1 /* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014 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"
25 #include "byte-order.h"
30 #include "dynamic-string.h"
35 #include "mac-learning.h"
36 #include "mcast-snooping.h"
37 #include "meta-flow.h"
38 #include "multipath.h"
39 #include "netdev-vport.h"
42 #include "odp-execute.h"
43 #include "ofp-actions.h"
44 #include "ofproto/ofproto-dpif-ipfix.h"
45 #include "ofproto/ofproto-dpif-mirror.h"
46 #include "ofproto/ofproto-dpif-monitor.h"
47 #include "ofproto/ofproto-dpif-sflow.h"
48 #include "ofproto/ofproto-dpif.h"
49 #include "ofproto/ofproto-provider.h"
50 #include "packet-dpif.h"
54 COVERAGE_DEFINE(xlate_actions
);
55 COVERAGE_DEFINE(xlate_actions_oversize
);
56 COVERAGE_DEFINE(xlate_actions_mpls_overflow
);
58 VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate
);
60 /* Maximum depth of flow table recursion (due to resubmit actions) in a
61 * flow translation. */
62 #define MAX_RESUBMIT_RECURSION 64
63 #define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in
66 /* Timeout for internal rules created to handle recirculation */
67 #define RECIRC_TIMEOUT 60
69 /* Maximum number of resubmit actions in a flow translation, whether they are
70 * recursive or not. */
71 #define MAX_RESUBMITS (MAX_RESUBMIT_RECURSION * MAX_RESUBMIT_RECURSION)
74 struct hmap_node hmap_node
; /* Node in global 'xbridges' map. */
75 struct ofproto_dpif
*ofproto
; /* Key in global 'xbridges' map. */
77 struct list xbundles
; /* Owned xbundles. */
78 struct hmap xports
; /* Indexed by ofp_port. */
80 char *name
; /* Name used in log messages. */
81 struct dpif
*dpif
; /* Datapath interface. */
82 struct mac_learning
*ml
; /* Mac learning handle. */
83 struct mcast_snooping
*ms
; /* Multicast Snooping handle. */
84 struct mbridge
*mbridge
; /* Mirroring. */
85 struct dpif_sflow
*sflow
; /* SFlow handle, or null. */
86 struct dpif_ipfix
*ipfix
; /* Ipfix handle, or null. */
87 struct netflow
*netflow
; /* Netflow handle, or null. */
88 struct stp
*stp
; /* STP or null if disabled. */
90 /* Special rules installed by ofproto-dpif. */
91 struct rule_dpif
*miss_rule
;
92 struct rule_dpif
*no_packet_in_rule
;
94 enum ofp_config_flags frag
; /* Fragmentation handling. */
95 bool has_in_band
; /* Bridge has in band control? */
96 bool forward_bpdu
; /* Bridge forwards STP BPDUs? */
98 /* True if the datapath supports recirculation. */
101 /* True if the datapath supports variable-length
102 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions.
103 * False if the datapath supports only 8-byte (or shorter) userdata. */
104 bool variable_length_userdata
;
106 /* Number of MPLS label stack entries that the datapath supports
108 size_t max_mpls_depth
;
112 struct hmap_node hmap_node
; /* In global 'xbundles' map. */
113 struct ofbundle
*ofbundle
; /* Key in global 'xbundles' map. */
115 struct list list_node
; /* In parent 'xbridges' list. */
116 struct xbridge
*xbridge
; /* Parent xbridge. */
118 struct list xports
; /* Contains "struct xport"s. */
120 char *name
; /* Name used in log messages. */
121 struct bond
*bond
; /* Nonnull iff more than one port. */
122 struct lacp
*lacp
; /* LACP handle or null. */
124 enum port_vlan_mode vlan_mode
; /* VLAN mode. */
125 int vlan
; /* -1=trunk port, else a 12-bit VLAN ID. */
126 unsigned long *trunks
; /* Bitmap of trunked VLANs, if 'vlan' == -1.
127 * NULL if all VLANs are trunked. */
128 bool use_priority_tags
; /* Use 802.1p tag for frames in VLAN 0? */
129 bool floodable
; /* No port has OFPUTIL_PC_NO_FLOOD set? */
133 struct hmap_node hmap_node
; /* Node in global 'xports' map. */
134 struct ofport_dpif
*ofport
; /* Key in global 'xports map. */
136 struct hmap_node ofp_node
; /* Node in parent xbridge 'xports' map. */
137 ofp_port_t ofp_port
; /* Key in parent xbridge 'xports' map. */
139 odp_port_t odp_port
; /* Datapath port number or ODPP_NONE. */
141 struct list bundle_node
; /* In parent xbundle (if it exists). */
142 struct xbundle
*xbundle
; /* Parent xbundle or null. */
144 struct netdev
*netdev
; /* 'ofport''s netdev. */
146 struct xbridge
*xbridge
; /* Parent bridge. */
147 struct xport
*peer
; /* Patch port peer or null. */
149 enum ofputil_port_config config
; /* OpenFlow port configuration. */
150 enum ofputil_port_state state
; /* OpenFlow port state. */
151 int stp_port_no
; /* STP port number or -1 if not in use. */
153 struct hmap skb_priorities
; /* Map of 'skb_priority_to_dscp's. */
155 bool may_enable
; /* May be enabled in bonds. */
156 bool is_tunnel
; /* Is a tunnel port. */
158 struct cfm
*cfm
; /* CFM handle or null. */
159 struct bfd
*bfd
; /* BFD handle or null. */
163 struct xlate_in
*xin
;
164 struct xlate_out
*xout
;
166 const struct xbridge
*xbridge
;
168 /* Flow at the last commit. */
169 struct flow base_flow
;
171 /* Tunnel IP destination address as received. This is stored separately
172 * as the base_flow.tunnel is cleared on init to reflect the datapath
173 * behavior. Used to make sure not to send tunneled output to ourselves,
174 * which might lead to an infinite loop. This could happen easily
175 * if a tunnel is marked as 'ip_remote=flow', and the flow does not
176 * actually set the tun_dst field. */
177 ovs_be32 orig_tunnel_ip_dst
;
179 /* Stack for the push and pop actions. Each stack element is of type
180 * "union mf_subvalue". */
181 union mf_subvalue init_stack
[1024 / sizeof(union mf_subvalue
)];
184 /* The rule that we are currently translating, or NULL. */
185 struct rule_dpif
*rule
;
187 /* Resubmit statistics, via xlate_table_action(). */
188 int recurse
; /* Current resubmit nesting depth. */
189 int resubmits
; /* Total number of resubmits. */
190 bool in_group
; /* Currently translating ofgroup, if true. */
192 uint32_t orig_skb_priority
; /* Priority when packet arrived. */
193 uint8_t table_id
; /* OpenFlow table ID where flow was found. */
194 uint32_t sflow_n_outputs
; /* Number of output ports. */
195 odp_port_t sflow_odp_port
; /* Output port for composing sFlow action. */
196 uint16_t user_cookie_offset
;/* Used for user_action_cookie fixup. */
197 bool exit
; /* No further actions should be processed. */
199 bool use_recirc
; /* Should generate recirc? */
200 struct xlate_recirc recirc
; /* Information used for generating
201 * recirculation actions */
203 /* True if a packet was but is no longer MPLS (due to an MPLS pop action).
204 * This is a trigger for recirculation in cases where translating an action
205 * or looking up a flow requires access to the fields of the packet after
206 * the MPLS label stack that was originally present. */
209 /* OpenFlow 1.1+ action set.
211 * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS.
212 * When translation is otherwise complete, ofpacts_execute_action_set()
213 * converts it to a set of "struct ofpact"s that can be translated into
214 * datapath actions. */
215 struct ofpbuf action_set
; /* Action set. */
216 uint64_t action_set_stub
[1024 / 8];
219 /* A controller may use OFPP_NONE as the ingress port to indicate that
220 * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for
221 * when an input bundle is needed for validation (e.g., mirroring or
222 * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have
223 * any 'port' structs, so care must be taken when dealing with it. */
224 static struct xbundle ofpp_none_bundle
= {
226 .vlan_mode
= PORT_VLAN_TRUNK
229 /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from
230 * 'priority' (the datapath's term for QoS queue) to the dscp bits which all
231 * traffic egressing the 'ofport' with that priority should be marked with. */
232 struct skb_priority_to_dscp
{
233 struct hmap_node hmap_node
; /* Node in 'ofport_dpif''s 'skb_priorities'. */
234 uint32_t skb_priority
; /* Priority of this queue (see struct flow). */
236 uint8_t dscp
; /* DSCP bits to mark outgoing traffic with. */
251 /* xlate_cache entries hold enough information to perform the side effects of
252 * xlate_actions() for a rule, without needing to perform rule translation
253 * from scratch. The primary usage of these is to submit statistics to objects
254 * that a flow relates to, although they may be used for other effects as well
255 * (for instance, refreshing hard timeouts for learned flows). */
259 struct rule_dpif
*rule
;
266 struct netflow
*netflow
;
271 struct mbridge
*mbridge
;
272 mirror_mask_t mirrors
;
280 struct ofproto_dpif
*ofproto
;
281 struct ofputil_flow_mod
*fm
;
282 struct ofpbuf
*ofpacts
;
285 struct ofproto_dpif
*ofproto
;
290 struct rule_dpif
*rule
;
295 struct group_dpif
*group
;
296 struct ofputil_bucket
*bucket
;
301 #define XC_ENTRY_FOR_EACH(entry, entries, xcache) \
302 entries = xcache->entries; \
303 for (entry = ofpbuf_try_pull(&entries, sizeof *entry); \
305 entry = ofpbuf_try_pull(&entries, sizeof *entry))
308 struct ofpbuf entries
;
311 /* Xlate config contains hash maps of all bridges, bundles and ports.
312 * Xcfgp contains the pointer to the current xlate configuration.
313 * When the main thread needs to change the configuration, it copies xcfgp to
314 * new_xcfg and edits new_xcfg. This enables the use of RCU locking which
315 * does not block handler and revalidator threads. */
317 struct hmap xbridges
;
318 struct hmap xbundles
;
321 static OVSRCU_TYPE(struct xlate_cfg
*) xcfgp
= OVSRCU_TYPE_INITIALIZER
;
322 static struct xlate_cfg
*new_xcfg
= NULL
;
324 static bool may_receive(const struct xport
*, struct xlate_ctx
*);
325 static void do_xlate_actions(const struct ofpact
*, size_t ofpacts_len
,
327 static void xlate_normal(struct xlate_ctx
*);
328 static void xlate_report(struct xlate_ctx
*, const char *);
329 static void xlate_table_action(struct xlate_ctx
*, ofp_port_t in_port
,
330 uint8_t table_id
, bool may_packet_in
,
331 bool honor_table_miss
);
332 static bool input_vid_is_valid(uint16_t vid
, struct xbundle
*, bool warn
);
333 static uint16_t input_vid_to_vlan(const struct xbundle
*, uint16_t vid
);
334 static void output_normal(struct xlate_ctx
*, const struct xbundle
*,
336 static void compose_output_action(struct xlate_ctx
*, ofp_port_t ofp_port
);
338 static struct xbridge
*xbridge_lookup(struct xlate_cfg
*,
339 const struct ofproto_dpif
*);
340 static struct xbundle
*xbundle_lookup(struct xlate_cfg
*,
341 const struct ofbundle
*);
342 static struct xport
*xport_lookup(struct xlate_cfg
*,
343 const struct ofport_dpif
*);
344 static struct xport
*get_ofp_port(const struct xbridge
*, ofp_port_t ofp_port
);
345 static struct skb_priority_to_dscp
*get_skb_priority(const struct xport
*,
346 uint32_t skb_priority
);
347 static void clear_skb_priorities(struct xport
*);
348 static bool dscp_from_skb_priority(const struct xport
*, uint32_t skb_priority
,
351 static struct xc_entry
*xlate_cache_add_entry(struct xlate_cache
*xc
,
353 static void xlate_xbridge_init(struct xlate_cfg
*, struct xbridge
*);
354 static void xlate_xbundle_init(struct xlate_cfg
*, struct xbundle
*);
355 static void xlate_xport_init(struct xlate_cfg
*, struct xport
*);
356 static void xlate_xbridge_set(struct xbridge
*xbridge
,
358 struct rule_dpif
*miss_rule
,
359 struct rule_dpif
*no_packet_in_rule
,
360 const struct mac_learning
*ml
, struct stp
*stp
,
361 const struct mcast_snooping
*ms
,
362 const struct mbridge
*mbridge
,
363 const struct dpif_sflow
*sflow
,
364 const struct dpif_ipfix
*ipfix
,
365 const struct netflow
*netflow
,
366 enum ofp_config_flags frag
,
367 bool forward_bpdu
, bool has_in_band
,
369 bool variable_length_userdata
,
370 size_t max_mpls_depth
);
371 static void xlate_xbundle_set(struct xbundle
*xbundle
,
372 enum port_vlan_mode vlan_mode
, int vlan
,
373 unsigned long *trunks
, bool use_priority_tags
,
374 const struct bond
*bond
, const struct lacp
*lacp
,
376 static void xlate_xport_set(struct xport
*xport
, odp_port_t odp_port
,
377 const struct netdev
*netdev
, const struct cfm
*cfm
,
378 const struct bfd
*bfd
, int stp_port_no
,
379 enum ofputil_port_config config
,
380 enum ofputil_port_state state
, bool is_tunnel
,
382 static void xlate_xbridge_remove(struct xlate_cfg
*, struct xbridge
*);
383 static void xlate_xbundle_remove(struct xlate_cfg
*, struct xbundle
*);
384 static void xlate_xport_remove(struct xlate_cfg
*, struct xport
*);
385 static void xlate_xbridge_copy(struct xbridge
*);
386 static void xlate_xbundle_copy(struct xbridge
*, struct xbundle
*);
387 static void xlate_xport_copy(struct xbridge
*, struct xbundle
*,
389 static void xlate_xcfg_free(struct xlate_cfg
*);
393 xlate_xbridge_init(struct xlate_cfg
*xcfg
, struct xbridge
*xbridge
)
395 list_init(&xbridge
->xbundles
);
396 hmap_init(&xbridge
->xports
);
397 hmap_insert(&xcfg
->xbridges
, &xbridge
->hmap_node
,
398 hash_pointer(xbridge
->ofproto
, 0));
402 xlate_xbundle_init(struct xlate_cfg
*xcfg
, struct xbundle
*xbundle
)
404 list_init(&xbundle
->xports
);
405 list_insert(&xbundle
->xbridge
->xbundles
, &xbundle
->list_node
);
406 hmap_insert(&xcfg
->xbundles
, &xbundle
->hmap_node
,
407 hash_pointer(xbundle
->ofbundle
, 0));
411 xlate_xport_init(struct xlate_cfg
*xcfg
, struct xport
*xport
)
413 hmap_init(&xport
->skb_priorities
);
414 hmap_insert(&xcfg
->xports
, &xport
->hmap_node
,
415 hash_pointer(xport
->ofport
, 0));
416 hmap_insert(&xport
->xbridge
->xports
, &xport
->ofp_node
,
417 hash_ofp_port(xport
->ofp_port
));
421 xlate_xbridge_set(struct xbridge
*xbridge
,
423 struct rule_dpif
*miss_rule
,
424 struct rule_dpif
*no_packet_in_rule
,
425 const struct mac_learning
*ml
, struct stp
*stp
,
426 const struct mcast_snooping
*ms
,
427 const struct mbridge
*mbridge
,
428 const struct dpif_sflow
*sflow
,
429 const struct dpif_ipfix
*ipfix
,
430 const struct netflow
*netflow
, enum ofp_config_flags frag
,
431 bool forward_bpdu
, bool has_in_band
,
433 bool variable_length_userdata
,
434 size_t max_mpls_depth
)
436 if (xbridge
->ml
!= ml
) {
437 mac_learning_unref(xbridge
->ml
);
438 xbridge
->ml
= mac_learning_ref(ml
);
441 if (xbridge
->ms
!= ms
) {
442 mcast_snooping_unref(xbridge
->ms
);
443 xbridge
->ms
= mcast_snooping_ref(ms
);
446 if (xbridge
->mbridge
!= mbridge
) {
447 mbridge_unref(xbridge
->mbridge
);
448 xbridge
->mbridge
= mbridge_ref(mbridge
);
451 if (xbridge
->sflow
!= sflow
) {
452 dpif_sflow_unref(xbridge
->sflow
);
453 xbridge
->sflow
= dpif_sflow_ref(sflow
);
456 if (xbridge
->ipfix
!= ipfix
) {
457 dpif_ipfix_unref(xbridge
->ipfix
);
458 xbridge
->ipfix
= dpif_ipfix_ref(ipfix
);
461 if (xbridge
->stp
!= stp
) {
462 stp_unref(xbridge
->stp
);
463 xbridge
->stp
= stp_ref(stp
);
466 if (xbridge
->netflow
!= netflow
) {
467 netflow_unref(xbridge
->netflow
);
468 xbridge
->netflow
= netflow_ref(netflow
);
471 xbridge
->dpif
= dpif
;
472 xbridge
->forward_bpdu
= forward_bpdu
;
473 xbridge
->has_in_band
= has_in_band
;
474 xbridge
->frag
= frag
;
475 xbridge
->miss_rule
= miss_rule
;
476 xbridge
->no_packet_in_rule
= no_packet_in_rule
;
477 xbridge
->enable_recirc
= enable_recirc
;
478 xbridge
->variable_length_userdata
= variable_length_userdata
;
479 xbridge
->max_mpls_depth
= max_mpls_depth
;
483 xlate_xbundle_set(struct xbundle
*xbundle
,
484 enum port_vlan_mode vlan_mode
, int vlan
,
485 unsigned long *trunks
, bool use_priority_tags
,
486 const struct bond
*bond
, const struct lacp
*lacp
,
489 ovs_assert(xbundle
->xbridge
);
491 xbundle
->vlan_mode
= vlan_mode
;
492 xbundle
->vlan
= vlan
;
493 xbundle
->trunks
= trunks
;
494 xbundle
->use_priority_tags
= use_priority_tags
;
495 xbundle
->floodable
= floodable
;
497 if (xbundle
->bond
!= bond
) {
498 bond_unref(xbundle
->bond
);
499 xbundle
->bond
= bond_ref(bond
);
502 if (xbundle
->lacp
!= lacp
) {
503 lacp_unref(xbundle
->lacp
);
504 xbundle
->lacp
= lacp_ref(lacp
);
509 xlate_xport_set(struct xport
*xport
, odp_port_t odp_port
,
510 const struct netdev
*netdev
, const struct cfm
*cfm
,
511 const struct bfd
*bfd
, int stp_port_no
,
512 enum ofputil_port_config config
, enum ofputil_port_state state
,
513 bool is_tunnel
, bool may_enable
)
515 xport
->config
= config
;
516 xport
->state
= state
;
517 xport
->stp_port_no
= stp_port_no
;
518 xport
->is_tunnel
= is_tunnel
;
519 xport
->may_enable
= may_enable
;
520 xport
->odp_port
= odp_port
;
522 if (xport
->cfm
!= cfm
) {
523 cfm_unref(xport
->cfm
);
524 xport
->cfm
= cfm_ref(cfm
);
527 if (xport
->bfd
!= bfd
) {
528 bfd_unref(xport
->bfd
);
529 xport
->bfd
= bfd_ref(bfd
);
532 if (xport
->netdev
!= netdev
) {
533 netdev_close(xport
->netdev
);
534 xport
->netdev
= netdev_ref(netdev
);
539 xlate_xbridge_copy(struct xbridge
*xbridge
)
541 struct xbundle
*xbundle
;
543 struct xbridge
*new_xbridge
= xzalloc(sizeof *xbridge
);
544 new_xbridge
->ofproto
= xbridge
->ofproto
;
545 new_xbridge
->name
= xstrdup(xbridge
->name
);
546 xlate_xbridge_init(new_xcfg
, new_xbridge
);
548 xlate_xbridge_set(new_xbridge
,
549 xbridge
->dpif
, xbridge
->miss_rule
,
550 xbridge
->no_packet_in_rule
, xbridge
->ml
, xbridge
->stp
,
551 xbridge
->ms
, xbridge
->mbridge
, xbridge
->sflow
,
552 xbridge
->ipfix
, xbridge
->netflow
, xbridge
->frag
,
553 xbridge
->forward_bpdu
, xbridge
->has_in_band
,
554 xbridge
->enable_recirc
, xbridge
->variable_length_userdata
,
555 xbridge
->max_mpls_depth
);
556 LIST_FOR_EACH (xbundle
, list_node
, &xbridge
->xbundles
) {
557 xlate_xbundle_copy(new_xbridge
, xbundle
);
560 /* Copy xports which are not part of a xbundle */
561 HMAP_FOR_EACH (xport
, ofp_node
, &xbridge
->xports
) {
562 if (!xport
->xbundle
) {
563 xlate_xport_copy(new_xbridge
, NULL
, xport
);
569 xlate_xbundle_copy(struct xbridge
*xbridge
, struct xbundle
*xbundle
)
572 struct xbundle
*new_xbundle
= xzalloc(sizeof *xbundle
);
573 new_xbundle
->ofbundle
= xbundle
->ofbundle
;
574 new_xbundle
->xbridge
= xbridge
;
575 new_xbundle
->name
= xstrdup(xbundle
->name
);
576 xlate_xbundle_init(new_xcfg
, new_xbundle
);
578 xlate_xbundle_set(new_xbundle
, xbundle
->vlan_mode
,
579 xbundle
->vlan
, xbundle
->trunks
,
580 xbundle
->use_priority_tags
, xbundle
->bond
, xbundle
->lacp
,
582 LIST_FOR_EACH (xport
, bundle_node
, &xbundle
->xports
) {
583 xlate_xport_copy(xbridge
, new_xbundle
, xport
);
588 xlate_xport_copy(struct xbridge
*xbridge
, struct xbundle
*xbundle
,
591 struct skb_priority_to_dscp
*pdscp
, *new_pdscp
;
592 struct xport
*new_xport
= xzalloc(sizeof *xport
);
593 new_xport
->ofport
= xport
->ofport
;
594 new_xport
->ofp_port
= xport
->ofp_port
;
595 new_xport
->xbridge
= xbridge
;
596 xlate_xport_init(new_xcfg
, new_xport
);
598 xlate_xport_set(new_xport
, xport
->odp_port
, xport
->netdev
, xport
->cfm
,
599 xport
->bfd
, xport
->stp_port_no
, xport
->config
, xport
->state
,
600 xport
->is_tunnel
, xport
->may_enable
);
603 struct xport
*peer
= xport_lookup(new_xcfg
, xport
->peer
->ofport
);
605 new_xport
->peer
= peer
;
606 new_xport
->peer
->peer
= new_xport
;
611 new_xport
->xbundle
= xbundle
;
612 list_insert(&new_xport
->xbundle
->xports
, &new_xport
->bundle_node
);
615 HMAP_FOR_EACH (pdscp
, hmap_node
, &xport
->skb_priorities
) {
616 new_pdscp
= xmalloc(sizeof *pdscp
);
617 new_pdscp
->skb_priority
= pdscp
->skb_priority
;
618 new_pdscp
->dscp
= pdscp
->dscp
;
619 hmap_insert(&new_xport
->skb_priorities
, &new_pdscp
->hmap_node
,
620 hash_int(new_pdscp
->skb_priority
, 0));
624 /* Sets the current xlate configuration to new_xcfg and frees the old xlate
625 * configuration in xcfgp.
