1 /* Copyright (c) 2009, 2010, 2011, 2012, 2013 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"
23 #include "byte-order.h"
28 #include "dynamic-string.h"
33 #include "mac-learning.h"
34 #include "meta-flow.h"
35 #include "multipath.h"
36 #include "netdev-vport.h"
39 #include "odp-execute.h"
40 #include "ofp-actions.h"
41 #include "ofproto/ofproto-dpif-ipfix.h"
42 #include "ofproto/ofproto-dpif-mirror.h"
43 #include "ofproto/ofproto-dpif-sflow.h"
44 #include "ofproto/ofproto-dpif.h"
48 COVERAGE_DEFINE(xlate_actions
);
50 VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate
);
52 /* Maximum depth of flow table recursion (due to resubmit actions) in a
53 * flow translation. */
54 #define MAX_RESUBMIT_RECURSION 64
57 struct hmap_node hmap_node
; /* Node in global 'xbridges' map. */
58 struct ofproto_dpif
*ofproto
; /* Key in global 'xbridges' map. */
60 struct list xbundles
; /* Owned xbundles. */
61 struct hmap xports
; /* Indexed by ofp_port. */
63 char *name
; /* Name used in log messages. */
64 struct mac_learning
*ml
; /* Mac learning handle. */
65 struct mbridge
*mbridge
; /* Mirroring. */
66 struct dpif_sflow
*sflow
; /* SFlow handle, or null. */
67 struct dpif_ipfix
*ipfix
; /* Ipfix handle, or null. */
68 struct stp
*stp
; /* STP or null if disabled. */
70 enum ofp_config_flags frag
; /* Fragmentation handling. */
71 bool has_netflow
; /* Bridge runs netflow? */
72 bool has_in_band
; /* Bridge has in band control? */
73 bool forward_bpdu
; /* Bridge forwards STP BPDUs? */
77 struct hmap_node hmap_node
; /* In global 'xbundles' map. */
78 struct ofbundle
*ofbundle
; /* Key in global 'xbundles' map. */
80 struct list list_node
; /* In parent 'xbridges' list. */
81 struct xbridge
*xbridge
; /* Parent xbridge. */
83 struct list xports
; /* Contains "struct xport"s. */
85 char *name
; /* Name used in log messages. */
86 struct bond
*bond
; /* Nonnull iff more than one port. */
87 struct lacp
*lacp
; /* LACP handle or null. */
89 enum port_vlan_mode vlan_mode
; /* VLAN mode. */
90 int vlan
; /* -1=trunk port, else a 12-bit VLAN ID. */
91 unsigned long *trunks
; /* Bitmap of trunked VLANs, if 'vlan' == -1.
92 * NULL if all VLANs are trunked. */
93 bool use_priority_tags
; /* Use 802.1p tag for frames in VLAN 0? */
94 bool floodable
; /* No port has OFPUTIL_PC_NO_FLOOD set? */
98 struct hmap_node hmap_node
; /* Node in global 'xports' map. */
99 struct ofport_dpif
*ofport
; /* Key in global 'xports map. */
101 struct hmap_node ofp_node
; /* Node in parent xbridge 'xports' map. */
102 ofp_port_t ofp_port
; /* Key in parent xbridge 'xports' map. */
104 odp_port_t odp_port
; /* Datapath port number or ODPP_NONE. */
106 struct list bundle_node
; /* In parent xbundle (if it exists). */
107 struct xbundle
*xbundle
; /* Parent xbundle or null. */
109 struct netdev
*netdev
; /* 'ofport''s netdev. */
111 struct xbridge
*xbridge
; /* Parent bridge. */
112 struct xport
*peer
; /* Patch port peer or null. */
114 enum ofputil_port_config config
; /* OpenFlow port configuration. */
115 int stp_port_no
; /* STP port number or 0 if not in use. */
117 struct hmap skb_priorities
; /* Map of 'skb_priority_to_dscp's. */
119 bool may_enable
; /* May be enabled in bonds. */
120 bool is_tunnel
; /* Is a tunnel port. */
122 struct cfm
*cfm
; /* CFM handle or null. */
123 struct bfd
*bfd
; /* BFD handle or null. */
127 struct xlate_in
*xin
;
128 struct xlate_out
*xout
;
130 const struct xbridge
*xbridge
;
132 /* Flow at the last commit. */
133 struct flow base_flow
;
135 /* Tunnel IP destination address as received. This is stored separately
136 * as the base_flow.tunnel is cleared on init to reflect the datapath
137 * behavior. Used to make sure not to send tunneled output to ourselves,
138 * which might lead to an infinite loop. This could happen easily
139 * if a tunnel is marked as 'ip_remote=flow', and the flow does not
140 * actually set the tun_dst field. */
141 ovs_be32 orig_tunnel_ip_dst
;
143 /* Stack for the push and pop actions. Each stack element is of type
144 * "union mf_subvalue". */
145 union mf_subvalue init_stack
[1024 / sizeof(union mf_subvalue
)];
148 /* The rule that we are currently translating, or NULL. */
149 struct rule_dpif
*rule
;
151 int recurse
; /* Recursion level, via xlate_table_action. */
152 bool max_resubmit_trigger
; /* Recursed too deeply during translation. */
153 uint32_t orig_skb_priority
; /* Priority when packet arrived. */
154 uint8_t table_id
; /* OpenFlow table ID where flow was found. */
155 uint32_t sflow_n_outputs
; /* Number of output ports. */
156 odp_port_t sflow_odp_port
; /* Output port for composing sFlow action. */
157 uint16_t user_cookie_offset
;/* Used for user_action_cookie fixup. */
158 bool exit
; /* No further actions should be processed. */
161 /* A controller may use OFPP_NONE as the ingress port to indicate that
162 * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for
163 * when an input bundle is needed for validation (e.g., mirroring or
164 * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have
165 * any 'port' structs, so care must be taken when dealing with it.
166 * The bundle's name and vlan mode are initialized in lookup_input_bundle() */
167 static struct xbundle ofpp_none_bundle
;
169 /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from
170 * 'priority' (the datapath's term for QoS queue) to the dscp bits which all
171 * traffic egressing the 'ofport' with that priority should be marked with. */
172 struct skb_priority_to_dscp
{
173 struct hmap_node hmap_node
; /* Node in 'ofport_dpif''s 'skb_priorities'. */
174 uint32_t skb_priority
; /* Priority of this queue (see struct flow). */
176 uint8_t dscp
; /* DSCP bits to mark outgoing traffic with. */
179 static struct hmap xbridges
= HMAP_INITIALIZER(&xbridges
);
180 static struct hmap xbundles
= HMAP_INITIALIZER(&xbundles
);
181 static struct hmap xports
= HMAP_INITIALIZER(&xports
);
183 static bool may_receive(const struct xport
*, struct xlate_ctx
*);
184 static void do_xlate_actions(const struct ofpact
*, size_t ofpacts_len
,
186 static void xlate_normal(struct xlate_ctx
*);
187 static void xlate_report(struct xlate_ctx
*, const char *);
188 static void xlate_table_action(struct xlate_ctx
*, ofp_port_t in_port
,
189 uint8_t table_id
, bool may_packet_in
);
190 static bool input_vid_is_valid(uint16_t vid
, struct xbundle
*, bool warn
);
191 static uint16_t input_vid_to_vlan(const struct xbundle
*, uint16_t vid
);
192 static void output_normal(struct xlate_ctx
*, const struct xbundle
*,
194 static void compose_output_action(struct xlate_ctx
*, ofp_port_t ofp_port
);
196 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
198 static struct xbridge
*xbridge_lookup(const struct ofproto_dpif
*);
199 static struct xbundle
*xbundle_lookup(const struct ofbundle
*);
200 static struct xport
*xport_lookup(struct ofport_dpif
*);
201 static struct xport
*get_ofp_port(const struct xbridge
*, ofp_port_t ofp_port
);
202 static struct skb_priority_to_dscp
*get_skb_priority(const struct xport
*,
203 uint32_t skb_priority
);
204 static void clear_skb_priorities(struct xport
*);
205 static bool dscp_from_skb_priority(const struct xport
*, uint32_t skb_priority
,
209 xlate_ofproto_set(struct ofproto_dpif
*ofproto
, const char *name
,
210 const struct mac_learning
*ml
, struct stp
*stp
,
211 const struct mbridge
*mbridge
,
212 const struct dpif_sflow
*sflow
,
213 const struct dpif_ipfix
*ipfix
, enum ofp_config_flags frag
,
214 bool forward_bpdu
, bool has_in_band
, bool has_netflow
)
216 struct xbridge
*xbridge
= xbridge_lookup(ofproto
);
219 xbridge
= xzalloc(sizeof *xbridge
);
220 xbridge
->ofproto
= ofproto
;
222 hmap_insert(&xbridges
, &xbridge
->hmap_node
, hash_pointer(ofproto
, 0));
223 hmap_init(&xbridge
->xports
);
224 list_init(&xbridge
->xbundles
);
227 if (xbridge
->ml
!= ml
) {
228 mac_learning_unref(xbridge
->ml
);
229 xbridge
->ml
= mac_learning_ref(ml
);
232 if (xbridge
->mbridge
!= mbridge
) {
233 mbridge_unref(xbridge
->mbridge
);
234 xbridge
->mbridge
= mbridge_ref(mbridge
);
237 if (xbridge
->sflow
!= sflow
) {
238 dpif_sflow_unref(xbridge
->sflow
);
239 xbridge
->sflow
= dpif_sflow_ref(sflow
);
242 if (xbridge
->ipfix
!= ipfix
) {
243 dpif_ipfix_unref(xbridge
->ipfix
);
244 xbridge
->ipfix
= dpif_ipfix_ref(ipfix
);
247 if (xbridge
->stp
!= stp
) {
248 stp_unref(xbridge
->stp
);
249 xbridge
->stp
= stp_ref(stp
);
253 xbridge
->name
= xstrdup(name
);
255 xbridge
->forward_bpdu
= forward_bpdu
;
256 xbridge
->has_in_band
= has_in_band
;
257 xbridge
->has_netflow
= has_netflow
;
258 xbridge
->frag
= frag
;
262 xlate_remove_ofproto(struct ofproto_dpif
*ofproto
)
264 struct xbridge
*xbridge
= xbridge_lookup(ofproto
);
265 struct xbundle
*xbundle
, *next_xbundle
;
266 struct xport
*xport
, *next_xport
;
272 HMAP_FOR_EACH_SAFE (xport
, next_xport
, ofp_node
, &xbridge
->xports
) {
273 xlate_ofport_remove(xport
->ofport
);
276 LIST_FOR_EACH_SAFE (xbundle
