u
.rule
= rule
;
3155 rule_dpif_ref(rule
);
3157 xlate_recursively(ctx
, rule
);
3160 ctx
->table_id
= old_table_id
;
3168 xlate_group_stats(struct xlate_ctx
*ctx
, struct group_dpif
*group
,
3169 struct ofputil_bucket
*bucket
)
3171 if (ctx
->xin
->resubmit_stats
) {
3172 group_dpif_credit_stats(group
, bucket
, ctx
->xin
->resubmit_stats
);
3174 if (ctx
->xin
->xcache
) {
3175 struct xc_entry
*entry
;
3177 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_GROUP
);
3178 entry
->u
.group
.group
= group_dpif_ref(group
);
3179 entry
->u
.group
.bucket
= bucket
;
3184 xlate_group_bucket(struct xlate_ctx
*ctx
, struct ofputil_bucket
*bucket
)
3186 uint64_t action_list_stub
[1024 / 8];
3187 struct ofpbuf action_list
, action_set
;
3188 struct flow old_flow
= ctx
->xin
->flow
;
3189 bool old_was_mpls
= ctx
->was_mpls
;
3191 ofpbuf_use_const(&action_set
, bucket
->ofpacts
, bucket
->ofpacts_len
);
3192 ofpbuf_use_stub(&action_list
, action_list_stub
, sizeof action_list_stub
);
3194 ofpacts_execute_action_set(&action_list
, &action_set
);
3196 do_xlate_actions(action_list
.data
, action_list
.size
, ctx
);
3199 ofpbuf_uninit(&action_set
);
3200 ofpbuf_uninit(&action_list
);
3202 /* Check if need to recirculate. */
3203 if (exit_recirculates(ctx
)) {
3204 compose_recirculate_action(ctx
);
3207 /* Roll back flow to previous state.
3208 * This is equivalent to cloning the packet for each bucket.
3210 * As a side effect any subsequently applied actions will
3211 * also effectively be applied to a clone of the packet taken
3212 * just before applying the all or indirect group.
3214 * Note that group buckets are action sets, hence they cannot modify the
3215 * main action set. Also any stack actions are ignored when executing an
3216 * action set, so group buckets cannot change the stack either.
3217 * However, we do allow resubmit actions in group buckets, which could
3218 * break the above assumptions. It is up to the controller to not mess up
3219 * with the action_set and stack in the tables resubmitted to from
3221 ctx
->xin
->flow
= old_flow
;
3223 /* The group bucket popping MPLS should have no effect after bucket
3225 ctx
->was_mpls
= old_was_mpls
;
3227 /* The fact that the group bucket exits (for any reason) does not mean that
3228 * the translation after the group action should exit. Specifically, if
3229 * the group bucket recirculates (which typically modifies the packet), the
3230 * actions after the group action must continue processing with the
3231 * original, not the recirculated packet! */
3236 xlate_all_group(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
3238 struct ofputil_bucket
*bucket
;
3239 const struct ovs_list
*buckets
;
3241 group_dpif_get_buckets(group
, &buckets
);
3243 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
3244 xlate_group_bucket(ctx
, bucket
);
3246 xlate_group_stats(ctx
, group
, NULL
);
3250 xlate_ff_group(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
3252 struct ofputil_bucket
*bucket
;
3254 bucket
= group_first_live_bucket(ctx
, group
, 0);
3256 xlate_group_bucket(ctx
, bucket
);
3257 xlate_group_stats(ctx
, group
, bucket
);
3262 xlate_default_select_group(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
3264 struct flow_wildcards
*wc
= ctx
->wc
;
3265 struct ofputil_bucket
*bucket
;
3268 basis
= flow_hash_symmetric_l4(&ctx
->xin
->flow
, 0);
3269 flow_mask_hash_fields(&ctx
->xin
->flow
, wc
, NX_HASH_FIELDS_SYMMETRIC_L4
);
3270 bucket
= group_best_live_bucket(ctx
, group
, basis
);
3272 xlate_group_bucket(ctx
, bucket
);
3273 xlate_group_stats(ctx
, group
, bucket
);
3278 xlate_hash_fields_select_group(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
3280 struct mf_bitmap hash_fields
= MF_BITMAP_INITIALIZER
;
3281 const struct field_array
*fields
;
3282 struct ofputil_bucket
*bucket
;
3286 fields
= group_dpif_get_fields(group
);
3287 basis
= hash_uint64(group_dpif_get_selection_method_param(group
));
3289 /* Determine which fields to hash */
3290 for (i
= 0; i
< MFF_N_IDS
; i
++) {
3291 if (bitmap_is_set(fields
->used
.bm
, i
)) {
3292 const struct mf_field
*mf
;
3294 /* If the field is already present in 'hash_fields' then
3295 * this loop has already checked that it and its pre-requisites
3296 * are present in the flow and its pre-requisites have
3297 * already been added to 'hash_fields'. There is nothing more
3298 * to do here and as an optimisation the loop can continue. */
3299 if (bitmap_is_set(hash_fields
.bm
, i
)) {
3305 /* Only hash a field if it and its pre-requisites are present
3307 if (!mf_are_prereqs_ok(mf
, &ctx
->xin
->flow
)) {
3311 /* Hash both the field and its pre-requisites */
3312 mf_bitmap_set_field_and_prereqs(mf
, &hash_fields
);
3316 /* Hash the fields */
3317 for (i
= 0; i
< MFF_N_IDS
; i
++) {
3318 if (bitmap_is_set(hash_fields
.bm
, i
)) {
3319 const struct mf_field
*mf
= mf_from_id(i
);
3320 union mf_value value
;
3323 mf_get_value(mf
, &ctx
->xin
->flow
, &value
);
3324 /* This seems inefficient but so does apply_mask() */
3325 for (j
= 0; j
< mf
->n_bytes
; j
++) {
3326 ((uint8_t *) &value
)[j
] &= ((uint8_t *) &fields
->value
[i
])[j
];
3328 basis
= hash_bytes(&value
, mf
->n_bytes
, basis
);
3330 mf_mask_field(mf
, &ctx
->wc
->masks
);
3334 bucket
= group_best_live_bucket(ctx
, group
, basis
);
3336 xlate_group_bucket(ctx
, bucket
);
3337 xlate_group_stats(ctx
, group
, bucket
);
3342 xlate_select_group(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
3344 const char *selection_method
= group_dpif_get_selection_method(group
);
3346 if (selection_method
[0] == '\0') {
3347 xlate_default_select_group(ctx
, group
);
3348 } else if (!strcasecmp("hash", selection_method
)) {
3349 xlate_hash_fields_select_group(ctx
, group
);
3351 /* Parsing of groups should ensure this never happens */
3357 xlate_group_action__(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
3359 bool was_in_group
= ctx
->in_group
;
3360 ctx
->in_group
= true;
3362 switch (group_dpif_get_type(group
)) {
3364 case OFPGT11_INDIRECT
:
3365 xlate_all_group(ctx
, group
);
3367 case OFPGT11_SELECT
:
3368 xlate_select_group(ctx
, group
);
3371 xlate_ff_group(ctx
, group
);
3376 group_dpif_unref(group
);
3378 ctx
->in_group
= was_in_group
;
3382 xlate_group_action(struct xlate_ctx
*ctx
, uint32_t group_id
)
3384 if (xlate_resubmit_resource_check(ctx
)) {
3385 struct group_dpif
*group
;
3388 got_group
= group_dpif_lookup(ctx
->xbridge
->ofproto
, group_id
, &group
);
3390 xlate_group_action__(ctx
, group
);
3400 xlate_ofpact_resubmit(struct xlate_ctx
*ctx
,
3401 const struct ofpact_resubmit
*resubmit
)
3405 bool may_packet_in
= false;
3406 bool honor_table_miss
= false;
3408 if (ctx
->rule
&& rule_dpif_is_internal(ctx
->rule
)) {
3409 /* Still allow missed packets to be sent to the controller
3410 * if resubmitting from an internal table. */
3411 may_packet_in
= true;
3412 honor_table_miss
= true;
3415 in_port
= resubmit
->in_port
;
3416 if (in_port
== OFPP_IN_PORT
) {
3417 in_port
= ctx
->xin
->flow
.in_port
.ofp_port
;
3420 table_id
= resubmit
->table_id
;
3421 if (table_id
== 255) {
3422 table_id
= ctx
->table_id
;
3425 xlate_table_action(ctx
, in_port
, table_id
, may_packet_in
,
3430 flood_packets(struct xlate_ctx
*ctx
, bool all
)
3432 const struct xport
*xport
;
3434 HMAP_FOR_EACH (xport
, ofp_node
, &ctx
->xbridge
->xports
) {
3435 if (xport
->ofp_port
== ctx
->xin
->flow
.in_port
.ofp_port
) {
3440 compose_output_action__(ctx
, xport
->ofp_port
, NULL
, false);
3441 } else if (!(xport
->config
& OFPUTIL_PC_NO_FLOOD
)) {
3442 compose_output_action(ctx
, xport
->ofp_port
, NULL
);
3446 ctx
->nf_output_iface
= NF_OUT_FLOOD
;
3450 execute_controller_action(struct xlate_ctx
*ctx
, int len
,
3451 enum ofp_packet_in_reason reason
,
3452 uint16_t controller_id
)
3454 struct ofproto_packet_in
*pin
;
3455 struct dp_packet
*packet
;
3458 ctx
->xout
->slow
|= SLOW_CONTROLLER
;
3459 if (!ctx
->xin
->packet
) {
3463 packet
= dp_packet_clone(ctx
->xin
->packet
);
3465 use_masked
= ctx
->xbridge
->support
.masked_set_action
;
3466 ctx
->xout
->slow
|= commit_odp_actions(&ctx
->xin
->flow
, &ctx
->base_flow
,
3468 ctx
->wc
, use_masked
);
3470 odp_execute_actions(NULL
, &packet
, 1, false,
3471 ctx
->odp_actions
->data
, ctx
->odp_actions
->size
, NULL
);
3473 pin
= xmalloc(sizeof *pin
);
3474 pin
->up
.packet_len
= dp_packet_size(packet
);
3475 pin
->up
.packet
= dp_packet_steal_data(packet
);
3476 pin
->up
.reason
= reason
;
3477 pin
->up
.table_id
= ctx
->table_id
;
3478 pin
->up
.cookie
= ctx
->rule_cookie
;
3480 flow_get_metadata(&ctx
->xin
->flow
, &pin
->up
.flow_metadata
);
3482 pin
->controller_id
= controller_id
;
3483 pin
->send_len
= len
;
3484 /* If a rule is a table-miss rule then this is
3485 * a table-miss handled by a table-miss rule.