627 * This needs to be called after editing the xlate configuration.
629 * Functions that edit the new xlate configuration are
630 * xlate_<ofport/bundle/ofport>_set and xlate_<ofport/bundle/ofport>_remove.
636 * edit_xlate_configuration();
638 * xlate_txn_commit(); */
640 xlate_txn_commit(void)
642 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
644 ovsrcu_set(&xcfgp
, new_xcfg
);
645 ovsrcu_postpone(xlate_xcfg_free
, xcfg
);
650 /* Copies the current xlate configuration in xcfgp to new_xcfg.
652 * This needs to be called prior to editing the xlate configuration. */
654 xlate_txn_start(void)
656 struct xbridge
*xbridge
;
657 struct xlate_cfg
*xcfg
;
659 ovs_assert(!new_xcfg
);
661 new_xcfg
= xmalloc(sizeof *new_xcfg
);
662 hmap_init(&new_xcfg
->xbridges
);
663 hmap_init(&new_xcfg
->xbundles
);
664 hmap_init(&new_xcfg
->xports
);
666 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
671 HMAP_FOR_EACH (xbridge
, hmap_node
, &xcfg
->xbridges
) {
672 xlate_xbridge_copy(xbridge
);
678 xlate_xcfg_free(struct xlate_cfg
*xcfg
)
680 struct xbridge
*xbridge
, *next_xbridge
;
686 HMAP_FOR_EACH_SAFE (xbridge
, next_xbridge
, hmap_node
, &xcfg
->xbridges
) {
687 xlate_xbridge_remove(xcfg
, xbridge
);
690 hmap_destroy(&xcfg
->xbridges
);
691 hmap_destroy(&xcfg
->xbundles
);
692 hmap_destroy(&xcfg
->xports
);
697 xlate_ofproto_set(struct ofproto_dpif
*ofproto
, const char *name
,
698 struct dpif
*dpif
, struct rule_dpif
*miss_rule
,
699 struct rule_dpif
*no_packet_in_rule
,
700 const struct mac_learning
*ml
, struct stp
*stp
,
701 const struct mcast_snooping
*ms
,
702 const struct mbridge
*mbridge
,
703 const struct dpif_sflow
*sflow
,
704 const struct dpif_ipfix
*ipfix
,
705 const struct netflow
*netflow
, enum ofp_config_flags frag
,
706 bool forward_bpdu
, bool has_in_band
,
708 bool variable_length_userdata
,
709 size_t max_mpls_depth
)
711 struct xbridge
*xbridge
;
713 ovs_assert(new_xcfg
);
715 xbridge
= xbridge_lookup(new_xcfg
, ofproto
);
717 xbridge
= xzalloc(sizeof *xbridge
);
718 xbridge
->ofproto
= ofproto
;
720 xlate_xbridge_init(new_xcfg
, xbridge
);
724 xbridge
->name
= xstrdup(name
);
726 xlate_xbridge_set(xbridge
, dpif
, miss_rule
, no_packet_in_rule
, ml
, stp
,
727 ms
, mbridge
, sflow
, ipfix
, netflow
, frag
, forward_bpdu
,
728 has_in_band
, enable_recirc
, variable_length_userdata
,
733 xlate_xbridge_remove(struct xlate_cfg
*xcfg
, struct xbridge
*xbridge
)
735 struct xbundle
*xbundle
, *next_xbundle
;
736 struct xport
*xport
, *next_xport
;
742 HMAP_FOR_EACH_SAFE (xport
, next_xport
, ofp_node
, &xbridge
->xports
) {
743 xlate_xport_remove(xcfg
, xport
);
746 LIST_FOR_EACH_SAFE (xbundle
, next_xbundle
, list_node
, &xbridge
->xbundles
) {
747 xlate_xbundle_remove(xcfg
, xbundle
);
750 hmap_remove(&xcfg
->xbridges
, &xbridge
->hmap_node
);
751 mac_learning_unref(xbridge
->ml
);
752 mcast_snooping_unref(xbridge
->ms
);
753 mbridge_unref(xbridge
->mbridge
);
754 dpif_sflow_unref(xbridge
->sflow
);
755 dpif_ipfix_unref(xbridge
->ipfix
);
756 stp_unref(xbridge
->stp
);
757 hmap_destroy(&xbridge
->xports
);
763 xlate_remove_ofproto(struct ofproto_dpif
*ofproto
)
765 struct xbridge
*xbridge
;
767 ovs_assert(new_xcfg
);
769 xbridge
= xbridge_lookup(new_xcfg
, ofproto
);
770 xlate_xbridge_remove(new_xcfg
, xbridge
);
774 xlate_bundle_set(struct ofproto_dpif
*ofproto
, struct ofbundle
*ofbundle
,
775 const char *name
, enum port_vlan_mode vlan_mode
, int vlan
,
776 unsigned long *trunks
, bool use_priority_tags
,
777 const struct bond
*bond
, const struct lacp
*lacp
,
780 struct xbundle
*xbundle
;
782 ovs_assert(new_xcfg
);
784 xbundle
= xbundle_lookup(new_xcfg
, ofbundle
);
786 xbundle
= xzalloc(sizeof *xbundle
);
787 xbundle
->ofbundle
= ofbundle
;
788 xbundle
->xbridge
= xbridge_lookup(new_xcfg
, ofproto
);
790 xlate_xbundle_init(new_xcfg
, xbundle
);
794 xbundle
->name
= xstrdup(name
);
796 xlate_xbundle_set(xbundle
, vlan_mode
, vlan
, trunks
,
797 use_priority_tags
, bond
, lacp
, floodable
);
801 xlate_xbundle_remove(struct xlate_cfg
*xcfg
, struct xbundle
*xbundle
)
803 struct xport
*xport
, *next
;
809 LIST_FOR_EACH_SAFE (xport
, next
, bundle_node
, &xbundle
->xports
) {
810 list_remove(&xport
->bundle_node
);
811 xport
->xbundle
= NULL
;
814 hmap_remove(&xcfg
->xbundles
, &xbundle
->hmap_node
);
815 list_remove(&xbundle
->list_node
);
816 bond_unref(xbundle
->bond
);
817 lacp_unref(xbundle
->lacp
);
823 xlate_bundle_remove(struct ofbundle
*ofbundle
)
825 struct xbundle
*xbundle
;
827 ovs_assert(new_xcfg
);
829 xbundle
= xbundle_lookup(new_xcfg
, ofbundle
);
830 xlate_xbundle_remove(new_xcfg
, xbundle
);
834 xlate_ofport_set(struct ofproto_dpif
*ofproto
, struct ofbundle
*ofbundle
,
835 struct ofport_dpif
*ofport
, ofp_port_t ofp_port
,
836 odp_port_t odp_port
, const struct netdev
*netdev
,
837 const struct cfm
*cfm
, const struct bfd
*bfd
,
838 struct ofport_dpif
*peer
, int stp_port_no
,
839 const struct ofproto_port_queue
*qdscp_list
, size_t n_qdscp
,
840 enum ofputil_port_config config
,
841 enum ofputil_port_state state
, bool is_tunnel
,
847 ovs_assert(new_xcfg
);
849 xport
= xport_lookup(new_xcfg
, ofport
);
851 xport
= xzalloc(sizeof *xport
);
852 xport
->ofport
= ofport
;
853 xport
->xbridge
= xbridge_lookup(new_xcfg
, ofproto
);
854 xport
->ofp_port
= ofp_port
;
856 xlate_xport_init(new_xcfg
, xport
);
859 ovs_assert(xport
->ofp_port
== ofp_port
);
861 xlate_xport_set(xport
, odp_port
, netdev
, cfm
, bfd
, stp_port_no
, config
,
862 state
, is_tunnel
, may_enable
);
865 xport
->peer
->peer
= NULL
;
867 xport
->peer
= xport_lookup(new_xcfg
, peer
);
869 xport
->peer
->peer
= xport
;
872 if (xport
->xbundle
) {
873 list_remove(&xport
->bundle_node
);
875 xport
->xbundle
= xbundle_lookup(new_xcfg
, ofbundle
);
876 if (xport
->xbundle
) {
877 list_insert(&xport
->xbundle
->xports
, &xport
->bundle_node
);
880 clear_skb_priorities(xport
);
881 for (i
= 0; i
< n_qdscp
; i
++) {
882 struct skb_priority_to_dscp
*pdscp
;
883 uint32_t skb_priority
;
885 if (dpif_queue_to_priority(xport
->xbridge
->dpif
, qdscp_list
[i
].queue
,
890 pdscp
= xmalloc(sizeof *pdscp
);
891 pdscp
->skb_priority
= skb_priority
;
892 pdscp
->dscp
= (qdscp_list
[i
].dscp
<< 2) & IP_DSCP_MASK
;
893 hmap_insert(&xport
->skb_priorities
, &pdscp
->hmap_node
,
894 hash_int(pdscp
->skb_priority
, 0));
899 xlate_xport_remove(struct xlate_cfg
*xcfg
, struct xport
*xport
)
906 xport
->peer
->peer
= NULL
;
910 if (xport
->xbundle
) {
911 list_remove(&xport
->bundle_node
);
914 clear_skb_priorities(xport
);
915 hmap_destroy(&xport
->skb_priorities
);
917 hmap_remove(&xcfg
->xports
, &xport
->hmap_node
);
918 hmap_remove(&xport
->xbridge
->xports
, &xport
->ofp_node
);
920 netdev_close(xport
->netdev
);
921 cfm_unref(xport
->cfm
);
922 bfd_unref(xport
->bfd
);
927 xlate_ofport_remove(struct ofport_dpif
*ofport
)
931 ovs_assert(new_xcfg
);
933 xport
= xport_lookup(new_xcfg
, ofport
);
934 xlate_xport_remove(new_xcfg
, xport
);
937 /* Given a datpath, packet, and flow metadata ('backer', 'packet', and 'key'
938 * respectively), populates 'flow' with the result of odp_flow_key_to_flow().
939 * Optionally populates 'ofproto' with the ofproto_dpif, 'odp_in_port' with
940 * the datapath in_port, that 'packet' ingressed, and 'ipfix', 'sflow', and
941 * 'netflow' with the appropriate handles for those protocols if they're
942 * enabled. Caller is responsible for unrefing them.
944 * If 'ofproto' is nonnull, requires 'flow''s in_port to exist. Otherwise sets
945 * 'flow''s in_port to OFPP_NONE.
947 * This function does post-processing on data returned from
948 * odp_flow_key_to_flow() to help make VLAN splinters transparent to the rest
949 * of the upcall processing logic. In particular, if the extracted in_port is
950 * a VLAN splinter port, it replaces flow->in_port by the "real" port, sets
951 * flow->vlan_tci correctly for the VLAN of the VLAN splinter port, and pushes
952 * a VLAN header onto 'packet' (if it is nonnull).
954 * Similarly, this function also includes some logic to help with tunnels. It
955 * may modify 'flow' as necessary to make the tunneling implementation
956 * transparent to the upcall processing logic.