, next_xbundle
, list_node
, &xbridge
->xbundles
) {
277 xlate_bundle_remove(xbundle
->ofbundle
);
280 hmap_remove(&xbridges
, &xbridge
->hmap_node
);
286 xlate_bundle_set(struct ofproto_dpif
*ofproto
, struct ofbundle
*ofbundle
,
287 const char *name
, enum port_vlan_mode vlan_mode
, int vlan
,
288 unsigned long *trunks
, bool use_priority_tags
,
289 const struct bond
*bond
, const struct lacp
*lacp
,
292 struct xbundle
*xbundle
= xbundle_lookup(ofbundle
);
295 xbundle
= xzalloc(sizeof *xbundle
);
296 xbundle
->ofbundle
= ofbundle
;
297 xbundle
->xbridge
= xbridge_lookup(ofproto
);
299 hmap_insert(&xbundles
, &xbundle
->hmap_node
, hash_pointer(ofbundle
, 0));
300 list_insert(&xbundle
->xbridge
->xbundles
, &xbundle
->list_node
);
301 list_init(&xbundle
->xports
);
304 ovs_assert(xbundle
->xbridge
);
307 xbundle
->name
= xstrdup(name
);
309 xbundle
->vlan_mode
= vlan_mode
;
310 xbundle
->vlan
= vlan
;
311 xbundle
->trunks
= trunks
;
312 xbundle
->use_priority_tags
= use_priority_tags
;
313 xbundle
->floodable
= floodable
;
315 if (xbundle
->bond
!= bond
) {
316 bond_unref(xbundle
->bond
);
317 xbundle
->bond
= bond_ref(bond
);
320 if (xbundle
->lacp
!= lacp
) {
321 lacp_unref(xbundle
->lacp
);
322 xbundle
->lacp
= lacp_ref(lacp
);
327 xlate_bundle_remove(struct ofbundle
*ofbundle
)
329 struct xbundle
*xbundle
= xbundle_lookup(ofbundle
);
330 struct xport
*xport
, *next
;
336 LIST_FOR_EACH_SAFE (xport
, next
, bundle_node
, &xbundle
->xports
) {
337 list_remove(&xport
->bundle_node
);
338 xport
->xbundle
= NULL
;
341 hmap_remove(&xbundles
, &xbundle
->hmap_node
);
342 list_remove(&xbundle
->list_node
);
343 bond_unref(xbundle
->bond
);
344 lacp_unref(xbundle
->lacp
);
350 xlate_ofport_set(struct ofproto_dpif
*ofproto
, struct ofbundle
*ofbundle
,
351 struct ofport_dpif
*ofport
, ofp_port_t ofp_port
,
352 odp_port_t odp_port
, const struct netdev
*netdev
,
353 const struct cfm
*cfm
, const struct bfd
*bfd
,
354 struct ofport_dpif
*peer
, int stp_port_no
,
355 const struct ofproto_port_queue
*qdscp_list
, size_t n_qdscp
,
356 enum ofputil_port_config config
, bool is_tunnel
,
359 struct xport
*xport
= xport_lookup(ofport
);
363 xport
= xzalloc(sizeof *xport
);
364 xport
->ofport
= ofport
;
365 xport
->xbridge
= xbridge_lookup(ofproto
);
366 xport
->ofp_port
= ofp_port
;
368 hmap_init(&xport
->skb_priorities
);
369 hmap_insert(&xports
, &xport
->hmap_node
, hash_pointer(ofport
, 0));
370 hmap_insert(&xport
->xbridge
->xports
, &xport
->ofp_node
,
371 hash_ofp_port(xport
->ofp_port
));
374 ovs_assert(xport
->ofp_port
== ofp_port
);
376 xport
->config
= config
;
377 xport
->stp_port_no
= stp_port_no
;
378 xport
->is_tunnel
= is_tunnel
;
379 xport
->may_enable
= may_enable
;
380 xport
->odp_port
= odp_port
;
382 if (xport
->netdev
!= netdev
) {
383 netdev_close(xport
->netdev
);
384 xport
->netdev
= netdev_ref(netdev
);
387 if (xport
->cfm
!= cfm
) {
388 cfm_unref(xport
->cfm
);
389 xport
->cfm
= cfm_ref(cfm
);
392 if (xport
->bfd
!= bfd
) {
393 bfd_unref(xport
->bfd
);
394 xport
->bfd
= bfd_ref(bfd
);
398 xport
->peer
->peer
= NULL
;
400 xport
->peer
= peer
? xport_lookup(peer
) : NULL
;
402 xport
->peer
->peer
= xport
;
405 if (xport
->xbundle
) {
406 list_remove(&xport
->bundle_node
);
408 xport
->xbundle
= ofbundle
? xbundle_lookup(ofbundle
) : NULL
;
409 if (xport
->xbundle
) {
410 list_insert(&xport
->xbundle
->xports
, &xport
->bundle_node
);
413 clear_skb_priorities(xport
);
414 for (i
= 0; i
< n_qdscp
; i
++) {
415 struct skb_priority_to_dscp
*pdscp
;
416 uint32_t skb_priority
;
418 if (ofproto_dpif_queue_to_priority(xport
->xbridge
->ofproto
,
424 pdscp
= xmalloc(sizeof *pdscp
);
425 pdscp
->skb_priority
= skb_priority
;
426 pdscp
->dscp
= (qdscp_list
[i
].dscp
<< 2) & IP_DSCP_MASK
;
427 hmap_insert(&xport
->skb_priorities
, &pdscp
->hmap_node
,
428 hash_int(pdscp
->skb_priority
, 0));
433 xlate_ofport_remove(struct ofport_dpif
*ofport
)
435 struct xport
*xport
= xport_lookup(ofport
);
442 xport
->peer
->peer
= NULL
;
446 if (xport
->xbundle
) {
447 list_remove(&xport
->bundle_node
);
450 clear_skb_priorities(xport
);
451 hmap_destroy(&xport
->skb_priorities
);
453 hmap_remove(&xports
, &xport
->hmap_node
);
454 hmap_remove(&xport
->xbridge
->xports
, &xport
->ofp_node
);
456 netdev_close(xport
->netdev
);
457 cfm_unref(xport
->cfm
);
458 bfd_unref(xport
->bfd
);
462 static struct xbridge
*
463 xbridge_lookup(const struct ofproto_dpif
*ofproto
)
465 struct xbridge
*xbridge
;
467 HMAP_FOR_EACH_IN_BUCKET (xbridge
, hmap_node
, hash_pointer(ofproto
, 0),
469 if (xbridge
->ofproto
== ofproto
) {
476 static struct xbundle
*
477 xbundle_lookup(const struct ofbundle
*ofbundle
)
479 struct xbundle
*xbundle
;
481 HMAP_FOR_EACH_IN_BUCKET (xbundle
, hmap_node
, hash_pointer(ofbundle
, 0),
483 if (xbundle
->ofbundle
== ofbundle
) {
490 static struct xport
*
491 xport_lookup(struct ofport_dpif
*ofport
)
495 HMAP_FOR_EACH_IN_BUCKET (xport
, hmap_node
, hash_pointer(ofport
, 0),
497 if (xport
->ofport
== ofport
) {
504 static struct stp_port
*
505 xport_get_stp_port(const struct xport
*xport
)
507 return xport
->xbridge
->stp
&& xport
->stp_port_no
508 ? stp_get_port(xport
->xbridge
->stp
, xport
->stp_port_no
)
512 static enum stp_state
513 xport_stp_learn_state(const struct xport
*xport
)
515 struct stp_port
*sp
= xport_get_stp_port(xport
);
516 return stp_learn_in_state(sp
? stp_port_get_state(sp
) : STP_DISABLED
);
520 xport_stp_forward_state(const struct xport
*xport
)
522 struct stp_port
*sp
= xport_get_stp_port(xport
);
523 return stp_forward_in_state(sp
? stp_port_get_state(sp
) : STP_DISABLED
);
526 /* Returns true if STP should process 'flow'. Sets fields in 'wc' that
527 * were used to make the determination.*/
529 stp_should_process_flow(const struct flow
*flow
, struct flow_wildcards
*wc
)
531 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
532 return eth_addr_equals(flow
->dl_dst
, eth_addr_stp
);
536 stp_process_packet(const struct xport
*xport
, const struct ofpbuf
*packet
)
538 struct stp_port
*sp
= xport_get_stp_port(xport
);
539 struct ofpbuf payload
= *packet
;
540 struct eth_header
*eth
= payload
.data
;
542 /* Sink packets on ports that have STP disabled when the bridge has
544 if (!sp
|| stp_port_get_state(sp
) == STP_DISABLED
) {
548 /* Trim off padding on payload. */
549 if (payload
.size
> ntohs(eth
->eth_type
) + ETH_HEADER_LEN
) {
550 payload
.size
= ntohs(eth
->eth_type
) + ETH_HEADER_LEN
;
553 if (ofpbuf_try_pull(&payload
, ETH_HEADER_LEN
+ LLC_HEADER_LEN
)) {
554 stp_received_bpdu(sp
, payload
.data
, payload
.size
);
558 static struct xport
*
559 get_ofp_port(const struct xbridge
*xbridge
, ofp_port_t ofp_port
)
563 HMAP_FOR_EACH_IN_BUCKET (xport
, ofp_node
, hash_ofp_port(ofp_port
),
565 if (xport
->ofp_port
== ofp_port
) {
573 ofp_port_to_odp_port(const struct xbridge
*xbridge
, ofp_port_t ofp_port
)
575 const struct xport
*xport
= get_ofp_port(xbridge
, ofp_port
);
576 return xport
? xport
->odp_port
: ODPP_NONE
;
580 xbundle_trunks_vlan(const struct xbundle
*bundle
, uint16_t vlan
)
582 return (bundle
->vlan_mode
!= PORT_VLAN_ACCESS
583 && (!bundle
->trunks
|| bitmap_is_set(bundle
->trunks
, vlan
)));
587 xbundle_includes_vlan(const struct xbundle
*xbundle
, uint16_t vlan
)
589 return vlan
== xbundle
->vlan
|| xbundle_trunks_vlan(xbundle
, vlan
);
593 xbundle_mirror_out(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
595 return xbundle
!= &ofpp_none_bundle
596 ? mirror_bundle_out(xbridge
->mbridge
, xbundle
->ofbundle
)
601 xbundle_mirror_src(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
603 return xbundle
!= &ofpp_none_bundle
604 ? mirror_bundle_src(xbridge
->mbridge
, xbundle
->ofbundle
)
609 xbundle_mirror_dst(const struct xbridge
*xbridge
, struct xbundle
*xbundle
)
611 return xbundle
!= &ofpp_none_bundle
612 ? mirror_bundle_dst(xbridge
->mbridge
, xbundle
->ofbundle
)
616 static struct xbundle
*
617 lookup_input_bundle(const struct xbridge
*xbridge
, ofp_port_t in_port
,
618 bool warn
, struct xport
**in_xportp
)
622 /* Find the port and bundle for the received packet. */
623 xport
= get_ofp_port(xbridge
, in_port
);
627 if (xport
&& xport
->xbundle
) {
628 return xport
->xbundle
;
631 /* Special-case OFPP_NONE, which a controller may use as the ingress
632 * port for traffic that it is sourcing. */
633 if (in_port
== OFPP_NONE
) {
634 ofpp_none_bundle
.name
= "OFPP_NONE";
635 ofpp_none_bundle
.vlan_mode
= PORT_VLAN_TRUNK
;
636 return &ofpp_none_bundle
;
639 /* Odd. A few possible reasons here:
641 * - We deleted a port but there are still a few packets queued up
644 * - Someone externally added a port (e.g. "ovs-dpctl add-if") that
645 * we don't know about.
647 * - The ofproto client didn't configure the port as part of a bundle.
648 * This is particularly likely to happen if a packet was received on the
649 * port after it was created, but before the client had a chance to
650 * configure its bundle.