3487 * Else, if rule is internal and has a controller action,
3488 * the later being implied by the rule being processed here,
3489 * then this is a table-miss handled without a table-miss rule.
3491 * Otherwise this is not a table-miss. */
3492 pin
->miss_type
= OFPROTO_PACKET_IN_NO_MISS
;
3494 if (rule_dpif_is_table_miss(ctx
->rule
)) {
3495 pin
->miss_type
= OFPROTO_PACKET_IN_MISS_FLOW
;
3496 } else if (rule_dpif_is_internal(ctx
->rule
)) {
3497 pin
->miss_type
= OFPROTO_PACKET_IN_MISS_WITHOUT_FLOW
;
3500 ofproto_dpif_send_packet_in(ctx
->xbridge
->ofproto
, pin
);
3501 dp_packet_delete(packet
);
3504 /* Called only when ctx->recirc_action_offset is set. */
3506 compose_recirculate_action(struct xlate_ctx
*ctx
)
3508 struct recirc_metadata md
;
3512 use_masked
= ctx
->xbridge
->support
.masked_set_action
;
3513 ctx
->xout
->slow
|= commit_odp_actions(&ctx
->xin
->flow
, &ctx
->base_flow
,
3515 ctx
->wc
, use_masked
);
3517 recirc_metadata_from_flow(&md
, &ctx
->xin
->flow
);
3519 ovs_assert(ctx
->recirc_action_offset
>= 0);
3521 struct recirc_state state
= {
3523 .ofproto
= ctx
->xbridge
->ofproto
,
3525 .stack
= &ctx
->stack
,
3526 .mirrors
= ctx
->mirrors
,
3527 .action_set_len
= ctx
->recirc_action_offset
,
3528 .ofpacts_len
= ctx
->action_set
.size
,
3529 .ofpacts
= ctx
->action_set
.data
,
3532 /* Only allocate recirculation ID if we have a packet. */
3533 if (ctx
->xin
->packet
) {
3534 /* Allocate a unique recirc id for the given metadata state in the
3535 * flow. The life-cycle of this recirc id is managed by associating it
3536 * with the udpif key ('ukey') created for each new datapath flow. */
3537 id
= recirc_alloc_id_ctx(&state
);
3539 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3540 VLOG_ERR_RL(&rl
, "Failed to allocate recirculation id");
3543 xlate_out_add_recirc(ctx
->xout
, id
);
3545 /* Look up an existing recirc id for the given metadata state in the
3546 * flow. No new reference is taken, as the ID is RCU protected and is
3547 * only required temporarily for verification.
3549 * This might fail and return 0. We let zero 'id' to be used in the
3550 * RECIRC action below, which will fail all revalidations as zero is
3551 * not a valid recirculation ID. */
3552 id
= recirc_find_id(&state
);
3555 nl_msg_put_u32(ctx
->odp_actions
, OVS_ACTION_ATTR_RECIRC
, id
);
3557 /* Undo changes done by recirculation. */
3558 ctx
->action_set
.size
= ctx
->recirc_action_offset
;
3559 ctx
->recirc_action_offset
= -1;
3560 ctx
->last_unroll_offset
= -1;
3564 compose_mpls_push_action(struct xlate_ctx
*ctx
, struct ofpact_push_mpls
*mpls
)
3566 struct flow
*flow
= &ctx
->xin
->flow
;
3569 ovs_assert(eth_type_mpls(mpls
->ethertype
));
3571 n
= flow_count_mpls_labels(flow
, ctx
->wc
);
3573 bool use_masked
= ctx
->xbridge
->support
.masked_set_action
;
3575 ctx
->xout
->slow
|= commit_odp_actions(flow
, &ctx
->base_flow
,
3577 ctx
->wc
, use_masked
);
3578 } else if (n
>= FLOW_MAX_MPLS_LABELS
) {
3579 if (ctx
->xin
->packet
!= NULL
) {
3580 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3581 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet on which an "
3582 "MPLS push action can't be performed as it would "
3583 "have more MPLS LSEs than the %d supported.",
3584 ctx
->xbridge
->name
, FLOW_MAX_MPLS_LABELS
);
3590 flow_push_mpls(flow
, n
, mpls
->ethertype
, ctx
->wc
);
3594 compose_mpls_pop_action(struct xlate_ctx
*ctx
, ovs_be16 eth_type
)
3596 struct flow
*flow
= &ctx
->xin
->flow
;
3597 int n
= flow_count_mpls_labels(flow
, ctx
->wc
);
3599 if (flow_pop_mpls(flow
, n
, eth_type
, ctx
->wc
)) {
3600 if (ctx
->xbridge
->support
.odp
.recirc
) {
3601 ctx
->was_mpls
= true;
3603 } else if (n
>= FLOW_MAX_MPLS_LABELS
) {
3604 if (ctx
->xin
->packet
!= NULL
) {
3605 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3606 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet on which an "
3607 "MPLS pop action can't be performed as it has "
3608 "more MPLS LSEs than the %d supported.",
3609 ctx
->xbridge
->name
, FLOW_MAX_MPLS_LABELS
);
3612 ofpbuf_clear(ctx
->odp_actions
);
3617 compose_dec_ttl(struct xlate_ctx
*ctx
, struct ofpact_cnt_ids
*ids
)
3619 struct flow
*flow
= &ctx
->xin
->flow
;
3621 if (!is_ip_any(flow
)) {
3625 ctx
->wc
->masks
.nw_ttl
= 0xff;
3626 if (flow
->nw_ttl
> 1) {
3632 for (i
= 0; i
< ids
->n_controllers
; i
++) {
3633 execute_controller_action(ctx
, UINT16_MAX
, OFPR_INVALID_TTL
,
3637 /* Stop processing for current table. */
3643 compose_set_mpls_label_action(struct xlate_ctx
*ctx
, ovs_be32 label
)
3645 if (eth_type_mpls(ctx
->xin
->flow
.dl_type
)) {
3646 ctx
->wc
->masks
.mpls_lse
[0] |= htonl(MPLS_LABEL_MASK
);
3647 set_mpls_lse_label(&ctx
->xin
->flow
.mpls_lse
[0], label
);
3652 compose_set_mpls_tc_action(struct xlate_ctx
*ctx
, uint8_t tc
)
3654 if (eth_type_mpls(ctx
->xin
->flow
.dl_type
)) {
3655 ctx
->wc
->masks
.mpls_lse
[0] |= htonl(MPLS_TC_MASK
);
3656 set_mpls_lse_tc(&ctx
->xin
->flow
.mpls_lse
[0], tc
);
3661 compose_set_mpls_ttl_action(struct xlate_ctx
*ctx
, uint8_t ttl
)
3663 if (eth_type_mpls(ctx
->xin
->flow
.dl_type
)) {
3664 ctx
->wc
->masks
.mpls_lse
[0] |= htonl(MPLS_TTL_MASK
);
3665 set_mpls_lse_ttl(&ctx
->xin
->flow
.mpls_lse
[0], ttl
);
3670 compose_dec_mpls_ttl_action(struct xlate_ctx
*ctx
)
3672 struct flow
*flow
= &ctx
->xin
->flow
;
3674 if (eth_type_mpls(flow
->dl_type
)) {
3675 uint8_t ttl
= mpls_lse_to_ttl(flow
->mpls_lse
[0]);
3677 ctx
->wc
->masks
.mpls_lse
[0] |= htonl(MPLS_TTL_MASK
);
3680 set_mpls_lse_ttl(&flow
->mpls_lse
[0], ttl
);
3683 execute_controller_action(ctx
, UINT16_MAX
, OFPR_INVALID_TTL
, 0);
3687 /* Stop processing for current table. */
3692 xlate_output_action(struct xlate_ctx
*ctx
,
3693 ofp_port_t port
, uint16_t max_len
, bool may_packet_in
)
3695 ofp_port_t prev_nf_output_iface
= ctx
->nf_output_iface
;
3697 ctx
->nf_output_iface
= NF_OUT_DROP
;
3701 compose_output_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
, NULL
);
3704 xlate_table_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
,
3705 0, may_packet_in
, true);
3711 flood_packets(ctx
, false);
3714 flood_packets(ctx
, true);
3716 case OFPP_CONTROLLER
:
3717 execute_controller_action(ctx
, max_len
,
3718 (ctx
->in_group
? OFPR_GROUP
3719 : ctx
->in_action_set
? OFPR_ACTION_SET
3727 if (port
!= ctx
->xin
->flow
.in_port
.ofp_port
) {
3728 compose_output_action(ctx
, port
, NULL
);
3730 xlate_report(ctx
, "skipping output to input port");
3735 if (prev_nf_output_iface
== NF_OUT_FLOOD
) {
3736 ctx
->nf_output_iface
= NF_OUT_FLOOD
;
3737 } else if (ctx
->nf_output_iface
== NF_OUT_DROP
) {
3738 ctx
->nf_output_iface
= prev_nf_output_iface
;
3739 } else if (prev_nf_output_iface