958 * Returns 0 if successful, ENODEV if the parsed flow has no associated ofport,
959 * or some other positive errno if there are other problems. */
961 xlate_receive(const struct dpif_backer
*backer
, struct ofpbuf
*packet
,
962 const struct nlattr
*key
, size_t key_len
, struct flow
*flow
,
963 struct ofproto_dpif
**ofproto
, struct dpif_ipfix
**ipfix
,
964 struct dpif_sflow
**sflow
, struct netflow
**netflow
,
965 odp_port_t
*odp_in_port
)
967 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
969 const struct xport
*xport
;
971 if (odp_flow_key_to_flow(key
, key_len
, flow
) == ODP_FIT_ERROR
) {
977 *odp_in_port
= flow
->in_port
.odp_port
;
980 xport
= xport_lookup(xcfg
, tnl_port_should_receive(flow
)
981 ? tnl_port_receive(flow
)
982 : odp_port_to_ofport(backer
, flow
->in_port
.odp_port
));
984 flow
->in_port
.ofp_port
= xport
? xport
->ofp_port
: OFPP_NONE
;
989 if (vsp_adjust_flow(xport
->xbridge
->ofproto
, flow
)) {
991 /* Make the packet resemble the flow, so that it gets sent to
992 * an OpenFlow controller properly, so that it looks correct
993 * for sFlow, and so that flow_extract() will get the correct
994 * vlan_tci if it is called on 'packet'. */
995 eth_push_vlan(packet
, htons(ETH_TYPE_VLAN
), flow
->vlan_tci
);
1001 *ofproto
= xport
->xbridge
->ofproto
;
1005 *ipfix
= dpif_ipfix_ref(xport
->xbridge
->ipfix
);
1009 *sflow
= dpif_sflow_ref(xport
->xbridge
->sflow
);
1013 *netflow
= netflow_ref(xport
->xbridge
->netflow
);
1018 static struct xbridge
*
1019 xbridge_lookup(struct xlate_cfg
*xcfg
, const struct ofproto_dpif
*ofproto
)
1021 struct hmap
*xbridges
;
1022 struct xbridge
*xbridge
;
1024 if (!ofproto
|| !xcfg
) {
1028 xbridges
= &xcfg
->xbridges
;
1030 HMAP_FOR_EACH_IN_BUCKET (xbridge
, hmap_node
, hash_pointer(ofproto
, 0),
1032 if (xbridge
->ofproto
== ofproto
) {
1039 static struct xbundle
*
1040 xbundle_lookup(struct xlate_cfg
*xcfg
, const struct ofbundle
*ofbundle
)
1042 struct hmap
*xbundles
;
1043 struct xbundle
*xbundle
;
1045 if (!ofbundle
|| !xcfg
) {
1049 xbundles
= &xcfg
->xbundles
;
1051 HMAP_FOR_EACH_IN_BUCKET (xbundle
, hmap_node
, hash_pointer(ofbundle
, 0),
1053 if (xbundle
->ofbundle
== ofbundle
) {
1060 static struct xport
*
1061 xport_lookup(struct xlate_cfg
*xcfg
, const struct ofport_dpif
*ofport
)
1063 struct hmap
*xports
;
1064 struct xport
*xport
;
1066 if (!ofport
|| !xcfg
) {
1070 xports
= &xcfg
->xports
;
1072 HMAP_FOR_EACH_IN_BUCKET (xport
, hmap_node
, hash_pointer(ofport
, 0),
1074 if (xport
->ofport
== ofport
) {
1081 static struct stp_port
*
1082 xport_get_stp_port(const struct xport
*xport
)
1084 return xport
->xbridge
->stp
&& xport
->stp_port_no
!= -1
1085 ? stp_get_port(xport
->xbridge
->stp
, xport
->stp_port_no
)
1090 xport_stp_learn_state(const struct xport
*xport
)
1092 struct stp_port
*sp
= xport_get_stp_port(xport
);
1093 return stp_learn_in_state(sp
? stp_port_get_state(sp
) : STP_DISABLED
);
1097 xport_stp_forward_state(const struct xport
*xport
)
1099 struct stp_port
*sp
= xport_get_stp_port(xport
);
1100 return stp_forward_in_state(sp
? stp_port_get_state(sp
) : STP_DISABLED
);
1104 xport_stp_should_forward_bpdu(const struct xport
*xport
)
1106 struct stp_port
*sp
= xport_get_stp_port(xport
);
1107 return stp_should_forward_bpdu(sp
? stp_port_get_state(sp
) : STP_DISABLED
);
1110 /* Returns true if STP should process 'flow'. Sets fields in 'wc' that
1111 * were used to make the determination.*/
1113 stp_should_process_flow(const struct flow
*flow
, struct flow_wildcards
*wc
)
1115 /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */
1116 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
1117 return is_stp(flow
);
1121 stp_process_packet(const struct xport
*xport
, const struct ofpbuf
*packet
)
1123 struct stp_port
*sp
= xport_get_stp_port(xport
);
1124 struct ofpbuf payload
= *packet
;
1125 struct eth_header
*eth
= ofpbuf_data(&payload
);
1127 /* Sink packets on ports that have STP disabled when the bridge has
1129 if (!sp
|| stp_port_get_state(sp
) == STP_DISABLED
) {
1133 /* Trim off padding on payload. */
1134 if (ofpbuf_size(&payload
) > ntohs(eth
->eth_type
) + ETH_HEADER_LEN
) {
1135 ofpbuf_set_size(&payload
, ntohs(eth
->eth_type
) + ETH_HEADER_LEN
);
1138 if (ofpbuf_try_pull(&payload
, ETH_HEADER_LEN
+ LLC_HEADER_LEN
)) {
1139 stp_received_bpdu(sp
, ofpbuf_data(&payload
), ofpbuf_size(&payload
));
1143 static struct xport
*
1144 get_ofp_port(const struct xbridge
*xbridge
, ofp_port_t ofp_port
)
1146 struct xport
*xport
;
1148 HMAP_FOR_EACH_IN_BUCKET (xport
, ofp_node
, hash_ofp_port(ofp_port
),
1150 if (xport
->ofp_port
== ofp_port
) {
1158 ofp_port_to_odp_port(const struct xbridge
*xbridge
, ofp_port_t ofp_port
)
1160 const struct xport
*xport
= get_ofp_port(xbridge
, ofp_port
);
1161 return xport
? xport
->odp_port
: ODPP_NONE
;
1165 odp_port_is_alive(const struct xlate_ctx
*ctx
, ofp_port_t ofp_port
)
1167 struct xport
*xport
;
1169 xport
= get_ofp_port(ctx
->xbridge
, ofp_port
);
1170 if (!xport
|| xport
->config
& OFPUTIL_PC_PORT_DOWN
||
1171 xport
->state
& OFPUTIL_PS_LINK_DOWN
) {
1178 static struct ofputil_bucket
*
1179 group_first_live_bucket(const struct xlate_ctx
*, const struct group_dpif
*,
1183 group_is_alive(const struct xlate_ctx
*ctx
, uint32_t group_id
, int depth
)
1185 struct group_dpif
*group
;
1187 if (group_dpif_lookup(ctx
->xbridge
->ofproto
, group_id
, &group
)) {
1188 struct ofputil_bucket
*bucket
;
1190 bucket
= group_first_live_bucket(ctx
, group
, depth
);
1191 group_dpif_unref(group
);
1192 return bucket
== NULL
;
1198 #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */
1201 bucket_is_alive(const struct xlate_ctx
*ctx
,
1202 struct ofputil_bucket
*bucket
, int depth
)
1204 if (depth
>= MAX_LIVENESS_RECURSION
) {
1205 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
1207 VLOG_WARN_RL(&rl
, "bucket chaining exceeded %d links",
1208 MAX_LIVENESS_RECURSION
);
1212 return (!ofputil_bucket_has_liveness(bucket
)
1213 || (bucket
->watch_port
!= OFPP_ANY
1214 && odp_port_is_alive(ctx
, bucket
->watch_port
))
1215 || (bucket
->watch_group
!= OFPG_ANY
1216 && group_is_alive(ctx
, bucket
->watch_group
, depth
+ 1)));
1219 static struct ofputil_bucket
*
1220 group_first_live_bucket(const struct xlate_ctx
*ctx
,
1221 const struct group_dpif
*group
, int depth
)
1223 struct ofputil_bucket
*bucket
;
1224 const struct list
*buckets
;
1226 group_dpif_get_buckets(group
, &buckets
);
1227 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
1228 if (bucket_is_alive(ctx
, bucket
, depth
)) {
1236 static struct ofputil_bucket
*
1237 group_best_live_bucket(const struct xlate_ctx
*ctx
,
1238 const struct group_dpif
*group
,
1241 struct ofputil_bucket
*best_bucket
= NULL
;
1242 uint32_t best_score
= 0;
1245 struct ofputil_bucket
*bucket
;
1246 const struct list
*buckets
;
1248 group_dpif_get_buckets(group
, &buckets
);
1249 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
1250 if (bucket_is_alive(ctx
, bucket
, 0)) {
1251 uint32_t score
= (hash_int(i
, basis
) & 0xffff) * bucket
->weight
;
1252 if (score
>= best_score
) {
1253 best_bucket
= bucket
;
1264 xbundle_trunks_vlan(const struct xbundle
*bundle
, uint16_t vlan
)
1266 return (bundle
->vlan_mode
!= PORT_VLAN_ACCESS
1267 && (!bundle
->trunks
|| bitmap_is_set(bundle
->trunks
, vlan
)));
1271 xbundle_includes_vlan(const struct xbundle
*xbundle
, uint16_t vlan
)
1273 return vlan
== xbundle
->vlan
|| xbundle_trunks_vlan(xbundle
, vlan
);
1276 static mirror_mask_t
1277 xbundle_mirror_out(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
1279 return xbundle
!= &ofpp_none_bundle
1280 ? mirror_bundle_out(xbridge
->mbridge
, xbundle
->ofbundle
)
1284 static mirror_mask_t
1285 xbundle_mirror_src(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
1287 return xbundle
!= &ofpp_none_bundle
1288 ? mirror_bundle_src(xbridge
->mbridge
, xbundle
->ofbundle
)
1292 static mirror_mask_t
1293 xbundle_mirror_dst(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
1295 return xbundle
!= &ofpp_none_bundle
1296 ? mirror_bundle_dst(xbridge
->mbridge
, xbundle
->ofbundle
)
1300 static struct xbundle
*
1301 lookup_input_bundle(const struct xbridge
*xbridge
, ofp_port_t in_port
,
1302 bool warn
, struct xport
**in_xportp
)
1304 struct xport
*xport
;
1306 /* Find the port and bundle for the received packet. */
1307 xport
= get_ofp_port(xbridge
, in_port
);
1311 if (xport
&& xport
->xbundle
) {
1312 return xport
->xbundle
;
1315 /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+),
1316 * which a controller may use as the ingress port for traffic that
1317 * it is sourcing. */
1318 if (in_port
== OFPP_CONTROLLER
|| in_port
== OFPP_NONE
) {
1319 return &ofpp_none_bundle
;
1322 /* Odd. A few possible reasons here:
1324 * - We deleted a port but there are still a few packets queued up
1327 * - Someone externally added a port (e.g. "ovs-dpctl add-if") that
1328 * we don't know about.
1330 * - The ofproto client didn't configure the port as part of a bundle.
1331 * This is particularly likely to happen if a packet was received on the
1332 * port after it was created, but before the client had a chance to
1333 * configure its bundle.
1336 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1338 VLOG_WARN_RL(&rl
, "bridge %s: received packet on unknown "
1339 "port %"PRIu16
, xbridge
->name
, in_port
);
1345 add_mirror_actions(struct xlate_ctx
*ctx
, const struct flow
*orig_flow
)
1347 const struct xbridge
*xbridge
= ctx
->xbridge
;
1348 mirror_mask_t mirrors
;
1349 struct xbundle
*in_xbundle
;
1353 mirrors
= ctx
->xout
->mirrors
;
1354 ctx
->xout
->mirrors
= 0;
1356 in_xbundle
= lookup_input_bundle(xbridge
, orig_flow
->in_port
.ofp_port
,
1357 ctx
->xin
->packet
!= NULL
, NULL
);
1361 mirrors
|= xbundle_mirror_src(xbridge
, in_xbundle
);
1363 /* Drop frames on bundles reserved for mirroring. */
1364 if (xbundle_mirror_out(xbridge
, in_xbundle
)) {
1365 if (ctx
->xin
->packet
!= NULL
) {
1366 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1367 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet received on port "
1368 "%s, which is reserved exclusively for mirroring",
1369 ctx
->xbridge
->name
, in_xbundle
->name
);
1371 ofpbuf_clear(&ctx
->xout
->odp_actions
);
1376 vid
= vlan_tci_to_vid(orig_flow
->vlan_tci
);
1377 if (!input_vid_is_valid(vid
, in_xbundle
, ctx
->xin
->packet
!= NULL
)) {
1380 vlan
= input_vid_to_vlan(in_xbundle
, vid
);
1386 /* Restore the original packet before adding the mirror actions. */
1387 ctx
->xin
->flow
= *orig_flow
;
1390 mirror_mask_t dup_mirrors
;
1391 struct ofbundle
*out
;
1392 unsigned long *vlans
;
1397 has_mirror
= mirror_get(xbridge
->mbridge
, raw_ctz(mirrors
),
1398 &vlans
, &dup_mirrors
, &out
, &out_vlan
);
1399 ovs_assert(has_mirror
);
1402 ctx
->xout
->wc
.masks
.vlan_tci
|= htons(VLAN_CFI
| VLAN_VID_MASK
);
1404 vlan_mirrored
= !vlans
|| bitmap_is_set(vlans
, vlan
);
1407 if (!vlan_mirrored
) {
1408 mirrors
= zero_rightmost_1bit(mirrors
);
1412 mirrors
&= ~dup_mirrors
;
1413 ctx
->xout
->mirrors
|= dup_mirrors
;
1415 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1416 struct xbundle
*out_xbundle
= xbundle_lookup(xcfg
, out
);
1418 output_normal(ctx
, out_xbundle
, vlan
);
1420 } else if (vlan
!= out_vlan
1421 && !eth_addr_is_reserved(orig_flow
->dl_dst
)) {
1422 struct xbundle
*xbundle
;
1424 LIST_FOR_EACH (xbundle
, list_node
, &xbridge
->xbundles
) {
1425 if (xbundle_includes_vlan(xbundle
, out_vlan
)
1426 && !xbundle_mirror_out(xbridge
, xbundle
)) {
1427 output_normal(ctx
, xbundle
, out_vlan
);
1434 /* Given 'vid', the VID obtained from the 802.1Q header that was received as
1435 * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle',
1436 * the bundle on which the packet was received, returns the VLAN to which the
1439 * Both 'vid' and the return value are in the range 0...4095. */
1441 input_vid_to_vlan(const struct xbundle
*in_xbundle
, uint16_t vid
)
1443 switch (in_xbundle
->vlan_mode
) {
1444 case PORT_VLAN_ACCESS
:
1445 return in_xbundle
->vlan
;
1448 case PORT_VLAN_TRUNK
:
1451 case PORT_VLAN_NATIVE_UNTAGGED
:
1452 case PORT_VLAN_NATIVE_TAGGED
:
1453 return vid
? vid
: in_xbundle
->vlan
;
1460 /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'.
1461 * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs
1464 * 'vid' should be the VID obtained from the 802.1Q header that was received as
1465 * part of a packet (specify 0 if there was no 802.1Q header), in the range
1468 input_vid_is_valid(uint16_t vid
, struct xbundle
*in_xbundle
, bool warn
)
1470 /* Allow any VID on the OFPP_NONE port. */
1471 if (in_xbundle
== &ofpp_none_bundle
) {
1475 switch (in_xbundle
->vlan_mode
) {
1476 case PORT_VLAN_ACCESS
:
1479 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1480 VLOG_WARN_RL(&rl
, "dropping VLAN %"PRIu16
" tagged "
1481 "packet received on port %s configured as VLAN "
1482 "%"PRIu16
" access port", vid
, in_xbundle
->name
,
1489 case PORT_VLAN_NATIVE_UNTAGGED
:
1490 case PORT_VLAN_NATIVE_TAGGED
:
1492 /* Port must always carry its native VLAN. */
1496 case PORT_VLAN_TRUNK
:
1497 if (!xbundle_includes_vlan(in_xbundle
, vid
)) {
1499 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1500 VLOG_WARN_RL(&rl
, "dropping VLAN %"PRIu16
" packet "
1501 "received on port %s not configured for trunking "
1502 "VLAN %"PRIu16
, vid
, in_xbundle
->name
, vid
);
1514 /* Given 'vlan', the VLAN that a packet belongs to, and
1515 * 'out_xbundle', a bundle on which the packet is to be output, returns the VID
1516 * that should be included in the 802.1Q header. (If the return value is 0,
1517 * then the 802.1Q header should only be included in the packet if there is a
1520 * Both 'vlan' and the return value are in the range 0...4095. */
1522 output_vlan_to_vid(const struct xbundle
*out_xbundle
, uint16_t vlan
)
1524 switch (out_xbundle
->vlan_mode
) {
1525 case PORT_VLAN_ACCESS
:
1528 case PORT_VLAN_TRUNK
:
1529 case PORT_VLAN_NATIVE_TAGGED
:
1532 case PORT_VLAN_NATIVE_UNTAGGED
:
1533 return vlan
== out_xbundle
->vlan
? 0 : vlan
;
1541 output_normal(struct xlate_ctx
*ctx
, const struct xbundle
*out_xbundle
,
1544 ovs_be16
*flow_tci
= &ctx
->xin
->flow
.vlan_tci
;
1546 ovs_be16 tci
, old_tci
;
1547 struct xport
*xport
;
1549 vid
= output_vlan_to_vid(out_xbundle
, vlan
);
1550 if (list_is_empty(&out_xbundle
->xports
)) {
1551 /* Partially configured bundle with no slaves. Drop the packet. */
1553 } else if (!out_xbundle
->bond
) {
1554 ctx
->use_recirc
= false;
1555 xport
= CONTAINER_OF(list_front(&out_xbundle
->xports
), struct xport
,
1558 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1559 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
1560 struct xlate_recirc
*xr
= &ctx
->recirc
;
1561 struct ofport_dpif
*ofport
;
1563 if (ctx
->xbridge
->enable_recirc
) {
1564 ctx
->use_recirc
= bond_may_recirc(
1565 out_xbundle
->bond
, &xr
->recirc_id
, &xr
->hash_basis
);
1567 if (ctx
->use_recirc
) {
1568 /* Only TCP mode uses recirculation. */
1569 xr
->hash_alg
= OVS_HASH_ALG_L4
;
1570 bond_update_post_recirc_rules(out_xbundle
->bond
, false);
1572 /* Recirculation does not require unmasking hash fields. */
1577 ofport
= bond_choose_output_slave(out_xbundle
->bond
,
1578 &ctx
->xin
->flow
, wc
, vid
);
1579 xport
= xport_lookup(xcfg
, ofport
);
1582 /* No slaves enabled, so drop packet. */
1586 /* If ctx->xout->use_recirc is set, the main thread will handle stats
1587 * accounting for this bond. */
1588 if (!ctx
->use_recirc
) {
1589 if (ctx
->xin
->resubmit_stats
) {
1590 bond_account(out_xbundle
->bond
, &ctx
->xin
->flow
, vid
,
1591 ctx
->xin
->resubmit_stats
->n_bytes
);
1593 if (ctx
->xin
->xcache
) {
1594 struct xc_entry
*entry
;
1597 flow
= &ctx
->xin
->flow
;
1598 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_BOND
);
1599 entry
->u
.bond
.bond
= bond_ref(out_xbundle
->bond
);
1600 entry
->u
.bond
.flow
= xmemdup(flow
, sizeof *flow
);
1601 entry
->u
.bond
.vid
= vid
;
1606 old_tci
= *flow_tci
;
1608 if (tci
|| out_xbundle
->use_priority_tags
) {
1609 tci
|= *flow_tci
& htons(VLAN_PCP_MASK
);
1611 tci
|= htons(VLAN_CFI
);
1616 compose_output_action(ctx
, xport
->ofp_port
);
1617 *flow_tci
= old_tci
;
1620 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
1621 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
1622 * indicate this; newer upstream kernels use gratuitous ARP requests. */
1624 is_gratuitous_arp(const struct flow
*flow
, struct flow_wildcards
*wc
)
1626 if (flow
->dl_type
!= htons(ETH_TYPE_ARP
)) {
1630 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
1631 if (!eth_addr_is_broadcast(flow
->dl_dst
)) {
1635 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
1636 if (flow
->nw_proto
== ARP_OP_REPLY
) {
1638 } else if (flow
->nw_proto
== ARP_OP_REQUEST
) {
1639 memset(&wc
->masks
.nw_src
, 0xff, sizeof wc
->masks
.nw_src
);
1640 memset(&wc
->masks
.nw_dst
, 0xff, sizeof wc
->masks
.nw_dst
);
1642 return flow
->nw_src
== flow
->nw_dst
;
1648 /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or
1649 * dropped. Returns true if they may be forwarded, false if they should be
1652 * 'in_port' must be the xport that corresponds to flow->in_port.
1653 * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull).
1655 * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as
1656 * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as
1657 * checked by input_vid_is_valid().
1659 * May also add tags to '*tags', although the current implementation only does
1660 * so in one special case.
1663 is_admissible(struct xlate_ctx
*ctx
, struct xport
*in_port
,
1666 struct xbundle
*in_xbundle
= in_port
->xbundle
;
1667 const struct xbridge
*xbridge
= ctx
->xbridge
;
1668 struct flow
*flow
= &ctx
->xin
->flow
;
1670 /* Drop frames for reserved multicast addresses
1671 * only if forward_bpdu option is absent. */
1672 if (!xbridge
->forward_bpdu
&& eth_addr_is_reserved(flow
->dl_dst
)) {
1673 xlate_report(ctx
, "packet has reserved destination MAC, dropping");
1677 if (in_xbundle
->bond
) {
1678 struct mac_entry
*mac
;
1680 switch (bond_check_admissibility(in_xbundle
->bond
, in_port
->ofport
,
1686 xlate_report(ctx
, "bonding refused admissibility, dropping");
1689 case BV_DROP_IF_MOVED
:
1690 ovs_rwlock_rdlock(&xbridge
->ml
->rwlock
);
1691 mac
= mac_learning_lookup(xbridge
->ml
, flow
->dl_src
, vlan
);
1692 if (mac
&& mac
->port
.p
!= in_xbundle
->ofbundle
&&
1693 (!is_gratuitous_arp(flow
, &ctx
->xout
->wc
)
1694 || mac_entry_is_grat_arp_locked(mac
))) {
1695 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
1696 xlate_report(ctx
, "SLB bond thinks this packet looped back, "
1700 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
1708 /* Checks whether a MAC learning update is necessary for MAC learning table
1709 * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in
1712 * Most packets processed through the MAC learning table do not actually
1713 * change it in any way. This function requires only a read lock on the MAC
1714 * learning table, so it is much cheaper in this common case.
1716 * Keep the code here synchronized with that in update_learning_table__()
1719 is_mac_learning_update_needed(const struct mac_learning
*ml
,
1720 const struct flow
*flow
,
1721 struct flow_wildcards
*wc
,
1722 int vlan
, struct xbundle
*in_xbundle
)
1723 OVS_REQ_RDLOCK(ml
->rwlock
)
1725 struct mac_entry
*mac
;
1727 if (!mac_learning_may_learn(ml
, flow
->dl_src
, vlan
)) {
1731 mac
= mac_learning_lookup(ml
, flow
->dl_src
, vlan
);
1732 if (!mac
|| mac_entry_age(ml
, mac
)) {
1736 if (is_gratuitous_arp(flow
, wc
)) {
1737 /* We don't want to learn from gratuitous ARP packets that are
1738 * reflected back over bond slaves so we lock the learning table. */
1739 if (!in_xbundle
->bond
) {
1741 } else if (mac_entry_is_grat_arp_locked(mac
)) {
1746 return mac
->port
.p
!= in_xbundle
->ofbundle
;
1750 /* Updates MAC learning table 'ml' given that a packet matching 'flow' was
1751 * received on 'in_xbundle' in 'vlan'.
1753 * This code repeats all the checks in is_mac_learning_update_needed() because
1754 * the lock was released between there and here and thus the MAC learning state
1755 * could have changed.