653 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
655 VLOG_WARN_RL(&rl
, "bridge %s: received packet on unknown "
656 "port %"PRIu16
, xbridge
->name
, in_port
);
662 add_mirror_actions(struct xlate_ctx
*ctx
, const struct flow
*orig_flow
)
664 const struct xbridge
*xbridge
= ctx
->xbridge
;
665 mirror_mask_t mirrors
;
666 struct xbundle
*in_xbundle
;
670 mirrors
= ctx
->xout
->mirrors
;
671 ctx
->xout
->mirrors
= 0;
673 in_xbundle
= lookup_input_bundle(xbridge
, orig_flow
->in_port
.ofp_port
,
674 ctx
->xin
->packet
!= NULL
, NULL
);
678 mirrors
|= xbundle_mirror_src(xbridge
, in_xbundle
);
680 /* Drop frames on bundles reserved for mirroring. */
681 if (xbundle_mirror_out(xbridge
, in_xbundle
)) {
682 if (ctx
->xin
->packet
!= NULL
) {
683 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
684 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet received on port "
685 "%s, which is reserved exclusively for mirroring",
686 ctx
->xbridge
->name
, in_xbundle
->name
);
688 ofpbuf_clear(&ctx
->xout
->odp_actions
);
693 vid
= vlan_tci_to_vid(orig_flow
->vlan_tci
);
694 if (!input_vid_is_valid(vid
, in_xbundle
, ctx
->xin
->packet
!= NULL
)) {
697 vlan
= input_vid_to_vlan(in_xbundle
, vid
);
703 /* Restore the original packet before adding the mirror actions. */
704 ctx
->xin
->flow
= *orig_flow
;
707 mirror_mask_t dup_mirrors
;
708 struct ofbundle
*out
;
709 unsigned long *vlans
;
714 has_mirror
= mirror_get(xbridge
->mbridge
, mirror_mask_ffs(mirrors
) - 1,
715 &vlans
, &dup_mirrors
, &out
, &out_vlan
);
716 ovs_assert(has_mirror
);
719 ctx
->xout
->wc
.masks
.vlan_tci
|= htons(VLAN_CFI
| VLAN_VID_MASK
);
721 vlan_mirrored
= !vlans
|| bitmap_is_set(vlans
, vlan
);
724 if (!vlan_mirrored
) {
725 mirrors
= zero_rightmost_1bit(mirrors
);
729 mirrors
&= ~dup_mirrors
;
730 ctx
->xout
->mirrors
|= dup_mirrors
;
732 struct xbundle
*out_xbundle
= xbundle_lookup(out
);
734 output_normal(ctx
, out_xbundle
, vlan
);
736 } else if (vlan
!= out_vlan
737 && !eth_addr_is_reserved(orig_flow
->dl_dst
)) {
738 struct xbundle
*xbundle
;
740 LIST_FOR_EACH (xbundle
, list_node
, &xbridge
->xbundles
) {
741 if (xbundle_includes_vlan(xbundle
, out_vlan
)
742 && !xbundle_mirror_out(xbridge
, xbundle
)) {
743 output_normal(ctx
, xbundle
, out_vlan
);
750 /* Given 'vid', the VID obtained from the 802.1Q header that was received as
751 * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle',
752 * the bundle on which the packet was received, returns the VLAN to which the
755 * Both 'vid' and the return value are in the range 0...4095. */
757 input_vid_to_vlan(const struct xbundle
*in_xbundle
, uint16_t vid
)
759 switch (in_xbundle
->vlan_mode
) {
760 case PORT_VLAN_ACCESS
:
761 return in_xbundle
->vlan
;
764 case PORT_VLAN_TRUNK
:
767 case PORT_VLAN_NATIVE_UNTAGGED
:
768 case PORT_VLAN_NATIVE_TAGGED
:
769 return vid
? vid
: in_xbundle
->vlan
;
776 /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'.
777 * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs
780 * 'vid' should be the VID obtained from the 802.1Q header that was received as
781 * part of a packet (specify 0 if there was no 802.1Q header), in the range
784 input_vid_is_valid(uint16_t vid
, struct xbundle
*in_xbundle
, bool warn
)
786 /* Allow any VID on the OFPP_NONE port. */
787 if (in_xbundle
== &ofpp_none_bundle
) {
791 switch (in_xbundle
->vlan_mode
) {
792 case PORT_VLAN_ACCESS
:
795 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
796 VLOG_WARN_RL(&rl
, "dropping VLAN %"PRIu16
" tagged "
797 "packet received on port %s configured as VLAN "
798 "%"PRIu16
" access port", vid
, in_xbundle
->name
,
805 case PORT_VLAN_NATIVE_UNTAGGED
:
806 case PORT_VLAN_NATIVE_TAGGED
:
808 /* Port must always carry its native VLAN. */
812 case PORT_VLAN_TRUNK
:
813 if (!xbundle_includes_vlan(in_xbundle
, vid
)) {
815 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
816 VLOG_WARN_RL(&rl
, "dropping VLAN %"PRIu16
" packet "
817 "received on port %s not configured for trunking "
818 "VLAN %"PRIu16
, vid
, in_xbundle
->name
, vid
);
830 /* Given 'vlan', the VLAN that a packet belongs to, and
831 * 'out_xbundle', a bundle on which the packet is to be output, returns the VID
832 * that should be included in the 802.1Q header. (If the return value is 0,
833 * then the 802.1Q header should only be included in the packet if there is a
836 * Both 'vlan' and the return value are in the range 0...4095. */
838 output_vlan_to_vid(const struct xbundle
*out_xbundle
, uint16_t vlan
)
840 switch (out_xbundle
->vlan_mode
) {
841 case PORT_VLAN_ACCESS
:
844 case PORT_VLAN_TRUNK
:
845 case PORT_VLAN_NATIVE_TAGGED
:
848 case PORT_VLAN_NATIVE_UNTAGGED
:
849 return vlan
== out_xbundle
->vlan
? 0 : vlan
;
857 output_normal(struct xlate_ctx
*ctx
, const struct xbundle
*out_xbundle
,
860 ovs_be16
*flow_tci
= &ctx
->xin
->flow
.vlan_tci
;
862 ovs_be16 tci
, old_tci
;
865 vid
= output_vlan_to_vid(out_xbundle
, vlan
);
866 if (list_is_empty(&out_xbundle
->xports
)) {
867 /* Partially configured bundle with no slaves. Drop the packet. */
869 } else if (!out_xbundle
->bond
) {
870 xport
= CONTAINER_OF(list_front(&out_xbundle
->xports
), struct xport
,
873 struct ofport_dpif
*ofport
;
875 ofport
= bond_choose_output_slave(out_xbundle
->bond
, &ctx
->xin
->flow
,
876 &ctx
->xout
->wc
, vid
);
877 xport
= ofport
? xport_lookup(ofport
) : NULL
;
880 /* No slaves enabled, so drop packet. */
887 if (tci
|| out_xbundle
->use_priority_tags
) {
888 tci
|= *flow_tci
& htons(VLAN_PCP_MASK
);
890 tci
|= htons(VLAN_CFI
);
895 compose_output_action(ctx
, xport
->ofp_port
);
899 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
900 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
901 * indicate this; newer upstream kernels use gratuitous ARP requests. */
903 is_gratuitous_arp(const struct flow
*flow
, struct flow_wildcards
*wc
)
905 if (flow
->dl_type
!= htons(ETH_TYPE_ARP
)) {
909 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
910 if (!eth_addr_is_broadcast(flow
->dl_dst
)) {
914 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
915 if (flow
->nw_proto
== ARP_OP_REPLY
) {
917 } else if (flow
->nw_proto
== ARP_OP_REQUEST
) {
918 memset(&wc
->masks
.nw_src
, 0xff, sizeof wc
->masks
.nw_src
);
919 memset(&wc
->masks
.nw_dst
, 0xff, sizeof wc
->masks
.nw_dst
);
921 return flow
->nw_src
== flow
->nw_dst
;
928 update_learning_table(const struct xbridge
*xbridge
,
929 const struct flow
*flow
, struct flow_wildcards
*wc
,
930 int vlan
, struct xbundle
*in_xbundle
)
932 struct mac_entry
*mac
;
934 /* Don't learn the OFPP_NONE port. */
935 if (in_xbundle
== &ofpp_none_bundle
) {
939 ovs_rwlock_wrlock(&xbridge
->ml
->rwlock
);
940 if (!mac_learning_may_learn(xbridge
->ml
, flow
->dl_src
, vlan
)) {
944 mac
= mac_learning_insert(xbridge
->ml
, flow
->dl_src
, vlan
);
945 if (is_gratuitous_arp(flow
, wc
)) {
946 /* We don't want to learn from gratuitous ARP packets that are
947 * reflected back over bond slaves so we lock the learning table. */
948 if (!in_xbundle
->bond
) {
949 mac_entry_set_grat_arp_lock(mac
);
950 } else if (mac_entry_is_grat_arp_locked(mac
)) {
955 if (mac
->port
.p
!= in_xbundle
->ofbundle
) {
956 /* The log messages here could actually be useful in debugging,
957 * so keep the rate limit relatively high. */
958 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(30, 300);
959 VLOG_DBG_RL(&rl
, "bridge %s: learned that "ETH_ADDR_FMT
" is "
960 "on port %s in VLAN %d",
961 xbridge
->name
, ETH_ADDR_ARGS(flow
->dl_src
),
962 in_xbundle
->name
, vlan
);
964 mac
->port
.p
= in_xbundle
->ofbundle
;
965 mac_learning_changed(xbridge
->ml
);
968 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
971 /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or
972 * dropped. Returns true if they may be forwarded, false if they should be
975 * 'in_port' must be the xport that corresponds to flow->in_port.
976 * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull).
978 * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as
979 * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as
980 * checked by input_vid_is_valid().
982 * May also add tags to '*tags', although the current implementation only does
983 * so in one special case.