!= NF_OUT_DROP
&&
3740 ctx
->nf_output_iface
!= NF_OUT_FLOOD
) {
3741 ctx
->nf_output_iface
= NF_OUT_MULTI
;
3746 xlate_output_reg_action(struct xlate_ctx
*ctx
,
3747 const struct ofpact_output_reg
*or)
3749 uint64_t port
= mf_get_subfield(&or->src
, &ctx
->xin
->flow
);
3750 if (port
<= UINT16_MAX
) {
3751 union mf_subvalue value
;
3753 memset(&value
, 0xff, sizeof value
);
3754 mf_write_subfield_flow(&or->src
, &value
, &ctx
->wc
->masks
);
3755 xlate_output_action(ctx
, u16_to_ofp(port
),
3756 or->max_len
, false);
3761 xlate_enqueue_action(struct xlate_ctx
*ctx
,
3762 const struct ofpact_enqueue
*enqueue
)
3764 ofp_port_t ofp_port
= enqueue
->port
;
3765 uint32_t queue_id
= enqueue
->queue
;
3766 uint32_t flow_priority
, priority
;
3769 /* Translate queue to priority. */
3770 error
= dpif_queue_to_priority(ctx
->xbridge
->dpif
, queue_id
, &priority
);
3772 /* Fall back to ordinary output action. */
3773 xlate_output_action(ctx
, enqueue
->port
, 0, false);
3777 /* Check output port. */
3778 if (ofp_port
== OFPP_IN_PORT
) {
3779 ofp_port
= ctx
->xin
->flow
.in_port
.ofp_port
;
3780 } else if (ofp_port
== ctx
->xin
->flow
.in_port
.ofp_port
) {
3784 /* Add datapath actions. */
3785 flow_priority
= ctx
->xin
->flow
.skb_priority
;
3786 ctx
->xin
->flow
.skb_priority
= priority
;
3787 compose_output_action(ctx
, ofp_port
, NULL
);
3788 ctx
->xin
->flow
.skb_priority
= flow_priority
;
3790 /* Update NetFlow output port. */
3791 if (ctx
->nf_output_iface
== NF_OUT_DROP
) {
3792 ctx
->nf_output_iface
= ofp_port
;
3793 } else if (ctx
->nf_output_iface
!= NF_OUT_FLOOD
) {
3794 ctx
->nf_output_iface
= NF_OUT_MULTI
;
3799 xlate_set_queue_action(struct xlate_ctx
*ctx
, uint32_t queue_id
)
3801 uint32_t skb_priority
;
3803 if (!dpif_queue_to_priority(ctx
->xbridge
->dpif
, queue_id
, &skb_priority
)) {
3804 ctx
->xin
->flow
.skb_priority
= skb_priority
;
3806 /* Couldn't translate queue to a priority. Nothing to do. A warning
3807 * has already been logged. */
3812 slave_enabled_cb(ofp_port_t ofp_port
, void *xbridge_
)
3814 const struct xbridge
*xbridge
= xbridge_
;
3825 case OFPP_CONTROLLER
: /* Not supported by the bundle action. */
3828 port
= get_ofp_port(xbridge
, ofp_port
);
3829 return port
? port
->may_enable
: false;
3834 xlate_bundle_action(struct xlate_ctx
*ctx
,
3835 const struct ofpact_bundle
*bundle
)
3839 port
= bundle_execute(bundle
, &ctx
->xin
->flow
, ctx
->wc
, slave_enabled_cb
,
3840 CONST_CAST(struct xbridge
*, ctx
->xbridge
));
3841 if (bundle
->dst
.field
) {
3842 nxm_reg_load(&bundle
->dst
, ofp_to_u16(port
), &ctx
->xin
->flow
, ctx
->wc
);
3844 xlate_output_action(ctx
, port
, 0, false);
3849 xlate_learn_action__(struct xlate_ctx
*ctx
, const struct ofpact_learn
*learn
,
3850 struct ofputil_flow_mod
*fm
, struct ofpbuf
*ofpacts
)
3852 learn_execute(learn
, &ctx
->xin
->flow
, fm
, ofpacts
);
3853 if (ctx
->xin
->may_learn
) {
3854 ofproto_dpif_flow_mod(ctx
->xbridge
->ofproto
, fm
);
3859 xlate_learn_action(struct xlate_ctx
*ctx
, const struct ofpact_learn
*learn
)
3861 learn_mask(learn
, ctx
->wc
);
3863 if (ctx
->xin
->xcache
) {
3864 struct xc_entry
*entry
;
3866 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_LEARN
);
3867 entry
->u
.learn
.ofproto
= ctx
->xbridge
->ofproto
;
3868 entry
->u
.learn
.fm
= xmalloc(sizeof *entry
->u
.learn
.fm
);
3869 entry
->u
.learn
.ofpacts
= ofpbuf_new(64);
3870 xlate_learn_action__(ctx
, learn
, entry
->u
.learn
.fm
,
3871 entry
->u
.learn
.ofpacts
);
3872 } else if (ctx
->xin
->may_learn
) {
3873 uint64_t ofpacts_stub
[1024 / 8];
3874 struct ofputil_flow_mod fm
;
3875 struct ofpbuf ofpacts
;
3877 ofpbuf_use_stub(&ofpacts
, ofpacts_stub
, sizeof ofpacts_stub
);
3878 xlate_learn_action__(ctx
, learn
, &fm
, &ofpacts
);
3879 ofpbuf_uninit(&ofpacts
);
3884 xlate_fin_timeout__(struct rule_dpif
*rule
, uint16_t tcp_flags
,
3885 uint16_t idle_timeout
, uint16_t hard_timeout
)
3887 if (tcp_flags
& (TCP_FIN
| TCP_RST
)) {
3888 rule_dpif_reduce_timeouts(rule
, idle_timeout
, hard_timeout
);
3893 xlate_fin_timeout(struct xlate_ctx
*ctx
,
3894 const struct ofpact_fin_timeout
*oft
)
3897 xlate_fin_timeout__(ctx
->rule
, ctx
->xin
->tcp_flags
,
3898 oft
->fin_idle_timeout
, oft
->fin_hard_timeout
);
3899 if (ctx
->xin
->xcache
) {
3900 struct xc_entry
*entry
;
3902 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_FIN_TIMEOUT
);
3903 /* XC_RULE already holds a reference on the rule, none is taken
3905 entry
->u
.fin
.rule
= ctx
->rule
;
3906 entry
->u
.fin
.idle
= oft
->fin_idle_timeout
;
3907 entry
->u
.fin
.hard
= oft
->fin_hard_timeout
;
3913 xlate_sample_action(struct xlate_ctx
*ctx
,
3914 const struct ofpact_sample
*os
)
3916 /* Scale the probability from 16-bit to 32-bit while representing
3917 * the same percentage. */
3918 uint32_t probability
= (os
->probability
<< 16) | os
->probability
;
3921 if (!ctx
->xbridge
->support
.variable_length_userdata
) {
3922 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
3924 VLOG_ERR_RL(&rl
, "ignoring NXAST_SAMPLE action because datapath "
3925 "lacks support (needs Linux 3.10+ or kernel module from "
3930 use_masked
= ctx
->xbridge
->support
.masked_set_action
;
3931 ctx
->xout
->slow
|= commit_odp_actions(&ctx
->xin
->flow
, &ctx
->base_flow
,
3933 ctx
->wc
, use_masked
);
3935 union user_action_cookie cookie
= {
3937 .type
= USER_ACTION_COOKIE_FLOW_SAMPLE
,
3938 .probability
= os
->probability
,
3939 .collector_set_id
= os
->collector_set_id
,
3940 .obs_domain_id
= os
->obs_domain_id
,
3941 .obs_point_id
= os
->obs_point_id
,
3944 compose_sample_action(ctx
, probability
, &cookie
, sizeof cookie
.flow_sample
,
3949 may_receive(const struct xport
*xport
, struct xlate_ctx
*ctx
)
3951 if (xport
->config
& (is_stp(&ctx
->xin
->flow
)
3952 ? OFPUTIL_PC_NO_RECV_STP
3953 : OFPUTIL_PC_NO_RECV
)) {
3957 /* Only drop packets here if both forwarding and learning are
3958 * disabled. If just learning is enabled, we need to have
3959 * OFPP_NORMAL and the learning action have a look at the packet
3960 * before we can drop it. */
3961 if ((!xport_stp_forward_state(xport
) && !xport_stp_learn_state(xport
)) ||
3962 (!xport_rstp_forward_state(xport
) && !xport_rstp_learn_state(xport
))) {
3970 xlate_write_actions(struct xlate_ctx
*ctx
, const struct ofpact
*a
)
3972 const struct ofpact_nest
*on
= ofpact_get_WRITE_ACTIONS(a
);
3973 size_t on_len
= ofpact_nest_get_action_len(on
);
3974 const struct ofpact
*inner
;
3976 /* Maintain actset_output depending on the contents of the action set:
3978 * - OFPP_UNSET, if there is no "output" action.
3980 * - The output port, if there is an "output" action and no "group"
3983 * - OFPP_UNSET, if there is a "group" action.
