]> git.proxmox.com Git - mirror_ovs.git/blame - ofproto/ofproto-dpif-xlate.c
ofproto-dpif-xlate: Cleanup lookup functions.
[mirror_ovs.git] / ofproto / ofproto-dpif-xlate.c
CommitLineData
9583bc14
EJ
1/* Copyright (c) 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
2 *
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:
6 *
7 * http://www.apache.org/licenses/LICENSE-2.0
8 *
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. */
14
15#include <config.h>
16
17#include "ofproto/ofproto-dpif-xlate.h"
18
db7d4e46 19#include "bfd.h"
9583bc14
EJ
20#include "bitmap.h"
21#include "bond.h"
22#include "bundle.h"
23#include "byte-order.h"
db7d4e46 24#include "cfm.h"
9583bc14
EJ
25#include "connmgr.h"
26#include "coverage.h"
27#include "dpif.h"
28#include "dynamic-string.h"
f7f1ea29 29#include "in-band.h"
db7d4e46 30#include "lacp.h"
9583bc14 31#include "learn.h"
46c88433 32#include "list.h"
9583bc14
EJ
33#include "mac-learning.h"
34#include "meta-flow.h"
35#include "multipath.h"
36#include "netdev-vport.h"
37#include "netlink.h"
38#include "nx-match.h"
39#include "odp-execute.h"
40#include "ofp-actions.h"
41#include "ofproto/ofproto-dpif-ipfix.h"
ec7ceaed 42#include "ofproto/ofproto-dpif-mirror.h"
9583bc14
EJ
43#include "ofproto/ofproto-dpif-sflow.h"
44#include "ofproto/ofproto-dpif.h"
45#include "tunnel.h"
46#include "vlog.h"
47
46c88433 48COVERAGE_DEFINE(xlate_actions);
9583bc14
EJ
49
50VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate);
51
8a553e9a
EJ
52/* Maximum depth of flow table recursion (due to resubmit actions) in a
53 * flow translation. */
54#define MAX_RESUBMIT_RECURSION 64
55
46c88433
EJ
56struct xbridge {
57 struct hmap_node hmap_node; /* Node in global 'xbridges' map. */
58 struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */
59
60 struct list xbundles; /* Owned xbundles. */
61 struct hmap xports; /* Indexed by ofp_port. */
62
63 char *name; /* Name used in log messages. */
89a8a7f0 64 struct dpif *dpif; /* Datapath interface. */
46c88433
EJ
65 struct mac_learning *ml; /* Mac learning handle. */
66 struct mbridge *mbridge; /* Mirroring. */
67 struct dpif_sflow *sflow; /* SFlow handle, or null. */
68 struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */
9d189a50 69 struct stp *stp; /* STP or null if disabled. */
46c88433
EJ
70
71 enum ofp_config_flags frag; /* Fragmentation handling. */
46c88433
EJ
72 bool has_netflow; /* Bridge runs netflow? */
73 bool has_in_band; /* Bridge has in band control? */
74 bool forward_bpdu; /* Bridge forwards STP BPDUs? */
75};
76
77struct xbundle {
78 struct hmap_node hmap_node; /* In global 'xbundles' map. */
79 struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */
80
81 struct list list_node; /* In parent 'xbridges' list. */
82 struct xbridge *xbridge; /* Parent xbridge. */
83
84 struct list xports; /* Contains "struct xport"s. */
85
86 char *name; /* Name used in log messages. */
87 struct bond *bond; /* Nonnull iff more than one port. */
88 struct lacp *lacp; /* LACP handle or null. */
89
90 enum port_vlan_mode vlan_mode; /* VLAN mode. */
91 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
92 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
93 * NULL if all VLANs are trunked. */
94 bool use_priority_tags; /* Use 802.1p tag for frames in VLAN 0? */
95 bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */
96};
97
98struct xport {
99 struct hmap_node hmap_node; /* Node in global 'xports' map. */
100 struct ofport_dpif *ofport; /* Key in global 'xports map. */
101
102 struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */
103 ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */
104
105 odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */
106
107 struct list bundle_node; /* In parent xbundle (if it exists). */
108 struct xbundle *xbundle; /* Parent xbundle or null. */
109
110 struct netdev *netdev; /* 'ofport''s netdev. */
111
112 struct xbridge *xbridge; /* Parent bridge. */
113 struct xport *peer; /* Patch port peer or null. */
114
115 enum ofputil_port_config config; /* OpenFlow port configuration. */
9d189a50 116 int stp_port_no; /* STP port number or 0 if not in use. */
46c88433 117
55954f6e
EJ
118 struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */
119
46c88433
EJ
120 bool may_enable; /* May be enabled in bonds. */
121 bool is_tunnel; /* Is a tunnel port. */
122
123 struct cfm *cfm; /* CFM handle or null. */
124 struct bfd *bfd; /* BFD handle or null. */
125};
126
4d0acc70
EJ
127struct xlate_ctx {
128 struct xlate_in *xin;
129 struct xlate_out *xout;
130
46c88433 131 const struct xbridge *xbridge;
4d0acc70
EJ
132
133 /* Flow at the last commit. */
134 struct flow base_flow;
135
136 /* Tunnel IP destination address as received. This is stored separately
137 * as the base_flow.tunnel is cleared on init to reflect the datapath
138 * behavior. Used to make sure not to send tunneled output to ourselves,
139 * which might lead to an infinite loop. This could happen easily
140 * if a tunnel is marked as 'ip_remote=flow', and the flow does not
141 * actually set the tun_dst field. */
142 ovs_be32 orig_tunnel_ip_dst;
143
144 /* Stack for the push and pop actions. Each stack element is of type
145 * "union mf_subvalue". */
146 union mf_subvalue init_stack[1024 / sizeof(union mf_subvalue)];
147 struct ofpbuf stack;
148
149 /* The rule that we are currently translating, or NULL. */
150 struct rule_dpif *rule;
151
152 int recurse; /* Recursion level, via xlate_table_action. */
4d0acc70
EJ
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. */
4e022ec0 156 odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */
4d0acc70
EJ
157 uint16_t user_cookie_offset;/* Used for user_action_cookie fixup. */
158 bool exit; /* No further actions should be processed. */
159};
160
9583bc14
EJ
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
3815d6c2
LS
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() */
167static struct xbundle ofpp_none_bundle;
9583bc14 168
55954f6e
EJ
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. */
172struct 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). */
175
176 uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */
177};
178
46c88433
EJ
179static struct hmap xbridges = HMAP_INITIALIZER(&xbridges);
180static struct hmap xbundles = HMAP_INITIALIZER(&xbundles);
181static struct hmap xports = HMAP_INITIALIZER(&xports);
182
183static bool may_receive(const struct xport *, struct xlate_ctx *);
9583bc14
EJ
184static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len,
185 struct xlate_ctx *);
186static void xlate_normal(struct xlate_ctx *);
187static void xlate_report(struct xlate_ctx *, const char *);
4e022ec0 188static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port,
9583bc14 189 uint8_t table_id, bool may_packet_in);
46c88433
EJ
190static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn);
191static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid);
192static void output_normal(struct xlate_ctx *, const struct xbundle *,
9583bc14 193 uint16_t vlan);
4e022ec0 194static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port);
9583bc14
EJ
195
196static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
197
46c88433
EJ
198static struct xbridge *xbridge_lookup(const struct ofproto_dpif *);
199static struct xbundle *xbundle_lookup(const struct ofbundle *);
5e6af486 200static struct xport *xport_lookup(const struct ofport_dpif *);
46c88433 201static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port);
55954f6e
EJ
202static struct skb_priority_to_dscp *get_skb_priority(const struct xport *,
203 uint32_t skb_priority);
204static void clear_skb_priorities(struct xport *);
205static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority,
206 uint8_t *dscp);
46c88433
EJ
207
208void
209xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name,
89a8a7f0
EJ
210 struct dpif *dpif, const struct mac_learning *ml,
211 struct stp *stp, const struct mbridge *mbridge,
46c88433
EJ
212 const struct dpif_sflow *sflow,
213 const struct dpif_ipfix *ipfix, enum ofp_config_flags frag,
9d189a50 214 bool forward_bpdu, bool has_in_band, bool has_netflow)
46c88433
EJ
215{
216 struct xbridge *xbridge = xbridge_lookup(ofproto);
217
218 if (!xbridge) {
219 xbridge = xzalloc(sizeof *xbridge);
220 xbridge->ofproto = ofproto;
221
222 hmap_insert(&xbridges, &xbridge->hmap_node, hash_pointer(ofproto, 0));
223 hmap_init(&xbridge->xports);
224 list_init(&xbridge->xbundles);
225 }
226
227 if (xbridge->ml != ml) {
228 mac_learning_unref(xbridge->ml);
229 xbridge->ml = mac_learning_ref(ml);
230 }
231
232 if (xbridge->mbridge != mbridge) {
233 mbridge_unref(xbridge->mbridge);
234 xbridge->mbridge = mbridge_ref(mbridge);
235 }
236
237 if (xbridge->sflow != sflow) {
238 dpif_sflow_unref(xbridge->sflow);
239 xbridge->sflow = dpif_sflow_ref(sflow);
240 }
241
242 if (xbridge->ipfix != ipfix) {
243 dpif_ipfix_unref(xbridge->ipfix);
244 xbridge->ipfix = dpif_ipfix_ref(ipfix);
245 }
246
9d189a50
EJ
247 if (xbridge->stp != stp) {
248 stp_unref(xbridge->stp);
249 xbridge->stp = stp_ref(stp);
250 }
251
46c88433
EJ
252 free(xbridge->name);
253 xbridge->name = xstrdup(name);
254
89a8a7f0 255 xbridge->dpif = dpif;
46c88433
EJ
256 xbridge->forward_bpdu = forward_bpdu;
257 xbridge->has_in_band = has_in_band;
258 xbridge->has_netflow = has_netflow;
46c88433
EJ
259 xbridge->frag = frag;
260}
261
262void
263xlate_remove_ofproto(struct ofproto_dpif *ofproto)
264{
265 struct xbridge *xbridge = xbridge_lookup(ofproto);
266 struct xbundle *xbundle, *next_xbundle;
267 struct xport *xport, *next_xport;
268
269 if (!xbridge) {
270 return;
271 }
272
273 HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) {
274 xlate_ofport_remove(xport->ofport);
275 }
276
277 LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) {
278 xlate_bundle_remove(xbundle->ofbundle);
279 }
280
281 hmap_remove(&xbridges, &xbridge->hmap_node);
282 free(xbridge->name);
283 free(xbridge);
284}
285
286void
287xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle,
288 const char *name, enum port_vlan_mode vlan_mode, int vlan,
289 unsigned long *trunks, bool use_priority_tags,
290 const struct bond *bond, const struct lacp *lacp,
291 bool floodable)
292{
293 struct xbundle *xbundle = xbundle_lookup(ofbundle);
294
295 if (!xbundle) {
296 xbundle = xzalloc(sizeof *xbundle);
297 xbundle->ofbundle = ofbundle;
298 xbundle->xbridge = xbridge_lookup(ofproto);
299
300 hmap_insert(&xbundles, &xbundle->hmap_node, hash_pointer(ofbundle, 0));
301 list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node);
302 list_init(&xbundle->xports);
303 }
304
305 ovs_assert(xbundle->xbridge);
306
307 free(xbundle->name);
308 xbundle->name = xstrdup(name);
309
310 xbundle->vlan_mode = vlan_mode;
311 xbundle->vlan = vlan;
312 xbundle->trunks = trunks;
313 xbundle->use_priority_tags = use_priority_tags;
314 xbundle->floodable = floodable;
315
316 if (xbundle->bond != bond) {
317 bond_unref(xbundle->bond);
318 xbundle->bond = bond_ref(bond);
319 }
320
321 if (xbundle->lacp != lacp) {
322 lacp_unref(xbundle->lacp);
323 xbundle->lacp = lacp_ref(lacp);
324 }
325}
326
327void
328xlate_bundle_remove(struct ofbundle *ofbundle)
329{
330 struct xbundle *xbundle = xbundle_lookup(ofbundle);
331 struct xport *xport, *next;
332
333 if (!xbundle) {
334 return;
335 }
336
337 LIST_FOR_EACH_SAFE (xport, next, bundle_node, &xbundle->xports) {
338 list_remove(&xport->bundle_node);
339 xport->xbundle = NULL;
340 }
341
342 hmap_remove(&xbundles, &xbundle->hmap_node);
343 list_remove(&xbundle->list_node);
344 bond_unref(xbundle->bond);
345 lacp_unref(xbundle->lacp);
346 free(xbundle->name);
347 free(xbundle);
348}
349
350void
351xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle,
352 struct ofport_dpif *ofport, ofp_port_t ofp_port,
353 odp_port_t odp_port, const struct netdev *netdev,