986 is_admissible(struct xlate_ctx
*ctx
, struct xport
*in_port
,
989 struct xbundle
*in_xbundle
= in_port
->xbundle
;
990 const struct xbridge
*xbridge
= ctx
->xbridge
;
991 struct flow
*flow
= &ctx
->xin
->flow
;
993 /* Drop frames for reserved multicast addresses
994 * only if forward_bpdu option is absent. */
995 if (!xbridge
->forward_bpdu
&& eth_addr_is_reserved(flow
->dl_dst
)) {
996 xlate_report(ctx
, "packet has reserved destination MAC, dropping");
1000 if (in_xbundle
->bond
) {
1001 struct mac_entry
*mac
;
1003 switch (bond_check_admissibility(in_xbundle
->bond
, in_port
->ofport
,
1009 xlate_report(ctx
, "bonding refused admissibility, dropping");
1012 case BV_DROP_IF_MOVED
:
1013 ovs_rwlock_rdlock(&xbridge
->ml
->rwlock
);
1014 mac
= mac_learning_lookup(xbridge
->ml
, flow
->dl_src
, vlan
);
1015 if (mac
&& mac
->port
.p
!= in_xbundle
->ofbundle
&&
1016 (!is_gratuitous_arp(flow
, &ctx
->xout
->wc
)
1017 || mac_entry_is_grat_arp_locked(mac
))) {
1018 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
1019 xlate_report(ctx
, "SLB bond thinks this packet looped back, "
1023 ovs_rwlock_unlock(&xbridge
->ml
->rwlock
);
1032 xlate_normal(struct xlate_ctx
*ctx
)
1034 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
1035 struct flow
*flow
= &ctx
->xin
->flow
;
1036 struct xbundle
*in_xbundle
;
1037 struct xport
*in_port
;
1038 struct mac_entry
*mac
;
1042 ctx
->xout
->has_normal
= true;
1044 memset(&wc
->masks
.dl_src
, 0xff, sizeof wc
->masks
.dl_src
);
1045 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
1046 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
1048 in_xbundle
= lookup_input_bundle(ctx
->xbridge
, flow
->in_port
.ofp_port
,
1049 ctx
->xin
->packet
!= NULL
, &in_port
);
1051 xlate_report(ctx
, "no input bundle, dropping");
1055 /* Drop malformed frames. */
1056 if (flow
->dl_type
== htons(ETH_TYPE_VLAN
) &&
1057 !(flow
->vlan_tci
& htons(VLAN_CFI
))) {
1058 if (ctx
->xin
->packet
!= NULL
) {
1059 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1060 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet with partial "
1061 "VLAN tag received on port %s",
1062 ctx
->xbridge
->name
, in_xbundle
->name
);
1064 xlate_report(ctx
, "partial VLAN tag, dropping");
1068 /* Drop frames on bundles reserved for mirroring. */
1069 if (xbundle_mirror_out(ctx
->xbridge
, in_xbundle
)) {
1070 if (ctx
->xin
->packet
!= NULL
) {
1071 static struct vlog_rate_limit rl
= VLOG_RATE_LIMIT_INIT(1, 5);
1072 VLOG_WARN_RL(&rl
, "bridge %s: dropping packet received on port "
1073 "%s, which is reserved exclusively for mirroring",
1074 ctx
->xbridge
->name
, in_xbundle
->name
);
1076 xlate_report(ctx
, "input port is mirror output port, dropping");
1081 vid
= vlan_tci_to_vid(flow
->vlan_tci
);
1082 if (!input_vid_is_valid(vid
, in_xbundle
, ctx
->xin
->packet
!= NULL
)) {
1083 xlate_report(ctx
, "disallowed VLAN VID for this input port, dropping");
1086 vlan
= input_vid_to_vlan(in_xbundle
, vid
);
1088 /* Check other admissibility requirements. */
1089 if (in_port
&& !is_admissible(ctx
, in_port
, vlan
)) {
1093 /* Learn source MAC. */
1094 if (ctx
->xin
->may_learn
) {
1095 update_learning_table(ctx
->xbridge
, flow
, wc
, vlan
, in_xbundle
);
1098 /* Determine output bundle. */
1099 ovs_rwlock_rdlock(&ctx
->xbridge
->ml
->rwlock
);
1100 mac
= mac_learning_lookup(ctx
->xbridge
->ml
, flow
->dl_dst
, vlan
);
1102 struct xbundle
*mac_xbundle
= xbundle_lookup(mac
->port
.p
);
1103 if (mac_xbundle
&& mac_xbundle
!= in_xbundle
) {
1104 xlate_report(ctx
, "forwarding to learned port");
1105 output_normal(ctx
, mac_xbundle
, vlan
);
1106 } else if (!mac_xbundle
) {
1107 xlate_report(ctx
, "learned port is unknown, dropping");
1109 xlate_report(ctx
, "learned port is input port, dropping");
1112 struct xbundle
*xbundle
;
1114 xlate_report(ctx
, "no learned MAC for destination, flooding");
1115 LIST_FOR_EACH (xbundle
, list_node
, &ctx
->xbridge
->xbundles
) {
1116 if (xbundle
!= in_xbundle
1117 && xbundle_includes_vlan(xbundle
, vlan
)
1118 && xbundle
->floodable
1119 && !xbundle_mirror_out(ctx
->xbridge
, xbundle
)) {
1120 output_normal(ctx
, xbundle
, vlan
);
1123 ctx
->xout
->nf_output_iface
= NF_OUT_FLOOD
;
1125 ovs_rwlock_unlock(&ctx
->xbridge
->ml
->rwlock
);
1128 /* Compose SAMPLE action for sFlow or IPFIX. The given probability is
1129 * the number of packets out of UINT32_MAX to sample. The given
1130 * cookie is passed back in the callback for each sampled packet.
1133 compose_sample_action(const struct xbridge
*xbridge
,
1134 struct ofpbuf
*odp_actions
,
1135 const struct flow
*flow
,
1136 const uint32_t probability
,
1137 const union user_action_cookie
*cookie
,
1138 const size_t cookie_size
)
1140 size_t sample_offset
, actions_offset
;
1143 sample_offset
= nl_msg_start_nested(odp_actions
, OVS_ACTION_ATTR_SAMPLE
);
1145 nl_msg_put_u32(odp_actions
, OVS_SAMPLE_ATTR_PROBABILITY
, probability
);
1147 actions_offset
= nl_msg_start_nested(odp_actions
, OVS_SAMPLE_ATTR_ACTIONS
);
1148 cookie_offset
= put_userspace_action(xbridge
->ofproto
, odp_actions
, flow
,
1149 cookie
, cookie_size
);
1151 nl_msg_end_nested(odp_actions
, actions_offset
);
1152 nl_msg_end_nested(odp_actions
, sample_offset
);
1153 return cookie_offset
;
1157 compose_sflow_cookie(const struct xbridge
*xbridge
, ovs_be16 vlan_tci
,
1158 odp_port_t odp_port
, unsigned int n_outputs
,
1159 union user_action_cookie
*cookie
)
1163 cookie
->type
= USER_ACTION_COOKIE_SFLOW
;
1164 cookie
->sflow
.vlan_tci
= vlan_tci
;
1166 /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output
1167 * port information") for the interpretation of cookie->output. */
1168 switch (n_outputs
) {
1170 /* 0x40000000 | 256 means "packet dropped for unknown reason". */
1171 cookie
->sflow
.output
= 0x40000000 | 256;
1175 ifindex
= dpif_sflow_odp_port_to_ifindex(xbridge
->sflow
, odp_port
);
1177 cookie
->sflow
.output
= ifindex
;
1182 /* 0x80000000 means "multiple output ports. */
1183 cookie
->sflow
.output
= 0x80000000 | n_outputs
;
1188 /* Compose SAMPLE action for sFlow bridge sampling. */
1190 compose_sflow_action(const struct xbridge
*xbridge
,
1191 struct ofpbuf
*odp_actions
,
1192 const struct flow
*flow
,
1193 odp_port_t odp_port
)
1195 uint32_t probability
;
1196 union user_action_cookie cookie
;
1198 if (!xbridge
->sflow
|| flow
->in_port
.ofp_port
== OFPP_NONE
) {
1202 probability
= dpif_sflow_get_probability(xbridge
->sflow
);
1203 compose_sflow_cookie(xbridge
, htons(0), odp_port
,
1204 odp_port
== ODPP_NONE
? 0 : 1, &cookie
);
1206 return compose_sample_action(xbridge
, odp_actions
, flow
, probability
,
1207 &cookie
, sizeof cookie
.sflow
);
1211 compose_flow_sample_cookie(uint16_t probability
, uint32_t collector_set_id
,
1212 uint32_t obs_domain_id
, uint32_t obs_point_id
,
1213 union user_action_cookie
*cookie
)
1215 cookie
->type
= USER_ACTION_COOKIE_FLOW_SAMPLE
;
1216 cookie
->flow_sample
.probability
= probability
;
1217 cookie
->flow_sample
.collector_set_id
= collector_set_id
;
1218 cookie
->flow_sample
.obs_domain_id
= obs_domain_id
;
1219 cookie
->flow_sample
.obs_point_id
= obs_point_id
;
1223 compose_ipfix_cookie(union user_action_cookie
*cookie
)
1225 cookie
->type
= USER_ACTION_COOKIE_IPFIX
;
1228 /* Compose SAMPLE action for IPFIX bridge sampling. */
1230 compose_ipfix_action(const struct xbridge
*xbridge
,
1231 struct ofpbuf
*odp_actions
,
1232 const struct flow
*flow
)
1234 uint32_t probability
;
1235 union user_action_cookie cookie
;
1237 if (!xbridge
->ipfix
|| flow
->in_port
.ofp_port
== OFPP_NONE
) {
1241 probability
= dpif_ipfix_get_bridge_exporter_probability(xbridge
->ipfix
);
1242 compose_ipfix_cookie(&cookie
);
1244 compose_sample_action(xbridge
, odp_actions
, flow
, probability
,
1245 &cookie
, sizeof cookie
.ipfix
);
1248 /* SAMPLE action for sFlow must be first action in any given list of
1249 * actions. At this point we do not have all information required to
1250 * build it. So try to build sample action as complete as possible. */
1252 add_sflow_action(struct xlate_ctx
*ctx
)
1254 ctx
->user_cookie_offset
= compose_sflow_action(ctx
->xbridge
,
1255 &ctx
->xout
->odp_actions
,
1256 &ctx
->xin
->flow
, ODPP_NONE
);
1257 ctx
->sflow_odp_port
= 0;
1258 ctx
->sflow_n_outputs
= 0;
1261 /* SAMPLE action for IPFIX must be 1st or 2nd action in any given list
1262 * of actions, eventually after the SAMPLE action for sFlow. */
1264 add_ipfix_action(struct xlate_ctx
*ctx
)
1266 compose_ipfix_action(ctx
->xbridge
, &ctx
->xout
->odp_actions
,
1270 /* Fix SAMPLE action according to data collected while composing ODP actions.