3985 if (!ctx
->action_set_has_group
) {
3986 OFPACT_FOR_EACH (inner
, on
->actions
, on_len
) {
3987 if (inner
->type
== OFPACT_OUTPUT
) {
3988 ctx
->xin
->flow
.actset_output
= ofpact_get_OUTPUT(inner
)->port
;
3989 } else if (inner
->type
== OFPACT_GROUP
) {
3990 ctx
->xin
->flow
.actset_output
= OFPP_UNSET
;
3991 ctx
->action_set_has_group
= true;
3997 ofpbuf_put(&ctx
->action_set
, on
->actions
, on_len
);
3998 ofpact_pad(&ctx
->action_set
);
4002 xlate_action_set(struct xlate_ctx
*ctx
)
4004 uint64_t action_list_stub
[1024 / 64];
4005 struct ofpbuf action_list
;
4007 ctx
->in_action_set
= true;
4008 ofpbuf_use_stub(&action_list
, action_list_stub
, sizeof action_list_stub
);
4009 ofpacts_execute_action_set(&action_list
, &ctx
->action_set
);
4010 /* Clear the action set, as it is not needed any more. */
4011 ofpbuf_clear(&ctx
->action_set
);
4012 do_xlate_actions(action_list
.data
, action_list
.size
, ctx
);
4013 ctx
->in_action_set
= false;
4014 ofpbuf_uninit(&action_list
);
4018 recirc_put_unroll_xlate(struct xlate_ctx
*ctx
)
4020 struct ofpact_unroll_xlate
*unroll
;
4022 unroll
= ctx
->last_unroll_offset
< 0
4024 : ALIGNED_CAST(struct ofpact_unroll_xlate
*,
4025 (char *)ctx
->action_set
.data
+ ctx
->last_unroll_offset
);
4027 /* Restore the table_id and rule cookie for a potential PACKET
4030 (ctx
->table_id
!= unroll
->rule_table_id
4031 || ctx
->rule_cookie
!= unroll
->rule_cookie
)) {
4033 ctx
->last_unroll_offset
= ctx
->action_set
.size
;
4034 unroll
= ofpact_put_UNROLL_XLATE(&ctx
->action_set
);
4035 unroll
->rule_table_id
= ctx
->table_id
;
4036 unroll
->rule_cookie
= ctx
->rule_cookie
;
4041 /* Copy remaining actions to the action_set to be executed after recirculation.
4042 * UNROLL_XLATE action is inserted, if not already done so, before actions that
4043 * may generate PACKET_INs from the current table and without matching another
4046 recirc_unroll_actions(const struct ofpact
*ofpacts
, size_t ofpacts_len
,
4047 struct xlate_ctx
*ctx
)
4049 const struct ofpact
*a
;
4051 OFPACT_FOR_EACH (a
, ofpacts
, ofpacts_len
) {
4053 /* May generate PACKET INs. */
4054 case OFPACT_OUTPUT_REG
:
4057 case OFPACT_CONTROLLER
:
4058 case OFPACT_DEC_MPLS_TTL
:
4059 case OFPACT_DEC_TTL
:
4060 recirc_put_unroll_xlate(ctx
);
4063 /* These may not generate PACKET INs. */
4064 case OFPACT_SET_TUNNEL
:
4065 case OFPACT_REG_MOVE
:
4066 case OFPACT_SET_FIELD
:
4067 case OFPACT_STACK_PUSH
:
4068 case OFPACT_STACK_POP
:
4070 case OFPACT_WRITE_METADATA
:
4071 case OFPACT_RESUBMIT
: /* May indirectly generate PACKET INs, */
4072 case OFPACT_GOTO_TABLE
: /* but from a different table and rule. */
4073 case OFPACT_ENQUEUE
:
4074 case OFPACT_SET_VLAN_VID
:
4075 case OFPACT_SET_VLAN_PCP
:
4076 case OFPACT_STRIP_VLAN
:
4077 case OFPACT_PUSH_VLAN
:
4078 case OFPACT_SET_ETH_SRC
:
4079 case OFPACT_SET_ETH_DST
:
4080 case OFPACT_SET_IPV4_SRC
:
4081 case OFPACT_SET_IPV4_DST
:
4082 case OFPACT_SET_IP_DSCP
:
4083 case OFPACT_SET_IP_ECN
:
4084 case OFPACT_SET_IP_TTL
:
4085 case OFPACT_SET_L4_SRC_PORT
:
4086 case OFPACT_SET_L4_DST_PORT
:
4087 case OFPACT_SET_QUEUE
:
4088 case OFPACT_POP_QUEUE
:
4089 case OFPACT_PUSH_MPLS
:
4090 case OFPACT_POP_MPLS
:
4091 case OFPACT_SET_MPLS_LABEL
:
4092 case OFPACT_SET_MPLS_TC
:
4093 case OFPACT_SET_MPLS_TTL
:
4094 case OFPACT_MULTIPATH
:
4097 case OFPACT_UNROLL_XLATE
:
4098 case OFPACT_FIN_TIMEOUT
:
4099 case OFPACT_CLEAR_ACTIONS
:
4100 case OFPACT_WRITE_ACTIONS
:
4103 case OFPACT_DEBUG_RECIRC
:
4106 /* These need not be copied for restoration. */
4108 case OFPACT_CONJUNCTION
:
4111 /* Copy the action over. */
4112 ofpbuf_put(&ctx
->action_set
, a
, OFPACT_ALIGN(a
->len
));
4116 #define CHECK_MPLS_RECIRCULATION() \
4117 if (ctx->was_mpls) { \
4118 ctx_trigger_recirculation(ctx); \
4121 #define CHECK_MPLS_RECIRCULATION_IF(COND) \
4123 CHECK_MPLS_RECIRCULATION(); \
4127 do_xlate_actions(const struct ofpact
*ofpacts
, size_t ofpacts_len
,
4128 struct xlate_ctx
*ctx
)
4130 struct flow_wildcards
*wc
= ctx
->wc
;
4131 struct flow
*flow
= &ctx
->xin
->flow
;
4132 const struct ofpact
*a
;
4134 if (ovs_native_tunneling_is_on(ctx
->xbridge
->ofproto
)) {
4135 tnl_arp_snoop(flow
, wc
, ctx
->xbridge
->name
);
4137 /* dl_type already in the mask, not set below. */
4139 OFPACT_FOR_EACH (a
, ofpacts
, ofpacts_len
) {
4140 struct ofpact_controller
*controller
;
4141 const struct ofpact_metadata
*metadata
;
4142 const struct ofpact_set_field
*set_field
;
4143 const struct mf_field
*mf
;
4146 /* Check if need to store the remaining actions for later
4148 if (exit_recirculates(ctx
)) {
4149 recirc_unroll_actions(a
, OFPACT_ALIGN(ofpacts_len
-
4151 (uint8_t *)ofpacts
)),
4159 xlate_output_action(ctx
, ofpact_get_OUTPUT(a
)->port
,
4160 ofpact_get_OUTPUT(a
)->max_len
, true);
4164 if (xlate_group_action(ctx
, ofpact_get_GROUP(a
)->group_id
)) {
4165 /* Group could not be found. */
4170 case OFPACT_CONTROLLER
:
4171 controller
= ofpact_get_CONTROLLER(a
);
4172 execute_controller_action(ctx
, controller
->max_len
,
4174 controller
->controller_id
);
4177 case OFPACT_ENQUEUE
:
4178 memset(&wc
->masks
.skb_priority
, 0xff,
4179 sizeof wc
->masks
.skb_priority
);
4180 xlate_enqueue_action(ctx
, ofpact_get_ENQUEUE(a
));
4183 case OFPACT_SET_VLAN_VID
:
4184 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
4185 if (flow
->vlan_tci
& htons(VLAN_CFI
) ||
4186 ofpact_get_SET_VLAN_VID(a
)->push_vlan_if_needed
) {
4187 flow
->vlan_tci
&= ~htons(VLAN_VID_MASK
);
4188 flow
->vlan_tci
|= (htons(ofpact_get_SET_VLAN_VID(a
)->vlan_vid
)
4193 case OFPACT_SET_VLAN_PCP
:
4194 wc
->masks
.vlan_tci
|= htons(VLAN_PCP_MASK
| VLAN_CFI
);
4195 if (flow
->vlan_tci
& htons(VLAN_CFI
) ||
4196 ofpact_get_SET_VLAN_PCP(a
)->push_vlan_if_needed
) {
4197 flow
->vlan_tci
&= ~htons(VLAN_PCP_MASK
);
4198 flow
->vlan_tci
|= htons((ofpact_get_SET_VLAN_PCP(a
)->vlan_pcp
4199 << VLAN_PCP_SHIFT
) | VLAN_CFI
);
4203 case OFPACT_STRIP_VLAN
:
4204 memset(&wc
->masks
.vlan_tci
, 0xff, sizeof wc
->masks
.vlan_tci
);
4205 flow
->vlan_tci
= htons(0);
4208 case OFPACT_PUSH_VLAN
:
4209 /* XXX 802.1AD(QinQ) */
4210 memset(&wc
->masks
.vlan_tci
, 0xff, sizeof wc
->masks
.vlan_tci
);
4211 flow
->vlan_tci
= htons(VLAN_CFI
);
4214 case OFPACT_SET_ETH_SRC
:
4215 WC_MASK_FIELD(wc
, dl_src
);
4216 flow
->dl_src
= ofpact_get_SET_ETH_SRC(a
)->mac
;
4219 case OFPACT_SET_ETH_DST
:
4220 WC_MASK_FIELD(wc
, dl_dst
);
4221 flow
->dl_dst
= ofpact_get_SET_ETH_DST(a
)->mac
;
4224 case OFPACT_SET_IPV4_SRC
:
4225 CHECK_MPLS_RECIRCULATION();
4226 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
4227 memset(&wc
->masks
.nw_src
, 0xff, sizeof wc
->masks
.nw_src
);
4228 flow
->nw_src
= ofpact_get_SET_IPV4_SRC(a
)->ipv4
;
4232 case OFPACT_SET_IPV4_DST
:
4233 CHECK_MPLS_RECIRCULATION();
4234 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
4235 memset(&wc
->masks
.nw_dst
, 0xff, sizeof wc
->masks
.nw_dst
);
4236 flow
->nw_dst
= ofpact_get_SET_IPV4_DST(a
)->ipv4
;
4240 case OFPACT_SET_IP_DSCP
:
4241 CHECK_MPLS_RECIRCULATION();
4242 if (is_ip_any(flow
)) {
4243 wc
->masks
.nw_tos
|= IP_DSCP_MASK
;
4244 flow
->nw_tos
&= ~IP_DSCP_MASK
;
4245 flow
->nw_tos
|= ofpact_get_SET_IP_DSCP(a
)->dscp
;
4249 case OFPACT_SET_IP_ECN
:
4250 CHECK_MPLS_RECIRCULATION();
4251 if (is_ip_any(flow
)) {
4252 wc
->masks
.nw_tos
|= IP_ECN_MASK
;
4253 flow
->nw_tos
&= ~IP_ECN_MASK
;
4254 flow
->nw_tos
|= ofpact_get_SET_IP_ECN(a
)->ecn
;