354 const struct cfm *cfm, const struct bfd *bfd,
9d189a50 355 struct ofport_dpif *peer, int stp_port_no,
55954f6e 356 const struct ofproto_port_queue *qdscp_list, size_t n_qdscp,
9d189a50
EJ
357 enum ofputil_port_config config, bool is_tunnel,
358 bool may_enable)
46c88433
EJ
359{
360 struct xport *xport = xport_lookup(ofport);
55954f6e 361 size_t i;
46c88433
EJ
362
363 if (!xport) {
364 xport = xzalloc(sizeof *xport);
365 xport->ofport = ofport;
366 xport->xbridge = xbridge_lookup(ofproto);
367 xport->ofp_port = ofp_port;
368
55954f6e 369 hmap_init(&xport->skb_priorities);
46c88433
EJ
370 hmap_insert(&xports, &xport->hmap_node, hash_pointer(ofport, 0));
371 hmap_insert(&xport->xbridge->xports, &xport->ofp_node,
372 hash_ofp_port(xport->ofp_port));
373 }
374
375 ovs_assert(xport->ofp_port == ofp_port);
376
377 xport->config = config;
9d189a50 378 xport->stp_port_no = stp_port_no;
46c88433
EJ
379 xport->is_tunnel = is_tunnel;
380 xport->may_enable = may_enable;
381 xport->odp_port = odp_port;
382
383 if (xport->netdev != netdev) {
384 netdev_close(xport->netdev);
385 xport->netdev = netdev_ref(netdev);
386 }
387
388 if (xport->cfm != cfm) {
389 cfm_unref(xport->cfm);
390 xport->cfm = cfm_ref(cfm);
391 }
392
393 if (xport->bfd != bfd) {
394 bfd_unref(xport->bfd);
395 xport->bfd = bfd_ref(bfd);
396 }
397
398 if (xport->peer) {
399 xport->peer->peer = NULL;
400 }
5e6af486 401 xport->peer = xport_lookup(peer);
46c88433
EJ
402 if (xport->peer) {
403 xport->peer->peer = xport;
404 }
405
406 if (xport->xbundle) {
407 list_remove(&xport->bundle_node);
408 }
5e6af486 409 xport->xbundle = xbundle_lookup(ofbundle);
46c88433
EJ
410 if (xport->xbundle) {
411 list_insert(&xport->xbundle->xports, &xport->bundle_node);
412 }
55954f6e
EJ
413
414 clear_skb_priorities(xport);
415 for (i = 0; i < n_qdscp; i++) {
416 struct skb_priority_to_dscp *pdscp;
417 uint32_t skb_priority;
418
89a8a7f0
EJ
419 if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue,
420 &skb_priority)) {
55954f6e
EJ
421 continue;
422 }
423
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));
429 }
46c88433
EJ
430}
431
432void
433xlate_ofport_remove(struct ofport_dpif *ofport)
434{
435 struct xport *xport = xport_lookup(ofport);
436
437 if (!xport) {
438 return;
439 }
440
441 if (xport->peer) {
442 xport->peer->peer = NULL;
443 xport->peer = NULL;
444 }
445
e621a12d
EJ
446 if (xport->xbundle) {
447 list_remove(&xport->bundle_node);
448 }
449
55954f6e
EJ
450 clear_skb_priorities(xport);
451 hmap_destroy(&xport->skb_priorities);
452
46c88433
EJ
453 hmap_remove(&xports, &xport->hmap_node);
454 hmap_remove(&xport->xbridge->xports, &xport->ofp_node);
455
456 netdev_close(xport->netdev);
457 cfm_unref(xport->cfm);
458 bfd_unref(xport->bfd);
459 free(xport);
460}
461
462static struct xbridge *
463xbridge_lookup(const struct ofproto_dpif *ofproto)
464{
465 struct xbridge *xbridge;
466
5e6af486
EJ
467 if (!ofproto) {
468 return NULL;
469 }
470
46c88433
EJ
471 HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0),
472 &xbridges) {
473 if (xbridge->ofproto == ofproto) {
474 return xbridge;
475 }
476 }
477 return NULL;
478}
479
480static struct xbundle *
481xbundle_lookup(const struct ofbundle *ofbundle)
482{
483 struct xbundle *xbundle;
484
5e6af486
EJ
485 if (!ofbundle) {
486 return NULL;
487 }
488
46c88433
EJ
489 HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0),
490 &xbundles) {
491 if (xbundle->ofbundle == ofbundle) {
492 return xbundle;
493 }
494 }
495 return NULL;
496}
497
498static struct xport *
5e6af486 499xport_lookup(const struct ofport_dpif *ofport)
46c88433
EJ
500{
501 struct xport *xport;
502
5e6af486
EJ
503 if (!ofport) {
504 return NULL;
505 }
506
46c88433
EJ
507 HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0),
508 &xports) {
509 if (xport->ofport == ofport) {
510 return xport;
511 }
512 }
513 return NULL;
514}
515
40085e56
EJ
516static struct stp_port *
517xport_get_stp_port(const struct xport *xport)
518{
519 return xport->xbridge->stp && xport->stp_port_no
520 ? stp_get_port(xport->xbridge->stp, xport->stp_port_no)
521 : NULL;
522}
9d189a50
EJ
523
524static enum stp_state
525xport_stp_learn_state(const struct xport *xport)
526{
40085e56
EJ
527 struct stp_port *sp = xport_get_stp_port(xport);
528 return stp_learn_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
9d189a50
EJ
529}
530
531static bool
532xport_stp_forward_state(const struct xport *xport)
533{
40085e56
EJ
534 struct stp_port *sp = xport_get_stp_port(xport);
535 return stp_forward_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
9d189a50
EJ
536}
537
538/* Returns true if STP should process 'flow'. Sets fields in 'wc' that
539 * were used to make the determination.*/
540static bool
541stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc)
542{
543 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
544 return eth_addr_equals(flow->dl_dst, eth_addr_stp);
545}
546
547static void
548stp_process_packet(const struct xport *xport, const struct ofpbuf *packet)
549{
40085e56 550 struct stp_port *sp = xport_get_stp_port(xport);
9d189a50
EJ
551 struct ofpbuf payload = *packet;
552 struct eth_header *eth = payload.data;
9d189a50
EJ
553
554 /* Sink packets on ports that have STP disabled when the bridge has
555 * STP enabled. */
556 if (!sp || stp_port_get_state(sp) == STP_DISABLED) {
557 return;
558 }
559
560 /* Trim off padding on payload. */
561 if (payload.size > ntohs(eth->eth_type) + ETH_HEADER_LEN) {
562 payload.size = ntohs(eth->eth_type) + ETH_HEADER_LEN;
563 }
564
565 if (ofpbuf_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) {
566 stp_received_bpdu(sp, payload.data, payload.size);
567 }
568}
569
46c88433
EJ
570static struct xport *
571get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port)
572{
573 struct xport *xport;
574
575 HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port),
576 &xbridge->xports) {
577 if (xport->ofp_port == ofp_port) {
578 return xport;
579 }
580 }
581 return NULL;
582}
583
584static odp_port_t
585ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port)
586{
587 const struct xport *xport = get_ofp_port(xbridge, ofp_port);
588 return xport ? xport->odp_port : ODPP_NONE;
589}
590
9583bc14 591static bool
46c88433 592xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan)
9583bc14
EJ
593{
594 return (bundle->vlan_mode != PORT_VLAN_ACCESS
595 && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan)));
596}
597
598static bool
46c88433
EJ
599xbundle_includes_vlan(const struct xbundle *xbundle, uint16_t vlan)
600{
601 return vlan == xbundle->vlan || xbundle_trunks_vlan(xbundle, vlan);
602}
603
604static mirror_mask_t
605xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle)
606{
607 return xbundle != &ofpp_none_bundle
608 ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle)
609 : 0;
610}
611
612static mirror_mask_t
613xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle)
9583bc14 614{
46c88433
EJ
615 return xbundle != &ofpp_none_bundle
616 ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle)
617 : 0;
9583bc14
EJ
618}
619
46c88433
EJ
620static mirror_mask_t
621xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle)
9583bc14 622{
46c88433
EJ
623 return xbundle != &ofpp_none_bundle
624 ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle)
625 : 0;
626}
627
628static struct xbundle *
629lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port,
630 bool warn, struct xport **in_xportp)
631{
632 struct xport *xport;
9583bc14
EJ
633
634 /* Find the port and bundle for the received packet. */
46c88433
EJ
635 xport = get_ofp_port(xbridge, in_port);
636 if (in_xportp) {
637 *in_xportp = xport;
9583bc14 638 }
46c88433
EJ
639 if (xport && xport->xbundle) {
640 return xport->xbundle;
9583bc14
EJ
641 }
642
643 /* Special-case OFPP_NONE, which a controller may use as the ingress
644 * port for traffic that it is sourcing. */
645 if (in_port == OFPP_NONE) {
3815d6c2
LS
646 ofpp_none_bundle.name = "OFPP_NONE";
647 ofpp_none_bundle.vlan_mode = PORT_VLAN_TRUNK;
9583bc14
EJ
648 return &ofpp_none_bundle;
649 }
650
651 /* Odd. A few possible reasons here:
652 *
653 * - We deleted a port but there are still a few packets queued up
654 * from it.
655 *
656 * - Someone externally added a port (e.g. "ovs-dpctl add-if") that
657 * we don't know about.
658 *
659 * - The ofproto client didn't configure the port as part of a bundle.
660 * This is particularly likely to happen if a packet was received on the
661 * port after it was created, but before the client had a chance to
662 * configure its bundle.
663 */
664 if (warn) {
665 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
666
667 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
46c88433 668 "port %"PRIu16, xbridge->name, in_port);
9583bc14
EJ
669 }
670 return NULL;
671}
672
673static void
674add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow)
675{
46c88433 676 const struct xbridge *xbridge = ctx->xbridge;
9583bc14 677 mirror_mask_t mirrors;
46c88433 678 struct xbundle *in_xbundle;
9583bc14
EJ
679 uint16_t vlan;
680 uint16_t vid;
cdf5d3a5
EJ
681
682 mirrors = ctx->xout->mirrors;
683 ctx->xout->mirrors = 0;
9583bc14 684
46c88433
EJ
685 in_xbundle = lookup_input_bundle(xbridge, orig_flow->in_port.ofp_port,
686 ctx->xin->packet != NULL, NULL);
687 if (!in_xbundle) {
9583bc14
EJ
688 return;
689 }
46c88433 690 mirrors |= xbundle_mirror_src(xbridge, in_xbundle);
9583bc14
EJ
691
692 /* Drop frames on bundles reserved for mirroring. */
46c88433 693 if (xbundle_mirror_out(xbridge, in_xbundle)) {
9583bc14
EJ
694 if (ctx->xin->packet != NULL) {
695 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
696 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
697 "%s, which is reserved exclusively for mirroring",
46c88433 698 ctx->xbridge->name, in_xbundle->name);
9583bc14 699 }
aaa0fbae 700 ofpbuf_clear(&ctx->xout->odp_actions);
9583bc14
EJ
701 return;
702 }
703
704 /* Check VLAN. */
705 vid = vlan_tci_to_vid(orig_flow->vlan_tci);
46c88433 706 if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) {
9583bc14
EJ
707 return;
708 }
46c88433 709 vlan = input_vid_to_vlan(in_xbundle, vid);
9583bc14 710
9583bc14
EJ
711 if (!mirrors) {
712 return;
713 }
714
715 /* Restore the original packet before adding the mirror actions. */
716 ctx->xin->flow = *orig_flow;
717
718 while (mirrors) {
ec7ceaed
EJ
719 mirror_mask_t dup_mirrors;
720 struct ofbundle *out;
721 unsigned long *vlans;
722 bool vlan_mirrored;
723 bool has_mirror;
724 int out_vlan;
725
46c88433 726 has_mirror = mirror_get(xbridge->mbridge, mirror_mask_ffs(mirrors) - 1,
ec7ceaed
EJ
727 &vlans, &dup_mirrors, &out, &out_vlan);
728 ovs_assert(has_mirror);
729
730 if (vlans) {
9583bc14
EJ
731 ctx->xout->wc.masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK);
732 }
ec7ceaed
EJ
733 vlan_mirrored = !vlans || bitmap_is_set(vlans, vlan);
734 free(vlans);
9583bc14 735
ec7ceaed 736 if (!vlan_mirrored) {
9583bc14
EJ
737 mirrors = zero_rightmost_1bit(mirrors);
738 continue;
739 }
740
ec7ceaed
EJ
741 mirrors &= ~dup_mirrors;
742 ctx->xout->mirrors |= dup_mirrors;
743 if (out) {
46c88433
EJ
744 struct xbundle *out_xbundle = xbundle_lookup(out);
745 if (out_xbundle) {
746 output_normal(ctx, out_xbundle, vlan);
747 }
ec7ceaed 748 } else if (vlan != out_vlan
9583bc14 749 && !eth_addr_is_reserved(orig_flow->dl_dst)) {
46c88433 750 struct xbundle *xbundle;
9583bc14 751
46c88433
EJ
752 LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) {
753 if (xbundle_includes_vlan(xbundle, out_vlan)
754 && !xbundle_mirror_out(xbridge, xbundle)) {
755 output_normal(ctx, xbundle, out_vlan);
9583bc14
EJ
756 }
757 }
758 }
759 }
760}
761
762/* Given 'vid', the VID obtained from the 802.1Q header that was received as
46c88433 763 * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle',
9583bc14
EJ
764 * the bundle on which the packet was received, returns the VLAN to which the
765 * packet belongs.