1271 * We need to fix SAMPLE actions OVS_SAMPLE_ATTR_ACTIONS attribute, i.e. nested
1272 * USERSPACE action's user-cookie which is required for sflow. */
1274 fix_sflow_action(struct xlate_ctx
*ctx
)
1276 const struct flow
*base
= &ctx
->base_flow
;
1277 union user_action_cookie
*cookie
;
1279 if (!ctx
->user_cookie_offset
) {
1283 cookie
= ofpbuf_at(&ctx
->xout
->odp_actions
, ctx
->user_cookie_offset
,
1284 sizeof cookie
->sflow
);
1285 ovs_assert(cookie
->type
== USER_ACTION_COOKIE_SFLOW
);
1287 compose_sflow_cookie(ctx
->xbridge
, base
->vlan_tci
,
1288 ctx
->sflow_odp_port
, ctx
->sflow_n_outputs
, cookie
);
1291 static enum slow_path_reason
1292 process_special(struct xlate_ctx
*ctx
, const struct flow
*flow
,
1293 const struct xport
*xport
, const struct ofpbuf
*packet
)
1295 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
1296 const struct xbridge
*xbridge
= ctx
->xbridge
;
1300 } else if (xport
->cfm
&& cfm_should_process_flow(xport
->cfm
, flow
, wc
)) {
1302 cfm_process_heartbeat(xport
->cfm
, packet
);
1305 } else if (xport
->bfd
&& bfd_should_process_flow(xport
->bfd
, flow
, wc
)) {
1307 bfd_process_packet(xport
->bfd
, flow
, packet
);
1310 } else if (xport
->xbundle
&& xport
->xbundle
->lacp
1311 && flow
->dl_type
== htons(ETH_TYPE_LACP
)) {
1313 lacp_process_packet(xport
->xbundle
->lacp
, xport
->ofport
, packet
);
1316 } else if (xbridge
->stp
&& stp_should_process_flow(flow
, wc
)) {
1318 stp_process_packet(xport
, packet
);
1327 compose_output_action__(struct xlate_ctx
*ctx
, ofp_port_t ofp_port
,
1330 const struct xport
*xport
= get_ofp_port(ctx
->xbridge
, ofp_port
);
1331 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
1332 struct flow
*flow
= &ctx
->xin
->flow
;
1333 ovs_be16 flow_vlan_tci
;
1334 uint32_t flow_skb_mark
;
1335 uint8_t flow_nw_tos
;
1336 odp_port_t out_port
, odp_port
;
1339 /* If 'struct flow' gets additional metadata, we'll need to zero it out
1340 * before traversing a patch port. */
1341 BUILD_ASSERT_DECL(FLOW_WC_SEQ
== 20);
1344 xlate_report(ctx
, "Nonexistent output port");
1346 } else if (xport
->config
& OFPUTIL_PC_NO_FWD
) {
1347 xlate_report(ctx
, "OFPPC_NO_FWD set, skipping output");
1349 } else if (check_stp
&& !xport_stp_forward_state(xport
)) {
1350 xlate_report(ctx
, "STP not in forwarding state, skipping output");
1354 if (mbridge_has_mirrors(ctx
->xbridge
->mbridge
) && xport
->xbundle
) {
1355 ctx
->xout
->mirrors
|= xbundle_mirror_dst(xport
->xbundle
->xbridge
,
1360 const struct xport
*peer
= xport
->peer
;
1361 struct flow old_flow
= ctx
->xin
->flow
;
1362 enum slow_path_reason special
;
1364 ctx
->xbridge
= peer
->xbridge
;
1365 flow
->in_port
.ofp_port
= peer
->ofp_port
;
1366 flow
->metadata
= htonll(0);
1367 memset(&flow
->tunnel
, 0, sizeof flow
->tunnel
);
1368 memset(flow
->regs
, 0, sizeof flow
->regs
);
1370 special
= process_special(ctx
, &ctx
->xin
->flow
, peer
,
1373 ctx
->xout
->slow
= special
;
1374 } else if (may_receive(peer
, ctx
)) {
1375 if (xport_stp_forward_state(peer
)) {
1376 xlate_table_action(ctx
, flow
->in_port
.ofp_port
, 0, true);
1378 /* Forwarding is disabled by STP. Let OFPP_NORMAL and the
1379 * learning action look at the packet, then drop it. */
1380 struct flow old_base_flow
= ctx
->base_flow
;
1381 size_t old_size
= ctx
->xout
->odp_actions
.size
;
1382 mirror_mask_t old_mirrors
= ctx
->xout
->mirrors
;
1383 xlate_table_action(ctx
, flow
->in_port
.ofp_port
, 0, true);
1384 ctx
->xout
->mirrors
= old_mirrors
;
1385 ctx
->base_flow
= old_base_flow
;
1386 ctx
->xout
->odp_actions
.size
= old_size
;
1390 ctx
->xin
->flow
= old_flow
;
1391 ctx
->xbridge
= xport
->xbundle
->xbridge
;
1393 if (ctx
->xin
->resubmit_stats
) {
1394 netdev_vport_inc_tx(xport
->netdev
, ctx
->xin
->resubmit_stats
);
1395 netdev_vport_inc_rx(peer
->netdev
, ctx
->xin
->resubmit_stats
);
1401 flow_vlan_tci
= flow
->vlan_tci
;
1402 flow_skb_mark
= flow
->skb_mark
;
1403 flow_nw_tos
= flow
->nw_tos
;
1405 if (dscp_from_skb_priority(xport
, flow
->skb_priority
, &dscp
)) {
1406 wc
->masks
.nw_tos
|= IP_ECN_MASK
;
1407 flow
->nw_tos
&= ~IP_DSCP_MASK
;
1408 flow
->nw_tos
|= dscp
;
1411 if (xport
->is_tunnel
) {
1412 /* Save tunnel metadata so that changes made due to
1413 * the Logical (tunnel) Port are not visible for any further
1414 * matches, while explicit set actions on tunnel metadata are.
1416 struct flow_tnl flow_tnl
= flow
->tunnel
;
1417 odp_port
= tnl_port_send(xport
->ofport
, flow
, &ctx
->xout
->wc
);
1418 if (odp_port
== ODPP_NONE
) {
1419 xlate_report(ctx
, "Tunneling decided against output");
1420 goto out
; /* restore flow_nw_tos */
1422 if (flow
->tunnel
.ip_dst
== ctx
->orig_tunnel_ip_dst
) {
1423 xlate_report(ctx
, "Not tunneling to our own address");
1424 goto out
; /* restore flow_nw_tos */
1426 if (ctx
->xin
->resubmit_stats
) {
1427 netdev_vport_inc_tx(xport
->netdev
, ctx
->xin
->resubmit_stats
);
1429 out_port
= odp_port
;
1430 commit_odp_tunnel_action(flow
, &ctx
->base_flow
,
1431 &ctx
->xout
->odp_actions
);
1432 flow
->tunnel
= flow_tnl
; /* Restore tunnel metadata */
1434 ofp_port_t vlandev_port
;
1436 odp_port
= xport
->odp_port
;
1437 if (ofproto_has_vlan_splinters(ctx
->xbridge
->ofproto
)) {
1438 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
1440 vlandev_port
= vsp_realdev_to_vlandev(ctx
->xbridge
->ofproto
, ofp_port
,
1442 if (vlandev_port
== ofp_port
) {
1443 out_port
= odp_port
;
1445 out_port
= ofp_port_to_odp_port(ctx
->xbridge
, vlandev_port
);
1446 flow
->vlan_tci
= htons(0);
1448 flow
->skb_mark
&= ~IPSEC_MARK
;
1451 if (out_port
!= ODPP_NONE
) {
1452 commit_odp_actions(flow
, &ctx
->base_flow
,
1453 &ctx
->xout
->odp_actions
, &ctx
->xout
->wc
);
1454 nl_msg_put_odp_port(&ctx
->xout
->odp_actions
, OVS_ACTION_ATTR_OUTPUT
,
1457 ctx
->sflow_odp_port
= odp_port
;
1458 ctx
->sflow_n_outputs
++;
1459 ctx
->xout
->nf_output_iface
= ofp_port
;
1464 flow
->vlan_tci
= flow_vlan_tci
;
1465 flow
->skb_mark
= flow_skb_mark
;
1466 flow
->nw_tos
= flow_nw_tos
;
1470 compose_output_action(struct xlate_ctx
*ctx
, ofp_port_t ofp_port
)
1472 compose_output_action__(ctx
, ofp_port
, true);
1475 /* Common rule processing in one place to avoid duplicating code. */
1476 static struct rule_dpif
*
1477 ctx_rule_hooks(struct xlate_ctx
*ctx
, struct rule_dpif
*rule
,
1480 if (ctx
->xin
->resubmit_hook
) {
1481 ctx
->xin
->resubmit_hook(ctx
->xin
, rule
, ctx
->recurse
);
1483 if (rule
== NULL
&& may_packet_in
) {
1485 * check if table configuration flags
1486 * OFPTC_TABLE_MISS_CONTROLLER, default.
1487 * OFPTC_TABLE_MISS_CONTINUE,
1488 * OFPTC_TABLE_MISS_DROP
1489 * When OF1.0, OFPTC_TABLE_MISS_CONTINUE is used. What to do?
1491 rule
= rule_dpif_miss_rule(ctx
->xbridge
->ofproto
, &ctx
->xin
->flow
);
1493 if (rule
&& ctx
->xin
->resubmit_stats
) {
1494 rule_credit_stats(rule
, ctx
->xin
->resubmit_stats
);
1500 xlate_table_action(struct xlate_ctx
*ctx
,
1501 ofp_port_t in_port
, uint8_t table_id
, bool may_packet_in
)
1503 if (ctx
->recurse
< MAX_RESUBMIT_RECURSION
) {
1504 struct rule_dpif
*rule
;
1505 ofp_port_t old_in_port
= ctx
->xin
->flow
.in_port
.ofp_port
;
1506 uint8_t old_table_id
= ctx
->table_id
;
1508 ctx
->table_id
= table_id
;
1510 /* Look up a flow with 'in_port' as the input port. */
1511 ctx
->xin
->flow
.in_port
.ofp_port
= in_port
;
1512 rule
= rule_dpif_lookup_in_table(ctx
->xbridge
->ofproto
,
1513 &ctx
->xin
->flow
, &ctx
->xout
->wc
,
1516 /* Restore the original input port. Otherwise OFPP_NORMAL and
1517 * OFPP_IN_PORT will have surprising behavior. */
1518 ctx
->xin
->flow
.in_port
.ofp_port
= old_in_port
;
1520 rule
= ctx_rule_hooks(ctx
, rule
, may_packet_in
);
1523 struct rule_dpif
*old_rule
= ctx
->rule
;
1527 do_xlate_actions(rule
->up
.ofpacts
, rule
->up
.ofpacts_len
, ctx
);
1528 ctx
->rule
= old_rule
;
1532 ctx
->table_id
= old_table_id
;
1534 static struct vlog_rate_limit recurse_rl
= VLOG_RATE_LIMIT_INIT(1, 1);
1536 VLOG_ERR_RL(&recurse_rl
, "resubmit actions recursed over %d times",
1537 MAX_RESUBMIT_RECURSION
);
1538 ctx
->max_resubmit_trigger
= true;
1543 xlate_ofpact_resubmit(struct xlate_ctx