1757 * Keep the code here synchronized with that in is_mac_learning_update_needed()
1760 update_learning_table__(const struct xbridge
*xbridge
,
1761 const struct flow
*flow
, struct flow_wildcards
*wc
,
1762 int vlan
, struct xbundle
*in_xbundle
)
1763 OVS_REQ_WRLOCK(xbridge
->ml
->rwlock
)
1765 struct mac_entry
*mac
;
1767 if (!mac_learning_may_learn(xbridge
->ml
, flow
->dl_src
, vlan
)) {
1771 mac
= mac_learning_insert(xbridge
->ml
, flow
->dl_src
, vlan
);
1772 if (is_gratuitous_arp(flow
, wc
)) {
1773 /* We don't want to learn from gratuitous ARP packets that are
1774 * reflected back over bond slaves so we lock the learning table. */
1775 if (!in_xbundle
->bond
) {
1776 mac_entry_set_grat_arp_lock(mac
);
1777 } else if (mac_entry_is_grat_arp_locked(mac
)) {
1782 if (mac
->port
.p
!= in_xbundle
->ofbundle
) {
1783 /* The log messages here could actually be useful in debugging,
1784 * so keep the rate limit relatively high. */
1785 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(30, 300);
1787 VLOG_DBG_RL(&rl
, "bridge %s: learned that "ETH_ADDR_FMT
" is "
1788 "on port %s in VLAN %d",
1789 xbridge
->name
, ETH_ADDR_ARGS(flow
->dl_src
),
1790 in_xbundle
->name
, vlan
);
1792 mac
->port
.p
= in_xbundle
->ofbundle
;
1793 mac_learning_changed(xbridge
->ml
);
1798 update_learning_table(const struct xbridge
*xbridge
,
1799 const struct flow
*flow
, struct flow_wildcards
*wc
,
1800 int vlan
, struct xbundle
*in_xbundle
)
1804 /* Don't learn the OFPP_NONE port. */
1805 if (in_xbundle
== &ofpp_none_bundle
) {
1809 /* First try the common case: no change to MAC learning table. */
1810 ovs_rwlock_rdlock(&xbridge
->ml
->rwlock
);
1811 need_update
= is_mac_learning_update_needed(xbridge
->ml
, flow
, wc
, vlan
,
1813 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
1816 /* Slow path: MAC learning table might need an update. */
1817 ovs_rwlock_wrlock(&xbridge
->ml
->rwlock
);
1818 update_learning_table__(xbridge
, flow
, wc
, vlan
, in_xbundle
);
1819 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
1823 /* Updates multicast snooping table 'ms' given that a packet matching 'flow'
1824 * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */
1826 update_mcast_snooping_table__(const struct xbridge
*xbridge
,
1827 const struct flow
*flow
,
1828 struct mcast_snooping
*ms
,
1829 ovs_be32 ip4
, int vlan
,
1830 struct xbundle
*in_xbundle
)
1831 OVS_REQ_WRLOCK(ms
->rwlock
)
1833 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(60, 30);
1835 switch (ntohs(flow
->tp_src
)) {
1836 case IGMP_HOST_MEMBERSHIP_REPORT
:
1837 case IGMPV2_HOST_MEMBERSHIP_REPORT
:
1838 if (mcast_snooping_add_group(ms
, ip4
, vlan
, in_xbundle
->ofbundle
)) {
1839 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping learned that "
1840 IP_FMT
" is on port %s in VLAN %d",
1841 xbridge
->name
, IP_ARGS(ip4
), in_xbundle
->name
, vlan
);
1844 case IGMP_HOST_LEAVE_MESSAGE
:
1845 if (mcast_snooping_leave_group(ms
, ip4
, vlan
, in_xbundle
->ofbundle
)) {
1846 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping leaving "
1847 IP_FMT
" is on port %s in VLAN %d",
1848 xbridge
->name
, IP_ARGS(ip4
), in_xbundle
->name
, vlan
);
1851 case IGMP_HOST_MEMBERSHIP_QUERY
:
1852 if (flow
->nw_src
&& mcast_snooping_add_mrouter(ms
, vlan
,
1853 in_xbundle
->ofbundle
)) {
1854 VLOG_DBG_RL(&rl
, "bridge %s: multicast snooping query from "
1855 IP_FMT
" is on port %s in VLAN %d",
1856 xbridge
->name
, IP_ARGS(flow
->nw_src
),
1857 in_xbundle
->name
, vlan
);
1863 /* Updates multicast snooping table 'ms' given that a packet matching 'flow'
1864 * was received on 'in_xbundle' in 'vlan'. */
1866 update_mcast_snooping_table(const struct xbridge
*xbridge
,
1867 const struct flow
*flow
, int vlan
,
1868 struct xbundle
*in_xbundle
)
1870 struct mcast_snooping
*ms
= xbridge
->ms
;
1871 struct xlate_cfg
*xcfg
;
1872 struct xbundle
*mcast_xbundle
;
1873 struct mcast_fport_bundle
*fport
;
1875 /* Don't learn the OFPP_NONE port. */
1876 if (in_xbundle
== &ofpp_none_bundle
) {
1880 /* Don't learn from flood ports */
1881 mcast_xbundle
= NULL
;
1882 ovs_rwlock_wrlock(&ms
->rwlock
);
1883 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1884 LIST_FOR_EACH(fport
, fport_node
, &ms
->fport_list
) {
1885 mcast_xbundle
= xbundle_lookup(xcfg
, fport
->port
);
1886 if (mcast_xbundle
== in_xbundle
) {
1891 if (!mcast_xbundle
|| mcast_xbundle
!= in_xbundle
) {
1892 update_mcast_snooping_table__(xbridge
, flow
, ms
, flow
->igmp_group_ip4
,
1895 ovs_rwlock_unlock(&ms
->rwlock
);
1898 /* send the packet to ports having the multicast group learned */
1900 xlate_normal_mcast_send_group(struct xlate_ctx
*ctx
,
1901 struct mcast_snooping
*ms OVS_UNUSED
,
1902 struct mcast_group
*grp
,
1903 struct xbundle
*in_xbundle
, uint16_t vlan
)
1904 OVS_REQ_RDLOCK(ms
->rwlock
)
1906 struct xlate_cfg
*xcfg
;
1907 struct mcast_group_bundle
*b
;
1908 struct xbundle
*mcast_xbundle
;
1910 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1911 LIST_FOR_EACH(b
, bundle_node
, &grp
->bundle_lru
) {
1912 mcast_xbundle
= xbundle_lookup(xcfg
, b
->port
);
1913 if (mcast_xbundle
&& mcast_xbundle
!= in_xbundle
) {
1914 xlate_report(ctx
, "forwarding to mcast group port");
1915 output_normal(ctx
, mcast_xbundle
, vlan
);
1916 } else if (!mcast_xbundle
) {
1917 xlate_report(ctx
, "mcast group port is unknown, dropping");
1919 xlate_report(ctx
, "mcast group port is input port, dropping");
1924 /* send the packet to ports connected to multicast routers */
1926 xlate_normal_mcast_send_mrouters(struct xlate_ctx
*ctx
,
1927 struct mcast_snooping
*ms
,
1928 struct xbundle
*in_xbundle
, uint16_t vlan
)
1929 OVS_REQ_RDLOCK(ms
->rwlock
)
1931 struct xlate_cfg
*xcfg
;
1932 struct mcast_mrouter_bundle
*mrouter
;
1933 struct xbundle
*mcast_xbundle
;
1935 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1936 LIST_FOR_EACH(mrouter
, mrouter_node
, &ms
->mrouter_lru
) {
1937 mcast_xbundle
= xbundle_lookup(xcfg
, mrouter
->port
);
1938 if (mcast_xbundle
&& mcast_xbundle
!= in_xbundle
) {
1939 xlate_report(ctx
, "forwarding to mcast router port");
1940 output_normal(ctx
, mcast_xbundle
, vlan
);
1941 } else if (!mcast_xbundle
) {
1942 xlate_report(ctx
, "mcast router port is unknown, dropping");
1944 xlate_report(ctx
, "mcast router port is input port, dropping");
1949 /* send the packet to ports flagged to be flooded */
1951 xlate_normal_mcast_send_fports(struct xlate_ctx
*ctx
,
1952 struct mcast_snooping
*ms
,
1953 struct xbundle
*in_xbundle
, uint16_t vlan
)
1954 OVS_REQ_RDLOCK(ms
->rwlock
)
1956 struct xlate_cfg
*xcfg
;
1957 struct mcast_fport_bundle
*fport
;
1958 struct xbundle
*mcast_xbundle
;
1960 xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
1961 LIST_FOR_EACH(fport
, fport_node
, &ms
->fport_list
) {
1962 mcast_xbundle
= xbundle_lookup(xcfg
, fport
->port
);
1963 if (mcast_xbundle
&& mcast_xbundle
!= in_xbundle
) {
1964 xlate_report(ctx
, "forwarding to mcast flood port");
1965 output_normal(ctx
, mcast_xbundle
, vlan
);
1966 } else if (!mcast_xbundle
) {
1967 xlate_report(ctx
, "mcast flood port is unknown, dropping");
1969 xlate_report(ctx
, "mcast flood port is input port, dropping");
1975 xlate_normal_flood(struct xlate_ctx
*ctx
, struct xbundle
*in_xbundle
,
1978 struct xbundle
*xbundle
;
1980 LIST_FOR_EACH (xbundle
, list_node
, &ctx
->xbridge
->xbundles
) {
1981 if (xbundle
!= in_xbundle
1982 && xbundle_includes_vlan(xbundle
, vlan
)
1983 && xbundle
->floodable
1984 && !xbundle_mirror_out(ctx
->xbridge
, xbundle
)) {
1985 output_normal(ctx
, xbundle
, vlan
);
1988 ctx
->xout
->nf_output_iface
= NF_OUT_FLOOD
;
1992 xlate_normal(struct xlate_ctx
*ctx
)
1994 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
1995 struct flow
*flow
= &ctx
->xin
->flow
;
1996 struct xbundle
*in_xbundle
;
1997 struct xport
*in_port
;
1998 struct mac_entry
*mac
;
2003 ctx
->xout
->has_normal
= true;
2005 memset(&wc
->masks
.dl_src
, 0xff, sizeof wc
->masks
.dl_src
);
2006 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
2007 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
2009 in_xbundle
= lookup_input_bundle(ctx
->xbridge
, flow
->in_port
.ofp_port
,
2010 ctx
->xin
->packet
!= NULL
, &in_port
);
2012 xlate_report(ctx
, "no input bundle, dropping");
2016 /* Drop malformed frames. */
2017 if (flow
->dl_type
== htons(ETH_TYPE_VLAN
) &&
2018 !(flow
->vlan_tci
& htons(VLAN_CFI
))) {
2019 if (ctx
->xin
->packet
!= NULL
) {
2020 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2021 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet with partial "
2022 "VLAN tag received on port %s",
2023 ctx
->xbridge
->name
, in_xbundle
->name
);
2025 xlate_report(ctx
, "partial VLAN tag, dropping");
2029 /* Drop frames on bundles reserved for mirroring. */
2030 if (xbundle_mirror_out(ctx
->xbridge
, in_xbundle
)) {
2031 if (ctx
->xin
->packet
!= NULL
) {
2032 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2033 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet received on port "
2034 "%s, which is reserved exclusively for mirroring",
2035 ctx
->xbridge
->name
, in_xbundle
->name
);
2037 xlate_report(ctx
, "input port is mirror output port, dropping");
2042 vid
= vlan_tci_to_vid(flow
->vlan_tci
);
2043 if (!input_vid_is_valid(vid
, in_xbundle
, ctx
->xin
->packet
!= NULL
)) {
2044 xlate_report(ctx
, "disallowed VLAN VID for this input port, dropping");
2047 vlan
= input_vid_to_vlan(in_xbundle
, vid
);
2049 /* Check other admissibility requirements. */
2050 if (in_port
&& !is_admissible(ctx
, in_port
, vlan
)) {
2054 /* Learn source MAC. */
2055 if (ctx
->xin
->may_learn
) {
2056 update_learning_table(ctx
->xbridge
, flow
, wc
, vlan
, in_xbundle
);
2058 if (ctx
->xin
->xcache
) {
2059 struct xc_entry
*entry
;
2061 /* Save enough info to update mac learning table later. */
2062 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_NORMAL
);
2063 entry
->u
.normal
.ofproto
= ctx
->xbridge
->ofproto
;
2064 entry
->u
.normal
.flow
= xmemdup(flow
, sizeof *flow
);
2065 entry
->u
.normal
.vlan
= vlan
;
2068 /* Determine output bundle. */
2069 if (mcast_snooping_enabled(ctx
->xbridge
->ms
)
2070 && !eth_addr_is_broadcast(flow
->dl_dst
)
2071 && eth_addr_is_multicast(flow
->dl_dst
)
2072 && flow
->dl_type
== htons(ETH_TYPE_IP
)) {
2073 struct mcast_snooping
*ms
= ctx
->xbridge
->ms
;
2074 struct mcast_group
*grp
;
2076 if (flow
->nw_proto
== IPPROTO_IGMP
) {
2077 if (ctx
->xin
->may_learn
) {
2078 if (mcast_snooping_is_membership(flow
->tp_src
) ||
2079 mcast_snooping_is_query(flow
->tp_src
)) {
2080 update_mcast_snooping_table(ctx
->xbridge
, flow
, vlan
,
2085 if (mcast_snooping_is_membership(flow
->tp_src
)) {
2086 ovs_rwlock_rdlock(&ms
->rwlock
);
2087 xlate_normal_mcast_send_mrouters(ctx
, ms
, in_xbundle
, vlan
);
2088 ovs_rwlock_unlock(&ms
->rwlock
);
2090 xlate_report(ctx
, "multicast traffic, flooding");
2091 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2095 if (ip_is_local_multicast(flow
->nw_dst
)) {
2096 /* RFC4541: section 2.1.2, item 2: Packets with a dst IP
2097 * address in the 224.0.0.x range which are not IGMP must
2098 * be forwarded on all ports */
2099 xlate_report(ctx
, "RFC4541: section 2.1.2, item 2, flooding");
2100 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2105 /* forwarding to group base ports */
2106 ovs_rwlock_rdlock(&ms
->rwlock
);
2107 grp
= mcast_snooping_lookup(ms
, flow
->nw_dst
, vlan
);
2109 xlate_normal_mcast_send_group(ctx
, ms
, grp
, in_xbundle
, vlan
);
2110 xlate_normal_mcast_send_fports(ctx
, ms
, in_xbundle
, vlan
);
2111 xlate_normal_mcast_send_mrouters(ctx
, ms
, in_xbundle
, vlan
);
2113 if (mcast_snooping_flood_unreg(ms
)) {
2114 xlate_report(ctx
, "unregistered multicast, flooding");
2115 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2117 xlate_normal_mcast_send_mrouters(ctx
, ms
, in_xbundle
, vlan
);
2118 xlate_normal_mcast_send_fports(ctx
, ms
, in_xbundle
, vlan
);
2121 ovs_rwlock_unlock(&ms
->rwlock
);
2123 ovs_rwlock_rdlock(&ctx
->xbridge
->ml
->rwlock
);
2124 mac
= mac_learning_lookup(ctx
->xbridge
->ml
, flow
->dl_dst
, vlan
);
2125 mac_port
= mac
? mac
->port
.p
: NULL
;
2126 ovs_rwlock_unlock(&ctx
->xbridge
->ml
->rwlock
);
2129 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
2130 struct xbundle
*mac_xbundle
= xbundle_lookup(xcfg
, mac_port
);
2131 if (mac_xbundle
&& mac_xbundle
!= in_xbundle
) {
2132 xlate_report(ctx
, "forwarding to learned port");
2133 output_normal(ctx
, mac_xbundle
, vlan
);
2134 } else if (!mac_xbundle
) {
2135 xlate_report(ctx
, "learned port is unknown, dropping");
2137 xlate_report(ctx
, "learned port is input port, dropping");
2140 xlate_report(ctx
, "no learned MAC for destination, flooding");
2141 xlate_normal_flood(ctx
, in_xbundle
, vlan
);
2146 /* Compose SAMPLE action for sFlow or IPFIX. The given probability is
2147 * the number of packets out of UINT32_MAX to sample. The given
2148 * cookie is passed back in the callback for each sampled packet.
2151 compose_sample_action(const struct xbridge
*xbridge
,
2152 struct ofpbuf
*odp_actions
,
2153 const struct flow
*flow
,
2154 const uint32_t probability
,
2155 const union user_action_cookie
*cookie
,
2156 const size_t cookie_size
)
2158 size_t sample_offset
, actions_offset
;
2159 odp_port_t odp_port
;
2163 sample_offset
= nl_msg_start_nested(odp_actions
, OVS_ACTION_ATTR_SAMPLE
);
2165 nl_msg_put_u32(odp_actions
, OVS_SAMPLE_ATTR_PROBABILITY
, probability
);
2167 actions_offset
= nl_msg_start_nested(odp_actions
, OVS_SAMPLE_ATTR_ACTIONS
);
2169 odp_port
= ofp_port_to_odp_port(xbridge
, flow
->in_port
.ofp_port
);
2170 pid
= dpif_port_get_pid(xbridge
->dpif
, odp_port
,
2171 flow_hash_5tuple(flow
, 0));
2172 cookie_offset
= odp_put_userspace_action(pid
, cookie
, cookie_size
,
2175 nl_msg_end_nested(odp_actions
, actions_offset
);
2176 nl_msg_end_nested(odp_actions
, sample_offset
);
2177 return cookie_offset
;
2181 compose_sflow_cookie(const struct xbridge
*xbridge
, ovs_be16 vlan_tci
,
2182 odp_port_t odp_port
, unsigned int n_outputs
,
2183 union user_action_cookie
*cookie
)
2187 cookie
->type
= USER_ACTION_COOKIE_SFLOW
;
2188 cookie
->sflow
.vlan_tci
= vlan_tci
;
2190 /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output
2191 * port information") for the interpretation of cookie->output. */
2192 switch (n_outputs
) {
2194 /* 0x40000000 | 256 means "packet dropped for unknown reason". */
2195 cookie
->sflow
.output
= 0x40000000 | 256;
2199 ifindex
= dpif_sflow_odp_port_to_ifindex(xbridge
->sflow
, odp_port
);
2201 cookie
->sflow
.output
= ifindex
;
2206 /* 0x80000000 means "multiple output ports. */
2207 cookie
->sflow
.output
= 0x80000000 | n_outputs
;
2212 /* Compose SAMPLE action for sFlow bridge sampling. */
2214 compose_sflow_action(const struct xbridge
*xbridge
,
2215 struct ofpbuf
*odp_actions
,
2216 const struct flow
*flow
,
2217 odp_port_t odp_port
)
2219 uint32_t probability
;
2220 union user_action_cookie cookie
;
2222 if (!xbridge
->sflow
|| flow
->in_port
.ofp_port
== OFPP_NONE
) {
2226 probability
= dpif_sflow_get_probability(xbridge
->sflow
);
2227 compose_sflow_cookie(xbridge
, htons(0), odp_port
,
2228 odp_port
== ODPP_NONE
? 0 : 1, &cookie
);
2230 return compose_sample_action(xbridge
, odp_actions
, flow
, probability
,
2231 &cookie
, sizeof cookie
.sflow
);
2235 compose_flow_sample_cookie(uint16_t probability
, uint32_t collector_set_id
,
2236 uint32_t obs_domain_id
, uint32_t obs_point_id
,
2237 union user_action_cookie
*cookie
)
2239 cookie
->type
= USER_ACTION_COOKIE_FLOW_SAMPLE
;
2240 cookie
->flow_sample
.probability
= probability
;
2241 cookie
->flow_sample
.collector_set_id
= collector_set_id
;
2242 cookie
->flow_sample
.obs_domain_id
= obs_domain_id
;
2243 cookie
->flow_sample
.obs_point_id
= obs_point_id
;
2247 compose_ipfix_cookie(union user_action_cookie
*cookie
)
2249 cookie
->type
= USER_ACTION_COOKIE_IPFIX
;
2252 /* Compose SAMPLE action for IPFIX bridge sampling. */
2254 compose_ipfix_action(const struct xbridge
*xbridge
,
2255 struct ofpbuf
*odp_actions
,
2256 const struct flow
*flow
)
2258 uint32_t probability
;
2259 union user_action_cookie cookie
;
2261 if (!xbridge
->ipfix
|| flow
->in_port
.ofp_port
== OFPP_NONE
) {
2265 probability
= dpif_ipfix_get_bridge_exporter_probability(xbridge
->ipfix
);
2266 compose_ipfix_cookie(&cookie
);
2268 compose_sample_action(xbridge
, odp_actions
, flow
, probability
,
2269 &cookie
, sizeof cookie
.ipfix
);
2272 /* SAMPLE action for sFlow must be first action in any given list of
2273 * actions. At this point we do not have all information required to
2274 * build it. So try to build sample action as complete as possible. */
2276 add_sflow_action(struct xlate_ctx
*ctx
)
2278 ctx
->user_cookie_offset
= compose_sflow_action(ctx
->xbridge
,
2279 &ctx
->xout
->odp_actions
,
2280 &ctx
->xin
->flow
, ODPP_NONE
);
2281 ctx
->sflow_odp_port
= 0;
2282 ctx
->sflow_n_outputs
= 0;
2285 /* SAMPLE action for IPFIX must be 1st or 2nd action in any given list
2286 * of actions, eventually after the SAMPLE action for sFlow. */
2288 add_ipfix_action(struct xlate_ctx
*ctx
)
2290 compose_ipfix_action(ctx
->xbridge
, &ctx
->xout
->odp_actions
,
2294 /* Fix SAMPLE action according to data collected while composing ODP actions.