766 *
767 * Both 'vid' and the return value are in the range 0...4095. */
768static uint16_t
46c88433 769input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid)
9583bc14 770{
46c88433 771 switch (in_xbundle->vlan_mode) {
9583bc14 772 case PORT_VLAN_ACCESS:
46c88433 773 return in_xbundle->vlan;
9583bc14
EJ
774 break;
775
776 case PORT_VLAN_TRUNK:
777 return vid;
778
779 case PORT_VLAN_NATIVE_UNTAGGED:
780 case PORT_VLAN_NATIVE_TAGGED:
46c88433 781 return vid ? vid : in_xbundle->vlan;
9583bc14
EJ
782
783 default:
784 NOT_REACHED();
785 }
786}
787
46c88433 788/* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'.
9583bc14
EJ
789 * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs
790 * a warning.
791 *
792 * 'vid' should be the VID obtained from the 802.1Q header that was received as
793 * part of a packet (specify 0 if there was no 802.1Q header), in the range
794 * 0...4095. */
795static bool
46c88433 796input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn)
9583bc14
EJ
797{
798 /* Allow any VID on the OFPP_NONE port. */
46c88433 799 if (in_xbundle == &ofpp_none_bundle) {
9583bc14
EJ
800 return true;
801 }
802
46c88433 803 switch (in_xbundle->vlan_mode) {
9583bc14
EJ
804 case PORT_VLAN_ACCESS:
805 if (vid) {
806 if (warn) {
807 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
46c88433 808 VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" tagged "
9583bc14 809 "packet received on port %s configured as VLAN "
46c88433
EJ
810 "%"PRIu16" access port", vid, in_xbundle->name,
811 in_xbundle->vlan);
9583bc14
EJ
812 }
813 return false;
814 }
815 return true;
816
817 case PORT_VLAN_NATIVE_UNTAGGED:
818 case PORT_VLAN_NATIVE_TAGGED:
819 if (!vid) {
820 /* Port must always carry its native VLAN. */
821 return true;
822 }
823 /* Fall through. */
824 case PORT_VLAN_TRUNK:
46c88433 825 if (!xbundle_includes_vlan(in_xbundle, vid)) {
9583bc14
EJ
826 if (warn) {
827 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
46c88433 828 VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" packet "
9583bc14 829 "received on port %s not configured for trunking "
46c88433 830 "VLAN %"PRIu16, vid, in_xbundle->name, vid);
9583bc14
EJ
831 }
832 return false;
833 }
834 return true;
835
836 default:
837 NOT_REACHED();
838 }
839
840}
841
842/* Given 'vlan', the VLAN that a packet belongs to, and
46c88433 843 * 'out_xbundle', a bundle on which the packet is to be output, returns the VID
9583bc14
EJ
844 * that should be included in the 802.1Q header. (If the return value is 0,
845 * then the 802.1Q header should only be included in the packet if there is a
846 * nonzero PCP.)
847 *
848 * Both 'vlan' and the return value are in the range 0...4095. */
849static uint16_t
46c88433 850output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan)
9583bc14 851{
46c88433 852 switch (out_xbundle->vlan_mode) {
9583bc14
EJ
853 case PORT_VLAN_ACCESS:
854 return 0;
855
856 case PORT_VLAN_TRUNK:
857 case PORT_VLAN_NATIVE_TAGGED:
858 return vlan;
859
860 case PORT_VLAN_NATIVE_UNTAGGED:
46c88433 861 return vlan == out_xbundle->vlan ? 0 : vlan;
9583bc14
EJ
862
863 default:
864 NOT_REACHED();
865 }
866}
867
868static void
46c88433 869output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle,
9583bc14
EJ
870 uint16_t vlan)
871{
33bf9176 872 ovs_be16 *flow_tci = &ctx->xin->flow.vlan_tci;
9583bc14
EJ
873 uint16_t vid;
874 ovs_be16 tci, old_tci;
46c88433 875 struct xport *xport;
9583bc14 876
46c88433
EJ
877 vid = output_vlan_to_vid(out_xbundle, vlan);
878 if (list_is_empty(&out_xbundle->xports)) {
879 /* Partially configured bundle with no slaves. Drop the packet. */
880 return;
881 } else if (!out_xbundle->bond) {
882 xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport,
883 bundle_node);
9583bc14 884 } else {
46c88433
EJ
885 struct ofport_dpif *ofport;
886
887 ofport = bond_choose_output_slave(out_xbundle->bond, &ctx->xin->flow,
4a1b8f30 888 &ctx->xout->wc, vid);
5e6af486 889 xport = xport_lookup(ofport);
46c88433
EJ
890
891 if (!xport) {
9583bc14
EJ
892 /* No slaves enabled, so drop packet. */
893 return;
894 }
895 }
896
33bf9176 897 old_tci = *flow_tci;
9583bc14 898 tci = htons(vid);
46c88433 899 if (tci || out_xbundle->use_priority_tags) {
33bf9176 900 tci |= *flow_tci & htons(VLAN_PCP_MASK);
9583bc14
EJ
901 if (tci) {
902 tci |= htons(VLAN_CFI);
903 }
904 }
33bf9176 905 *flow_tci = tci;
9583bc14 906
46c88433 907 compose_output_action(ctx, xport->ofp_port);
33bf9176 908 *flow_tci = old_tci;
9583bc14
EJ
909}
910
911/* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
912 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
913 * indicate this; newer upstream kernels use gratuitous ARP requests. */
914static bool
915is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc)
916{
917 if (flow->dl_type != htons(ETH_TYPE_ARP)) {
918 return false;
919 }
920
921 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
922 if (!eth_addr_is_broadcast(flow->dl_dst)) {
923 return false;
924 }
925
926 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
927 if (flow->nw_proto == ARP_OP_REPLY) {
928 return true;
929 } else if (flow->nw_proto == ARP_OP_REQUEST) {
930 memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src);
931 memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst);
932
933 return flow->nw_src == flow->nw_dst;
934 } else {
935 return false;
936 }
937}
938
939static void
46c88433 940update_learning_table(const struct xbridge *xbridge,
9583bc14 941 const struct flow *flow, struct flow_wildcards *wc,
46c88433 942 int vlan, struct xbundle *in_xbundle)
9583bc14
EJ
943{
944 struct mac_entry *mac;
945
946 /* Don't learn the OFPP_NONE port. */
46c88433 947 if (in_xbundle == &ofpp_none_bundle) {
9583bc14
EJ
948 return;
949 }
950
509c0149 951 ovs_rwlock_wrlock(&xbridge->ml->rwlock);
46c88433 952 if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) {
509c0149 953 goto out;
9583bc14
EJ
954 }
955
46c88433 956 mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan);
9583bc14
EJ
957 if (is_gratuitous_arp(flow, wc)) {
958 /* We don't want to learn from gratuitous ARP packets that are
959 * reflected back over bond slaves so we lock the learning table. */
46c88433 960 if (!in_xbundle->bond) {
9583bc14
EJ
961 mac_entry_set_grat_arp_lock(mac);
962 } else if (mac_entry_is_grat_arp_locked(mac)) {
509c0149 963 goto out;
9583bc14
EJ
964 }
965 }
966
30618594 967 if (mac->port.p != in_xbundle->ofbundle) {
9583bc14
EJ
968 /* The log messages here could actually be useful in debugging,
969 * so keep the rate limit relatively high. */
970 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
971 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
972 "on port %s in VLAN %d",
46c88433
EJ
973 xbridge->name, ETH_ADDR_ARGS(flow->dl_src),
974 in_xbundle->name, vlan);
9583bc14 975
46c88433 976 mac->port.p = in_xbundle->ofbundle;
30618594 977 mac_learning_changed(xbridge->ml);
9583bc14 978 }
509c0149
EJ
979out:
980 ovs_rwlock_unlock(&xbridge->ml->rwlock);
9583bc14
EJ
981}
982
46c88433 983/* Determines whether packets in 'flow' within 'xbridge' should be forwarded or
9583bc14
EJ
984 * dropped. Returns true if they may be forwarded, false if they should be
985 * dropped.
986 *
46c88433 987 * 'in_port' must be the xport that corresponds to flow->in_port.
9583bc14
EJ
988 * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull).
989 *
990 * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as
991 * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as
992 * checked by input_vid_is_valid().
993 *
994 * May also add tags to '*tags', although the current implementation only does
995 * so in one special case.