4258 case OFPACT_SET_IP_TTL
:
4259 CHECK_MPLS_RECIRCULATION();
4260 if (is_ip_any(flow
)) {
4261 wc
->masks
.nw_ttl
= 0xff;
4262 flow
->nw_ttl
= ofpact_get_SET_IP_TTL(a
)->ttl
;
4266 case OFPACT_SET_L4_SRC_PORT
:
4267 CHECK_MPLS_RECIRCULATION();
4268 if (is_ip_any(flow
) && !(flow
->nw_frag
& FLOW_NW_FRAG_LATER
)) {
4269 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
4270 memset(&wc
->masks
.tp_src
, 0xff, sizeof wc
->masks
.tp_src
);
4271 flow
->tp_src
= htons(ofpact_get_SET_L4_SRC_PORT(a
)->port
);
4275 case OFPACT_SET_L4_DST_PORT
:
4276 CHECK_MPLS_RECIRCULATION();
4277 if (is_ip_any(flow
) && !(flow
->nw_frag
& FLOW_NW_FRAG_LATER
)) {
4278 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
4279 memset(&wc
->masks
.tp_dst
, 0xff, sizeof wc
->masks
.tp_dst
);
4280 flow
->tp_dst
= htons(ofpact_get_SET_L4_DST_PORT(a
)->port
);
4284 case OFPACT_RESUBMIT
:
4285 xlate_ofpact_resubmit(ctx
, ofpact_get_RESUBMIT(a
));
4288 case OFPACT_SET_TUNNEL
:
4289 flow
->tunnel
.tun_id
= htonll(ofpact_get_SET_TUNNEL(a
)->tun_id
);
4292 case OFPACT_SET_QUEUE
:
4293 memset(&wc
->masks
.skb_priority
, 0xff,
4294 sizeof wc
->masks
.skb_priority
);
4295 xlate_set_queue_action(ctx
, ofpact_get_SET_QUEUE(a
)->queue_id
);
4298 case OFPACT_POP_QUEUE
:
4299 memset(&wc
->masks
.skb_priority
, 0xff,
4300 sizeof wc
->masks
.skb_priority
);
4301 flow
->skb_priority
= ctx
->orig_skb_priority
;
4304 case OFPACT_REG_MOVE
:
4305 CHECK_MPLS_RECIRCULATION_IF(
4306 mf_is_l3_or_higher(ofpact_get_REG_MOVE(a
)->dst
.field
) ||
4307 mf_is_l3_or_higher(ofpact_get_REG_MOVE(a
)->src
.field
));
4308 nxm_execute_reg_move(ofpact_get_REG_MOVE(a
), flow
, wc
);
4311 case OFPACT_SET_FIELD
:
4312 CHECK_MPLS_RECIRCULATION_IF(
4313 mf_is_l3_or_higher(ofpact_get_SET_FIELD(a
)->field
));
4314 set_field
= ofpact_get_SET_FIELD(a
);
4315 mf
= set_field
->field
;
4317 /* Set field action only ever overwrites packet's outermost
4318 * applicable header fields. Do nothing if no header exists. */
4319 if (mf
->id
== MFF_VLAN_VID
) {
4320 wc
->masks
.vlan_tci
|= htons(VLAN_CFI
);
4321 if (!(flow
->vlan_tci
& htons(VLAN_CFI
))) {
4324 } else if ((mf
->id
== MFF_MPLS_LABEL
|| mf
->id
== MFF_MPLS_TC
)
4325 /* 'dl_type' is already unwildcarded. */
4326 && !eth_type_mpls(flow
->dl_type
)) {
4329 /* A flow may wildcard nw_frag. Do nothing if setting a trasport
4330 * header field on a packet that does not have them. */
4331 mf_mask_field_and_prereqs(mf
, wc
);
4332 if (mf_are_prereqs_ok(mf
, flow
)) {
4333 mf_set_flow_value_masked(mf
, &set_field
->value
,
4334 &set_field
->mask
, flow
);
4338 case OFPACT_STACK_PUSH
:
4339 CHECK_MPLS_RECIRCULATION_IF(
4340 mf_is_l3_or_higher(ofpact_get_STACK_PUSH(a
)->subfield
.field
));
4341 nxm_execute_stack_push(ofpact_get_STACK_PUSH(a
), flow
, wc
,
4345 case OFPACT_STACK_POP
:
4346 CHECK_MPLS_RECIRCULATION_IF(
4347 mf_is_l3_or_higher(ofpact_get_STACK_POP(a
)->subfield
.field
));
4348 nxm_execute_stack_pop(ofpact_get_STACK_POP(a
), flow
, wc
,
4352 case OFPACT_PUSH_MPLS
:
4353 /* Recirculate if it is an IP packet with a zero ttl. This may
4354 * indicate that the packet was previously MPLS and an MPLS pop
4355 * action converted it to IP. In this case recirculating should
4356 * reveal the IP TTL which is used as the basis for a new MPLS
4358 CHECK_MPLS_RECIRCULATION_IF(
4359 !flow_count_mpls_labels(flow
, wc
)
4360 && flow
->nw_ttl
== 0
4361 && is_ip_any(flow
));
4362 compose_mpls_push_action(ctx
, ofpact_get_PUSH_MPLS(a
));
4365 case OFPACT_POP_MPLS
:
4366 CHECK_MPLS_RECIRCULATION();
4367 compose_mpls_pop_action(ctx
, ofpact_get_POP_MPLS(a
)->ethertype
);
4370 case OFPACT_SET_MPLS_LABEL
:
4371 CHECK_MPLS_RECIRCULATION();
4372 compose_set_mpls_label_action(
4373 ctx
, ofpact_get_SET_MPLS_LABEL(a
)->label
);
4376 case OFPACT_SET_MPLS_TC
:
4377 CHECK_MPLS_RECIRCULATION();
4378 compose_set_mpls_tc_action(ctx
, ofpact_get_SET_MPLS_TC(a
)->tc
);
4381 case OFPACT_SET_MPLS_TTL
:
4382 CHECK_MPLS_RECIRCULATION();
4383 compose_set_mpls_ttl_action(ctx
, ofpact_get_SET_MPLS_TTL(a
)->ttl
);
4386 case OFPACT_DEC_MPLS_TTL
:
4387 CHECK_MPLS_RECIRCULATION();
4388 if (compose_dec_mpls_ttl_action(ctx
)) {
4393 case OFPACT_DEC_TTL
:
4394 CHECK_MPLS_RECIRCULATION();
4395 wc
->masks
.nw_ttl
= 0xff;
4396 if (compose_dec_ttl(ctx
, ofpact_get_DEC_TTL(a
))) {
4402 /* Nothing to do. */
4405 case OFPACT_MULTIPATH
:
4406 CHECK_MPLS_RECIRCULATION();
4407 multipath_execute(ofpact_get_MULTIPATH(a
), flow
, wc
);
4411 CHECK_MPLS_RECIRCULATION();
4412 xlate_bundle_action(ctx
, ofpact_get_BUNDLE(a
));
4415 case OFPACT_OUTPUT_REG
:
4416 xlate_output_reg_action(ctx
, ofpact_get_OUTPUT_REG(a
));
4420 CHECK_MPLS_RECIRCULATION();
4421 xlate_learn_action(ctx
, ofpact_get_LEARN(a
));
4424 case OFPACT_CONJUNCTION
: {
4425 /* A flow with a "conjunction" action represents part of a special
4426 * kind of "set membership match". Such a flow should not actually
4427 * get executed, but it could via, say, a "packet-out", even though
4428 * that wouldn't be useful. Log it to help debugging. */
4429 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
4430 VLOG_INFO_RL(&rl
, "executing no-op conjunction action");
4438 case OFPACT_UNROLL_XLATE
: {
4439 struct ofpact_unroll_xlate
*unroll
= ofpact_get_UNROLL_XLATE(a
);
4441 /* Restore translation context data that was stored earlier. */
4442 ctx
->table_id
= unroll
->rule_table_id
;
4443 ctx
->rule_cookie
= unroll
->rule_cookie
;
4446 case OFPACT_FIN_TIMEOUT
:
4447 CHECK_MPLS_RECIRCULATION();
4448 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
4449 xlate_fin_timeout(ctx
, ofpact_get_FIN_TIMEOUT(a
));
4452 case OFPACT_CLEAR_ACTIONS
:
4453 ofpbuf_clear(&ctx
->action_set
);
4454 ctx
->xin
->flow
.actset_output
= OFPP_UNSET
;
4455 ctx
->action_set_has_group
= false;
4458 case OFPACT_WRITE_ACTIONS
:
4459 xlate_write_actions(ctx
, a
);
4462 case OFPACT_WRITE_METADATA
:
4463 metadata
= ofpact_get_WRITE_METADATA(a
);
4464 flow
->metadata
&= ~metadata
->mask
;
4465 flow
->metadata
|= metadata
->metadata
& metadata
->mask
;
4469 /* Not implemented yet. */
4472 case OFPACT_GOTO_TABLE
: {
4473 struct ofpact_goto_table
*ogt
= ofpact_get_GOTO_TABLE(a
);
4475 /* Allow ctx->table_id == TBL_INTERNAL, which will be greater
4476 * than ogt->table_id. This is to allow goto_table actions that
4477 * triggered recirculation: ctx->table_id will be TBL_INTERNAL
4478 * after recirculation. */
4479 ovs_assert(ctx
->table_id
== TBL_INTERNAL
4480 || ctx
->table_id
< ogt
->table_id
);
4481 xlate_table_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
,
4482 ogt
->table_id
, true, true);
4487 xlate_sample_action(ctx
, ofpact_get_SAMPLE(a
));
4490 case OFPACT_DEBUG_RECIRC
:
4491 ctx_trigger_recirculation(ctx
);
4496 /* Check if need to store this and the remaining actions for later
4498 if (ctx
->exit
&& ctx_first_recirculation_action(ctx
)) {
4499 recirc_unroll_actions(a
, OFPACT_ALIGN(ofpacts_len
-
4501 (uint8_t *)ofpacts
)),
4509 xlate_in_init(struct xlate_in
*xin
, struct ofproto_dpif
*ofproto
,
4510 const struct flow
*flow
, ofp_port_t in_port
,
4511 struct rule_dpif
*rule
, uint16_t tcp_flags
,
4512 const struct dp_packet
*packet
, struct flow_wildcards
*wc
,
4513 struct ofpbuf
*odp_actions
)
4515 xin
->ofproto
= ofproto
;
4517 xin
->flow
.in_port
.ofp_port
= in_port
;
4518 xin
->flow
.actset_output
= OFPP_UNSET
;
4519 xin
->packet
= packet
;
4520 xin
->may_learn
= packet
!= NULL
;
4523 xin
->ofpacts
= NULL