2295 * We need to fix SAMPLE actions OVS_SAMPLE_ATTR_ACTIONS attribute, i.e. nested
2296 * USERSPACE action's user-cookie which is required for sflow. */
2298 fix_sflow_action(struct xlate_ctx
*ctx
)
2300 const struct flow
*base
= &ctx
->base_flow
;
2301 union user_action_cookie
*cookie
;
2303 if (!ctx
->user_cookie_offset
) {
2307 cookie
= ofpbuf_at(&ctx
->xout
->odp_actions
, ctx
->user_cookie_offset
,
2308 sizeof cookie
->sflow
);
2309 ovs_assert(cookie
->type
== USER_ACTION_COOKIE_SFLOW
);
2311 compose_sflow_cookie(ctx
->xbridge
, base
->vlan_tci
,
2312 ctx
->sflow_odp_port
, ctx
->sflow_n_outputs
, cookie
);
2315 static enum slow_path_reason
2316 process_special(struct xlate_ctx
*ctx
, const struct flow
*flow
,
2317 const struct xport
*xport
, const struct ofpbuf
*packet
)
2319 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
2320 const struct xbridge
*xbridge
= ctx
->xbridge
;
2324 } else if (xport
->cfm
&& cfm_should_process_flow(xport
->cfm
, flow
, wc
)) {
2326 cfm_process_heartbeat(xport
->cfm
, packet
);
2329 } else if (xport
->bfd
&& bfd_should_process_flow(xport
->bfd
, flow
, wc
)) {
2331 bfd_process_packet(xport
->bfd
, flow
, packet
);
2332 /* If POLL received, immediately sends FINAL back. */
2333 if (bfd_should_send_packet(xport
->bfd
)) {
2334 ofproto_dpif_monitor_port_send_soon(xport
->ofport
);
2338 } else if (xport
->xbundle
&& xport
->xbundle
->lacp
2339 && flow
->dl_type
== htons(ETH_TYPE_LACP
)) {
2341 lacp_process_packet(xport
->xbundle
->lacp
, xport
->ofport
, packet
);
2344 } else if (xbridge
->stp
&& stp_should_process_flow(flow
, wc
)) {
2346 stp_process_packet(xport
, packet
);
2355 compose_output_action__(struct xlate_ctx
*ctx
, ofp_port_t ofp_port
,
2358 const struct xport
*xport
= get_ofp_port(ctx
->xbridge
, ofp_port
);
2359 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
2360 struct flow
*flow
= &ctx
->xin
->flow
;
2361 ovs_be16 flow_vlan_tci
;
2362 uint32_t flow_pkt_mark
;
2363 uint8_t flow_nw_tos
;
2364 odp_port_t out_port
, odp_port
;
2367 /* If 'struct flow' gets additional metadata, we'll need to zero it out
2368 * before traversing a patch port. */
2369 BUILD_ASSERT_DECL(FLOW_WC_SEQ
== 27);
2372 xlate_report(ctx
, "Nonexistent output port");
2374 } else if (xport
->config
& OFPUTIL_PC_NO_FWD
) {
2375 xlate_report(ctx
, "OFPPC_NO_FWD set, skipping output");
2377 } else if (check_stp
) {
2378 if (is_stp(&ctx
->base_flow
)) {
2379 if (!xport_stp_should_forward_bpdu(xport
)) {
2380 xlate_report(ctx
, "STP not in listening state, "
2381 "skipping bpdu output");
2384 } else if (!xport_stp_forward_state(xport
)) {
2385 xlate_report(ctx
, "STP not in forwarding state, "
2391 if (mbridge_has_mirrors(ctx
->xbridge
->mbridge
) && xport
->xbundle
) {
2392 ctx
->xout
->mirrors
|= xbundle_mirror_dst(xport
->xbundle
->xbridge
,
2397 const struct xport
*peer
= xport
->peer
;
2398 struct flow old_flow
= ctx
->xin
->flow
;
2399 enum slow_path_reason special
;
2401 ctx
->xbridge
= peer
->xbridge
;
2402 flow
->in_port
.ofp_port
= peer
->ofp_port
;
2403 flow
->metadata
= htonll(0);
2404 memset(&flow
->tunnel
, 0, sizeof flow
->tunnel
);
2405 memset(flow
->regs
, 0, sizeof flow
->regs
);
2407 special
= process_special(ctx
, &ctx
->xin
->flow
, peer
,
2410 ctx
->xout
->slow
|= special
;
2411 } else if (may_receive(peer
, ctx
)) {
2412 if (xport_stp_forward_state(peer
)) {
2413 xlate_table_action(ctx
, flow
->in_port
.ofp_port
, 0, true, true);
2415 /* Forwarding is disabled by STP. Let OFPP_NORMAL and the
2416 * learning action look at the packet, then drop it. */
2417 struct flow old_base_flow
= ctx
->base_flow
;
2418 size_t old_size
= ofpbuf_size(&ctx
->xout
->odp_actions
);
2419 mirror_mask_t old_mirrors
= ctx
->xout
->mirrors
;
2420 xlate_table_action(ctx
, flow
->in_port
.ofp_port
, 0, true, true);
2421 ctx
->xout
->mirrors
= old_mirrors
;
2422 ctx
->base_flow
= old_base_flow
;
2423 ofpbuf_set_size(&ctx
->xout
->odp_actions
, old_size
);
2427 ctx
->xin
->flow
= old_flow
;
2428 ctx
->xbridge
= xport
->xbridge
;
2430 if (ctx
->xin
->resubmit_stats
) {
2431 netdev_vport_inc_tx(xport
->netdev
, ctx
->xin
->resubmit_stats
);
2432 netdev_vport_inc_rx(peer
->netdev
, ctx
->xin
->resubmit_stats
);
2434 bfd_account_rx(peer
->bfd
, ctx
->xin
->resubmit_stats
);
2437 if (ctx
->xin
->xcache
) {
2438 struct xc_entry
*entry
;
2440 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_NETDEV
);
2441 entry
->u
.dev
.tx
= netdev_ref(xport
->netdev
);
2442 entry
->u
.dev
.rx
= netdev_ref(peer
->netdev
);
2443 entry
->u
.dev
.bfd
= bfd_ref(peer
->bfd
);
2448 flow_vlan_tci
= flow
->vlan_tci
;
2449 flow_pkt_mark
= flow
->pkt_mark
;
2450 flow_nw_tos
= flow
->nw_tos
;
2452 if (dscp_from_skb_priority(xport
, flow
->skb_priority
, &dscp
)) {
2453 wc
->masks
.nw_tos
|= IP_DSCP_MASK
;
2454 flow
->nw_tos
&= ~IP_DSCP_MASK
;
2455 flow
->nw_tos
|= dscp
;
2458 if (xport
->is_tunnel
) {
2459 /* Save tunnel metadata so that changes made due to
2460 * the Logical (tunnel) Port are not visible for any further
2461 * matches, while explicit set actions on tunnel metadata are.
2463 struct flow_tnl flow_tnl
= flow
->tunnel
;
2464 odp_port
= tnl_port_send(xport
->ofport
, flow
, &ctx
->xout
->wc
);
2465 if (odp_port
== ODPP_NONE
) {
2466 xlate_report(ctx
, "Tunneling decided against output");
2467 goto out
; /* restore flow_nw_tos */
2469 if (flow
->tunnel
.ip_dst
== ctx
->orig_tunnel_ip_dst
) {
2470 xlate_report(ctx
, "Not tunneling to our own address");
2471 goto out
; /* restore flow_nw_tos */
2473 if (ctx
->xin
->resubmit_stats
) {
2474 netdev_vport_inc_tx(xport
->netdev
, ctx
->xin
->resubmit_stats
);
2476 if (ctx
->xin
->xcache
) {
2477 struct xc_entry
*entry
;
2479 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_NETDEV
);
2480 entry
->u
.dev
.tx
= netdev_ref(xport
->netdev
);
2482 out_port
= odp_port
;
2483 commit_odp_tunnel_action(flow
, &ctx
->base_flow
,
2484 &ctx
->xout
->odp_actions
);
2485 flow
->tunnel
= flow_tnl
; /* Restore tunnel metadata */
2487 odp_port
= xport
->odp_port
;
2488 out_port
= odp_port
;
2489 if (ofproto_has_vlan_splinters(ctx
->xbridge
->ofproto
)) {
2490 ofp_port_t vlandev_port
;
2492 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
2493 vlandev_port
= vsp_realdev_to_vlandev(ctx
->xbridge
->ofproto
,
2494 ofp_port
, flow
->vlan_tci
);
2495 if (vlandev_port
!= ofp_port
) {
2496 out_port
= ofp_port_to_odp_port(ctx
->xbridge
, vlandev_port
);
2497 flow
->vlan_tci
= htons(0);
2502 if (out_port
!= ODPP_NONE
) {
2503 ctx
->xout
->slow
|= commit_odp_actions(flow
, &ctx
->base_flow
,
2504 &ctx
->xout
->odp_actions
,
2507 if (ctx
->use_recirc
) {
2508 struct ovs_action_hash
*act_hash
;
2509 struct xlate_recirc
*xr
= &ctx
->recirc
;
2512 act_hash
= nl_msg_put_unspec_uninit(&ctx
->xout
->odp_actions
,
2513 OVS_ACTION_ATTR_HASH
,
2515 act_hash
->hash_alg
= xr
->hash_alg
;
2516 act_hash
->hash_basis
= xr
->hash_basis
;
2518 /* Recirc action. */
2519 nl_msg_put_u32(&ctx
->xout
->odp_actions
, OVS_ACTION_ATTR_RECIRC
,
2522 nl_msg_put_odp_port(&ctx
->xout
->odp_actions
, OVS_ACTION_ATTR_OUTPUT
,
2526 ctx
->sflow_odp_port
= odp_port
;
2527 ctx
->sflow_n_outputs
++;
2528 ctx
->xout
->nf_output_iface
= ofp_port
;
2533 flow
->vlan_tci
= flow_vlan_tci
;
2534 flow
->pkt_mark
= flow_pkt_mark
;
2535 flow
->nw_tos
= flow_nw_tos
;
2539 compose_output_action(struct xlate_ctx
*ctx
, ofp_port_t ofp_port
)
2541 compose_output_action__(ctx
, ofp_port
, true);
2545 xlate_recursively(struct xlate_ctx
*ctx
, struct rule_dpif
*rule
)
2547 struct rule_dpif
*old_rule
= ctx
->rule
;
2548 const struct rule_actions
*actions
;
2550 if (ctx
->xin
->resubmit_stats
) {
2551 rule_dpif_credit_stats(rule
, ctx
->xin
->resubmit_stats
);
2557 actions
= rule_dpif_get_actions(rule
);
2558 do_xlate_actions(actions
->ofpacts
, actions
->ofpacts_len
, ctx
);
2559 ctx
->rule
= old_rule
;
2564 xlate_resubmit_resource_check(struct xlate_ctx
*ctx
)
2566 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2568 if (ctx
->recurse
>= MAX_RESUBMIT_RECURSION
+ MAX_INTERNAL_RESUBMITS
) {
2569 VLOG_ERR_RL(&rl
, "resubmit actions recursed over %d times",
2570 MAX_RESUBMIT_RECURSION
);
2571 } else if (ctx
->resubmits
>= MAX_RESUBMITS
+ MAX_INTERNAL_RESUBMITS
) {
2572 VLOG_ERR_RL(&rl
, "over %d resubmit actions", MAX_RESUBMITS
);
2573 } else if (ofpbuf_size(&ctx
->xout
->odp_actions
) > UINT16_MAX
) {
2574 VLOG_ERR_RL(&rl
, "resubmits yielded over 64 kB of actions");
2575 } else if (ofpbuf_size(&ctx
->stack
) >= 65536) {
2576 VLOG_ERR_RL(&rl
, "resubmits yielded over 64 kB of stack");
2585 xlate_table_action(struct xlate_ctx
*ctx
, ofp_port_t in_port
, uint8_t table_id
,
2586 bool may_packet_in
, bool honor_table_miss
)
2588 if (xlate_resubmit_resource_check(ctx
)) {
2589 ofp_port_t old_in_port
= ctx
->xin
->flow
.in_port
.ofp_port
;
2590 bool skip_wildcards
= ctx
->xin
->skip_wildcards
;
2591 uint8_t old_table_id
= ctx
->table_id
;
2592 struct rule_dpif
*rule
;
2593 enum rule_dpif_lookup_verdict verdict
;
2594 enum ofputil_port_config config
= 0;
2596 ctx
->table_id
= table_id
;
2598 /* Look up a flow with 'in_port' as the input port. Then restore the
2599 * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
2600 * have surprising behavior). */
2601 ctx
->xin
->flow
.in_port
.ofp_port
= in_port
;
2602 verdict
= rule_dpif_lookup_from_table(ctx
->xbridge
->ofproto
,
2605 ? &ctx
->xout
->wc
: NULL
,
2607 &ctx
->table_id
, &rule
,
2608 ctx
->xin
->xcache
!= NULL
,
2609 ctx
->xin
->resubmit_stats
);
2610 ctx
->xin
->flow
.in_port
.ofp_port
= old_in_port
;
2612 if (ctx
->xin
->resubmit_hook
) {
2613 ctx
->xin
->resubmit_hook(ctx
->xin
, rule
, ctx
->recurse
);
2617 case RULE_DPIF_LOOKUP_VERDICT_MATCH
:
2619 case RULE_DPIF_LOOKUP_VERDICT_CONTROLLER
:
2620 if (may_packet_in
) {
2621 struct xport
*xport
;
2623 xport
= get_ofp_port(ctx
->xbridge
,
2624 ctx
->xin
->flow
.in_port
.ofp_port
);
2625 config
= xport
? xport
->config
: 0;
2628 /* Fall through to drop */
2629 case RULE_DPIF_LOOKUP_VERDICT_DROP
:
2630 config
= OFPUTIL_PC_NO_PACKET_IN
;
2632 case RULE_DPIF_LOOKUP_VERDICT_DEFAULT
:
2633 if (!ofproto_dpif_wants_packet_in_on_miss(ctx
->xbridge
->ofproto
)) {
2634 config
= OFPUTIL_PC_NO_PACKET_IN
;
2641 choose_miss_rule(config
, ctx
->xbridge
->miss_rule
,
2642 ctx
->xbridge
->no_packet_in_rule
, &rule
,
2643 ctx
->xin
->xcache
!= NULL
);
2647 /* Fill in the cache entry here instead of xlate_recursively
2648 * to make the reference counting more explicit. We take a
2649 * reference in the lookups above if we are going to cache the
2651 if (ctx
->xin
->xcache
) {
2652 struct xc_entry
*entry
;
2654 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_RULE
);
2655 entry
->u
.rule
= rule
;
2657 xlate_recursively(ctx
, rule
);
2660 ctx
->table_id
= old_table_id
;
2668 xlate_group_stats(struct xlate_ctx
*ctx
, struct group_dpif
*group
,
2669 struct ofputil_bucket
*bucket
)
2671 if (ctx
->xin
->resubmit_stats
) {
2672 group_dpif_credit_stats(group
, bucket
, ctx
->xin
->resubmit_stats
);
2674 if (ctx
->xin
->xcache
) {
2675 struct xc_entry
*entry
;
2677 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_GROUP
);
2678 entry
->u
.group
.group
= group_dpif_ref(group
);
2679 entry
->u
.group
.bucket
= bucket
;
2684 xlate_group_bucket(struct xlate_ctx
*ctx
, struct ofputil_bucket
*bucket
)
2686 uint64_t action_list_stub
[1024 / 8];
2687 struct ofpbuf action_list
, action_set
;
2689 ofpbuf_use_const(&action_set
, bucket
->ofpacts
, bucket
->ofpacts_len
);
2690 ofpbuf_use_stub(&action_list
, action_list_stub
, sizeof action_list_stub
);
2692 ofpacts_execute_action_set(&action_list
, &action_set
);
2694 do_xlate_actions(ofpbuf_data(&action_list
), ofpbuf_size(&action_list
), ctx
);
2697 ofpbuf_uninit(&action_set
);
2698 ofpbuf_uninit(&action_list
);
2702 xlate_all_group(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
2704 struct ofputil_bucket
*bucket
;
2705 const struct list
*buckets
;
2706 struct flow old_flow
= ctx
->xin
->flow
;
2708 group_dpif_get_buckets(group
, &buckets
);
2710 LIST_FOR_EACH (bucket
, list_node
, buckets
) {
2711 xlate_group_bucket(ctx
, bucket
);
2712 /* Roll back flow to previous state.
2713 * This is equivalent to cloning the packet for each bucket.
2715 * As a side effect any subsequently applied actions will
2716 * also effectively be applied to a clone of the packet taken
2717 * just before applying the all or indirect group. */
2718 ctx
->xin
->flow
= old_flow
;
2720 xlate_group_stats(ctx
, group
, NULL
);
2724 xlate_ff_group(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
2726 struct ofputil_bucket
*bucket
;
2728 bucket
= group_first_live_bucket(ctx
, group
, 0);
2730 xlate_group_bucket(ctx
, bucket
);
2731 xlate_group_stats(ctx
, group
, bucket
);
2736 xlate_select_group(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
2738 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
2739 struct ofputil_bucket
*bucket
;
2742 basis
= hash_mac(ctx
->xin
->flow
.dl_dst
, 0, 0);
2743 bucket
= group_best_live_bucket(ctx
, group
, basis
);
2745 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
2746 xlate_group_bucket(ctx
, bucket
);
2747 xlate_group_stats(ctx
, group
, bucket
);
2752 xlate_group_action__(struct xlate_ctx
*ctx
, struct group_dpif
*group
)
2754 ctx
->in_group
= true;
2756 switch (group_dpif_get_type(group
)) {
2758 case OFPGT11_INDIRECT
:
2759 xlate_all_group(ctx
, group
);
2761 case OFPGT11_SELECT
:
2762 xlate_select_group(ctx
, group
);
2765 xlate_ff_group(ctx
, group
);
2770 group_dpif_unref(group
);
2772 ctx
->in_group
= false;
2776 xlate_group_resource_check(struct xlate_ctx
*ctx
)
2778 if (!xlate_resubmit_resource_check(ctx
)) {
2780 } else if (ctx
->in_group
) {
2781 /* Prevent nested translation of OpenFlow groups.