996 */
997static bool
46c88433 998is_admissible(struct xlate_ctx *ctx, struct xport *in_port,
9583bc14
EJ
999 uint16_t vlan)
1000{
46c88433
EJ
1001 struct xbundle *in_xbundle = in_port->xbundle;
1002 const struct xbridge *xbridge = ctx->xbridge;
9583bc14 1003 struct flow *flow = &ctx->xin->flow;
9583bc14
EJ
1004
1005 /* Drop frames for reserved multicast addresses
1006 * only if forward_bpdu option is absent. */
46c88433 1007 if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) {
9583bc14
EJ
1008 xlate_report(ctx, "packet has reserved destination MAC, dropping");
1009 return false;
1010 }
1011
46c88433 1012 if (in_xbundle->bond) {
9583bc14
EJ
1013 struct mac_entry *mac;
1014
46c88433 1015 switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport,
4a1b8f30 1016 flow->dl_dst)) {
9583bc14
EJ
1017 case BV_ACCEPT:
1018 break;
1019
1020 case BV_DROP:
1021 xlate_report(ctx, "bonding refused admissibility, dropping");
1022 return false;
1023
1024 case BV_DROP_IF_MOVED:
509c0149 1025 ovs_rwlock_rdlock(&xbridge->ml->rwlock);
30618594 1026 mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan);
46c88433 1027 if (mac && mac->port.p != in_xbundle->ofbundle &&
9583bc14
EJ
1028 (!is_gratuitous_arp(flow, &ctx->xout->wc)
1029 || mac_entry_is_grat_arp_locked(mac))) {
509c0149 1030 ovs_rwlock_unlock(&xbridge->ml->rwlock);
9583bc14
EJ
1031 xlate_report(ctx, "SLB bond thinks this packet looped back, "
1032 "dropping");
1033 return false;
1034 }
509c0149 1035 ovs_rwlock_unlock(&xbridge->ml->rwlock);
9583bc14
EJ
1036 break;
1037 }
1038 }
1039
1040 return true;
1041}
1042
1043static void
1044xlate_normal(struct xlate_ctx *ctx)
1045{
33bf9176
BP
1046 struct flow_wildcards *wc = &ctx->xout->wc;
1047 struct flow *flow = &ctx->xin->flow;
46c88433
EJ
1048 struct xbundle *in_xbundle;
1049 struct xport *in_port;
9583bc14
EJ
1050 struct mac_entry *mac;
1051 uint16_t vlan;
1052 uint16_t vid;
1053
1054 ctx->xout->has_normal = true;
1055
33bf9176
BP
1056 memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src);
1057 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
1dd35f8a 1058 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
9583bc14 1059
46c88433
EJ
1060 in_xbundle = lookup_input_bundle(ctx->xbridge, flow->in_port.ofp_port,
1061 ctx->xin->packet != NULL, &in_port);
1062 if (!in_xbundle) {
9583bc14
EJ
1063 xlate_report(ctx, "no input bundle, dropping");
1064 return;
1065 }
1066
1067 /* Drop malformed frames. */
33bf9176
BP
1068 if (flow->dl_type == htons(ETH_TYPE_VLAN) &&
1069 !(flow->vlan_tci & htons(VLAN_CFI))) {
9583bc14
EJ
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 with partial "
1073 "VLAN tag received on port %s",
46c88433 1074 ctx->xbridge->name, in_xbundle->name);
9583bc14
EJ
1075 }
1076 xlate_report(ctx, "partial VLAN tag, dropping");
1077 return;
1078 }
1079
1080 /* Drop frames on bundles reserved for mirroring. */
46c88433 1081 if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) {
9583bc14
EJ
1082 if (ctx->xin->packet != NULL) {
1083 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1084 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
1085 "%s, which is reserved exclusively for mirroring",
46c88433 1086 ctx->xbridge->name, in_xbundle->name);
9583bc14
EJ
1087 }
1088 xlate_report(ctx, "input port is mirror output port, dropping");
1089 return;
1090 }
1091
1092 /* Check VLAN. */
33bf9176 1093 vid = vlan_tci_to_vid(flow->vlan_tci);
46c88433 1094 if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) {
9583bc14
EJ
1095 xlate_report(ctx, "disallowed VLAN VID for this input port, dropping");
1096 return;
1097 }
46c88433 1098 vlan = input_vid_to_vlan(in_xbundle, vid);
9583bc14
EJ
1099
1100 /* Check other admissibility requirements. */
1101 if (in_port && !is_admissible(ctx, in_port, vlan)) {
1102 return;
1103 }
1104
1105 /* Learn source MAC. */
1106 if (ctx->xin->may_learn) {
46c88433 1107 update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle);
9583bc14
EJ
1108 }
1109
1110 /* Determine output bundle. */
509c0149 1111 ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock);
30618594 1112 mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan);
9583bc14 1113 if (mac) {
46c88433
EJ
1114 struct xbundle *mac_xbundle = xbundle_lookup(mac->port.p);
1115 if (mac_xbundle && mac_xbundle != in_xbundle) {
9583bc14 1116 xlate_report(ctx, "forwarding to learned port");
46c88433
EJ
1117 output_normal(ctx, mac_xbundle, vlan);
1118 } else if (!mac_xbundle) {
1119 xlate_report(ctx, "learned port is unknown, dropping");
9583bc14
EJ
1120 } else {
1121 xlate_report(ctx, "learned port is input port, dropping");
1122 }
1123 } else {
46c88433 1124 struct xbundle *xbundle;
9583bc14
EJ
1125
1126 xlate_report(ctx, "no learned MAC for destination, flooding");
46c88433
EJ
1127 LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) {
1128 if (xbundle != in_xbundle
1129 && xbundle_includes_vlan(xbundle, vlan)
1130 && xbundle->floodable
1131 && !xbundle_mirror_out(ctx->xbridge, xbundle)) {
1132 output_normal(ctx, xbundle, vlan);
9583bc14
EJ
1133 }
1134 }
1135 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
1136 }
509c0149 1137 ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock);
9583bc14
EJ
1138}
1139
1140/* Compose SAMPLE action for sFlow or IPFIX. The given probability is
1141 * the number of packets out of UINT32_MAX to sample. The given
1142 * cookie is passed back in the callback for each sampled packet.
1143 */
1144static size_t
46c88433 1145compose_sample_action(const struct xbridge *xbridge,
9583bc14
EJ
1146 struct ofpbuf *odp_actions,
1147 const struct flow *flow,
1148 const uint32_t probability,
1149 const union user_action_cookie *cookie,
1150 const size_t cookie_size)
1151{
1152 size_t sample_offset, actions_offset;
89a8a7f0 1153 odp_port_t odp_port;
9583bc14 1154 int cookie_offset;
89a8a7f0 1155 uint32_t pid;
9583bc14
EJ
1156
1157 sample_offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SAMPLE);
1158
1159 nl_msg_put_u32(odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability);
1160
1161 actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS);
89a8a7f0
EJ
1162
1163 odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port);
1164 pid = dpif_port_get_pid(xbridge->dpif, odp_port);
1165 cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, odp_actions);
9583bc14
EJ
1166
1167 nl_msg_end_nested(odp_actions, actions_offset);
1168 nl_msg_end_nested(odp_actions, sample_offset);
1169 return cookie_offset;
1170}
1171
1172static void
46c88433
EJ
1173compose_sflow_cookie(const struct xbridge *xbridge, ovs_be16 vlan_tci,
1174 odp_port_t odp_port, unsigned int n_outputs,
1175 union user_action_cookie *cookie)
9583bc14
EJ
1176{
1177 int ifindex;
1178
1179 cookie->type = USER_ACTION_COOKIE_SFLOW;
1180 cookie->sflow.vlan_tci = vlan_tci;
1181
1182 /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output
1183 * port information") for the interpretation of cookie->output. */
1184 switch (n_outputs) {
1185 case 0:
1186 /* 0x40000000 | 256 means "packet dropped for unknown reason". */
1187 cookie->sflow.output = 0x40000000 | 256;
1188 break;
1189
1190 case 1:
46c88433 1191 ifindex = dpif_sflow_odp_port_to_ifindex(xbridge->sflow, odp_port);
9583bc14
EJ
1192 if (ifindex) {
1193 cookie->sflow.output = ifindex;
1194 break;
1195 }
1196 /* Fall through. */
1197 default:
1198 /* 0x80000000 means "multiple output ports. */
1199 cookie->sflow.output = 0x80000000 | n_outputs;
1200 break;
1201 }
1202}
1203
1204/* Compose SAMPLE action for sFlow bridge sampling. */
1205static size_t
46c88433 1206compose_sflow_action(const struct xbridge *xbridge,
9583bc14
EJ
1207 struct ofpbuf *odp_actions,
1208 const struct flow *flow,
4e022ec0 1209 odp_port_t odp_port)
9583bc14
EJ
1210{
1211 uint32_t probability;
1212 union user_action_cookie cookie;
1213
46c88433 1214 if (!xbridge->sflow || flow->in_port.ofp_port == OFPP_NONE) {
9583bc14
EJ
1215 return 0;
1216 }
1217
46c88433
EJ
1218 probability = dpif_sflow_get_probability(xbridge->sflow);
1219 compose_sflow_cookie(xbridge, htons(0), odp_port,
4e022ec0 1220 odp_port == ODPP_NONE ? 0 : 1, &cookie);
9583bc14 1221
46c88433 1222 return compose_sample_action(xbridge, odp_actions, flow, probability,
9583bc14
EJ
1223 &cookie, sizeof cookie.sflow);
1224}
1225
1226static void
1227compose_flow_sample_cookie(uint16_t probability, uint32_t collector_set_id,
1228 uint32_t obs_domain_id, uint32_t obs_point_id,
1229 union user_action_cookie *cookie)
1230{
1231 cookie->type = USER_ACTION_COOKIE_FLOW_SAMPLE;
1232 cookie->flow_sample.probability = probability;
1233 cookie->flow_sample.collector_set_id = collector_set_id;
1234 cookie->flow_sample.obs_domain_id = obs_domain_id;
1235 cookie->flow_sample.obs_point_id = obs_point_id;
1236}
1237
1238static void
1239compose_ipfix_cookie(union user_action_cookie *cookie)
1240{
1241 cookie->type = USER_ACTION_COOKIE_IPFIX;
1242}
1243
1244/* Compose SAMPLE action for IPFIX bridge sampling. */
1245static void
46c88433 1246compose_ipfix_action(const struct xbridge *xbridge,
9583bc14
EJ
1247 struct ofpbuf *odp_actions,
1248 const struct flow *flow)
1249{
1250 uint32_t probability;
1251 union user_action_cookie cookie;
1252
46c88433 1253 if (!xbridge->ipfix || flow->in_port.ofp_port == OFPP_NONE) {
9583bc14
EJ
1254 return;
1255 }
1256
46c88433 1257 probability = dpif_ipfix_get_bridge_exporter_probability(xbridge->ipfix);
9583bc14
EJ
1258 compose_ipfix_cookie(&cookie);
1259
46c88433 1260 compose_sample_action(xbridge, odp_actions, flow, probability,
9583bc14
EJ
1261 &cookie, sizeof cookie.ipfix);
1262}
1263
1264/* SAMPLE action for sFlow must be first action in any given list of
1265 * actions. At this point we do not have all information required to
1266 * build it. So try to build sample action as complete as possible. */
1267static void
1268add_sflow_action(struct xlate_ctx *ctx)
1269{
46c88433 1270 ctx->user_cookie_offset = compose_sflow_action(ctx->xbridge,
9583bc14 1271 &ctx->xout->odp_actions,
4e022ec0 1272 &ctx->xin->flow, ODPP_NONE);
9583bc14
EJ
1273 ctx->sflow_odp_port = 0;
1274 ctx->sflow_n_outputs = 0;
1275}
1276
1277/* SAMPLE action for IPFIX must be 1st or 2nd action in any given list
1278 * of actions, eventually after the SAMPLE action for sFlow. */
1279static void
1280add_ipfix_action(struct xlate_ctx *ctx)
1281{
46c88433 1282 compose_ipfix_action(ctx->xbridge, &ctx->xout->odp_actions,
9583bc14
EJ
1283 &ctx->xin->flow);
1284}
1285
1286/* Fix SAMPLE action according to data collected while composing ODP actions.