*ctx
,
1544 const struct ofpact_resubmit
*resubmit
)
1549 in_port
= resubmit
->in_port
;
1550 if (in_port
== OFPP_IN_PORT
) {
1551 in_port
= ctx
->xin
->flow
.in_port
.ofp_port
;
1554 table_id
= resubmit
->table_id
;
1555 if (table_id
== 255) {
1556 table_id
= ctx
->table_id
;
1559 xlate_table_action(ctx
, in_port
, table_id
, false);
1563 flood_packets(struct xlate_ctx
*ctx
, bool all
)
1565 const struct xport
*xport
;
1567 HMAP_FOR_EACH (xport
, ofp_node
, &ctx
->xbridge
->xports
) {
1568 if (xport
->ofp_port
== ctx
->xin
->flow
.in_port
.ofp_port
) {
1573 compose_output_action__(ctx
, xport
->ofp_port
, false);
1574 } else if (!(xport
->config
& OFPUTIL_PC_NO_FLOOD
)) {
1575 compose_output_action(ctx
, xport
->ofp_port
);
1579 ctx
->xout
->nf_output_iface
= NF_OUT_FLOOD
;
1583 execute_controller_action(struct xlate_ctx
*ctx
, int len
,
1584 enum ofp_packet_in_reason reason
,
1585 uint16_t controller_id
)
1587 struct ofputil_packet_in pin
;
1588 struct ofpbuf
*packet
;
1591 ovs_assert(!ctx
->xout
->slow
|| ctx
->xout
->slow
== SLOW_CONTROLLER
);
1592 ctx
->xout
->slow
= SLOW_CONTROLLER
;
1593 if (!ctx
->xin
->packet
) {
1597 packet
= ofpbuf_clone(ctx
->xin
->packet
);
1599 key
.skb_priority
= 0;
1601 memset(&key
.tunnel
, 0, sizeof key
.tunnel
);
1603 commit_odp_actions(&ctx
->xin
->flow
, &ctx
->base_flow
,
1604 &ctx
->xout
->odp_actions
, &ctx
->xout
->wc
);
1606 odp_execute_actions(NULL
, packet
, &key
, ctx
->xout
->odp_actions
.data
,
1607 ctx
->xout
->odp_actions
.size
, NULL
, NULL
);
1609 pin
.packet
= packet
->data
;
1610 pin
.packet_len
= packet
->size
;
1611 pin
.reason
= reason
;
1612 pin
.controller_id
= controller_id
;
1613 pin
.table_id
= ctx
->table_id
;
1614 pin
.cookie
= ctx
->rule
? ctx
->rule
->up
.flow_cookie
: 0;
1617 flow_get_metadata(&ctx
->xin
->flow
, &pin
.fmd
);
1619 ofproto_dpif_send_packet_in(ctx
->xbridge
->ofproto
, &pin
);
1620 ofpbuf_delete(packet
);
1624 compose_mpls_push_action(struct xlate_ctx
*ctx
, ovs_be16 eth_type
)
1626 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
1627 struct flow
*flow
= &ctx
->xin
->flow
;
1629 ovs_assert(eth_type_mpls(eth_type
));
1631 memset(&wc
->masks
.mpls_lse
, 0xff, sizeof wc
->masks
.mpls_lse
);
1632 memset(&wc
->masks
.mpls_depth
, 0xff, sizeof wc
->masks
.mpls_depth
);
1634 if (flow
->mpls_depth
) {
1635 flow
->mpls_lse
&= ~htonl(MPLS_BOS_MASK
);
1641 if (flow
->dl_type
== htons(ETH_TYPE_IPV6
)) {
1642 label
= htonl(0x2); /* IPV6 Explicit Null. */
1644 label
= htonl(0x0); /* IPV4 Explicit Null. */
1646 wc
->masks
.nw_tos
|= IP_DSCP_MASK
;
1647 wc
->masks
.nw_ttl
= 0xff;
1648 tc
= (flow
->nw_tos
& IP_DSCP_MASK
) >> 2;
1649 ttl
= flow
->nw_ttl
? flow
->nw_ttl
: 0x40;
1650 flow
->mpls_lse
= set_mpls_lse_values(ttl
, tc
, 1, label
);
1651 flow
->mpls_depth
= 1;
1653 flow
->dl_type
= eth_type
;
1657 compose_mpls_pop_action(struct xlate_ctx
*ctx
, ovs_be16 eth_type
)
1659 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
1660 struct flow
*flow
= &ctx
->xin
->flow
;
1662 ovs_assert(eth_type_mpls(ctx
->xin
->flow
.dl_type
));
1663 ovs_assert(!eth_type_mpls(eth_type
));
1665 memset(&wc
->masks
.mpls_lse
, 0xff, sizeof wc
->masks
.mpls_lse
);
1666 memset(&wc
->masks
.mpls_depth
, 0xff, sizeof wc
->masks
.mpls_depth
);
1668 if (flow
->mpls_depth
) {
1670 flow
->mpls_lse
= htonl(0);
1671 if (!flow
->mpls_depth
) {
1672 flow
->dl_type
= eth_type
;
1678 compose_dec_ttl(struct xlate_ctx
*ctx
, struct ofpact_cnt_ids
*ids
)
1680 struct flow
*flow
= &ctx
->xin
->flow
;
1682 if (!is_ip_any(flow
)) {
1686 ctx
->xout
->wc
.masks
.nw_ttl
= 0xff;
1687 if (flow
->nw_ttl
> 1) {
1693 for (i
= 0; i
< ids
->n_controllers
; i
++) {
1694 execute_controller_action(ctx
, UINT16_MAX
, OFPR_INVALID_TTL
,
1698 /* Stop processing for current table. */
1704 compose_set_mpls_ttl_action(struct xlate_ctx
*ctx
, uint8_t ttl
)
1706 if (!eth_type_mpls(ctx
->xin
->flow
.dl_type
)) {
1710 ctx
->xout
->wc
.masks
.mpls_lse
|= htonl(MPLS_TTL_MASK
);
1711 set_mpls_lse_ttl(&ctx
->xin
->flow
.mpls_lse
, ttl
);
1716 compose_dec_mpls_ttl_action(struct xlate_ctx
*ctx
)
1718 struct flow
*flow
= &ctx
->xin
->flow
;
1719 uint8_t ttl
= mpls_lse_to_ttl(flow
->mpls_lse
);
1720 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
1722 memset(&wc
->masks
.mpls_lse
, 0xff, sizeof wc
->masks
.mpls_lse
);
1724 if (!eth_type_mpls(flow
->dl_type
)) {
1730 set_mpls_lse_ttl(&flow
->mpls_lse
, ttl
);
1733 execute_controller_action(ctx
, UINT16_MAX
, OFPR_INVALID_TTL
, 0);
1735 /* Stop processing for current table. */
1741 xlate_output_action(struct xlate_ctx
*ctx
,
1742 ofp_port_t port
, uint16_t max_len
, bool may_packet_in
)
1744 ofp_port_t prev_nf_output_iface
= ctx
->xout
->nf_output_iface
;
1746 ctx
->xout
->nf_output_iface
= NF_OUT_DROP
;
1750 compose_output_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
);
1753 xlate_table_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
,
1760 flood_packets(ctx
, false);
1763 flood_packets(ctx
, true);
1765 case OFPP_CONTROLLER
:
1766 execute_controller_action(ctx
, max_len
, OFPR_ACTION
, 0);
1772 if (port
!= ctx
->xin
->flow
.in_port
.ofp_port
) {
1773 compose_output_action(ctx
, port
);
1775 xlate_report(ctx
, "skipping output to input port");
1780 if (prev_nf_output_iface
== NF_OUT_FLOOD
) {
1781 ctx
->xout
->nf_output_iface
= NF_OUT_FLOOD
;
1782 } else if (ctx
->xout
->nf_output_iface
== NF_OUT_DROP
) {
1783 ctx
->xout
->nf_output_iface
= prev_nf_output_iface
;
1784 } else if (prev_nf_output_iface
!= NF_OUT_DROP
&&
1785 ctx
->xout
->nf_output_iface
!= NF_OUT_FLOOD
) {
1786 ctx
->xout
->nf_output_iface
= NF_OUT_MULTI
;
1791 xlate_output_reg_action(struct xlate_ctx
*ctx
,
1792 const struct ofpact_output_reg
*or)
1794 uint64_t port
= mf_get_subfield(&or->src
, &ctx
->xin
->flow
);
1795 if (port
<= UINT16_MAX
) {
1796 union mf_subvalue value
;
1798 memset(&value
, 0xff, sizeof value
);
1799 mf_write_subfield_flow(&or->src
, &value
, &ctx
->xout
->wc
.masks
);
1800 xlate_output_action(ctx
, u16_to_ofp(port
),
1801 or->max_len
, false);
1806 xlate_enqueue_action(struct xlate_ctx
*ctx
,
1807 const struct ofpact_enqueue
*enqueue
)
1809 ofp_port_t ofp_port
= enqueue
->port
;
1810 uint32_t queue_id
= enqueue
->queue
;
1811 uint32_t flow_priority
, priority
;
1814 /* Translate queue to priority. */
1815 error
= ofproto_dpif_queue_to_priority(ctx
->xbridge
->ofproto
, queue_id
,
1818 /* Fall back to ordinary output action. */
1819 xlate_output_action(ctx
, enqueue
->port
, 0, false);
1823 /* Check output port. */
1824 if (ofp_port
== OFPP_IN_PORT
) {
1825 ofp_port
= ctx
->xin
->flow
.in_port
.ofp_port
;
1826 } else if (ofp_port
== ctx
->xin
->flow
.in_port
.ofp_port
) {
1830 /* Add datapath actions. */
1831 flow_priority
= ctx
->xin
->flow
.skb_priority
;
1832 ctx
->xin
->flow
.skb_priority
= priority
;
1833 compose_output_action(ctx
, ofp_port
);
1834 ctx
->xin
->flow
.skb_priority
= flow_priority
;
1836 /* Update NetFlow output port. */
1837 if (ctx
->xout
->nf_output_iface
== NF_OUT_DROP
) {
1838 ctx
->xout
->nf_output_iface
= ofp_port
;
1839 } else if (ctx
->xout
->nf_output_iface
!= NF_OUT_FLOOD
) {
1840 ctx
->xout
->nf_output_iface
= NF_OUT_MULTI
;
1845 xlate_set_queue_action(struct xlate_ctx
*ctx
, uint32_t queue_id
)
1847 uint32_t skb_priority
;
1849 if (!ofproto_dpif_queue_to_priority(ctx
->xbridge
->ofproto
, queue_id
,
1851 ctx
->xin
->flow
.skb_priority
= skb_priority
;
1853 /* Couldn't translate queue to a priority. Nothing to do. A warning
1854 * has already been logged. */
1859 slave_enabled_cb(ofp_port_t ofp_port
, void *xbridge_
)
1861 const struct xbridge
*xbridge
= xbridge_
;
1872 case OFPP_CONTROLLER
: /* Not supported by the bundle action. */
1875 port
= get_ofp_port(xbridge
, ofp_port
);
1876 return port
? port
->may_enable
: false;
1881 xlate_bundle_action(struct xlate_ctx
*ctx
,
1882 const struct ofpact_bundle
*bundle
)
1886 port
= bundle_execute(bundle
, &ctx
->xin
->flow
, &ctx
->xout
->wc
,
1888 CONST_CAST(struct xbridge
*, ctx
->xbridge
));
1889 if (bundle
->dst
.field
) {
1890 nxm_reg_load(&bundle
->dst
, ofp_to_u16(port
), &ctx