2783 * OpenFlow allows this restriction. We enforce this restriction only
2784 * because, with the current architecture, we would otherwise have to
2785 * take a possibly recursive read lock on the ofgroup rwlock, which is
2786 * unsafe given that POSIX allows taking a read lock to block if there
2787 * is a thread blocked on taking the write lock. Other solutions
2788 * without this restriction are also possible, but seem unwarranted
2789 * given the current limited use of groups. */
2790 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2792 VLOG_ERR_RL(&rl
, "cannot recursively translate OpenFlow group");
2800 xlate_group_action(struct xlate_ctx
*ctx
, uint32_t group_id
)
2802 if (xlate_group_resource_check(ctx
)) {
2803 struct group_dpif
*group
;
2806 got_group
= group_dpif_lookup(ctx
->xbridge
->ofproto
, group_id
, &group
);
2808 xlate_group_action__(ctx
, group
);
2818 xlate_ofpact_resubmit(struct xlate_ctx
*ctx
,
2819 const struct ofpact_resubmit
*resubmit
)
2823 bool may_packet_in
= false;
2824 bool honor_table_miss
= false;
2826 if (ctx
->rule
&& rule_dpif_is_internal(ctx
->rule
)) {
2827 /* Still allow missed packets to be sent to the controller
2828 * if resubmitting from an internal table. */
2829 may_packet_in
= true;
2830 honor_table_miss
= true;
2833 in_port
= resubmit
->in_port
;
2834 if (in_port
== OFPP_IN_PORT
) {
2835 in_port
= ctx
->xin
->flow
.in_port
.ofp_port
;
2838 table_id
= resubmit
->table_id
;
2839 if (table_id
== 255) {
2840 table_id
= ctx
->table_id
;
2843 xlate_table_action(ctx
, in_port
, table_id
, may_packet_in
,
2848 flood_packets(struct xlate_ctx
*ctx
, bool all
)
2850 const struct xport
*xport
;
2852 HMAP_FOR_EACH (xport
, ofp_node
, &ctx
->xbridge
->xports
) {
2853 if (xport
->ofp_port
== ctx
->xin
->flow
.in_port
.ofp_port
) {
2858 compose_output_action__(ctx
, xport
->ofp_port
, false);
2859 } else if (!(xport
->config
& OFPUTIL_PC_NO_FLOOD
)) {
2860 compose_output_action(ctx
, xport
->ofp_port
);
2864 ctx
->xout
->nf_output_iface
= NF_OUT_FLOOD
;
2868 execute_controller_action(struct xlate_ctx
*ctx
, int len
,
2869 enum ofp_packet_in_reason reason
,
2870 uint16_t controller_id
)
2872 struct ofproto_packet_in
*pin
;
2873 struct dpif_packet
*packet
;
2874 struct pkt_metadata md
= PKT_METADATA_INITIALIZER(0);
2876 ctx
->xout
->slow
|= SLOW_CONTROLLER
;
2877 if (!ctx
->xin
->packet
) {
2881 packet
= dpif_packet_clone_from_ofpbuf(ctx
->xin
->packet
);
2883 ctx
->xout
->slow
|= commit_odp_actions(&ctx
->xin
->flow
, &ctx
->base_flow
,
2884 &ctx
->xout
->odp_actions
,
2887 odp_execute_actions(NULL
, &packet
, 1, false, &md
,
2888 ofpbuf_data(&ctx
->xout
->odp_actions
),
2889 ofpbuf_size(&ctx
->xout
->odp_actions
), NULL
);
2891 pin
= xmalloc(sizeof *pin
);
2892 pin
->up
.packet_len
= ofpbuf_size(&packet
->ofpbuf
);
2893 pin
->up
.packet
= ofpbuf_steal_data(&packet
->ofpbuf
);
2894 pin
->up
.reason
= reason
;
2895 pin
->up
.table_id
= ctx
->table_id
;
2896 pin
->up
.cookie
= (ctx
->rule
2897 ? rule_dpif_get_flow_cookie(ctx
->rule
)
2900 flow_get_metadata(&ctx
->xin
->flow
, &pin
->up
.fmd
);
2902 pin
->controller_id
= controller_id
;
2903 pin
->send_len
= len
;
2904 /* If a rule is a table-miss rule then this is
2905 * a table-miss handled by a table-miss rule.
2907 * Else, if rule is internal and has a controller action,
2908 * the later being implied by the rule being processed here,
2909 * then this is a table-miss handled without a table-miss rule.
2911 * Otherwise this is not a table-miss. */
2912 pin
->miss_type
= OFPROTO_PACKET_IN_NO_MISS
;
2914 if (rule_dpif_is_table_miss(ctx
->rule
)) {
2915 pin
->miss_type
= OFPROTO_PACKET_IN_MISS_FLOW
;
2916 } else if (rule_dpif_is_internal(ctx
->rule
)) {
2917 pin
->miss_type
= OFPROTO_PACKET_IN_MISS_WITHOUT_FLOW
;
2920 ofproto_dpif_send_packet_in(ctx
->xbridge
->ofproto
, pin
);
2921 dpif_packet_delete(packet
);
2925 compose_recirculate_action(struct xlate_ctx
*ctx
,
2926 const struct ofpact
*ofpacts_base
,
2927 const struct ofpact
*ofpact_current
,
2928 size_t ofpacts_base_len
)
2932 unsigned ofpacts_len
;
2935 struct ofpbuf ofpacts
;
2939 ofpacts_len
= ofpacts_base_len
-
2940 ((uint8_t *)ofpact_current
- (uint8_t *)ofpacts_base
);
2943 id
= rule_dpif_get_recirc_id(ctx
->rule
);
2945 /* In the case where ctx has no rule then allocate a recirc id.
2946 * The life-cycle of this recirc id is managed by associating it
2947 * with the internal rule that is created to to handle
2948 * recirculation below.
2950 * The known use-case of this is packet_out which
2951 * translates actions without a rule */
2952 id
= ofproto_dpif_alloc_recirc_id(ctx
->xbridge
->ofproto
);
2955 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2956 VLOG_ERR_RL(&rl
, "Failed to allocate recirculation id");
2960 match_init_catchall(&match
);
2961 match_set_recirc_id(&match
, id
);
2962 ofpbuf_use_const(&ofpacts
, ofpact_current
, ofpacts_len
);
2963 error
= ofproto_dpif_add_internal_flow(ctx
->xbridge
->ofproto
, &match
,
2964 RECIRC_RULE_PRIORITY
,
2965 RECIRC_TIMEOUT
, &ofpacts
, &rule
);
2967 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
2968 VLOG_ERR_RL(&rl
, "Failed to add post recirculation flow %s",
2969 match_to_string(&match
, 0));
2972 /* If ctx has no rule then associate the recirc id, which
2973 * was allocated above, with the internal rule. This allows
2974 * the recirc id to be released when the internal rule times out. */
2976 rule_set_recirc_id(rule
, id
);
2979 ctx
->xout
->slow
|= commit_odp_actions(&ctx
->xin
->flow
, &ctx
->base_flow
,
2980 &ctx
->xout
->odp_actions
,
2982 nl_msg_put_u32(&ctx
->xout
->odp_actions
, OVS_ACTION_ATTR_RECIRC
, id
);
2986 compose_mpls_push_action(struct xlate_ctx
*ctx
, struct ofpact_push_mpls
*mpls
)
2988 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
2989 struct flow
*flow
= &ctx
->xin
->flow
;
2992 ovs_assert(eth_type_mpls(mpls
->ethertype
));
2994 n
= flow_count_mpls_labels(flow
, wc
);
2996 ctx
->xout
->slow
|= commit_odp_actions(flow
, &ctx
->base_flow
,
2997 &ctx
->xout
->odp_actions
,
2999 } else if (n
>= FLOW_MAX_MPLS_LABELS
) {
3000 if (ctx
->xin
->packet
!= NULL
) {
3001 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3002 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet on which an "
3003 "MPLS push action can't be performed as it would "
3004 "have more MPLS LSEs than the %d supported.",
3005 ctx
->xbridge
->name
, FLOW_MAX_MPLS_LABELS
);
3009 } else if (n
>= ctx
->xbridge
->max_mpls_depth
) {
3010 COVERAGE_INC(xlate_actions_mpls_overflow
);
3011 ctx
->xout
->slow
|= SLOW_ACTION
;
3014 flow_push_mpls(flow
, n
, mpls
->ethertype
, wc
);
3018 compose_mpls_pop_action(struct xlate_ctx
*ctx
, ovs_be16 eth_type
)
3020 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
3021 struct flow
*flow
= &ctx
->xin
->flow
;
3022 int n
= flow_count_mpls_labels(flow
, wc
);
3024 if (flow_pop_mpls(flow
, n
, eth_type
, wc
)) {
3025 if (ctx
->xbridge
->enable_recirc
&& !eth_type_mpls(eth_type
)) {
3026 ctx
->was_mpls
= true;
3028 } else if (n
>= FLOW_MAX_MPLS_LABELS
) {
3029 if (ctx
->xin
->packet
!= NULL
) {
3030 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
3031 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet on which an "
3032 "MPLS pop action can't be performed as it has "
3033 "more MPLS LSEs than the %d supported.",
3034 ctx
->xbridge
->name
, FLOW_MAX_MPLS_LABELS
);
3037 ofpbuf_clear(&ctx
->xout
->odp_actions
);
3042 compose_dec_ttl(struct xlate_ctx
*ctx
, struct ofpact_cnt_ids
*ids
)
3044 struct flow
*flow
= &ctx
->xin
->flow
;
3046 if (!is_ip_any(flow
)) {
3050 ctx
->xout
->wc
.masks
.nw_ttl
= 0xff;
3051 if (flow
->nw_ttl
> 1) {
3057 for (i
= 0; i
< ids
->n_controllers
; i
++) {
3058 execute_controller_action(ctx
, UINT16_MAX
, OFPR_INVALID_TTL
,
3062 /* Stop processing for current table. */
3068 compose_set_mpls_label_action(struct xlate_ctx
*ctx
, ovs_be32 label
)
3070 if (eth_type_mpls(ctx
->xin
->flow
.dl_type
)) {
3071 ctx
->xout
->wc
.masks
.mpls_lse
[0] |= htonl(MPLS_LABEL_MASK
);
3072 set_mpls_lse_label(&ctx
->xin
->flow
.mpls_lse
[0], label
);
3077 compose_set_mpls_tc_action(struct xlate_ctx
*ctx
, uint8_t tc
)
3079 if (eth_type_mpls(ctx
->xin
->flow
.dl_type
)) {
3080 ctx
->xout
->wc
.masks
.mpls_lse
[0] |= htonl(MPLS_TC_MASK
);
3081 set_mpls_lse_tc(&ctx
->xin
->flow
.mpls_lse
[0], tc
);
3086 compose_set_mpls_ttl_action(struct xlate_ctx
*ctx
, uint8_t ttl
)
3088 if (eth_type_mpls(ctx
->xin
->flow
.dl_type
)) {
3089 ctx
->xout
->wc
.masks
.mpls_lse
[0] |= htonl(MPLS_TTL_MASK
);
3090 set_mpls_lse_ttl(&ctx
->xin
->flow
.mpls_lse
[0], ttl
);
3095 compose_dec_mpls_ttl_action(struct xlate_ctx
*ctx
)
3097 struct flow
*flow
= &ctx
->xin
->flow
;
3098 uint8_t ttl
= mpls_lse_to_ttl(flow
->mpls_lse
[0]);
3099 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
3101 memset(&wc
->masks
.mpls_lse
, 0xff, sizeof wc
->masks
.mpls_lse
);
3102 if (eth_type_mpls(flow
->dl_type
)) {
3105 set_mpls_lse_ttl(&flow
->mpls_lse
[0], ttl
);
3108 execute_controller_action(ctx
, UINT16_MAX
, OFPR_INVALID_TTL
, 0);
3110 /* Stop processing for current table. */
3119 xlate_output_action(struct xlate_ctx
*ctx
,
3120 ofp_port_t port
, uint16_t max_len
, bool may_packet_in
)
3122 ofp_port_t prev_nf_output_iface
= ctx
->xout
->nf_output_iface
;
3124 ctx
->xout
->nf_output_iface
= NF_OUT_DROP
;
3128 compose_output_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
);
3131 xlate_table_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
,
3132 0, may_packet_in
, true);
3138 flood_packets(ctx
, false);
3141 flood_packets(ctx
, true);
3143 case OFPP_CONTROLLER
:
3144 execute_controller_action(ctx
, max_len
, OFPR_ACTION
, 0);
3150 if (port
!= ctx
->xin
->flow
.in_port
.ofp_port
) {
3151 compose_output_action(ctx
, port
);
3153 xlate_report(ctx
, "skipping output to input port");
3158 if (prev_nf_output_iface
== NF_OUT_FLOOD
) {
3159 ctx
->xout
->nf_output_iface
= NF_OUT_FLOOD
;
3160 } else if (ctx
->xout
->nf_output_iface
== NF_OUT_DROP
) {
3161 ctx
->xout
->nf_output_iface
= prev_nf_output_iface
;
3162 } else if (prev_nf_output_iface
!= NF_OUT_DROP
&&
3163 ctx
->xout
->nf_output_iface
!= NF_OUT_FLOOD
) {
3164 ctx
->xout
->nf_output_iface
= NF_OUT_MULTI
;
3169 xlate_output_reg_action(struct xlate_ctx
*ctx
,
3170 const struct ofpact_output_reg
*or)
3172 uint64_t port
= mf_get_subfield(&or->src
, &ctx
->xin
->flow
);
3173 if (port
<= UINT16_MAX
) {
3174 union mf_subvalue value
;
3176 memset(&value
, 0xff, sizeof value
);
3177 mf_write_subfield_flow(&or->src
, &value
, &ctx
->xout
->wc
.masks
);
3178 xlate_output_action(ctx
, u16_to_ofp(port
),
3179 or->max_len
, false);
3184 xlate_enqueue_action(struct xlate_ctx
*ctx
,
3185 const struct ofpact_enqueue
*enqueue
)
3187 ofp_port_t ofp_port
= enqueue
->port
;
3188 uint32_t queue_id
= enqueue
->queue
;
3189 uint32_t flow_priority
, priority
;
3192 /* Translate queue to priority. */
3193 error
= dpif_queue_to_priority(ctx
->xbridge
->dpif
, queue_id
, &priority
);
3195 /* Fall back to ordinary output action. */
3196 xlate_output_action(ctx
, enqueue
->port
, 0, false);
3200 /* Check output port. */
3201 if (ofp_port
== OFPP_IN_PORT
) {
3202 ofp_port
= ctx
->xin
->flow
.in_port
.ofp_port
;
3203 } else if (ofp_port
== ctx
->xin
->flow
.in_port
.ofp_port
) {
3207 /* Add datapath actions. */
3208 flow_priority
= ctx
->xin
->flow
.skb_priority
;
3209 ctx
->xin
->flow
.skb_priority
= priority
;
3210 compose_output_action(ctx
, ofp_port
);
3211 ctx
->xin
->flow
.skb_priority
= flow_priority
;
3213 /* Update NetFlow output port. */
3214 if (ctx
->xout
->nf_output_iface
== NF_OUT_DROP
) {
3215 ctx
->xout
->nf_output_iface
= ofp_port
;
3216 } else if (ctx
->xout
->nf_output_iface
!= NF_OUT_FLOOD
) {
3217 ctx
->xout
->nf_output_iface
= NF_OUT_MULTI
;
3222 xlate_set_queue_action(struct xlate_ctx
*ctx
, uint32_t queue_id
)
3224 uint32_t skb_priority
;
3226 if (!dpif_queue_to_priority(ctx
->xbridge
->dpif
, queue_id
, &skb_priority
)) {
3227 ctx
->xin
->flow
.skb_priority
= skb_priority
;
3229 /* Couldn't translate queue to a priority. Nothing to do. A warning
3230 * has already been logged. */
3235 slave_enabled_cb(ofp_port_t ofp_port
, void *xbridge_
)
3237 const struct xbridge
*xbridge
= xbridge_
;
3248 case OFPP_CONTROLLER
: /* Not supported by the bundle action. */
3251 port
= get_ofp_port(xbridge
, ofp_port
);
3252 return port
? port
->may_enable
: false;
3257 xlate_bundle_action(struct xlate_ctx
*ctx
,
3258 const struct ofpact_bundle
*bundle
)
3262 port
= bundle_execute(bundle
, &ctx
->xin
->flow
, &ctx
->xout
->wc
,
3264 CONST_CAST(struct xbridge
*, ctx
->xbridge
));
3265 if (bundle
->dst
.field
) {
3266 nxm_reg_load(&bundle
->dst
, ofp_to_u16(port
), &ctx
->xin
->flow
,
3269 xlate_output_action(ctx
, port
, 0, false);
3274 xlate_learn_action__(struct xlate_ctx
*ctx
, const struct ofpact_learn
*learn
,
3275 struct ofputil_flow_mod
*fm
, struct ofpbuf
*ofpacts
)
3277 learn_execute(learn
, &ctx
->xin
->flow
, fm
, ofpacts
);
3278 if (ctx
->xin
->may_learn
) {
3279 ofproto_dpif_flow_mod(ctx
->xbridge
->ofproto
, fm
);
3284 xlate_learn_action(struct xlate_ctx
*ctx
, const struct ofpact_learn
*learn
)
3286 ctx
->xout
->has_learn
= true;
3287 learn_mask(learn
, &ctx
->xout
->wc
);
3289 if (ctx
->xin
->xcache
) {
3290 struct xc_entry
*entry
;
3292 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_LEARN
);
3293 entry
->u
.learn
.ofproto
= ctx
->xbridge
->ofproto
;
3294 entry
->u
.learn
.fm
= xmalloc(sizeof *entry
->u
.learn
.fm
);
3295 entry
->u
.learn
.ofpacts
= ofpbuf_new(64);
3296 xlate_learn_action__(ctx
, learn
, entry
->u
.learn
.fm
,
3297 entry
->u
.learn
.ofpacts
);
3298 } else if (ctx
->xin
->may_learn
) {
3299 uint64_t ofpacts_stub
[1024 / 8];
3300 struct ofputil_flow_mod fm
;
3301 struct ofpbuf ofpacts
;
3303 ofpbuf_use_stub(&ofpacts
, ofpacts_stub
, sizeof ofpacts_stub
);
3304 xlate_learn_action__(ctx
, learn
, &fm
, &ofpacts
);
3305 ofpbuf_uninit(&ofpacts
);
3310 xlate_fin_timeout__(struct rule_dpif
*rule
, uint16_t tcp_flags
,
3311 uint16_t idle_timeout
, uint16_t hard_timeout
)
3313 if (tcp_flags
& (TCP_FIN
| TCP_RST
)) {
3314 rule_dpif_reduce_timeouts(rule
, idle_timeout
, hard_timeout
);
3319 xlate_fin_timeout(struct xlate_ctx
*ctx
,
3320 const struct ofpact_fin_timeout
*oft
)
3323 xlate_fin_timeout__(ctx
->rule
, ctx
->xin
->tcp_flags
,
3324 oft
->fin_idle_timeout
, oft
->fin_hard_timeout
);
3325 if (ctx
->xin
->xcache
) {
3326 struct xc_entry
*entry
;
3328 entry
= xlate_cache_add_entry(ctx
->xin
->xcache
, XC_FIN_TIMEOUT
);
3329 /* XC_RULE already holds a reference on the rule, none is taken
3331 entry
->u
.fin
.rule
= ctx
->rule
;
3332 entry
->u
.fin