1287 * We need to fix SAMPLE actions OVS_SAMPLE_ATTR_ACTIONS attribute, i.e. nested
1288 * USERSPACE action's user-cookie which is required for sflow. */
1289static void
1290fix_sflow_action(struct xlate_ctx *ctx)
1291{
1292 const struct flow *base = &ctx->base_flow;
1293 union user_action_cookie *cookie;
1294
1295 if (!ctx->user_cookie_offset) {
1296 return;
1297 }
1298
1299 cookie = ofpbuf_at(&ctx->xout->odp_actions, ctx->user_cookie_offset,
1300 sizeof cookie->sflow);
1301 ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW);
1302
46c88433 1303 compose_sflow_cookie(ctx->xbridge, base->vlan_tci,
9583bc14
EJ
1304 ctx->sflow_odp_port, ctx->sflow_n_outputs, cookie);
1305}
1306
db7d4e46 1307static enum slow_path_reason
642dc74d 1308process_special(struct xlate_ctx *ctx, const struct flow *flow,
46c88433 1309 const struct xport *xport, const struct ofpbuf *packet)
db7d4e46 1310{
642dc74d 1311 struct flow_wildcards *wc = &ctx->xout->wc;
46c88433 1312 const struct xbridge *xbridge = ctx->xbridge;
642dc74d 1313
46c88433 1314 if (!xport) {
db7d4e46 1315 return 0;
46c88433 1316 } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) {
db7d4e46 1317 if (packet) {
46c88433 1318 cfm_process_heartbeat(xport->cfm, packet);
db7d4e46
JP
1319 }
1320 return SLOW_CFM;
fab52e16 1321 } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) {
db7d4e46 1322 if (packet) {
46c88433 1323 bfd_process_packet(xport->bfd, flow, packet);
db7d4e46
JP
1324 }
1325 return SLOW_BFD;
46c88433 1326 } else if (xport->xbundle && xport->xbundle->lacp
db7d4e46
JP
1327 && flow->dl_type == htons(ETH_TYPE_LACP)) {
1328 if (packet) {
46c88433 1329 lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet);
db7d4e46
JP
1330 }
1331 return SLOW_LACP;
9d189a50 1332 } else if (xbridge->stp && stp_should_process_flow(flow, wc)) {
db7d4e46 1333 if (packet) {
9d189a50 1334 stp_process_packet(xport, packet);
db7d4e46
JP
1335 }
1336 return SLOW_STP;
1337 } else {
1338 return 0;
1339 }
1340}
1341
9583bc14 1342static void
4e022ec0 1343compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port,
9583bc14
EJ
1344 bool check_stp)
1345{
46c88433 1346 const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port);
1dd35f8a 1347 struct flow_wildcards *wc = &ctx->xout->wc;
33bf9176 1348 struct flow *flow = &ctx->xin->flow;
9583bc14
EJ
1349 ovs_be16 flow_vlan_tci;
1350 uint32_t flow_skb_mark;
1351 uint8_t flow_nw_tos;
4e022ec0 1352 odp_port_t out_port, odp_port;
ca077186 1353 uint8_t dscp;
9583bc14
EJ
1354
1355 /* If 'struct flow' gets additional metadata, we'll need to zero it out
1356 * before traversing a patch port. */
1357 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 20);
1358
46c88433 1359 if (!xport) {
9583bc14
EJ
1360 xlate_report(ctx, "Nonexistent output port");
1361 return;
46c88433 1362 } else if (xport->config & OFPUTIL_PC_NO_FWD) {
9583bc14
EJ
1363 xlate_report(ctx, "OFPPC_NO_FWD set, skipping output");
1364 return;
9d189a50 1365 } else if (check_stp && !xport_stp_forward_state(xport)) {
9583bc14
EJ
1366 xlate_report(ctx, "STP not in forwarding state, skipping output");
1367 return;
1368 }
1369
46c88433
EJ
1370 if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) {
1371 ctx->xout->mirrors |= xbundle_mirror_dst(xport->xbundle->xbridge,
1372 xport->xbundle);
cdf5d3a5
EJ
1373 }
1374
46c88433
EJ
1375 if (xport->peer) {
1376 const struct xport *peer = xport->peer;
9583bc14 1377 struct flow old_flow = ctx->xin->flow;
9583bc14 1378 enum slow_path_reason special;
9583bc14 1379
46c88433
EJ
1380 ctx->xbridge = peer->xbridge;
1381 flow->in_port.ofp_port = peer->ofp_port;
33bf9176
BP
1382 flow->metadata = htonll(0);
1383 memset(&flow->tunnel, 0, sizeof flow->tunnel);
1384 memset(flow->regs, 0, sizeof flow->regs);
9583bc14 1385
642dc74d 1386 special = process_special(ctx, &ctx->xin->flow, peer,
9583bc14
EJ
1387 ctx->xin->packet);
1388 if (special) {
1389 ctx->xout->slow = special;
ddd3c975 1390 } else if (may_receive(peer, ctx)) {
9d189a50 1391 if (xport_stp_forward_state(peer)) {
4e022ec0 1392 xlate_table_action(ctx, flow->in_port.ofp_port, 0, true);
9583bc14
EJ
1393 } else {
1394 /* Forwarding is disabled by STP. Let OFPP_NORMAL and the
1395 * learning action look at the packet, then drop it. */
1396 struct flow old_base_flow = ctx->base_flow;
1397 size_t old_size = ctx->xout->odp_actions.size;
cdf5d3a5 1398 mirror_mask_t old_mirrors = ctx->xout->mirrors;
4e022ec0 1399 xlate_table_action(ctx, flow->in_port.ofp_port, 0, true);
cdf5d3a5 1400 ctx->xout->mirrors = old_mirrors;
9583bc14
EJ
1401 ctx->base_flow = old_base_flow;
1402 ctx->xout->odp_actions.size = old_size;
1403 }
1404 }
1405
1406 ctx->xin->flow = old_flow;
46c88433 1407 ctx->xbridge = xport->xbundle->xbridge;
9583bc14
EJ
1408
1409 if (ctx->xin->resubmit_stats) {
46c88433
EJ
1410 netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats);
1411 netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats);
9583bc14
EJ
1412 }
1413
1414 return;
1415 }
1416
33bf9176
BP
1417 flow_vlan_tci = flow->vlan_tci;
1418 flow_skb_mark = flow->skb_mark;
1419 flow_nw_tos = flow->nw_tos;
9583bc14 1420
55954f6e 1421 if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) {
1dd35f8a 1422 wc->masks.nw_tos |= IP_ECN_MASK;
33bf9176 1423 flow->nw_tos &= ~IP_DSCP_MASK;
ca077186 1424 flow->nw_tos |= dscp;
9583bc14
EJ
1425 }
1426
46c88433 1427 if (xport->is_tunnel) {
9583bc14
EJ
1428 /* Save tunnel metadata so that changes made due to
1429 * the Logical (tunnel) Port are not visible for any further
1430 * matches, while explicit set actions on tunnel metadata are.
1431 */
33bf9176 1432 struct flow_tnl flow_tnl = flow->tunnel;
46c88433 1433 odp_port = tnl_port_send(xport->ofport, flow, &ctx->xout->wc);
4e022ec0 1434 if (odp_port == ODPP_NONE) {
9583bc14
EJ
1435 xlate_report(ctx, "Tunneling decided against output");
1436 goto out; /* restore flow_nw_tos */
1437 }
33bf9176 1438 if (flow->tunnel.ip_dst == ctx->orig_tunnel_ip_dst) {
9583bc14
EJ
1439 xlate_report(ctx, "Not tunneling to our own address");
1440 goto out; /* restore flow_nw_tos */
1441 }
1442 if (ctx->xin->resubmit_stats) {
46c88433 1443 netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats);
9583bc14
EJ
1444 }
1445 out_port = odp_port;
33bf9176 1446 commit_odp_tunnel_action(flow, &ctx->base_flow,
9583bc14 1447 &ctx->xout->odp_actions);
33bf9176 1448 flow->tunnel = flow_tnl; /* Restore tunnel metadata */
9583bc14 1449 } else {
4e022ec0 1450 ofp_port_t vlandev_port;
1dd35f8a 1451
46c88433
EJ
1452 odp_port = xport->odp_port;
1453 if (ofproto_has_vlan_splinters(ctx->xbridge->ofproto)) {
1dd35f8a
JP
1454 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
1455 }
46c88433 1456 vlandev_port = vsp_realdev_to_vlandev(ctx->xbridge->ofproto, ofp_port,
33bf9176 1457 flow->vlan_tci);
9583bc14
EJ
1458 if (vlandev_port == ofp_port) {
1459 out_port = odp_port;
1460 } else {
46c88433 1461 out_port = ofp_port_to_odp_port(ctx->xbridge, vlandev_port);
33bf9176 1462 flow->vlan_tci = htons(0);
9583bc14 1463 }
33bf9176 1464 flow->skb_mark &= ~IPSEC_MARK;
9583bc14 1465 }
9583bc14 1466
4e022ec0 1467 if (out_port != ODPP_NONE) {
1dd35f8a
JP
1468 commit_odp_actions(flow, &ctx->base_flow,
1469 &ctx->xout->odp_actions, &ctx->xout->wc);
4e022ec0
AW
1470 nl_msg_put_odp_port(&ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
1471 out_port);
9583bc14 1472
6cbbf4fa
EJ
1473 ctx->sflow_odp_port = odp_port;
1474 ctx->sflow_n_outputs++;
1475 ctx->xout->nf_output_iface = ofp_port;
1476 }
1477
1478 out:
9583bc14 1479 /* Restore flow */
33bf9176
BP
1480 flow->vlan_tci = flow_vlan_tci;
1481 flow->skb_mark = flow_skb_mark;
33bf9176 1482 flow->nw_tos = flow_nw_tos;
9583bc14
EJ
1483}
1484
1485static void
4e022ec0 1486compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port)
9583bc14
EJ
1487{
1488 compose_output_action__(ctx, ofp_port, true);
1489}
1490
9583bc14
EJ
1491/* Common rule processing in one place to avoid duplicating code. */
1492static struct rule_dpif *
1493ctx_rule_hooks(struct xlate_ctx *ctx, struct rule_dpif *rule,
1494 bool may_packet_in)
1495{
1496 if (ctx->xin->resubmit_hook) {
4d0acc70 1497 ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse);
9583bc14
EJ
1498 }
1499 if (rule == NULL && may_packet_in) {
1500 /* XXX
1501 * check if table configuration flags
1502 * OFPTC_TABLE_MISS_CONTROLLER, default.
1503 * OFPTC_TABLE_MISS_CONTINUE,
1504 * OFPTC_TABLE_MISS_DROP
1505 * When OF1.0, OFPTC_TABLE_MISS_CONTINUE is used. What to do?