;
4524 xin
->ofpacts_len
= 0;
4525 xin
->tcp_flags
= tcp_flags
;
4526 xin
->resubmit_hook
= NULL
;
4527 xin
->report_hook
= NULL
;
4528 xin
->resubmit_stats
= NULL
;
4530 xin
->odp_actions
= odp_actions
;
4532 /* Do recirc lookup. */
4533 xin
->recirc
= flow
->recirc_id
4534 ? recirc_id_node_find(flow
->recirc_id
)
4539 xlate_out_uninit(struct xlate_out
*xout
)
4542 xlate_out_free_recircs(xout
);
4546 /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts'
4547 * into datapath actions, using 'ctx', and discards the datapath actions. */
4549 xlate_actions_for_side_effects(struct xlate_in
*xin
)
4551 struct xlate_out xout
;
4553 xlate_actions(xin
, &xout
);
4554 xlate_out_uninit(&xout
);
4557 static struct skb_priority_to_dscp
*
4558 get_skb_priority(const struct xport
*xport
, uint32_t skb_priority
)
4560 struct skb_priority_to_dscp
*pdscp
;
4563 hash
= hash_int(skb_priority
, 0);
4564 HMAP_FOR_EACH_IN_BUCKET (pdscp
, hmap_node
, hash
, &xport
->skb_priorities
) {
4565 if (pdscp
->skb_priority
== skb_priority
) {
4573 dscp_from_skb_priority(const struct xport
*xport
, uint32_t skb_priority
,
4576 struct skb_priority_to_dscp
*pdscp
= get_skb_priority(xport
, skb_priority
);
4577 *dscp
= pdscp
? pdscp
->dscp
: 0;
4578 return pdscp
!= NULL
;
4582 count_skb_priorities(const struct xport
*xport
)
4584 return hmap_count(&xport
->skb_priorities
);
4588 clear_skb_priorities(struct xport
*xport
)
4590 struct skb_priority_to_dscp
*pdscp
, *next
;
4592 HMAP_FOR_EACH_SAFE (pdscp
, next
, hmap_node
, &xport
->skb_priorities
) {
4593 hmap_remove(&xport
->skb_priorities
, &pdscp
->hmap_node
);
4599 actions_output_to_local_port(const struct xlate_ctx
*ctx
)
4601 odp_port_t local_odp_port
= ofp_port_to_odp_port(ctx
->xbridge
, OFPP_LOCAL
);
4602 const struct nlattr
*a
;
4605 NL_ATTR_FOR_EACH_UNSAFE (a
, left
, ctx
->odp_actions
->data
,
4606 ctx
->odp_actions
->size
) {
4607 if (nl_attr_type(a
) == OVS_ACTION_ATTR_OUTPUT
4608 && nl_attr_get_odp_port(a
) == local_odp_port
) {
4615 #if defined(__linux__)
4616 /* Returns the maximum number of packets that the Linux kernel is willing to
4617 * queue up internally to certain kinds of software-implemented ports, or the
4618 * default (and rarely modified) value if it cannot be determined. */
4620 netdev_max_backlog(void)
4622 static struct ovsthread_once once
= OVSTHREAD_ONCE_INITIALIZER
;
4623 static int max_backlog
= 1000; /* The normal default value. */
4625 if (ovsthread_once_start(&once
)) {
4626 static const char filename
[] = "/proc/sys/net/core/netdev_max_backlog";
4630 stream
= fopen(filename
, "r");
4632 VLOG_WARN("%s: open failed (%s)", filename
, ovs_strerror(errno
));
4634 if (fscanf(stream
, "%d", &n
) != 1) {
4635 VLOG_WARN("%s: read error", filename
);
4636 } else if (n
<= 100) {
4637 VLOG_WARN("%s: unexpectedly small value %d", filename
, n
);
4643 ovsthread_once_done(&once
);
4645 VLOG_DBG("%s: using %d max_backlog", filename
, max_backlog
);
4651 /* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in
4654 count_output_actions(const struct ofpbuf
*odp_actions
)
4656 const struct nlattr
*a
;
4660 NL_ATTR_FOR_EACH_UNSAFE (a
, left
, odp_actions
->data
, odp_actions
->size
) {
4661 if (a
->nla_type
== OVS_ACTION_ATTR_OUTPUT
) {
4667 #endif /* defined(__linux__) */
4669 /* Returns true if 'odp_actions' contains more output actions than the datapath
4670 * can reliably handle in one go. On Linux, this is the value of the
4671 * net.core.netdev_max_backlog sysctl, which limits the maximum number of
4672 * packets that the kernel is willing to queue up for processing while the
4673 * datapath is processing a set of actions. */
4675 too_many_output_actions(const struct ofpbuf
*odp_actions OVS_UNUSED
)
4678 return (odp_actions
->size
/ NL_A_U32_SIZE
> netdev_max_backlog()
4679 && count_output_actions(odp_actions
) > netdev_max_backlog());
4681 /* OSes other than Linux might have similar limits, but we don't know how
4682 * to determine them.*/
4688 xlate_wc_init(struct xlate_ctx
*ctx
)
4690 flow_wildcards_init_catchall(ctx
->wc
);
4692 /* Some fields we consider to always be examined. */
4693 WC_MASK_FIELD(ctx
->wc
, in_port
);
4694 WC_MASK_FIELD(ctx
->wc
, dl_type
);
4695 if (is_ip_any(&ctx
->xin
->flow
)) {
4696 WC_MASK_FIELD_MASK(ctx
->wc
, nw_frag
, FLOW_NW_FRAG_MASK
);
4699 if (ctx
->xbridge
->support
.odp
.recirc
) {
4700 /* Always exactly match recirc_id when datapath supports
4702 WC_MASK_FIELD(ctx
->wc
, recirc_id
);
4705 if (ctx
->xbridge
->netflow
) {
4706 netflow_mask_wc(&ctx
->xin
->flow
, ctx
->wc
);
4709 tnl_wc_init(&ctx
->xin
->flow
, ctx
->wc
);
4713 xlate_wc_finish(struct xlate_ctx
*ctx
)
4715 /* Clear the metadata and register wildcard masks, because we won't
4716 * use non-header fields as part of the cache. */
4717 flow_wildcards_clear_non_packet_fields(ctx
->wc
);
4719 /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow
4720 * uses the low 8 bits of the 16-bit tp_src and tp_dst members to
4721 * represent these fields. The datapath interface, on the other hand,
4722 * represents them with just 8 bits each. This means that if the high
4723 * 8 bits of the masks for these fields somehow become set, then they
4724 * will get chopped off by a round trip through the datapath, and
4725 * revalidation will spot that as an inconsistency and delete the flow.
4726 * Avoid the problem here by making sure that only the low 8 bits of
4727 * either field can be unwildcarded for ICMP.
4729 if (is_icmpv4(&ctx
->xin
->flow
) || is_icmpv6(&ctx
->xin
->flow
)) {
4730 ctx
->wc
->masks
.tp_src
&= htons(UINT8_MAX
);
4731 ctx
->wc
->masks
.tp_dst
&= htons(UINT8_MAX
);
4733 /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */
4734 if (ctx
->wc
->masks
.vlan_tci
) {
4735 ctx
->wc
->masks
.vlan_tci
|= htons(VLAN_CFI
);
4739 /* Translates the flow, actions, or rule in 'xin' into datapath actions in
4741 * The caller must take responsibility for eventually freeing 'xout', with
4742 * xlate_out_uninit(). */
4744 xlate_actions(struct xlate_in
*xin
, struct xlate_out
*xout
)
4746 *xout
= (struct xlate_out
) {
4752 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
4753 struct xbridge
*xbridge
= xbridge_lookup(xcfg
, xin
->ofproto
);
4758 struct flow
*flow
= &xin
->flow
;
4760 union mf_subvalue stack_stub
[1024 / sizeof(union mf_subvalue
)];
4761 uint64_t action_set_stub
[1024 / 8];
4762 struct flow_wildcards scratch_wc
;
4763 uint64_t actions_stub
[256 / 8];
4764 struct ofpbuf scratch_actions
= OFPBUF_STUB_INITIALIZER(actions_stub
);
4765 struct xlate_ctx ctx
= {
4769 .orig_tunnel_ip_dst
= flow
->tunnel
.ip_dst
,
4771 .stack
= OFPBUF_STUB_INITIALIZER(stack_stub
),
4773 .wc
= xin
->wc
? xin
->wc
: &scratch_wc
,
4774 .odp_actions
= xin
->odp_actions
? xin
->odp_actions
: &scratch_actions
,
4779 .in_action_set
= false,
4782 .rule_cookie
= OVS_BE64_MAX
,
4783 .orig_skb_priority
= flow
->skb_priority
,
4784 .sflow_n_outputs
= 0,
4785 .sflow_odp_port
= 0,
4786 .nf_output_iface
= NF_OUT_DROP
,
4790 .recirc_action_offset
= -1,
4791 .last_unroll_offset
= -1,
4795 .action_set_has_group
= false,
4796 .action_set
= OFPBUF_STUB_INITIALIZER(action_set_stub
),
4799 /* 'base_flow' reflects the packet as it came in, but we need it to reflect
4800 * the packet as the datapath will treat it for output actions:
4802 * - Our datapath doesn't retain tunneling information without us
4803 * re-setting it, so clear the tunnel data.