.idle
= oft
->fin_idle_timeout
;
3333 entry
->u
.fin
.hard
= oft
->fin_hard_timeout
;
3339 xlate_sample_action(struct xlate_ctx
*ctx
,
3340 const struct ofpact_sample
*os
)
3342 union user_action_cookie cookie
;
3343 /* Scale the probability from 16-bit to 32-bit while representing
3344 * the same percentage. */
3345 uint32_t probability
= (os
->probability
<< 16) | os
->probability
;
3347 if (!ctx
->xbridge
->variable_length_userdata
) {
3348 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 1);
3350 VLOG_ERR_RL(&rl
, "ignoring NXAST_SAMPLE action because datapath "
3351 "lacks support (needs Linux 3.10+ or kernel module from "
3356 ctx
->xout
->slow
|= commit_odp_actions(&ctx
->xin
->flow
, &ctx
->base_flow
,
3357 &ctx
->xout
->odp_actions
,
3360 compose_flow_sample_cookie(os
->probability
, os
->collector_set_id
,
3361 os
->obs_domain_id
, os
->obs_point_id
, &cookie
);
3362 compose_sample_action(ctx
->xbridge
, &ctx
->xout
->odp_actions
, &ctx
->xin
->flow
,
3363 probability
, &cookie
, sizeof cookie
.flow_sample
);
3367 may_receive(const struct xport
*xport
, struct xlate_ctx
*ctx
)
3369 if (xport
->config
& (is_stp(&ctx
->xin
->flow
)
3370 ? OFPUTIL_PC_NO_RECV_STP
3371 : OFPUTIL_PC_NO_RECV
)) {
3375 /* Only drop packets here if both forwarding and learning are
3376 * disabled. If just learning is enabled, we need to have
3377 * OFPP_NORMAL and the learning action have a look at the packet
3378 * before we can drop it. */
3379 if (!xport_stp_forward_state(xport
) && !xport_stp_learn_state(xport
)) {
3387 xlate_write_actions(struct xlate_ctx
*ctx
, const struct ofpact
*a
)
3389 struct ofpact_nest
*on
= ofpact_get_WRITE_ACTIONS(a
);
3390 ofpbuf_put(&ctx
->action_set
, on
->actions
, ofpact_nest_get_action_len(on
));
3391 ofpact_pad(&ctx
->action_set
);
3395 xlate_action_set(struct xlate_ctx
*ctx
)
3397 uint64_t action_list_stub
[1024 / 64];
3398 struct ofpbuf action_list
;
3400 ofpbuf_use_stub(&action_list
, action_list_stub
, sizeof action_list_stub
);
3401 ofpacts_execute_action_set(&action_list
, &ctx
->action_set
);
3402 do_xlate_actions(ofpbuf_data(&action_list
), ofpbuf_size(&action_list
), ctx
);
3403 ofpbuf_uninit(&action_list
);
3407 ofpact_needs_recirculation_after_mpls(const struct xlate_ctx
*ctx
,
3408 const struct ofpact
*a
)
3410 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
3411 struct flow
*flow
= &ctx
->xin
->flow
;
3416 case OFPACT_CONTROLLER
:
3417 case OFPACT_STRIP_VLAN
:
3418 case OFPACT_SET_VLAN_PCP
:
3419 case OFPACT_SET_VLAN_VID
:
3420 case OFPACT_ENQUEUE
:
3421 case OFPACT_PUSH_VLAN
:
3422 case OFPACT_SET_ETH_SRC
:
3423 case OFPACT_SET_ETH_DST
:
3424 case OFPACT_SET_TUNNEL
:
3425 case OFPACT_SET_QUEUE
:
3426 case OFPACT_POP_QUEUE
:
3427 case OFPACT_POP_MPLS
:
3428 case OFPACT_DEC_MPLS_TTL
:
3429 case OFPACT_SET_MPLS_TTL
:
3430 case OFPACT_SET_MPLS_TC
:
3431 case OFPACT_SET_MPLS_LABEL
:
3433 case OFPACT_OUTPUT_REG
:
3436 case OFPACT_WRITE_METADATA
:
3437 case OFPACT_WRITE_ACTIONS
:
3438 case OFPACT_CLEAR_ACTIONS
:
3442 case OFPACT_SET_IPV4_SRC
:
3443 case OFPACT_SET_IPV4_DST
:
3444 case OFPACT_SET_IP_DSCP
:
3445 case OFPACT_SET_IP_ECN
:
3446 case OFPACT_SET_IP_TTL
:
3447 case OFPACT_SET_L4_SRC_PORT
:
3448 case OFPACT_SET_L4_DST_PORT
:
3449 case OFPACT_RESUBMIT
:
3450 case OFPACT_STACK_PUSH
:
3451 case OFPACT_STACK_POP
:
3452 case OFPACT_DEC_TTL
:
3453 case OFPACT_MULTIPATH
:
3456 case OFPACT_FIN_TIMEOUT
:
3457 case OFPACT_GOTO_TABLE
:
3460 case OFPACT_REG_MOVE
:
3461 return (mf_is_l3_or_higher(ofpact_get_REG_MOVE(a
)->dst
.field
) ||
3462 mf_is_l3_or_higher(ofpact_get_REG_MOVE(a
)->src
.field
));
3464 case OFPACT_REG_LOAD
:
3465 return mf_is_l3_or_higher(ofpact_get_REG_LOAD(a
)->dst
.field
);
3467 case OFPACT_SET_FIELD
:
3468 return mf_is_l3_or_higher(ofpact_get_SET_FIELD(a
)->field
);
3470 case OFPACT_PUSH_MPLS
:
3471 /* Recirculate if it is an IP packet with a zero ttl. This may
3472 * indicate that the packet was previously MPLS and an MPLS pop action
3473 * converted it to IP. In this case recirculating should reveal the IP
3474 * TTL which is used as the basis for a new MPLS LSE. */
3475 return (!flow_count_mpls_labels(flow
, wc
)
3476 && flow
->nw_ttl
== 0
3477 && is_ip_any(flow
));
3484 do_xlate_actions(const struct ofpact
*ofpacts
, size_t ofpacts_len
,
3485 struct xlate_ctx
*ctx
)
3487 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
3488 struct flow
*flow
= &ctx
->xin
->flow
;
3489 const struct ofpact
*a
;
3491 /* dl_type already in the mask, not set below. */
3493 OFPACT_FOR_EACH (a
, ofpacts
, ofpacts_len
) {
3494 struct ofpact_controller
*controller
;
3495 const struct ofpact_metadata
*metadata
;
3496 const struct ofpact_set_field
*set_field
;
3497 const struct mf_field
*mf
;
3503 if (ctx
->was_mpls
&& ofpact_needs_recirculation_after_mpls(ctx
, a
)) {
3504 compose_recirculate_action(ctx
, ofpacts
, a
, ofpacts_len
);
3510 xlate_output_action(ctx
, ofpact_get_OUTPUT(a
)->port
,
3511 ofpact_get_OUTPUT(a
)->max_len
, true);
3515 if (xlate_group_action(ctx
, ofpact_get_GROUP(a
)->group_id
)) {
3520 case OFPACT_CONTROLLER
:
3521 controller
= ofpact_get_CONTROLLER(a
);
3522 execute_controller_action(ctx
, controller
->max_len
,
3524 controller
->controller_id
);
3527 case OFPACT_ENQUEUE
:
3528 xlate_enqueue_action(ctx
, ofpact_get_ENQUEUE(a
));
3531 case OFPACT_SET_VLAN_VID
:
3532 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
3533 if (flow
->vlan_tci
& htons(VLAN_CFI
) ||
3534 ofpact_get_SET_VLAN_VID(a
)->push_vlan_if_needed
) {
3535 flow
->vlan_tci
&= ~htons(VLAN_VID_MASK
);
3536 flow
->vlan_tci
|= (htons(ofpact_get_SET_VLAN_VID(a
)->vlan_vid
)
3541 case OFPACT_SET_VLAN_PCP
:
3542 wc
->masks
.vlan_tci
|= htons(VLAN_PCP_MASK
| VLAN_CFI
);
3543 if (flow
->vlan_tci
& htons(VLAN_CFI
) ||
3544 ofpact_get_SET_VLAN_PCP(a
)->push_vlan_if_needed
) {
3545 flow
->vlan_tci
&= ~htons(VLAN_PCP_MASK
);
3546 flow
->vlan_tci
|= htons((ofpact_get_SET_VLAN_PCP(a
)->vlan_pcp
3547 << VLAN_PCP_SHIFT
) | VLAN_CFI
);
3551 case OFPACT_STRIP_VLAN
:
3552 memset(&wc
->masks
.vlan_tci
, 0xff, sizeof wc
->masks
.vlan_tci
);
3553 flow
->vlan_tci
= htons(0);
3556 case OFPACT_PUSH_VLAN
:
3557 /* XXX 802.1AD(QinQ) */
3558 memset(&wc
->masks
.vlan_tci
, 0xff, sizeof wc
->masks
.vlan_tci
);
3559 flow
->vlan_tci
= htons(VLAN_CFI
);
3562 case OFPACT_SET_ETH_SRC
:
3563 memset(&wc
->masks
.dl_src
, 0xff, sizeof wc
->masks
.dl_src
);
3564 memcpy(flow
->dl_src
, ofpact_get_SET_ETH_SRC(a
)->mac
, ETH_ADDR_LEN
);
3567 case OFPACT_SET_ETH_DST
:
3568 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
3569 memcpy(flow
->dl_dst
, ofpact_get_SET_ETH_DST(a
)->mac
, ETH_ADDR_LEN
);
3572 case OFPACT_SET_IPV4_SRC
:
3573 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
3574 memset(&wc
->masks
.nw_src
, 0xff, sizeof wc
->masks
.nw_src
);
3575 flow
->nw_src
= ofpact_get_SET_IPV4_SRC(a
)->ipv4
;
3579 case OFPACT_SET_IPV4_DST
:
3580 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
3581 memset(&wc
->masks
.nw_dst
, 0xff, sizeof wc
->masks
.nw_dst
);
3582 flow
->nw_dst
= ofpact_get_SET_IPV4_DST(a
)->ipv4
;
3586 case OFPACT_SET_IP_DSCP
:
3587 if (is_ip_any(flow
)) {
3588 wc
->masks
.nw_tos
|= IP_DSCP_MASK
;
3589 flow
->nw_tos
&= ~IP_DSCP_MASK
;
3590 flow
->nw_tos
|= ofpact_get_SET_IP_DSCP(a
)->dscp
;
3594 case OFPACT_SET_IP_ECN
:
3595 if (is_ip_any(flow
)) {
3596 wc
->masks
.nw_tos
|= IP_ECN_MASK
;
3597 flow
->nw_tos
&= ~IP_ECN_MASK
;
3598 flow
->nw_tos
|= ofpact_get_SET_IP_ECN(a
)->ecn
;
3602 case OFPACT_SET_IP_TTL
:
3603 if (is_ip_any(flow
)) {
3604 wc
->masks
.nw_ttl
= 0xff;
3605 flow
->nw_ttl
= ofpact_get_SET_IP_TTL(a
)->ttl
;
3609 case OFPACT_SET_L4_SRC_PORT
:
3610 if (is_ip_any(flow
)) {
3611 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
3612 memset(&wc
->masks
.tp_src
, 0xff, sizeof wc
->masks
.tp_src
);
3613 flow
->tp_src
= htons(ofpact_get_SET_L4_SRC_PORT(a
)->port
);
3617 case OFPACT_SET_L4_DST_PORT
:
3618 if (is_ip_any(flow
)) {
3619 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
3620 memset(&wc
->masks
.tp_dst
, 0xff, sizeof wc
->masks
.tp_dst
);
3621 flow
->tp_dst
= htons(ofpact_get_SET_L4_DST_PORT(a
)->port
);
3625 case OFPACT_RESUBMIT
:
3626 xlate_ofpact_resubmit(ctx
, ofpact_get_RESUBMIT(a
));
3629 case OFPACT_SET_TUNNEL
:
3630 flow
->tunnel
.tun_id
= htonll(ofpact_get_SET_TUNNEL(a
)->tun_id
);
3633 case OFPACT_SET_QUEUE
:
3634 xlate_set_queue_action(ctx
, ofpact_get_SET_QUEUE(a
)->queue_id
);
3637 case OFPACT_POP_QUEUE
:
3638 flow
->skb_priority
= ctx
->orig_skb_priority
;
3641 case OFPACT_REG_MOVE
:
3642 nxm_execute_reg_move(ofpact_get_REG_MOVE(a
), flow
, wc
);
3645 case OFPACT_REG_LOAD
:
3646 nxm_execute_reg_load(ofpact_get_REG_LOAD(a
), flow
, wc
);
3649 case OFPACT_SET_FIELD
:
3650 set_field
= ofpact_get_SET_FIELD(a
);
3651 mf
= set_field
->field
;
3653 /* Set field action only ever overwrites packet's outermost
3654 * applicable header fields. Do nothing if no header exists. */
3655 if (mf
->id
== MFF_VLAN_VID
) {
3656 wc
->masks
.vlan_tci
|= htons(VLAN_CFI
);
3657 if (!(flow
->vlan_tci
& htons(VLAN_CFI
))) {
3660 } else if ((mf
->id
== MFF_MPLS_LABEL
|| mf
->id
== MFF_MPLS_TC
)
3661 /* 'dl_type' is already unwildcarded. */
3662 && !eth_type_mpls(flow
->dl_type
)) {
3666 mf_mask_field_and_prereqs(mf
, &wc
->masks
);
3667 mf_set_flow_value(mf
, &set_field
->value
, flow
);
3670 case OFPACT_STACK_PUSH
:
3671 nxm_execute_stack_push(ofpact_get_STACK_PUSH(a
), flow
, wc
,
3675 case OFPACT_STACK_POP
:
3676 nxm_execute_stack_pop(ofpact_get_STACK_POP(a
), flow
, wc
,
3680 case OFPACT_PUSH_MPLS
:
3681 compose_mpls_push_action(ctx
, ofpact_get_PUSH_MPLS(a
));
3684 case OFPACT_POP_MPLS
:
3685 compose_mpls_pop_action(ctx
, ofpact_get_POP_MPLS(a
)->ethertype
);
3688 case OFPACT_SET_MPLS_LABEL
:
3689 compose_set_mpls_label_action(
3690 ctx
, ofpact_get_SET_MPLS_LABEL(a
)->label
);
3693 case OFPACT_SET_MPLS_TC
:
3694 compose_set_mpls_tc_action(ctx
, ofpact_get_SET_MPLS_TC(a
)->tc
);
3697 case OFPACT_SET_MPLS_TTL
:
3698 compose_set_mpls_ttl_action(ctx
, ofpact_get_SET_MPLS_TTL(a
)->ttl
);
3701 case OFPACT_DEC_MPLS_TTL
:
3702 if (compose_dec_mpls_ttl_action(ctx
)) {
3707 case OFPACT_DEC_TTL
:
3708 wc
->masks
.nw_ttl
= 0xff;
3709 if (compose_dec_ttl(ctx
, ofpact_get_DEC_TTL(a
))) {
3715 /* Nothing to do. */
3718 case OFPACT_MULTIPATH
:
3719 multipath_execute(ofpact_get_MULTIPATH(a
), flow
, wc
);
3723 xlate_bundle_action(ctx
, ofpact_get_BUNDLE(a
));
3726 case OFPACT_OUTPUT_REG
:
3727 xlate_output_reg_action(ctx
, ofpact_get_OUTPUT_REG(a
));
3731 xlate_learn_action(ctx
, ofpact_get_LEARN(a
));
3738 case OFPACT_FIN_TIMEOUT
:
3739 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
3740 ctx
->xout
->has_fin_timeout
= true;
3741 xlate_fin_timeout(ctx
, ofpact_get_FIN_TIMEOUT(a
));
3744 case OFPACT_CLEAR_ACTIONS
:
3745 ofpbuf_clear(&ctx
->action_set
);
3748 case OFPACT_WRITE_ACTIONS
:
3749 xlate_write_actions(ctx
, a
);
3752 case OFPACT_WRITE_METADATA
:
3753 metadata
= ofpact_get_WRITE_METADATA(a
);
3754 flow
->metadata
&= ~metadata
->mask
;
3755 flow
->metadata
|= metadata
->metadata
& metadata
->mask
;
3759 /* Not implemented yet. */
3762 case OFPACT_GOTO_TABLE
: {
3763 struct ofpact_goto_table
*ogt
= ofpact_get_GOTO_TABLE(a
);
3765 /* Allow ctx->table_id == TBL_INTERNAL, which will be greater
3766 * than ogt->table_id. This is to allow goto_table actions that
3767 * triggered recirculation: ctx->table_id will be TBL_INTERNAL
3768 * after recirculation. */
3769 ovs_assert(ctx
->table_id
== TBL_INTERNAL
3770 || ctx
->table_id
< ogt
->table_id
);
3771 xlate_table_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
,
3772 ogt
->table_id
, true, true);
3777 xlate_sample_action(ctx
, ofpact_get_SAMPLE(a
));
3784 xlate_in_init(struct xlate_in
*xin
, struct ofproto_dpif
*ofproto
,
3785 const struct flow
*flow
, struct rule_dpif
*rule
,
3786 uint16_t tcp_flags
, const struct ofpbuf
*packet
)
3788 xin
->ofproto
= ofproto
;
3790 xin
->packet
= packet
;
3791 xin
->may_learn
= packet
!= NULL
;
3794 xin
->ofpacts
= NULL
;
3795 xin
->ofpacts_len
= 0;
3796 xin
->tcp_flags
= tcp_flags
;
3797 xin
->resubmit_hook
= NULL
;
3798 xin
->report_hook
= NULL
;
3799 xin
->resubmit_stats
= NULL
;
3800 xin
->skip_wildcards
= false;
3804 xlate_out_uninit(struct xlate_out
*xout
)
3807 ofpbuf_uninit(&xout
->odp_actions
);
3811 /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts'
3812 * into datapath actions, using 'ctx', and discards the datapath actions. */
3814 xlate_actions_for_side_effects(struct xlate_in
*xin
)
3816 struct xlate_out xout
;
3818 xlate_actions(xin
, &xout
);
3819 xlate_out_uninit(&xout
);
3823 xlate_report(struct xlate_ctx
*ctx
, const char *s
)
3825 if (ctx
->xin
->report_hook
) {
3826 ctx
->xin
->report_hook(ctx
->xin
, s
, ctx
->recurse
);
3831 xlate_out_copy(struct xlate_out
*dst
, const struct xlate_out
*src
)
3834 dst
->slow
= src
->slow
;
3835 dst
->has_learn
= src
->has_learn
;
3836 dst
->has_normal
= src
->has_normal
;
3837 dst
->has_fin_timeout
= src
->has_fin_timeout
;
3838 dst
->nf_output_iface
= src
->nf_output_iface
;
3839 dst
->mirrors
= src
->mirrors
;
3841 ofpbuf_use_stub(&dst
->odp_actions
, dst
->odp_actions_stub
,
3842 sizeof dst
->odp_actions_stub
);
3843 ofpbuf_put(&dst
->odp_actions
, ofpbuf_data(&src
->odp_actions
),
3844 ofpbuf_size(&src
->odp_actions
));
3847 static struct skb_priority_to_dscp
*
3848 get_skb_priority(const struct xport
*xport
, uint32_t skb_priority
)
3850 struct skb_priority_to_dscp
*pdscp
;
3853 hash
= hash_int(skb_priority
, 0);
3854 HMAP_FOR_EACH_IN_BUCKET (pdscp
, hmap_node
, hash
, &xport
->skb_priorities
) {
3855 if (pdscp
->skb_priority
== skb_priority
) {
3863 dscp_from_skb_priority(const struct xport
*xport
, uint32_t skb_priority
,
3866 struct skb_priority_to_dscp
*pdscp
= get_skb_priority(xport
, skb_priority
);
3867 *dscp
= pdscp
? pdscp
->dscp
: 0;
3868 return pdscp
!= NULL
;
3872 clear_skb_priorities(struct xport
*xport
)
3874 struct skb_priority_to_dscp
*pdscp
, *next
;
3876 HMAP_FOR_EACH_SAFE (pdscp
, next
, hmap_node
, &xport
->skb_priorities
) {
3877 hmap_remove(&xport
->skb_priorities
, &pdscp
->hmap_node
);
3883 actions_output_to_local_port(const struct xlate_ctx
*ctx
)
3885 odp_port_t local_odp_port
= ofp_port_to_odp_port(ctx
->xbridge
, OFPP_LOCAL
);
3886 const struct nlattr
*a
;
3889 NL_ATTR_FOR_EACH_UNSAFE (a
, left
, ofpbuf_data(&ctx
->xout
->odp_actions
),
3890 ofpbuf_size(&ctx
->xout
->odp_actions
)) {
3891 if (nl_attr_type(a
) == OVS_ACTION_ATTR_OUTPUT
3892 && nl_attr_get_odp_port(a
) == local_odp_port
) {
3899 /* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts'
3900 * into datapath actions in 'odp_actions', using 'ctx'.