1506 */
46c88433 1507 rule = rule_dpif_miss_rule(ctx->xbridge->ofproto, &ctx->xin->flow);
9583bc14
EJ
1508 }
1509 if (rule && ctx->xin->resubmit_stats) {
1510 rule_credit_stats(rule, ctx->xin->resubmit_stats);
1511 }
1512 return rule;
1513}
1514
1515static void
1516xlate_table_action(struct xlate_ctx *ctx,
4e022ec0 1517 ofp_port_t in_port, uint8_t table_id, bool may_packet_in)
9583bc14
EJ
1518{
1519 if (ctx->recurse < MAX_RESUBMIT_RECURSION) {
1520 struct rule_dpif *rule;
4e022ec0 1521 ofp_port_t old_in_port = ctx->xin->flow.in_port.ofp_port;
9583bc14
EJ
1522 uint8_t old_table_id = ctx->table_id;
1523
1524 ctx->table_id = table_id;
1525
1526 /* Look up a flow with 'in_port' as the input port. */
4e022ec0 1527 ctx->xin->flow.in_port.ofp_port = in_port;
46c88433
EJ
1528 rule = rule_dpif_lookup_in_table(ctx->xbridge->ofproto,
1529 &ctx->xin->flow, &ctx->xout->wc,
1530 table_id);
9583bc14 1531
9583bc14
EJ
1532 /* Restore the original input port. Otherwise OFPP_NORMAL and
1533 * OFPP_IN_PORT will have surprising behavior. */
4e022ec0 1534 ctx->xin->flow.in_port.ofp_port = old_in_port;
9583bc14
EJ
1535
1536 rule = ctx_rule_hooks(ctx, rule, may_packet_in);
1537
1538 if (rule) {
1539 struct rule_dpif *old_rule = ctx->rule;
1540
1541 ctx->recurse++;
1542 ctx->rule = rule;
1543 do_xlate_actions(rule->up.ofpacts, rule->up.ofpacts_len, ctx);
1544 ctx->rule = old_rule;
1545 ctx->recurse--;
1546 }
1547
1548 ctx->table_id = old_table_id;
1549 } else {
1550 static struct vlog_rate_limit recurse_rl = VLOG_RATE_LIMIT_INIT(1, 1);
1551
1552 VLOG_ERR_RL(&recurse_rl, "resubmit actions recursed over %d times",
1553 MAX_RESUBMIT_RECURSION);
9583bc14
EJ
1554 }
1555}
1556
1557static void
1558xlate_ofpact_resubmit(struct xlate_ctx *ctx,
1559 const struct ofpact_resubmit *resubmit)
1560{
4e022ec0 1561 ofp_port_t in_port;
9583bc14
EJ
1562 uint8_t table_id;
1563
1564 in_port = resubmit->in_port;
1565 if (in_port == OFPP_IN_PORT) {
4e022ec0 1566 in_port = ctx->xin->flow.in_port.ofp_port;
9583bc14
EJ
1567 }
1568
1569 table_id = resubmit->table_id;
1570 if (table_id == 255) {
1571 table_id = ctx->table_id;
1572 }
1573
1574 xlate_table_action(ctx, in_port, table_id, false);
1575}
1576
1577static void
1578flood_packets(struct xlate_ctx *ctx, bool all)
1579{
46c88433 1580 const struct xport *xport;
9583bc14 1581
46c88433
EJ
1582 HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) {
1583 if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) {
9583bc14
EJ
1584 continue;
1585 }
1586
1587 if (all) {
46c88433
EJ
1588 compose_output_action__(ctx, xport->ofp_port, false);
1589 } else if (!(xport->config & OFPUTIL_PC_NO_FLOOD)) {
1590 compose_output_action(ctx, xport->ofp_port);
9583bc14
EJ
1591 }
1592 }
1593
1594 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
1595}
1596
1597static void
1598execute_controller_action(struct xlate_ctx *ctx, int len,
1599 enum ofp_packet_in_reason reason,
1600 uint16_t controller_id)
1601{
1602 struct ofputil_packet_in pin;
1603 struct ofpbuf *packet;
1604 struct flow key;
1605
1606 ovs_assert(!ctx->xout->slow || ctx->xout->slow == SLOW_CONTROLLER);
1607 ctx->xout->slow = SLOW_CONTROLLER;
1608 if (!ctx->xin->packet) {
1609 return;
1610 }
1611
1612 packet = ofpbuf_clone(ctx->xin->packet);
1613
1614 key.skb_priority = 0;
1615 key.skb_mark = 0;
1616 memset(&key.tunnel, 0, sizeof key.tunnel);
1617
1618 commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
1dd35f8a 1619 &ctx->xout->odp_actions, &ctx->xout->wc);
9583bc14
EJ
1620
1621 odp_execute_actions(NULL, packet, &key, ctx->xout->odp_actions.data,
1622 ctx->xout->odp_actions.size, NULL, NULL);
1623
1624 pin.packet = packet->data;
1625 pin.packet_len = packet->size;
1626 pin.reason = reason;
1627 pin.controller_id = controller_id;
1628 pin.table_id = ctx->table_id;
1629 pin.cookie = ctx->rule ? ctx->rule->up.flow_cookie : 0;
1630
1631 pin.send_len = len;
1632 flow_get_metadata(&ctx->xin->flow, &pin.fmd);
1633
46c88433 1634 ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, &pin);
9583bc14
EJ
1635 ofpbuf_delete(packet);
1636}
1637
1638static void
9cfef3d0 1639compose_mpls_push_action(struct xlate_ctx *ctx, ovs_be16 eth_type)
9583bc14 1640{
33bf9176
BP
1641 struct flow_wildcards *wc = &ctx->xout->wc;
1642 struct flow *flow = &ctx->xin->flow;
1643
9583bc14
EJ
1644 ovs_assert(eth_type_mpls(eth_type));
1645
33bf9176
BP
1646 memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
1647 memset(&wc->masks.mpls_depth, 0xff, sizeof wc->masks.mpls_depth);
9583bc14 1648
33bf9176
BP
1649 if (flow->mpls_depth) {
1650 flow->mpls_lse &= ~htonl(MPLS_BOS_MASK);
1651 flow->mpls_depth++;
9583bc14
EJ
1652 } else {
1653 ovs_be32 label;
1654 uint8_t tc, ttl;
1655
33bf9176 1656 if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
9583bc14
EJ
1657 label = htonl(0x2); /* IPV6 Explicit Null. */
1658 } else {
1659 label = htonl(0x0); /* IPV4 Explicit Null. */
1660 }
1dd35f8a
JP
1661 wc->masks.nw_tos |= IP_DSCP_MASK;
1662 wc->masks.nw_ttl = 0xff;
33bf9176
BP
1663 tc = (flow->nw_tos & IP_DSCP_MASK) >> 2;
1664 ttl = flow->nw_ttl ? flow->nw_ttl : 0x40;
1665 flow->mpls_lse = set_mpls_lse_values(ttl, tc, 1, label);
1666 flow->mpls_depth = 1;
9583bc14 1667 }
33bf9176 1668 flow->dl_type = eth_type;
9583bc14
EJ
1669}
1670
1671static void
9cfef3d0 1672compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type)
9583bc14 1673{
33bf9176
BP
1674 struct flow_wildcards *wc = &ctx->xout->wc;
1675 struct flow *flow = &ctx->xin->flow;
1676
9583bc14
EJ
1677 ovs_assert(eth_type_mpls(ctx->xin->flow.dl_type));
1678 ovs_assert(!eth_type_mpls(eth_type));
1679
33bf9176
BP
1680 memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
1681 memset(&wc->masks.mpls_depth, 0xff, sizeof wc->masks.mpls_depth);
1682
1683 if (flow->mpls_depth) {
1684 flow->mpls_depth--;
1685 flow->mpls_lse = htonl(0);
1686 if (!flow->mpls_depth) {
1687 flow->dl_type = eth_type;
9583bc14
EJ
1688 }
1689 }
1690}
1691
1692static bool
1693compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids)
1694{
33bf9176
BP
1695 struct flow *flow = &ctx->xin->flow;
1696
1697 if (!is_ip_any(flow)) {
9583bc14
EJ
1698 return false;
1699 }
1700
1dd35f8a 1701 ctx->xout->wc.masks.nw_ttl = 0xff;
33bf9176
BP
1702 if (flow->nw_ttl > 1) {
1703 flow->nw_ttl--;
9583bc14
EJ
1704 return false;
1705 } else {
1706 size_t i;
1707
1708 for (i = 0; i < ids->n_controllers; i++) {
1709 execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL,
1710 ids->cnt_ids[i]);
1711 }
1712
1713 /* Stop processing for current table. */
1714 return true;
1715 }
1716}
1717
1718static bool
9cfef3d0 1719compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl)
9583bc14
EJ
1720{
1721 if (!eth_type_mpls(ctx->xin->flow.dl_type)) {
1722 return true;
1723 }
1724
f74e7df7 1725 ctx->xout->wc.masks.mpls_lse |= htonl(MPLS_TTL_MASK);
9583bc14
EJ
1726 set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse, ttl);
1727 return false;
1728}
1729
1730static bool
9cfef3d0 1731compose_dec_mpls_ttl_action(struct xlate_ctx *ctx)
9583bc14 1732{
33bf9176
BP
1733 struct flow *flow = &ctx->xin->flow;
1734 uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse);
1dd35f8a
JP
1735 struct flow_wildcards *wc = &ctx->xout->wc;
1736
1dd35f8a 1737 memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
9583bc14 1738
33bf9176 1739 if (!eth_type_mpls(flow->dl_type)) {
9583bc14
EJ
1740 return false;
1741 }
1742
1743 if (ttl > 1) {
1744 ttl--;
33bf9176 1745 set_mpls_lse_ttl(&flow->mpls_lse, ttl);
9583bc14
EJ
1746 return false;
1747 } else {
1748 execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0);
1749
1750 /* Stop processing for current table. */
1751 return true;
1752 }
1753}
1754
1755static void
1756xlate_output_action(struct xlate_ctx *ctx,
4e022ec0 1757 ofp_port_t port, uint16_t max_len, bool may_packet_in)
9583bc14 1758{
4e022ec0 1759 ofp_port_t prev_nf_output_iface = ctx->xout->nf_output_iface;
9583bc14
EJ
1760
1761 ctx->xout->nf_output_iface = NF_OUT_DROP;
1762
1763 switch (port) {
1764 case OFPP_IN_PORT:
4e022ec0 1765 compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port);
9583bc14
EJ
1766 break;
1767 case OFPP_TABLE:
4e022ec0
AW
1768 xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
1769 0, may_packet_in);
9583bc14
EJ
1770 break;
1771 case OFPP_NORMAL:
1772 xlate_normal(ctx);
1773 break;
1774 case OFPP_FLOOD:
1775 flood_packets(ctx, false);
1776 break;
1777 case OFPP_ALL:
1778 flood_packets(ctx, true);
1779 break;
1780 case OFPP_CONTROLLER:
1781 execute_controller_action(ctx, max_len, OFPR_ACTION, 0);
1782 break;
1783 case OFPP_NONE:
1784 break;
1785 case OFPP_LOCAL:
1786 default:
4e022ec0 1787 if (port != ctx->xin->flow.in_port.ofp_port) {
9583bc14
EJ
1788 compose_output_action(ctx, port);
1789 } else {
1790 xlate_report(ctx, "skipping output to input port");
1791 }
1792 break;
1793 }
1794
1795 if (prev_nf_output_iface == NF_OUT_FLOOD) {
1796 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
1797 } else if (ctx->xout->nf_output_iface == NF_OUT_DROP) {
1798 ctx->xout->nf_output_iface = prev_nf_output_iface;
1799 } else if (prev_nf_output_iface != NF_OUT_DROP &&
1800 ctx->xout->nf_output_iface != NF_OUT_FLOOD) {
1801 ctx->xout->nf_output_iface = NF_OUT_MULTI;
1802 }
1803}
1804
1805static void
1806xlate_output_reg_action(struct xlate_ctx *ctx,
1807 const struct ofpact_output_reg *or)
1808{
1809 uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow);
1810 if (port <= UINT16_MAX) {
1811 union mf_subvalue value;
1812
1813 memset(&value, 0xff, sizeof value);
1814 mf_write_subfield_flow(&or->src, &value, &ctx->xout->wc.masks);
4e022ec0
AW
1815 xlate_output_action(ctx, u16_to_ofp(port),
1816 or->max_len, false);
9583bc14
EJ
1817 }
1818}
1819
1820static void
1821xlate_enqueue_action(struct xlate_ctx *ctx,
1822 const struct ofpact_enqueue *enqueue)
1823{
4e022ec0 1824 ofp_port_t ofp_port = enqueue->port;
9583bc14
EJ
1825 uint32_t queue_id = enqueue->queue;
1826 uint32_t flow_priority, priority;
1827 int error;
1828
1829 /* Translate queue to priority. */
89a8a7f0 1830 error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority);
9583bc14
EJ
1831 if (error) {
1832 /* Fall back to ordinary output action. */
1833 xlate_output_action(ctx, enqueue->port, 0, false);
1834 return;
1835 }
1836
1837 /* Check output port. */
1838 if (ofp_port == OFPP_IN_PORT) {
4e022ec0
AW
1839 ofp_port = ctx->xin->flow.in_port.ofp_port;
1840 } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) {
9583bc14
EJ
1841 return;
1842 }
1843
1844 /* Add datapath actions. */
1845 flow_priority = ctx->xin->flow.skb_priority;
1846 ctx->xin->flow.skb_priority = priority;
1847 compose_output_action(ctx, ofp_port);
1848 ctx->xin->flow.skb_priority = flow_priority;
1849
1850 /* Update NetFlow output port. */
1851 if (ctx->xout->nf_output_iface == NF_OUT_DROP) {
1852 ctx->xout->nf_output_iface = ofp_port;
1853 } else if (ctx->xout->nf_output_iface != NF_OUT_FLOOD) {
1854 ctx->xout->nf_output_iface = NF_OUT_MULTI;
1855 }
1856}
1857
1858static void
1859xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id)
1860{
1861 uint32_t skb_priority;
1862
89a8a7f0 1863 if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) {
9583bc14
EJ
1864 ctx->xin->flow.skb_priority = skb_priority;
1865 } else {
1866 /* Couldn't translate queue to a priority. Nothing to do. A warning