4805 * - For VLAN splinters, a higher layer may pretend that the packet
4806 * came in on 'flow->in_port.ofp_port' with 'flow->vlan_tci'
4807 * attached, because that's how we want to treat it from an OpenFlow
4808 * perspective. But from the datapath's perspective it actually came
4809 * in on a VLAN device without any VLAN attached. So here we put the
4810 * datapath's view of the VLAN information in 'base_flow' to ensure
4811 * correct treatment.
4813 memset(&ctx
.base_flow
.tunnel
, 0, sizeof ctx
.base_flow
.tunnel
);
4814 if (flow
->in_port
.ofp_port
4815 != vsp_realdev_to_vlandev(xbridge
->ofproto
,
4816 flow
->in_port
.ofp_port
,
4818 ctx
.base_flow
.vlan_tci
= 0;
4821 ofpbuf_reserve(ctx
.odp_actions
, NL_A_U32_SIZE
);
4823 xlate_wc_init(&ctx
);
4826 COVERAGE_INC(xlate_actions
);
4829 const struct recirc_state
*state
= &xin
->recirc
->state
;
4831 xlate_report(&ctx
, "Restoring state post-recirculation:");
4833 if (xin
->ofpacts_len
> 0 || ctx
.rule
) {
4834 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
4835 const char *conflict
= xin
->ofpacts_len
? "actions" : "rule";
4837 VLOG_WARN_RL(&rl
, "Recirculation conflict (%s)!", conflict
);
4838 xlate_report(&ctx
, "- Recirculation conflict (%s)!", conflict
);
4842 /* Set the bridge for post-recirculation processing if needed. */
4843 if (ctx
.xbridge
->ofproto
!= state
->ofproto
) {
4844 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
4845 const struct xbridge
*new_bridge
4846 = xbridge_lookup(xcfg
, state
->ofproto
);
4848 if (OVS_UNLIKELY(!new_bridge
)) {
4849 /* Drop the packet if the bridge cannot be found. */
4850 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
4851 VLOG_WARN_RL(&rl
, "Recirculation bridge no longer exists.");
4852 xlate_report(&ctx
, "- Recirculation bridge no longer exists.");
4855 ctx
.xbridge
= new_bridge
;
4858 /* Set the post-recirculation table id. Note: A table lookup is done
4859 * only if there are no post-recirculation actions. */
4860 ctx
.table_id
= state
->table_id
;
4861 xlate_report(&ctx
, "- Resuming from table %"PRIu8
, ctx
.table_id
);
4863 /* Restore pipeline metadata. May change flow's in_port and other
4864 * metadata to the values that existed when recirculation was
4866 recirc_metadata_to_flow(&state
->metadata
, flow
);
4868 /* Restore stack, if any. */
4870 ofpbuf_put(&ctx
.stack
, state
->stack
->data
, state
->stack
->size
);
4873 /* Restore mirror state. */
4874 ctx
.mirrors
= state
->mirrors
;
4876 /* Restore action set, if any. */
4877 if (state
->action_set_len
) {
4878 const struct ofpact
*a
;
4880 xlate_report_actions(&ctx
, "- Restoring action set",
4881 state
->ofpacts
, state
->action_set_len
);
4883 ofpbuf_put(&ctx
.action_set
, state
->ofpacts
, state
->action_set_len
);
4885 OFPACT_FOR_EACH(a
, state
->ofpacts
, state
->action_set_len
) {
4886 if (a
->type
== OFPACT_GROUP
) {
4887 ctx
.action_set_has_group
= true;
4893 /* Restore recirculation actions. If there are no actions, processing
4894 * will start with a lookup in the table set above. */
4895 if (state
->ofpacts_len
> state
->action_set_len
) {
4896 xin
->ofpacts_len
= state
->ofpacts_len
- state
->action_set_len
;
4897 xin
->ofpacts
= state
->ofpacts
+
4898 state
->action_set_len
/ sizeof *state
->ofpacts
;
4900 xlate_report_actions(&ctx
, "- Restoring actions",
4901 xin
->ofpacts
, xin
->ofpacts_len
);
4903 } else if (OVS_UNLIKELY(flow
->recirc_id
)) {
4904 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
4906 VLOG_WARN_RL(&rl
, "Recirculation context not found for ID %"PRIx32
,
4910 /* The bridge is now known so obtain its table version. */
4911 ctx
.tables_version
= ofproto_dpif_get_tables_version(ctx
.xbridge
->ofproto
);
4913 if (!xin
->ofpacts
&& !ctx
.rule
) {
4914 ctx
.rule
= rule_dpif_lookup_from_table(
4915 ctx
.xbridge
->ofproto
, ctx
.tables_version
, flow
, xin
->wc
,
4916 ctx
.xin
->resubmit_stats
, &ctx
.table_id
,
4917 flow
->in_port
.ofp_port
, true, true);
4918 if (ctx
.xin
->resubmit_stats
) {
4919 rule_dpif_credit_stats(ctx
.rule
, ctx
.xin
->resubmit_stats
);
4921 if (ctx
.xin
->xcache
) {
4922 struct xc_entry
*entry
;
4924 entry
= xlate_cache_add_entry(ctx
.xin
->xcache
, XC_RULE
);
4925 entry
->u
.rule
= ctx
.rule
;
4926 rule_dpif_ref(ctx
.rule
);
4929 if (OVS_UNLIKELY(ctx
.xin
->resubmit_hook
)) {
4930 ctx
.xin
->resubmit_hook(ctx
.xin
, ctx
.rule
, 0);
4933 xout
->fail_open
= ctx
.rule
&& rule_dpif_is_fail_open(ctx
.rule
);
4935 /* Get the proximate input port of the packet. (If xin->recirc,
4936 * flow->in_port is the ultimate input port of the packet.) */
4937 struct xport
*in_port
= get_ofp_port(xbridge
,
4938 ctx
.base_flow
.in_port
.ofp_port
);
4940 /* Tunnel stats only for non-recirculated packets. */
4941 if (!xin
->recirc
&& in_port
&& in_port
->is_tunnel
) {
4942 if (ctx
.xin
->resubmit_stats
) {
4943 netdev_vport_inc_rx(in_port
->netdev
, ctx
.xin
->resubmit_stats
);
4945 bfd_account_rx(in_port
->bfd
, ctx
.xin
->resubmit_stats
);
4948 if (ctx
.xin
->xcache
) {
4949 struct xc_entry
*entry
;
4951 entry
= xlate_cache_add_entry(ctx
.xin
->xcache
, XC_NETDEV
);
4952 entry
->u
.dev
.rx
= netdev_ref(in_port
->netdev
);
4953 entry
->u
.dev
.bfd
= bfd_ref(in_port
->bfd
);
4957 if (!xin
->recirc
&& process_special(&ctx
, in_port
)) {
4958 /* process_special() did all the processing for this packet.