3902 * The caller must take responsibility for eventually freeing 'xout', with
3903 * xlate_out_uninit(). */
3905 xlate_actions(struct xlate_in
*xin
, struct xlate_out
*xout
)
3907 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
3908 struct flow_wildcards
*wc
= &xout
->wc
;
3909 struct flow
*flow
= &xin
->flow
;
3910 struct rule_dpif
*rule
= NULL
;
3912 const struct rule_actions
*actions
= NULL
;
3913 enum slow_path_reason special
;
3914 const struct ofpact
*ofpacts
;
3915 struct xport
*in_port
;
3916 struct flow orig_flow
;
3917 struct xlate_ctx ctx
;
3922 COVERAGE_INC(xlate_actions
);
3924 /* Flow initialization rules:
3925 * - 'base_flow' must match the kernel's view of the packet at the
3926 * time that action processing starts. 'flow' represents any
3927 * transformations we wish to make through actions.
3928 * - By default 'base_flow' and 'flow' are the same since the input
3929 * packet matches the output before any actions are applied.
3930 * - When using VLAN splinters, 'base_flow''s VLAN is set to the value
3931 * of the received packet as seen by the kernel. If we later output
3932 * to another device without any modifications this will cause us to
3933 * insert a new tag since the original one was stripped off by the
3935 * - Tunnel metadata as received is retained in 'flow'. This allows
3936 * tunnel metadata matching also in later tables.
3937 * Since a kernel action for setting the tunnel metadata will only be
3938 * generated with actual tunnel output, changing the tunnel metadata
3939 * values in 'flow' (such as tun_id) will only have effect with a later
3940 * tunnel output action.
3941 * - Tunnel 'base_flow' is completely cleared since that is what the
3942 * kernel does. If we wish to maintain the original values an action
3943 * needs to be generated. */
3948 ctx
.xout
->has_learn
= false;
3949 ctx
.xout
->has_normal
= false;
3950 ctx
.xout
->has_fin_timeout
= false;
3951 ctx
.xout
->nf_output_iface
= NF_OUT_DROP
;
3952 ctx
.xout
->mirrors
= 0;
3953 ofpbuf_use_stub(&ctx
.xout
->odp_actions
, ctx
.xout
->odp_actions_stub
,
3954 sizeof ctx
.xout
->odp_actions_stub
);
3955 ofpbuf_reserve(&ctx
.xout
->odp_actions
, NL_A_U32_SIZE
);
3957 ctx
.xbridge
= xbridge_lookup(xcfg
, xin
->ofproto
);
3962 ctx
.rule
= xin
->rule
;
3964 ctx
.base_flow
= *flow
;
3965 memset(&ctx
.base_flow
.tunnel
, 0, sizeof ctx
.base_flow
.tunnel
);
3966 ctx
.orig_tunnel_ip_dst
= flow
->tunnel
.ip_dst
;
3968 flow_wildcards_init_catchall(wc
);
3969 memset(&wc
->masks
.in_port
, 0xff, sizeof wc
->masks
.in_port
);
3970 memset(&wc
->masks
.skb_priority
, 0xff, sizeof wc
->masks
.skb_priority
);
3971 memset(&wc
->masks
.dl_type
, 0xff, sizeof wc
->masks
.dl_type
);
3972 if (is_ip_any(flow
)) {
3973 wc
->masks
.nw_frag
|= FLOW_NW_FRAG_MASK
;
3975 is_icmp
= is_icmpv4(flow
) || is_icmpv6(flow
);
3977 tnl_may_send
= tnl_xlate_init(&ctx
.base_flow
, flow
, wc
);
3978 if (ctx
.xbridge
->netflow
) {
3979 netflow_mask_wc(flow
, wc
);
3984 ctx
.in_group
= false;
3985 ctx
.orig_skb_priority
= flow
->skb_priority
;
3988 ctx
.use_recirc
= false;
3989 ctx
.was_mpls
= false;
3991 if (!xin
->ofpacts
&& !ctx
.rule
) {
3992 ctx
.table_id
= rule_dpif_lookup(ctx
.xbridge
->ofproto
, flow
,
3993 !xin
->skip_wildcards
? wc
: NULL
,
3994 &rule
, ctx
.xin
->xcache
!= NULL
,
3995 ctx
.xin
->resubmit_stats
);
3996 if (ctx
.xin
->resubmit_stats
) {
3997 rule_dpif_credit_stats(rule
, ctx
.xin
->resubmit_stats
);
3999 if (ctx
.xin
->xcache
) {
4000 struct xc_entry
*entry
;
4002 entry
= xlate_cache_add_entry(ctx
.xin
->xcache
, XC_RULE
);
4003 entry
->u
.rule
= rule
;
4007 xout
->fail_open
= ctx
.rule
&& rule_dpif_is_fail_open(ctx
.rule
);
4010 ofpacts
= xin
->ofpacts
;
4011 ofpacts_len
= xin
->ofpacts_len
;
4012 } else if (ctx
.rule
) {
4013 actions
= rule_dpif_get_actions(ctx
.rule
);
4014 ofpacts
= actions
->ofpacts
;
4015 ofpacts_len
= actions
->ofpacts_len
;
4020 ofpbuf_use_stub(&ctx
.stack
, ctx
.init_stack
, sizeof ctx
.init_stack
);
4021 ofpbuf_use_stub(&ctx
.action_set
,
4022 ctx
.action_set_stub
, sizeof ctx
.action_set_stub
);
4024 if (mbridge_has_mirrors(ctx
.xbridge
->mbridge
)) {
4025 /* Do this conditionally because the copy is expensive enough that it
4026 * shows up in profiles. */
4030 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
) {
4031 switch (ctx
.xbridge
->frag
) {
4032 case OFPC_FRAG_NORMAL
:
4033 /* We must pretend that transport ports are unavailable. */
4034 flow
->tp_src
= ctx
.base_flow
.tp_src
= htons(0);
4035 flow
->tp_dst
= ctx
.base_flow
.tp_dst
= htons(0);
4038 case OFPC_FRAG_DROP
:
4041 case OFPC_FRAG_REASM
:
4044 case OFPC_FRAG_NX_MATCH
:
4045 /* Nothing to do. */
4048 case OFPC_INVALID_TTL_TO_CONTROLLER
:
4053 in_port
= get_ofp_port(ctx
.xbridge
, flow
->in_port
.ofp_port
);
4054 if (in_port
&& in_port
->is_tunnel
) {
4055 if (ctx
.xin
->resubmit_stats
) {
4056 netdev_vport_inc_rx(in_port
->netdev
, ctx
.xin
->resubmit_stats
);
4058 bfd_account_rx(in_port
->bfd
, ctx
.xin
->resubmit_stats
);
4061 if (ctx
.xin
->xcache
) {
4062 struct xc_entry
*entry
;
4064 entry
= xlate_cache_add_entry(ctx
.xin
->xcache
, XC_NETDEV
);
4065 entry
->u
.dev
.rx
= netdev_ref(in_port
->netdev
);
4066 entry
->u
.dev
.bfd
= bfd_ref(in_port
->bfd
);
4070 special
= process_special(&ctx
, flow
, in_port
, ctx
.xin
->packet
);
4072 ctx
.xout
->slow
|= special
;
4074 size_t sample_actions_len
;
4076 if (flow
->in_port
.ofp_port
4077 != vsp_realdev_to_vlandev(ctx
.xbridge
->ofproto
,
4078 flow
->in_port
.ofp_port
,
4080 ctx
.base_flow
.vlan_tci
= 0;
4083 add_sflow_action(&ctx
);
4084 add_ipfix_action(&ctx
);
4085 sample_actions_len
= ofpbuf_size(&ctx
.xout
->odp_actions
);
4087 if (tnl_may_send
&& (!in_port
|| may_receive(in_port
, &ctx
))) {
4088 do_xlate_actions(ofpacts
, ofpacts_len
, &ctx
);
4090 /* We've let OFPP_NORMAL and the learning action look at the
4091 * packet, so drop it now if forwarding is disabled. */
4092 if (in_port
&& !xport_stp_forward_state(in_port
)) {
4093 ofpbuf_set_size(&ctx
.xout
->odp_actions
, sample_actions_len
);
4097 if (ofpbuf_size(&ctx
.action_set
)) {
4098 xlate_action_set(&ctx
);
4101 if (ctx
.xbridge
->has_in_band
4102 && in_band_must_output_to_local_port(flow
)
4103 && !actions_output_to_local_port(&ctx
)) {
4104 compose_output_action(&ctx
, OFPP_LOCAL
);
4107 fix_sflow_action(&ctx
);
4109 if (mbridge_has_mirrors(ctx
.xbridge
->mbridge
)) {
4110 add_mirror_actions(&ctx
, &orig_flow
);
4114 if (nl_attr_oversized(ofpbuf_size(&ctx
.xout
->odp_actions
))) {
4115 /* These datapath actions are too big for a Netlink attribute, so we
4116 * can't hand them to the kernel directly. dpif_execute() can execute
4117 * them one by one with help, so just mark the result as SLOW_ACTION to
4118 * prevent the flow from being installed. */
4119 COVERAGE_INC(xlate_actions_oversize
);
4120 ctx
.xout
->slow
|= SLOW_ACTION
;
4123 if (mbridge_has_mirrors(ctx
.xbridge
->mbridge
)) {
4124 if (ctx
.xin
->resubmit_stats
) {
4125 mirror_update_stats(ctx
.xbridge
->mbridge
, xout
->mirrors
,
4126 ctx
.xin
->resubmit_stats
->n_packets
,
4127 ctx
.xin
->resubmit_stats
->n_bytes
);
4129 if (ctx
.xin
->xcache
) {
4130 struct xc_entry
*entry
;
4132 entry
= xlate_cache_add_entry(ctx
.xin
->xcache
, XC_MIRROR
);
4133 entry
->u
.mirror
.mbridge
= mbridge_ref(ctx
.xbridge
->mbridge
);
4134 entry
->u
.mirror
.mirrors
= xout
->mirrors
;
4138 if (ctx
.xbridge
->netflow
) {
4139 /* Only update netflow if we don't have controller flow. We don't
4140 * report NetFlow expiration messages for such facets because they
4141 * are just part of the control logic for the network, not real
4143 if (ofpacts_len
== 0
4144 || ofpacts
->type
!= OFPACT_CONTROLLER
4145 || ofpact_next(ofpacts
) < ofpact_end(ofpacts
, ofpacts_len
)) {
4146 if (ctx
.xin
->resubmit_stats
) {
4147 netflow_flow_update(ctx
.xbridge
->netflow
, flow
,
4148 xout
->nf_output_iface
,
4149 ctx
.xin
->resubmit_stats
);
4151 if (ctx
.xin
->xcache
) {
4152 struct xc_entry
*entry
;
4154 entry
= xlate_cache_add_entry(ctx
.xin
->xcache
, XC_NETFLOW
);
4155 entry
->u
.nf
.netflow
= netflow_ref(ctx
.xbridge
->netflow
);
4156 entry
->u
.nf
.flow
= xmemdup(flow
, sizeof *flow
);
4157 entry
->u
.nf
.iface
= xout
->nf_output_iface
;
4162 ofpbuf_uninit(&ctx
.stack
);
4163 ofpbuf_uninit(&ctx
.action_set
);
4165 /* Clear the metadata and register wildcard masks, because we won't
4166 * use non-header fields as part of the cache. */
4167 flow_wildcards_clear_non_packet_fields(wc
);
4169 /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow uses
4170 * the low 8 bits of the 16-bit tp_src and tp_dst members to represent
4171 * these fields. The datapath interface, on the other hand, represents
4172 * them with just 8 bits each. This means that if the high 8 bits of the
4173 * masks for these fields somehow become set, then they will get chopped
4174 * off by a round trip through the datapath, and revalidation will spot
4175 * that as an inconsistency and delete the flow. Avoid the problem here by
4176 * making sure that only the low 8 bits of either field can be unwildcarded
4180 wc
->masks
.tp_src
&= htons(UINT8_MAX
);
4181 wc
->masks
.tp_dst
&= htons(UINT8_MAX
);
4185 /* Sends 'packet' out 'ofport'.
4186 * May modify 'packet'.
4187 * Returns 0 if successful, otherwise a positive errno value. */
4189 xlate_send_packet(const struct ofport_dpif
*ofport
, struct ofpbuf
*packet
)
4191 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
4192 struct xport
*xport
;
4193 struct ofpact_output output
;
4196 ofpact_init(&output
.ofpact
, OFPACT_OUTPUT
, sizeof output
);
4197 /* Use OFPP_NONE as the in_port to avoid special packet processing. */
4198 flow_extract(packet
, NULL
, &flow
);
4199 flow
.in_port
.ofp_port
= OFPP_NONE
;
4201 xport
= xport_lookup(xcfg
, ofport
);
4205 output
.port
= xport
->ofp_port
;
4208 return ofproto_dpif_execute_actions(xport
->xbridge
->ofproto
, &flow
, NULL
,
4209 &output
.ofpact
, sizeof output
,
4213 struct xlate_cache
*
4214 xlate_cache_new(void)
4216 struct xlate_cache
*xcache
= xmalloc(sizeof *xcache
);
4218 ofpbuf_init(&xcache
->entries
, 512);
4222 static struct xc_entry
*
4223 xlate_cache_add_entry(struct xlate_cache
*xcache
, enum xc_type type
)
4225 struct xc_entry
*entry
;
4227 entry
= ofpbuf_put_zeros(&xcache
->entries
, sizeof *entry
);
4234 xlate_cache_netdev(struct xc_entry
*entry
, const struct dpif_flow_stats
*stats
)
4236 if (entry
->u
.dev
.tx
) {
4237 netdev_vport_inc_tx(entry
->u
.dev
.tx
, stats
);
4239 if (entry
->u
.dev
.rx
) {
4240 netdev_vport_inc_rx(entry
->u
.dev
.rx
, stats
);
4242 if (entry
->u
.dev
.bfd
) {
4243 bfd_account_rx(entry
->u
.dev
.bfd
, stats
);
4248 xlate_cache_normal(struct ofproto_dpif
*ofproto
, struct flow
*flow
, int vlan
)
4250 struct xlate_cfg
*xcfg
= ovsrcu_get(struct xlate_cfg
*, &xcfgp
);
4251 struct xbridge
*xbridge
;
4252 struct xbundle
*xbundle
;
4253 struct flow_wildcards wc
;
4255 xbridge
= xbridge_lookup(xcfg
, ofproto
);
4260 xbundle
= lookup_input_bundle(xbridge
, flow
->in_port
.ofp_port
, false,
4266 update_learning_table(xbridge
, flow
, &wc
, vlan
, xbundle
);
4269 /* Push stats and perform side effects of flow translation. */
4271 xlate_push_stats(struct xlate_cache
*xcache
, bool may_learn
,
4272 const struct dpif_flow_stats
*stats
)
4274 struct xc_entry
*entry
;
4275 struct ofpbuf entries
= xcache
->entries
;
4277 XC_ENTRY_FOR_EACH (entry
, entries
, xcache
) {
4278 switch (entry
->type
) {
4280 rule_dpif_credit_stats(entry
->u
.rule
, stats
);
4283 bond_account(entry
->u
.bond
.bond
, entry
->u
.bond
.flow
,
4284 entry
->u
.bond
.vid
, stats
->n_bytes
);
4287 xlate_cache_netdev(entry
, stats
);
4290 netflow_flow_update(entry
->u
.nf
.netflow
, entry
->u
.nf
.flow
,
4291 entry
->u
.nf
.iface
, stats
);
4294 mirror_update_stats(entry
->u
.mirror
.mbridge
,
4295 entry
->u
.mirror
.mirrors
,
4296 stats
->n_packets
, stats
->n_bytes
);
4300 ofproto_dpif_flow_mod(entry
->u
.learn
.ofproto
,
4305 xlate_cache_normal(entry
->u
.normal
.ofproto
, entry
->u
.normal
.flow
,
4306 entry
->u
.normal
.vlan
);
4308 case XC_FIN_TIMEOUT
:
4309 xlate_fin_timeout__(entry
->u
.fin
.rule
, stats
->tcp_flags
,
4310 entry
->u
.fin
.idle
, entry
->u
.fin
.hard
);
4313 group_dpif_credit_stats(entry
->u
.group
.group
, entry
->u
.group
.bucket
,
4323 xlate_dev_unref(struct xc_entry
*entry
)
4325 if (entry
->u
.dev
.tx
) {
4326 netdev_close(entry
->u
.dev
.tx
);
4328 if (entry
->u
.dev
.rx
) {
4329 netdev_close(entry
->u
.dev
.rx
);
4331 if (entry
->u
.dev
.bfd
) {
4332 bfd_unref(entry
->u
.dev
.bfd
);
4337 xlate_cache_clear_netflow(struct netflow
*netflow
, struct flow
*flow
)
4339 netflow_flow_clear(netflow
, flow
);
4340 netflow_unref(netflow
);
4345 xlate_cache_clear(struct xlate_cache
*xcache
)
4347 struct xc_entry
*entry
;
4348 struct ofpbuf entries
;
4354 XC_ENTRY_FOR_EACH (entry
, entries
, xcache
) {
4355 switch (entry
->type
) {
4357 rule_dpif_unref(entry
->u
.rule
);
4360 free(entry
->u
.bond
.flow
);
4361 bond_unref(entry
->u
.bond
.bond
);
4364 xlate_dev_unref(entry
);
4367 xlate_cache_clear_netflow(entry
->u
.nf
.netflow
, entry
->u
.nf
.flow
);
4370 mbridge_unref(entry
->u
.mirror
.mbridge
);
4373 free(entry
->u
.learn
.fm
);
4374 ofpbuf_delete(entry
->u
.learn
.ofpacts
);
4377 free(entry
->u
.normal
.flow
);
4379 case XC_FIN_TIMEOUT
:
4380 /* 'u.fin.rule' is always already held as a XC_RULE, which
4381 * has already released it's reference above. */
4384 group_dpif_unref(entry
->u
.group
.group
);
4391 ofpbuf_clear(&xcache
->entries
);
4395 xlate_cache_delete(struct xlate_cache
*xcache
)
4397 xlate_cache_clear(xcache
);
4398 ofpbuf_uninit(&xcache
->entries
);