->xin
->flow
,
1893 xlate_output_action(ctx
, port
, 0, false);
1898 xlate_learn_action(struct xlate_ctx
*ctx
,
1899 const struct ofpact_learn
*learn
)
1901 struct ofputil_flow_mod
*fm
;
1902 struct ofpbuf ofpacts
;
1904 ctx
->xout
->has_learn
= true;
1906 learn_mask(learn
, &ctx
->xout
->wc
);
1908 if (!ctx
->xin
->may_learn
) {
1912 fm
= xmalloc(sizeof *fm
);
1913 ofpbuf_init(&ofpacts
, 0);
1914 learn_execute(learn
, &ctx
->xin
->flow
, fm
, &ofpacts
);
1916 ofproto_dpif_flow_mod(ctx
->xbridge
->ofproto
, fm
);
1919 /* Reduces '*timeout' to no more than 'max'. A value of zero in either case
1920 * means "infinite". */
1922 reduce_timeout(uint16_t max
, uint16_t *timeout
)
1924 if (max
&& (!*timeout
|| *timeout
> max
)) {
1930 xlate_fin_timeout(struct xlate_ctx
*ctx
,
1931 const struct ofpact_fin_timeout
*oft
)
1933 if (ctx
->xin
->tcp_flags
& (TCP_FIN
| TCP_RST
) && ctx
->rule
) {
1934 struct rule_dpif
*rule
= ctx
->rule
;
1936 if (list_is_empty(&rule
->up
.expirable
)) {
1937 list_insert(&rule
->up
.ofproto
->expirable
, &rule
->up
.expirable
);
1940 reduce_timeout(oft
->fin_idle_timeout
, &rule
->up
.idle_timeout
);
1941 reduce_timeout(oft
->fin_hard_timeout
, &rule
->up
.hard_timeout
);
1946 xlate_sample_action(struct xlate_ctx
*ctx
,
1947 const struct ofpact_sample
*os
)
1949 union user_action_cookie cookie
;
1950 /* Scale the probability from 16-bit to 32-bit while representing
1951 * the same percentage. */
1952 uint32_t probability
= (os
->probability
<< 16) | os
->probability
;
1954 commit_odp_actions(&ctx
->xin
->flow
, &ctx
->base_flow
,
1955 &ctx
->xout
->odp_actions
, &ctx
->xout
->wc
);
1957 compose_flow_sample_cookie(os
->probability
, os
->collector_set_id
,
1958 os
->obs_domain_id
, os
->obs_point_id
, &cookie
);
1959 compose_sample_action(ctx
->xbridge
, &ctx
->xout
->odp_actions
, &ctx
->xin
->flow
,
1960 probability
, &cookie
, sizeof cookie
.flow_sample
);
1964 may_receive(const struct xport
*xport
, struct xlate_ctx
*ctx
)
1966 if (xport
->config
& (eth_addr_equals(ctx
->xin
->flow
.dl_dst
, eth_addr_stp
)
1967 ? OFPUTIL_PC_NO_RECV_STP
1968 : OFPUTIL_PC_NO_RECV
)) {
1972 /* Only drop packets here if both forwarding and learning are
1973 * disabled. If just learning is enabled, we need to have
1974 * OFPP_NORMAL and the learning action have a look at the packet
1975 * before we can drop it. */
1976 if (!xport_stp_forward_state(xport
) && !xport_stp_learn_state(xport
)) {
1984 tunnel_ecn_ok(struct xlate_ctx
*ctx
)
1986 if (is_ip_any(&ctx
->base_flow
)
1987 && (ctx
->xin
->flow
.tunnel
.ip_tos
& IP_ECN_MASK
) == IP_ECN_CE
) {
1988 if ((ctx
->base_flow
.nw_tos
& IP_ECN_MASK
) == IP_ECN_NOT_ECT
) {
1989 VLOG_WARN_RL(&rl
, "dropping tunnel packet marked ECN CE"
1990 " but is not ECN capable");
1993 /* Set the ECN CE value in the tunneled packet. */
1994 ctx
->xin
->flow
.nw_tos
|= IP_ECN_CE
;
2002 do_xlate_actions(const struct ofpact
*ofpacts
, size_t ofpacts_len
,
2003 struct xlate_ctx
*ctx
)
2005 struct flow_wildcards
*wc
= &ctx
->xout
->wc
;
2006 struct flow
*flow
= &ctx
->xin
->flow
;
2007 bool was_evictable
= true;
2008 const struct ofpact
*a
;
2011 /* Don't let the rule we're working on get evicted underneath us. */
2012 was_evictable
= ctx
->rule
->up
.evictable
;
2013 ctx
->rule
->up
.evictable
= false;
2016 OFPACT_FOR_EACH (a
, ofpacts
, ofpacts_len
) {
2017 struct ofpact_controller
*controller
;
2018 const struct ofpact_metadata
*metadata
;
2026 xlate_output_action(ctx
, ofpact_get_OUTPUT(a
)->port
,
2027 ofpact_get_OUTPUT(a
)->max_len
, true);
2030 case OFPACT_CONTROLLER
:
2031 controller
= ofpact_get_CONTROLLER(a
);
2032 execute_controller_action(ctx
, controller
->max_len
,
2034 controller
->controller_id
);
2037 case OFPACT_ENQUEUE
:
2038 xlate_enqueue_action(ctx
, ofpact_get_ENQUEUE(a
));
2041 case OFPACT_SET_VLAN_VID
:
2042 wc
->masks
.vlan_tci
|= htons(VLAN_VID_MASK
| VLAN_CFI
);
2043 flow
->vlan_tci
&= ~htons(VLAN_VID_MASK
);
2044 flow
->vlan_tci
|= (htons(ofpact_get_SET_VLAN_VID(a
)->vlan_vid
)
2048 case OFPACT_SET_VLAN_PCP
:
2049 wc
->masks
.vlan_tci
|= htons(VLAN_PCP_MASK
| VLAN_CFI
);
2050 flow
->vlan_tci
&= ~htons(VLAN_PCP_MASK
);
2052 htons((ofpact_get_SET_VLAN_PCP(a
)->vlan_pcp
<< VLAN_PCP_SHIFT
)
2056 case OFPACT_STRIP_VLAN
:
2057 memset(&wc
->masks
.vlan_tci
, 0xff, sizeof wc
->masks
.vlan_tci
);
2058 flow
->vlan_tci
= htons(0);
2061 case OFPACT_PUSH_VLAN
:
2062 /* XXX 802.1AD(QinQ) */
2063 memset(&wc
->masks
.vlan_tci
, 0xff, sizeof wc
->masks
.vlan_tci
);
2064 flow
->vlan_tci
= htons(VLAN_CFI
);
2067 case OFPACT_SET_ETH_SRC
:
2068 memset(&wc
->masks
.dl_src
, 0xff, sizeof wc
->masks
.dl_src
);
2069 memcpy(flow
->dl_src
, ofpact_get_SET_ETH_SRC(a
)->mac
, ETH_ADDR_LEN
);
2072 case OFPACT_SET_ETH_DST
:
2073 memset(&wc
->masks
.dl_dst
, 0xff, sizeof wc
->masks
.dl_dst
);
2074 memcpy(flow
->dl_dst
, ofpact_get_SET_ETH_DST(a
)->mac
, ETH_ADDR_LEN
);
2077 case OFPACT_SET_IPV4_SRC
:
2078 memset(&wc
->masks
.nw_src
, 0xff, sizeof wc
->masks
.nw_src
);
2079 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
2080 flow
->nw_src
= ofpact_get_SET_IPV4_SRC(a
)->ipv4
;
2084 case OFPACT_SET_IPV4_DST
:
2085 memset(&wc
->masks
.nw_dst
, 0xff, sizeof wc
->masks
.nw_dst
);
2086 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
2087 flow
->nw_dst
= ofpact_get_SET_IPV4_DST(a
)->ipv4
;
2091 case OFPACT_SET_IPV4_DSCP
:
2092 wc
->masks
.nw_tos
|= IP_DSCP_MASK
;
2093 /* OpenFlow 1.0 only supports IPv4. */
2094 if (flow
->dl_type
== htons(ETH_TYPE_IP
)) {
2095 flow
->nw_tos
&= ~IP_DSCP_MASK
;
2096 flow
->nw_tos
|= ofpact_get_SET_IPV4_DSCP(a
)->dscp
;
2100 case OFPACT_SET_L4_SRC_PORT
:
2101 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
2102 memset(&wc
->masks
.tp_src
, 0xff, sizeof wc
->masks
.tp_src
);
2103 if (is_ip_any(flow
)) {
2104 flow
->tp_src
= htons(ofpact_get_SET_L4_SRC_PORT(a
)->port
);
2108 case OFPACT_SET_L4_DST_PORT
:
2109 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
2110 memset(&wc
->masks
.tp_dst
, 0xff, sizeof wc
->masks
.tp_dst
);
2111 if (is_ip_any(flow
)) {
2112 flow
->tp_dst
= htons(ofpact_get_SET_L4_DST_PORT(a
)->port
);
2116 case OFPACT_RESUBMIT
:
2117 xlate_ofpact_resubmit(ctx
, ofpact_get_RESUBMIT(a
));
2120 case OFPACT_SET_TUNNEL
:
2121 flow
->tunnel
.tun_id
= htonll(ofpact_get_SET_TUNNEL(a
)->tun_id
);
2124 case OFPACT_SET_QUEUE
:
2125 xlate_set_queue_action(ctx
, ofpact_get_SET_QUEUE(a
)->queue_id
);
2128 case OFPACT_POP_QUEUE
:
2129 flow
->skb_priority
= ctx
->orig_skb_priority
;
2132 case OFPACT_REG_MOVE
:
2133 nxm_execute_reg_move(ofpact_get_REG_MOVE(a
), flow
, wc
);
2136 case OFPACT_REG_LOAD
:
2137 nxm_execute_reg_load(ofpact_get_REG_LOAD(a
), flow
);
2140 case OFPACT_STACK_PUSH
:
2141 nxm_execute_stack_push(ofpact_get_STACK_PUSH(a
), flow
, wc
,
2145 case OFPACT_STACK_POP
:
2146 nxm_execute_stack_pop(ofpact_get_STACK_POP(a
), flow
, wc
,
2150 case OFPACT_PUSH_MPLS
:
2151 compose_mpls_push_action(ctx
, ofpact_get_PUSH_MPLS(a
)->ethertype
);
2154 case OFPACT_POP_MPLS
:
2155 compose_mpls_pop_action(ctx
, ofpact_get_POP_MPLS(a
)->ethertype
);
2158 case OFPACT_SET_MPLS_TTL
:
2159 if (compose_set_mpls_ttl_action(ctx
,
2160 ofpact_get_SET_MPLS_TTL(a
)->ttl
)) {
2165 case OFPACT_DEC_MPLS_TTL
:
2166 if (compose_dec_mpls_ttl_action(ctx
)) {
2171 case OFPACT_DEC_TTL
:
2172 wc
->masks
.nw_ttl
= 0xff;
2173 if (compose_dec_ttl(ctx
, ofpact_get_DEC_TTL(a
))) {
2179 /* Nothing to do. */
2182 case OFPACT_MULTIPATH
:
2183 multipath_execute(ofpact_get_MULTIPATH(a
), flow
, wc
);
2187 xlate_bundle_action(ctx
, ofpact_get_BUNDLE(a
));
2190 case OFPACT_OUTPUT_REG
:
2191 xlate_output_reg_action(ctx
, ofpact_get_OUTPUT_REG(a
));
2195 xlate_learn_action(ctx
, ofpact_get_LEARN(a
));
2202 case OFPACT_FIN_TIMEOUT
:
2203 memset(&wc
->masks
.nw_proto
, 0xff, sizeof wc
->masks
.nw_proto
);
2204 ctx
->xout
->has_fin_timeout
= true;
2205 xlate_fin_timeout(ctx
, ofpact_get_FIN_TIMEOUT(a
));
2208 case OFPACT_CLEAR_ACTIONS
:
2210 * Nothing to do because writa-actions is not supported for now.
2211 * When writa-actions is supported, clear-actions also must
2212 * be supported at the same time.