4960 * We do not perform special processing on recirculated packets, as
4961 * recirculated packets are not really received by the bridge.*/
4962 } else if (in_port
&& in_port
->xbundle
4963 && xbundle_mirror_out(xbridge
, in_port
->xbundle
)) {
4964 if (ctx
.xin
->packet
!= NULL
) {
4965 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
4966 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet received on port "
4967 "%s, which is reserved exclusively for mirroring",
4968 ctx
.xbridge
->name
, in_port
->xbundle
->name
);
4971 /* Sampling is done only for packets really received by the bridge. */
4972 unsigned int user_cookie_offset
= 0;
4974 user_cookie_offset
= compose_sflow_action(&ctx
);
4975 compose_ipfix_action(&ctx
, ODPP_NONE
);
4977 size_t sample_actions_len
= ctx
.odp_actions
->size
;
4979 if (tnl_process_ecn(flow
)
4980 && (!in_port
|| may_receive(in_port
, &ctx
))) {
4981 const struct ofpact
*ofpacts
;
4985 ofpacts
= xin
->ofpacts
;
4986 ofpacts_len
= xin
->ofpacts_len
;
4987 } else if (ctx
.rule
) {
4988 const struct rule_actions
*actions
4989 = rule_dpif_get_actions(ctx
.rule
);
4990 ofpacts
= actions
->ofpacts
;
4991 ofpacts_len
= actions
->ofpacts_len
;
4992 ctx
.rule_cookie
= rule_dpif_get_flow_cookie(ctx
.rule
);
4997 mirror_ingress_packet(&ctx
);
4998 do_xlate_actions(ofpacts
, ofpacts_len
, &ctx
);
5000 /* We've let OFPP_NORMAL and the learning action look at the
5001 * packet, so drop it now if forwarding is disabled. */
5002 if (in_port
&& (!xport_stp_forward_state(in_port
) ||
5003 !xport_rstp_forward_state(in_port
))) {
5004 /* Drop all actions added by do_xlate_actions() above. */
5005 ctx
.odp_actions
->size
= sample_actions_len
;
5007 /* Undo changes that may have been done for recirculation. */
5008 if (exit_recirculates(&ctx
)) {
5009 ctx
.action_set
.size
= ctx
.recirc_action_offset
;
5010 ctx
.recirc_action_offset
= -1;
5011 ctx
.last_unroll_offset
= -1;
5013 } else if (ctx
.action_set
.size
) {
5014 /* Translate action set only if not dropping the packet and
5015 * not recirculating. */
5016 if (!exit_recirculates(&ctx
)) {
5017 xlate_action_set(&ctx
);
5020 /* Check if need to recirculate. */
5021 if (exit_recirculates(&ctx
)) {
5022 compose_recirculate_action(&ctx
);
5026 /* Output only fully processed packets. */
5027 if (!exit_recirculates(&ctx
)
5028 && xbridge
->has_in_band
5029 && in_band_must_output_to_local_port(flow
)
5030 && !actions_output_to_local_port(&ctx
)) {
5031 compose_output_action(&ctx
, OFPP_LOCAL
, NULL
);
5034 if (user_cookie_offset
) {
5035 fix_sflow_action(&ctx
, user_cookie_offset
);
5039 if (nl_attr_oversized(ctx
.odp_actions
->size
)) {
5040 /* These datapath actions are too big for a Netlink attribute, so we
5041 * can't hand them to the kernel directly. dpif_execute() can execute
5042 * them one by one with help, so just mark the result as SLOW_ACTION to
5043 * prevent the flow from being installed. */
5044 COVERAGE_INC(xlate_actions_oversize
);
5045 ctx
.xout
->slow
|= SLOW_ACTION
;
5046 } else if (too_many_output_actions(ctx
.odp_actions
)) {
5047 COVERAGE_INC(xlate_actions_too_many_output
);
5048 ctx
.xout
->slow
|= SLOW_ACTION
;
5051 /* Do netflow only for packets really received by the bridge and not sent
5052 * to the controller. We consider packets sent to the controller to be
5053 * part of the control plane rather than the data plane. */
5054 if (!xin
->recirc
&& xbridge
->netflow
&& !(xout
->slow
& SLOW_CONTROLLER
)) {
5055 if (ctx
.xin
->resubmit_stats
) {
5056 netflow_flow_update(xbridge
->netflow
, flow
,
5057 ctx
.nf_output_iface
,
5058 ctx
.xin
->resubmit_stats
);
5060 if (ctx
.xin
->xcache
) {
5061 struct xc_entry
*entry
;
5063 entry
= xlate_cache_add_entry(ctx
.xin
->xcache
, XC_NETFLOW
);
5064 entry
->u
.nf
.netflow
= netflow_ref(xbridge
->netflow
);
5065 entry
->u
.nf
.flow
= xmemdup(flow
, sizeof *flow
);
5066 entry
->u
.nf
.iface
= ctx
.nf_output_iface
;
5071 xlate_wc_finish(&ctx
);
5075 ofpbuf_uninit(&ctx
.stack
);
5076 ofpbuf_uninit(&ctx
.action_set
);
5077 ofpbuf_uninit(&scratch_actions
);
5080 /* Sends 'packet' out 'ofport'.
5081 * May modify 'packet'.
5082 * Returns 0 if successful, otherwise a positive errno value. */
5084 xlate_send_packet(const struct ofport_dpif
*ofport
, struct dp_packet
*packet
)
5086 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
5087 struct xport
*xport
;
5088 struct ofpact_output output
;
5091 ofpact_init(&output
.ofpact
, OFPACT_OUTPUT
, sizeof output
);
5092 /* Use OFPP_NONE as the in_port to avoid special packet processing. */
5093 flow_extract(packet
, &flow
);
5094 flow
.in_port
.ofp_port
= OFPP_NONE
;
5096 xport
= xport_lookup(xcfg
, ofport
);
5100 output
.port
= xport
->ofp_port
;
5103 return ofproto_dpif_execute_actions(xport
->xbridge
->ofproto
, &flow
, NULL
,
5104 &output
.ofpact
, sizeof output
,
5108 struct xlate_cache
*
5109 xlate_cache_new(void)
5111 struct xlate_cache
*xcache
= xmalloc(sizeof *xcache
);
5113 ofpbuf_init(&xcache
->entries
, 512);
5117 static struct xc_entry
*
5118 xlate_cache_add_entry(struct xlate_cache
*xcache
, enum xc_type type
)
5120 struct xc_entry
*entry
;
5122 entry
= ofpbuf_put_zeros(&xcache
->entries
, sizeof *entry
);
5129 xlate_cache_netdev(struct xc_entry
*entry
, const struct dpif_flow_stats
*stats
)
5131 if (entry
->u
.dev
.tx
) {
5132 netdev_vport_inc_tx(entry
->u
.dev
.tx
, stats
);
5134 if (entry
->u
.dev
.rx
) {
5135 netdev_vport_inc_rx(entry
->u
.dev
.rx
, stats
);
5137 if (entry
->u
.dev
.bfd
) {
5138 bfd_account_rx(entry
->u
.dev
.bfd
, stats
);
5143 xlate_cache_normal(struct ofproto_dpif
*ofproto
, struct flow
*flow
, int vlan
)
5145 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
5146 struct xbridge
*xbridge
;
5147 struct xbundle
*xbundle
;
5148 struct flow_wildcards wc
;
5150 xbridge
= xbridge_lookup(xcfg
, ofproto
);
5155 xbundle
= lookup_input_bundle(xbridge
, flow
->in_port
.ofp_port
, false,
5161 update_learning_table(xbridge
, flow
, &wc
, vlan
, xbundle
);
5164 /* Push stats and perform side effects of flow translation. */
5166 xlate_push_stats(struct xlate_cache
*xcache
,
5167 const struct dpif_flow_stats
*stats
)
5169 struct xc_entry
*entry
;
5170 struct ofpbuf entries
= xcache
->entries
;
5171 struct eth_addr dmac
;
5173 if (!stats
->n_packets
) {
5177 XC_ENTRY_FOR_EACH (entry
, entries
, xcache
) {
5178 switch (entry
->type
) {
5180 rule_dpif_credit_stats(entry
->u
.rule
, stats
);
5183 bond_account(entry
->u
.bond
.bond
, entry
->u
.bond
.flow
,
5184 entry
->u
.bond
.vid
, stats
->n_bytes
);
5187 xlate_cache_netdev(entry
, stats
);
5190 netflow_flow_update(entry
->u
.nf
.netflow
, entry
->u
.nf
.flow
,
5191 entry
->u
.nf
.iface
, stats
);
5194 mirror_update_stats(entry
->u
.mirror
.mbridge
,
5195 entry
->u
.mirror
.mirrors
,
5196 stats
->n_packets
, stats
->n_bytes
);
5199 ofproto_dpif_flow_mod(entry
->u
.learn
.ofproto
, entry
->u
.learn
.fm
);
5202 xlate_cache_normal(entry
->u
.normal
.ofproto
, entry
->u
.normal
.flow
,
5203 entry
->u
.normal
.vlan
);
5205 case XC_FIN_TIMEOUT
:
5206 xlate_fin_timeout__(entry
->u
.fin
.rule
, stats
->tcp_flags
,
5207 entry
->u
.fin
.idle
, entry
->u
.fin
.hard
);
5210 group_dpif_credit_stats(entry
->u
.group
.group
, entry
->u
.group
.bucket
,
5214 /* Lookup arp to avoid arp timeout. */
5215 tnl_arp_lookup(entry
->u
.tnl_arp_cache
.br_name
,
5216 entry
->u
.tnl_arp_cache
.d_ip
, &dmac
);
5225 xlate_dev_unref(struct xc_entry
*entry
)
5227 if (entry
->u
.dev
.tx
) {
5228 netdev_close(entry
->u
.dev
.tx
);
5230 if (entry
->u
.dev
.rx
) {
5231 netdev_close(entry
->u
.dev
.rx
);
5233 if (entry
->u
.dev
.bfd
) {
5234 bfd_unref(entry
->u
.dev
.bfd
);
5239 xlate_cache_clear_netflow(struct netflow
*netflow
, struct flow
*flow
)
5241 netflow_flow_clear(netflow
, flow
);
5242 netflow_unref(netflow
);
5247 xlate_cache_clear(struct xlate_cache
*xcache
)
5249 struct xc_entry
*entry
;
5250 struct ofpbuf entries
;
5256 XC_ENTRY_FOR_EACH (entry
, entries
, xcache
) {
5257 switch (entry
->type
) {
5259 rule_dpif_unref(entry
->u
.rule
);
5262 free(entry
->u
.bond
.flow
);
5263 bond_unref(entry
->u
.bond
.bond
);
5266 xlate_dev_unref(entry
);
5269 xlate_cache_clear_netflow(entry
->u
.nf
.netflow
, entry
->u
.nf
.flow
);
5272 mbridge_unref(entry
->u
.mirror
.mbridge
);
5275 free(entry
->u
.learn
.fm
);
5276 ofpbuf_delete(entry
->u
.learn
.ofpacts
);
5279 free(entry
->u
.normal
.flow
);
5281 case XC_FIN_TIMEOUT
:
5282 /* 'u.fin.rule' is always already held as a XC_RULE, which
5283 * has already released it's reference above. */
5286 group_dpif_unref(entry
->u
.group
.group
);
5295 ofpbuf_clear(&xcache
->entries
);
5299 xlate_cache_delete(struct xlate_cache
*xcache
)
5301 xlate_cache_clear(xcache
);
5302 ofpbuf_uninit(&xcache
->entries
);