1867 * has already been logged. */
1868 }
1869}
1870
1871static bool
46c88433 1872slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_)
9583bc14 1873{
46c88433
EJ
1874 const struct xbridge *xbridge = xbridge_;
1875 struct xport *port;
9583bc14
EJ
1876
1877 switch (ofp_port) {
1878 case OFPP_IN_PORT:
1879 case OFPP_TABLE:
1880 case OFPP_NORMAL:
1881 case OFPP_FLOOD:
1882 case OFPP_ALL:
1883 case OFPP_NONE:
1884 return true;
1885 case OFPP_CONTROLLER: /* Not supported by the bundle action. */
1886 return false;
1887 default:
46c88433 1888 port = get_ofp_port(xbridge, ofp_port);
9583bc14
EJ
1889 return port ? port->may_enable : false;
1890 }
1891}
1892
1893static void
1894xlate_bundle_action(struct xlate_ctx *ctx,
1895 const struct ofpact_bundle *bundle)
1896{
4e022ec0 1897 ofp_port_t port;
9583bc14
EJ
1898
1899 port = bundle_execute(bundle, &ctx->xin->flow, &ctx->xout->wc,
46c88433
EJ
1900 slave_enabled_cb,
1901 CONST_CAST(struct xbridge *, ctx->xbridge));
9583bc14 1902 if (bundle->dst.field) {
f74e7df7
JP
1903 nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow,
1904 &ctx->xout->wc);
9583bc14
EJ
1905 } else {
1906 xlate_output_action(ctx, port, 0, false);
1907 }
1908}
1909
1910static void
1911xlate_learn_action(struct xlate_ctx *ctx,
1912 const struct ofpact_learn *learn)
1913{
3d9c5e58 1914 struct ofputil_flow_mod *fm;
9583bc14 1915 struct ofpbuf ofpacts;
9583bc14
EJ
1916
1917 ctx->xout->has_learn = true;
1918
1919 learn_mask(learn, &ctx->xout->wc);
1920
1921 if (!ctx->xin->may_learn) {
1922 return;
1923 }
1924
3d9c5e58
EJ
1925 fm = xmalloc(sizeof *fm);
1926 ofpbuf_init(&ofpacts, 0);
1927 learn_execute(learn, &ctx->xin->flow, fm, &ofpacts);
9583bc14 1928
3d9c5e58 1929 ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm);
9583bc14
EJ
1930}
1931
1932/* Reduces '*timeout' to no more than 'max'. A value of zero in either case
1933 * means "infinite". */
1934static void
1935reduce_timeout(uint16_t max, uint16_t *timeout)
1936{
1937 if (max && (!*timeout || *timeout > max)) {
1938 *timeout = max;
1939 }
1940}
1941
1942static void
1943xlate_fin_timeout(struct xlate_ctx *ctx,
1944 const struct ofpact_fin_timeout *oft)
1945{
1946 if (ctx->xin->tcp_flags & (TCP_FIN | TCP_RST) && ctx->rule) {
1947 struct rule_dpif *rule = ctx->rule;
1948
2a5183ac
JP
1949 if (list_is_empty(&rule->up.expirable)) {
1950 list_insert(&rule->up.ofproto->expirable, &rule->up.expirable);
1951 }
1952
9583bc14
EJ
1953 reduce_timeout(oft->fin_idle_timeout, &rule->up.idle_timeout);
1954 reduce_timeout(oft->fin_hard_timeout, &rule->up.hard_timeout);
1955 }
1956}
1957
1958static void
1959xlate_sample_action(struct xlate_ctx *ctx,
1960 const struct ofpact_sample *os)
1961{
1962 union user_action_cookie cookie;
1963 /* Scale the probability from 16-bit to 32-bit while representing
1964 * the same percentage. */
1965 uint32_t probability = (os->probability << 16) | os->probability;
1966
1967 commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
1dd35f8a 1968 &ctx->xout->odp_actions, &ctx->xout->wc);
9583bc14
EJ
1969
1970 compose_flow_sample_cookie(os->probability, os->collector_set_id,
1971 os->obs_domain_id, os->obs_point_id, &cookie);
46c88433 1972 compose_sample_action(ctx->xbridge, &ctx->xout->odp_actions, &ctx->xin->flow,
9583bc14
EJ
1973 probability, &cookie, sizeof cookie.flow_sample);
1974}
1975
1976static bool
46c88433 1977may_receive(const struct xport *xport, struct xlate_ctx *ctx)
9583bc14 1978{
46c88433
EJ
1979 if (xport->config & (eth_addr_equals(ctx->xin->flow.dl_dst, eth_addr_stp)
1980 ? OFPUTIL_PC_NO_RECV_STP
1981 : OFPUTIL_PC_NO_RECV)) {
9583bc14
EJ
1982 return false;
1983 }
1984
1985 /* Only drop packets here if both forwarding and learning are
1986 * disabled. If just learning is enabled, we need to have
1987 * OFPP_NORMAL and the learning action have a look at the packet
1988 * before we can drop it. */
9d189a50 1989 if (!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) {
9583bc14
EJ
1990 return false;
1991 }
1992
1993 return true;
1994}
1995
1996static bool
1997tunnel_ecn_ok(struct xlate_ctx *ctx)
1998{
1999 if (is_ip_any(&ctx->base_flow)
2000 && (ctx->xin->flow.tunnel.ip_tos & IP_ECN_MASK) == IP_ECN_CE) {
2001 if ((ctx->base_flow.nw_tos & IP_ECN_MASK) == IP_ECN_NOT_ECT) {
2002 VLOG_WARN_RL(&rl, "dropping tunnel packet marked ECN CE"
2003 " but is not ECN capable");
2004 return false;
2005 } else {
2006 /* Set the ECN CE value in the tunneled packet. */
2007 ctx->xin->flow.nw_tos |= IP_ECN_CE;
2008 }
2009 }
2010
2011 return true;
2012}
2013
2014static void
2015do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len,
2016 struct xlate_ctx *ctx)
2017{
33bf9176
BP
2018 struct flow_wildcards *wc = &ctx->xout->wc;
2019 struct flow *flow = &ctx->xin->flow;
9583bc14
EJ
2020 bool was_evictable = true;
2021 const struct ofpact *a;
2022
2023 if (ctx->rule) {
2024 /* Don't let the rule we're working on get evicted underneath us. */
2025 was_evictable = ctx->rule->up.evictable;
2026 ctx->rule->up.evictable = false;
2027 }
2028
9583bc14
EJ
2029 OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) {
2030 struct ofpact_controller *controller;
2031 const struct ofpact_metadata *metadata;
2032
2033 if (ctx->exit) {
2034 break;
2035 }
2036
2037 switch (a->type) {
2038 case OFPACT_OUTPUT:
2039 xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port,
2040 ofpact_get_OUTPUT(a)->max_len, true);
2041 break;
2042
2043 case OFPACT_CONTROLLER:
2044 controller = ofpact_get_CONTROLLER(a);
2045 execute_controller_action(ctx, controller->max_len,
2046 controller->reason,
2047 controller->controller_id);
2048 break;
2049
2050 case OFPACT_ENQUEUE:
2051 xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a));
2052 break;
2053
2054 case OFPACT_SET_VLAN_VID:
f74e7df7 2055 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
33bf9176
BP
2056 flow->vlan_tci &= ~htons(VLAN_VID_MASK);
2057 flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid)
2058 | htons(VLAN_CFI));
9583bc14
EJ
2059 break;
2060
2061 case OFPACT_SET_VLAN_PCP:
f74e7df7 2062 wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI);
33bf9176
BP
2063 flow->vlan_tci &= ~htons(VLAN_PCP_MASK);
2064 flow->vlan_tci |=
9583bc14
EJ
2065 htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp << VLAN_PCP_SHIFT)
2066 | VLAN_CFI);
2067 break;
2068
2069 case OFPACT_STRIP_VLAN:
f74e7df7 2070 memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
33bf9176 2071 flow->vlan_tci = htons(0);
9583bc14
EJ
2072 break;
2073
2074 case OFPACT_PUSH_VLAN:
2075 /* XXX 802.1AD(QinQ) */
f74e7df7 2076 memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
33bf9176 2077 flow->vlan_tci = htons(VLAN_CFI);
9583bc14
EJ
2078 break;
2079
2080 case OFPACT_SET_ETH_SRC:
f74e7df7 2081 memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src);
33bf9176 2082 memcpy(flow->dl_src, ofpact_get_SET_ETH_SRC(a)->mac, ETH_ADDR_LEN);
9583bc14
EJ
2083 break;
2084
2085 case OFPACT_SET_ETH_DST:
f74e7df7 2086 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
33bf9176 2087 memcpy(flow->dl_dst, ofpact_get_SET_ETH_DST(a)->mac, ETH_ADDR_LEN);
9583bc14
EJ
2088 break;
2089
2090 case OFPACT_SET_IPV4_SRC:
f74e7df7 2091 memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src);
33bf9176
BP
2092 if (flow->dl_type == htons(ETH_TYPE_IP)) {
2093 flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4;
9583bc14
EJ
2094 }
2095 break;
2096
2097 case OFPACT_SET_IPV4_DST:
f74e7df7 2098 memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst);
33bf9176
BP
2099 if (flow->dl_type == htons(ETH_TYPE_IP)) {
2100 flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4;
9583bc14
EJ
2101 }
2102 break;
2103
2104 case OFPACT_SET_IPV4_DSCP:
f74e7df7 2105 wc->masks.nw_tos |= IP_DSCP_MASK;
9583bc14 2106 /* OpenFlow 1.0 only supports IPv4. */
33bf9176
BP
2107 if (flow->dl_type == htons(ETH_TYPE_IP)) {
2108 flow->nw_tos &= ~IP_DSCP_MASK;
2109 flow->nw_tos |= ofpact_get_SET_IPV4_DSCP(a)->dscp;
9583bc14
EJ
2110 }
2111 break;
2112
2113 case OFPACT_SET_L4_SRC_PORT:
33bf9176 2114 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
f74e7df7 2115 memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src);
33bf9176
BP
2116 if (is_ip_any(flow)) {
2117 flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port);
9583bc14
EJ
2118 }
2119 break;
2120
2121 case OFPACT_SET_L4_DST_PORT:
33bf9176 2122 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
f74e7df7 2123 memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
33bf9176
BP
2124 if (is_ip_any(flow)) {
2125 flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port);
9583bc14
EJ
2126 }
2127 break;
2128
2129 case OFPACT_RESUBMIT:
2130 xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a));
2131 break;
2132
2133 case OFPACT_SET_TUNNEL:
33bf9176 2134 flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id);
9583bc14
EJ
2135 break;
2136
2137 case OFPACT_SET_QUEUE:
2138 xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id);
2139 break;
2140
2141 case OFPACT_POP_QUEUE:
33bf9176 2142 flow->skb_priority = ctx->orig_skb_priority;
9583bc14
EJ
2143 break;
2144
2145 case OFPACT_REG_MOVE:
33bf9176 2146 nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc);
9583bc14
EJ
2147 break;
2148
2149 case OFPACT_REG_LOAD:
33bf9176 2150 nxm_execute_reg_load(ofpact_get_REG_LOAD(a), flow);
9583bc14
EJ
2151 break;
2152
2153 case OFPACT_STACK_PUSH:
33bf9176
BP
2154 nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc,
2155 &ctx->stack);
9583bc14
EJ
2156 break;
2157
2158 case OFPACT_STACK_POP:
f74e7df7
JP
2159 nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc,
2160 &ctx->stack);
9583bc14
EJ
2161 break;
2162
2163 case OFPACT_PUSH_MPLS:
9cfef3d0 2164 compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)->ethertype);
9583bc14
EJ
2165 break;
2166
2167 case OFPACT_POP_MPLS:
9cfef3d0 2168 compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype);
9583bc14
EJ
2169 break;
2170
2171 case OFPACT_SET_MPLS_TTL:
9cfef3d0 2172 if (compose_set_mpls_ttl_action(ctx,
9583bc14
EJ
2173 ofpact_get_SET_MPLS_TTL(a)->ttl)) {
2174 goto out;
2175 }
2176 break;
2177
2178 case OFPACT_DEC_MPLS_TTL:
9cfef3d0 2179 if (compose_dec_mpls_ttl_action(ctx)) {
9583bc14
EJ
2180 goto out;
2181 }
2182 break;
2183
2184 case OFPACT_DEC_TTL:
f74e7df7 2185 wc->masks.nw_ttl = 0xff;
9583bc14
EJ
2186 if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) {
2187 goto out;
2188 }
2189 break;
2190
2191 case OFPACT_NOTE:
2192 /* Nothing to do. */
2193 break;
2194
2195 case OFPACT_MULTIPATH:
33bf9176 2196 multipath_execute(ofpact_get_MULTIPATH(a), flow, wc);
9583bc14
EJ
2197 break;
2198
2199 case OFPACT_BUNDLE:
9583bc14
EJ
2200 xlate_bundle_action(ctx, ofpact_get_BUNDLE(a));
2201 break;
2202
2203 case OFPACT_OUTPUT_REG:
2204 xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a));
2205 break;
2206
2207 case OFPACT_LEARN:
2208 xlate_learn_action(ctx, ofpact_get_LEARN(a));
2209 break;
2210
2211 case OFPACT_EXIT:
2212 ctx->exit = true;
2213 break;
2214
2215 case OFPACT_FIN_TIMEOUT:
33bf9176 2216 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
9583bc14
EJ
2217 ctx->xout->has_fin_timeout = true;
2218 xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a));
2219 break;
2220
2221 case OFPACT_CLEAR_ACTIONS:
2222 /* XXX
2223 * Nothing to do because writa-actions is not supported for now.