2216 case OFPACT_WRITE_METADATA
:
2217 metadata
= ofpact_get_WRITE_METADATA(a
);
2218 flow
->metadata
&= ~metadata
->mask
;
2219 flow
->metadata
|= metadata
->metadata
& metadata
->mask
;
2223 /* Not implemented yet. */
2226 case OFPACT_GOTO_TABLE
: {
2227 /* It is assumed that goto-table is the last action. */
2228 struct ofpact_goto_table
*ogt
= ofpact_get_GOTO_TABLE(a
);
2230 ovs_assert(ctx
->table_id
< ogt
->table_id
);
2231 xlate_table_action(ctx
, ctx
->xin
->flow
.in_port
.ofp_port
,
2232 ogt
->table_id
, true);
2237 xlate_sample_action(ctx
, ofpact_get_SAMPLE(a
));
2244 ctx
->rule
->up
.evictable
= was_evictable
;
2249 xlate_in_init(struct xlate_in
*xin
, struct ofproto_dpif
*ofproto
,
2250 const struct flow
*flow
, struct rule_dpif
*rule
,
2251 uint8_t tcp_flags
, const struct ofpbuf
*packet
)
2253 xin
->ofproto
= ofproto
;
2255 xin
->packet
= packet
;
2256 xin
->may_learn
= packet
!= NULL
;
2258 xin
->ofpacts
= NULL
;
2259 xin
->ofpacts_len
= 0;
2260 xin
->tcp_flags
= tcp_flags
;
2261 xin
->resubmit_hook
= NULL
;
2262 xin
->report_hook
= NULL
;
2263 xin
->resubmit_stats
= NULL
;
2267 xlate_out_uninit(struct xlate_out
*xout
)
2270 ofpbuf_uninit(&xout
->odp_actions
);
2274 /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts'
2275 * into datapath actions, using 'ctx', and discards the datapath actions. */
2277 xlate_actions_for_side_effects(struct xlate_in
*xin
)
2279 struct xlate_out xout
;
2281 xlate_actions(xin
, &xout
);
2282 xlate_out_uninit(&xout
);
2286 xlate_report(struct xlate_ctx
*ctx
, const char *s
)
2288 if (ctx
->xin
->report_hook
) {
2289 ctx
->xin
->report_hook(ctx
->xin
, s
, ctx
->recurse
);
2294 xlate_out_copy(struct xlate_out
*dst
, const struct xlate_out
*src
)
2297 dst
->slow
= src
->slow
;
2298 dst
->has_learn
= src
->has_learn
;
2299 dst
->has_normal
= src
->has_normal
;
2300 dst
->has_fin_timeout
= src
->has_fin_timeout
;
2301 dst
->nf_output_iface
= src
->nf_output_iface
;
2302 dst
->mirrors
= src
->mirrors
;
2304 ofpbuf_use_stub(&dst
->odp_actions
, dst
->odp_actions_stub
,
2305 sizeof dst
->odp_actions_stub
);
2306 ofpbuf_put(&dst
->odp_actions
, src
->odp_actions
.data
,
2307 src
->odp_actions
.size
);
2310 static struct skb_priority_to_dscp
*
2311 get_skb_priority(const struct xport
*xport
, uint32_t skb_priority
)
2313 struct skb_priority_to_dscp
*pdscp
;
2316 hash
= hash_int(skb_priority
, 0);
2317 HMAP_FOR_EACH_IN_BUCKET (pdscp
, hmap_node
, hash
, &xport
->skb_priorities
) {
2318 if (pdscp
->skb_priority
== skb_priority
) {
2326 dscp_from_skb_priority(const struct xport
*xport
, uint32_t skb_priority
,
2329 struct skb_priority_to_dscp
*pdscp
= get_skb_priority(xport
, skb_priority
);
2330 *dscp
= pdscp
? pdscp
->dscp
: 0;
2331 return pdscp
!= NULL
;
2335 clear_skb_priorities(struct xport
*xport
)
2337 struct skb_priority_to_dscp
*pdscp
, *next
;
2339 HMAP_FOR_EACH_SAFE (pdscp
, next
, hmap_node
, &xport
->skb_priorities
) {
2340 hmap_remove(&xport
->skb_priorities
, &pdscp
->hmap_node
);
2346 actions_output_to_local_port(const struct xlate_ctx
*ctx
)
2348 odp_port_t local_odp_port
= ofp_port_to_odp_port(ctx
->xbridge
, OFPP_LOCAL
);
2349 const struct nlattr
*a
;
2352 NL_ATTR_FOR_EACH_UNSAFE (a
, left
, ctx
->xout
->odp_actions
.data
,
2353 ctx
->xout
->odp_actions
.size
) {
2354 if (nl_attr_type(a
) == OVS_ACTION_ATTR_OUTPUT
2355 && nl_attr_get_odp_port(a
) == local_odp_port
) {
2362 /* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts'
2363 * into datapath actions in 'odp_actions', using 'ctx'. */
2365 xlate_actions(struct xlate_in
*xin
, struct xlate_out
*xout
)
2367 /* Normally false. Set to true if we ever hit MAX_RESUBMIT_RECURSION, so
2368 * that in the future we always keep a copy of the original flow for
2369 * tracing purposes. */
2370 static bool hit_resubmit_limit
;
2372 struct flow_wildcards
*wc
= &xout
->wc
;
2373 struct flow
*flow
= &xin
->flow
;
2375 enum slow_path_reason special
;
2376 const struct ofpact
*ofpacts
;
2377 struct xport
*in_port
;
2378 struct flow orig_flow
;
2379 struct xlate_ctx ctx
;
2382 COVERAGE_INC(xlate_actions
);
2384 /* Flow initialization rules:
2385 * - 'base_flow' must match the kernel's view of the packet at the
2386 * time that action processing starts. 'flow' represents any
2387 * transformations we wish to make through actions.
2388 * - By default 'base_flow' and 'flow' are the same since the input
2389 * packet matches the output before any actions are applied.
2390 * - When using VLAN splinters, 'base_flow''s VLAN is set to the value
2391 * of the received packet as seen by the kernel. If we later output
2392 * to another device without any modifications this will cause us to
2393 * insert a new tag since the original one was stripped off by the
2395 * - Tunnel metadata as received is retained in 'flow'. This allows
2396 * tunnel metadata matching also in later tables.
2397 * Since a kernel action for setting the tunnel metadata will only be
2398 * generated with actual tunnel output, changing the tunnel metadata
2399 * values in 'flow' (such as tun_id) will only have effect with a later
2400 * tunnel output action.
2401 * - Tunnel 'base_flow' is completely cleared since that is what the
2402 * kernel does. If we wish to maintain the original values an action
2403 * needs to be generated. */
2408 ctx
.xout
->has_learn
= false;
2409 ctx
.xout
->has_normal
= false;
2410 ctx
.xout
->has_fin_timeout
= false;
2411 ctx
.xout
->nf_output_iface
= NF_OUT_DROP
;
2412 ctx
.xout
->mirrors
= 0;
2413 ofpbuf_use_stub(&ctx
.xout
->odp_actions
, ctx
.xout
->odp_actions_stub
,
2414 sizeof ctx
.xout
->odp_actions_stub
);
2415 ofpbuf_reserve(&ctx
.xout
->odp_actions
, NL_A_U32_SIZE
);
2417 ctx
.xbridge
= xbridge_lookup(xin
->ofproto
);
2422 ctx
.rule
= xin
->rule
;
2424 ctx
.base_flow
= *flow
;
2425 memset(&ctx
.base_flow
.tunnel
, 0, sizeof ctx
.base_flow
.tunnel
);
2426 ctx
.orig_tunnel_ip_dst
= flow
->tunnel
.ip_dst
;
2428 flow_wildcards_init_catchall(wc
);
2429 memset(&wc
->masks
.in_port
, 0xff, sizeof wc
->masks
.in_port
);
2430 memset(&wc
->masks
.skb_priority
, 0xff, sizeof wc
->masks
.skb_priority
);
2431 memset(&wc
->masks
.dl_type
, 0xff, sizeof wc
->masks
.dl_type
);
2432 wc
->masks
.nw_frag
|= FLOW_NW_FRAG_MASK
;
2434 if (tnl_port_should_receive(&ctx
.xin
->flow
)) {
2435 memset(&wc
->masks
.tunnel
, 0xff, sizeof wc
->masks
.tunnel
);
2436 /* skb_mark is currently used only by tunnels but that will likely
2437 * change in the future. */
2438 memset(&wc
->masks
.skb_mark
, 0xff, sizeof wc
->masks
.skb_mark
);
2440 if (ctx
.xbridge
->has_netflow
) {
2441 netflow_mask_wc(flow
, wc
);
2445 ctx
.max_resubmit_trigger
= false;
2446 ctx
.orig_skb_priority
= flow
->skb_priority
;
2451 ofpacts
= xin
->ofpacts
;
2452 ofpacts_len
= xin
->ofpacts_len
;
2453 } else if (xin
->rule
) {
2454 ofpacts
= xin
->rule
->up
.ofpacts
;
2455 ofpacts_len
= xin
->rule
->up
.ofpacts_len
;
2460 ofpbuf_use_stub(&ctx
.stack
, ctx
.init_stack
, sizeof ctx
.init_stack
);
2462 if (mbridge_has_mirrors(ctx
.xbridge
->mbridge
) || hit_resubmit_limit
) {
2463 /* Do this conditionally because the copy is expensive enough that it
2464 * shows up in profiles. */
2468 if (flow
->nw_frag
& FLOW_NW_FRAG_ANY
) {
2469 switch (ctx
.xbridge
->frag
) {
2470 case OFPC_FRAG_NORMAL
:
2471 /* We must pretend that transport ports are unavailable. */
2472 flow
->tp_src
= ctx
.base_flow
.tp_src
= htons(0);
2473 flow
->tp_dst
= ctx
.base_flow
.tp_dst
= htons(0);
2476 case OFPC_FRAG_DROP
:
2479 case OFPC_FRAG_REASM
:
2482 case OFPC_FRAG_NX_MATCH
:
2483 /* Nothing to do. */
2486 case OFPC_INVALID_TTL_TO_CONTROLLER
:
2491 in_port
= get_ofp_port(ctx
.xbridge
, flow
->in_port
.ofp_port
);
2492 special
= process_special(&ctx
, flow
, in_port
, ctx
.xin
->packet
);
2494 ctx
.xout
->slow
= special
;
2496 static struct vlog_rate_limit trace_rl
= VLOG_RATE_LIMIT_INIT(1, 1);
2497 size_t sample_actions_len
;
2499 if (flow
->in_port
.ofp_port
2500 != vsp_realdev_to_vlandev(ctx
.xbridge
->ofproto
,
2501 flow
->in_port
.ofp_port
,
2503 ctx
.base_flow
.vlan_tci
= 0;
2506 add_sflow_action(&ctx
);
2507 add_ipfix_action(&ctx
);
2508 sample_actions_len
= ctx
.xout
->odp_actions
.size
;
2510 if (tunnel_ecn_ok(&ctx
) && (!in_port
|| may_receive(in_port
, &ctx
))) {
2511 do_xlate_actions(ofpacts
, ofpacts_len
, &ctx
);
2513 /* We've let OFPP_NORMAL and the learning action look at the
2514 * packet, so drop it now if forwarding is disabled. */
2515 if (in_port
&& !xport_stp_forward_state(in_port
)) {
2516 ctx
.xout
->odp_actions
.size
= sample_actions_len
;
2520 if (ctx
.max_resubmit_trigger
&& !ctx
.xin
->resubmit_hook
) {
2521 if (!hit_resubmit_limit
) {
2522 /* We didn't record the original flow. Make sure we do from
2524 hit_resubmit_limit
= true;
2525 } else if (!VLOG_DROP_ERR(&trace_rl
)) {
2526 struct ds ds
= DS_EMPTY_INITIALIZER
;
2528 ofproto_trace(ctx
.xbridge
->ofproto
, &orig_flow
,
2529 ctx
.xin
->packet
, &ds
);
2530 VLOG_ERR("Trace triggered by excessive resubmit "
2531 "recursion:\n%s", ds_cstr(&ds
));
2536 if (ctx
.xbridge
->has_in_band
2537 && in_band_must_output_to_local_port(flow
)
2538 && !actions_output_to_local_port(&ctx
)) {
2539 compose_output_action(&ctx
, OFPP_LOCAL
);
2542 fix_sflow_action(&ctx
);
2544 if (mbridge_has_mirrors(ctx
.xbridge
->mbridge
)) {
2545 add_mirror_actions(&ctx
, &orig_flow
);
2549 ofpbuf_uninit(&ctx
.stack
);
2551 /* Clear the metadata and register wildcard masks, because we won't
2552 * use non-header fields as part of the cache. */
2553 memset(&wc
->masks
.metadata
, 0, sizeof wc
->masks
.metadata
);
2554 memset(&wc
->masks
.regs
, 0, sizeof wc
->masks
.regs
);