2224 * When writa-actions is supported, clear-actions also must
2225 * be supported at the same time.
2226 */
2227 break;
2228
2229 case OFPACT_WRITE_METADATA:
2230 metadata = ofpact_get_WRITE_METADATA(a);
33bf9176
BP
2231 flow->metadata &= ~metadata->mask;
2232 flow->metadata |= metadata->metadata & metadata->mask;
9583bc14
EJ
2233 break;
2234
638a19b0
JR
2235 case OFPACT_METER:
2236 /* Not implemented yet. */
2237 break;
2238
9583bc14
EJ
2239 case OFPACT_GOTO_TABLE: {
2240 /* It is assumed that goto-table is the last action. */
2241 struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a);
9583bc14
EJ
2242
2243 ovs_assert(ctx->table_id < ogt->table_id);
4468099e
EJ
2244 xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
2245 ogt->table_id, true);
9583bc14
EJ
2246 break;
2247 }
2248
2249 case OFPACT_SAMPLE:
2250 xlate_sample_action(ctx, ofpact_get_SAMPLE(a));
2251 break;
2252 }
2253 }
2254
2255out:
2256 if (ctx->rule) {
2257 ctx->rule->up.evictable = was_evictable;
2258 }
2259}
2260
2261void
2262xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto,
2263 const struct flow *flow, struct rule_dpif *rule,
2264 uint8_t tcp_flags, const struct ofpbuf *packet)
2265{
2266 xin->ofproto = ofproto;
2267 xin->flow = *flow;
2268 xin->packet = packet;
2269 xin->may_learn = packet != NULL;
2270 xin->rule = rule;
2271 xin->ofpacts = NULL;
2272 xin->ofpacts_len = 0;
2273 xin->tcp_flags = tcp_flags;
2274 xin->resubmit_hook = NULL;
2275 xin->report_hook = NULL;
2276 xin->resubmit_stats = NULL;
2277}
2278
2279void
2280xlate_out_uninit(struct xlate_out *xout)
2281{
2282 if (xout) {
2283 ofpbuf_uninit(&xout->odp_actions);
2284 }
2285}
2286
2287/* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts'
2288 * into datapath actions, using 'ctx', and discards the datapath actions. */
2289void
2290xlate_actions_for_side_effects(struct xlate_in *xin)
2291{
2292 struct xlate_out xout;
2293
2294 xlate_actions(xin, &xout);
2295 xlate_out_uninit(&xout);
2296}
2297
2298static void
2299xlate_report(struct xlate_ctx *ctx, const char *s)
2300{
2301 if (ctx->xin->report_hook) {
4d0acc70 2302 ctx->xin->report_hook(ctx->xin, s, ctx->recurse);
9583bc14
EJ
2303 }
2304}
2305
2306void
2307xlate_out_copy(struct xlate_out *dst, const struct xlate_out *src)
2308{
2309 dst->wc = src->wc;
9583bc14
EJ
2310 dst->slow = src->slow;
2311 dst->has_learn = src->has_learn;
2312 dst->has_normal = src->has_normal;
2313 dst->has_fin_timeout = src->has_fin_timeout;
2314 dst->nf_output_iface = src->nf_output_iface;
2315 dst->mirrors = src->mirrors;
2316
2317 ofpbuf_use_stub(&dst->odp_actions, dst->odp_actions_stub,
2318 sizeof dst->odp_actions_stub);
2319 ofpbuf_put(&dst->odp_actions, src->odp_actions.data,
2320 src->odp_actions.size);
2321}
2322\f
55954f6e
EJ
2323static struct skb_priority_to_dscp *
2324get_skb_priority(const struct xport *xport, uint32_t skb_priority)
2325{
2326 struct skb_priority_to_dscp *pdscp;
2327 uint32_t hash;
2328
2329 hash = hash_int(skb_priority, 0);
2330 HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) {
2331 if (pdscp->skb_priority == skb_priority) {
2332 return pdscp;
2333 }
2334 }
2335 return NULL;
2336}
2337
2338static bool
2339dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority,
2340 uint8_t *dscp)
2341{
2342 struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority);
2343 *dscp = pdscp ? pdscp->dscp : 0;
2344 return pdscp != NULL;
2345}
2346
2347static void
2348clear_skb_priorities(struct xport *xport)
2349{
2350 struct skb_priority_to_dscp *pdscp, *next;
2351
2352 HMAP_FOR_EACH_SAFE (pdscp, next, hmap_node, &xport->skb_priorities) {
2353 hmap_remove(&xport->skb_priorities, &pdscp->hmap_node);
2354 free(pdscp);
2355 }
2356}
2357
ce4a6b76
BP
2358static bool
2359actions_output_to_local_port(const struct xlate_ctx *ctx)
2360{
46c88433 2361 odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL);
ce4a6b76
BP
2362 const struct nlattr *a;
2363 unsigned int left;
2364
2365 NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->xout->odp_actions.data,
2366 ctx->xout->odp_actions.size) {
2367 if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT
2368 && nl_attr_get_odp_port(a) == local_odp_port) {
2369 return true;
2370 }
2371 }
2372 return false;
2373}
9583bc14
EJ
2374
2375/* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts'
2376 * into datapath actions in 'odp_actions', using 'ctx'. */
2377void
2378xlate_actions(struct xlate_in *xin, struct xlate_out *xout)
2379{
33bf9176
BP
2380 struct flow_wildcards *wc = &xout->wc;
2381 struct flow *flow = &xin->flow;
2382
9583bc14
EJ
2383 enum slow_path_reason special;
2384 const struct ofpact *ofpacts;
46c88433 2385 struct xport *in_port;
9583bc14
EJ
2386 struct flow orig_flow;
2387 struct xlate_ctx ctx;
2388 size_t ofpacts_len;
2389
46c88433 2390 COVERAGE_INC(xlate_actions);
9583bc14
EJ
2391
2392 /* Flow initialization rules:
2393 * - 'base_flow' must match the kernel's view of the packet at the
2394 * time that action processing starts. 'flow' represents any
2395 * transformations we wish to make through actions.
2396 * - By default 'base_flow' and 'flow' are the same since the input
2397 * packet matches the output before any actions are applied.
2398 * - When using VLAN splinters, 'base_flow''s VLAN is set to the value
2399 * of the received packet as seen by the kernel. If we later output
2400 * to another device without any modifications this will cause us to
2401 * insert a new tag since the original one was stripped off by the
2402 * VLAN device.
2403 * - Tunnel metadata as received is retained in 'flow'. This allows
2404 * tunnel metadata matching also in later tables.
2405 * Since a kernel action for setting the tunnel metadata will only be
2406 * generated with actual tunnel output, changing the tunnel metadata
2407 * values in 'flow' (such as tun_id) will only have effect with a later
2408 * tunnel output action.
2409 * - Tunnel 'base_flow' is completely cleared since that is what the
2410 * kernel does. If we wish to maintain the original values an action
2411 * needs to be generated. */
2412
2413 ctx.xin = xin;
2414 ctx.xout = xout;
46c88433
EJ
2415 ctx.xout->slow = 0;
2416 ctx.xout->has_learn = false;
2417 ctx.xout->has_normal = false;
2418 ctx.xout->has_fin_timeout = false;
2419 ctx.xout->nf_output_iface = NF_OUT_DROP;
2420 ctx.xout->mirrors = 0;
2421 ofpbuf_use_stub(&ctx.xout->odp_actions, ctx.xout->odp_actions_stub,
2422 sizeof ctx.xout->odp_actions_stub);
2423 ofpbuf_reserve(&ctx.xout->odp_actions, NL_A_U32_SIZE);
2424
2425 ctx.xbridge = xbridge_lookup(xin->ofproto);
2426 if (!ctx.xbridge) {
2427 return;
2428 }
9583bc14 2429
9583bc14
EJ
2430 ctx.rule = xin->rule;
2431
33bf9176 2432 ctx.base_flow = *flow;
9583bc14 2433 memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel);
33bf9176 2434 ctx.orig_tunnel_ip_dst = flow->tunnel.ip_dst;
9583bc14 2435
33bf9176
BP
2436 flow_wildcards_init_catchall(wc);
2437 memset(&wc->masks.in_port, 0xff, sizeof wc->masks.in_port);
1dd35f8a 2438 memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority);
7431e171 2439 memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type);
1dd35f8a 2440 wc->masks.nw_frag |= FLOW_NW_FRAG_MASK;
9583bc14
EJ
2441
2442 if (tnl_port_should_receive(&ctx.xin->flow)) {
33bf9176 2443 memset(&wc->masks.tunnel, 0xff, sizeof wc->masks.tunnel);
8ba2ee85
JG
2444 /* skb_mark is currently used only by tunnels but that will likely
2445 * change in the future. */
2446 memset(&wc->masks.skb_mark, 0xff, sizeof wc->masks.skb_mark);
9583bc14 2447 }
46c88433 2448 if (ctx.xbridge->has_netflow) {
9b658910 2449 netflow_mask_wc(flow, wc);
9583bc14
EJ
2450 }
2451
9583bc14 2452 ctx.recurse = 0;
33bf9176 2453 ctx.orig_skb_priority = flow->skb_priority;
9583bc14
EJ
2454 ctx.table_id = 0;
2455 ctx.exit = false;
2456
2457 if (xin->ofpacts) {
2458 ofpacts = xin->ofpacts;
2459 ofpacts_len = xin->ofpacts_len;
2460 } else if (xin->rule) {
2461 ofpacts = xin->rule->up.ofpacts;
2462 ofpacts_len = xin->rule->up.ofpacts_len;
2463 } else {
2464 NOT_REACHED();
2465 }
2466
2467 ofpbuf_use_stub(&ctx.stack, ctx.init_stack, sizeof ctx.init_stack);
2468
ade6ad9c 2469 if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
9583bc14
EJ
2470 /* Do this conditionally because the copy is expensive enough that it
2471 * shows up in profiles. */
33bf9176 2472 orig_flow = *flow;
9583bc14
EJ
2473 }
2474
33bf9176 2475 if (flow->nw_frag & FLOW_NW_FRAG_ANY) {
46c88433 2476 switch (ctx.xbridge->frag) {
9583bc14
EJ
2477 case OFPC_FRAG_NORMAL:
2478 /* We must pretend that transport ports are unavailable. */
33bf9176
BP
2479 flow->tp_src = ctx.base_flow.tp_src = htons(0);
2480 flow->tp_dst = ctx.base_flow.tp_dst = htons(0);
9583bc14
EJ
2481 break;
2482
2483 case OFPC_FRAG_DROP:
2484 return;
2485
2486 case OFPC_FRAG_REASM:
2487 NOT_REACHED();
2488
2489 case OFPC_FRAG_NX_MATCH:
2490 /* Nothing to do. */
2491 break;
2492
2493 case OFPC_INVALID_TTL_TO_CONTROLLER:
2494 NOT_REACHED();
2495 }
2496 }
2497
46c88433 2498 in_port = get_ofp_port(ctx.xbridge, flow->in_port.ofp_port);
642dc74d 2499 special = process_special(&ctx, flow, in_port, ctx.xin->packet);
9583bc14
EJ
2500 if (special) {
2501 ctx.xout->slow = special;
2502 } else {
9583bc14 2503 size_t sample_actions_len;
9583bc14 2504
4e022ec0 2505 if (flow->in_port.ofp_port
46c88433
EJ
2506 != vsp_realdev_to_vlandev(ctx.xbridge->ofproto,
2507 flow->in_port.ofp_port,
33bf9176 2508 flow->vlan_tci)) {
9583bc14
EJ
2509 ctx.base_flow.vlan_tci = 0;
2510 }
2511
2512 add_sflow_action(&ctx);
2513 add_ipfix_action(&ctx);
2514 sample_actions_len = ctx.xout->odp_actions.size;
2515
2516 if (tunnel_ecn_ok(&ctx) && (!in_port || may_receive(in_port, &ctx))) {
2517 do_xlate_actions(ofpacts, ofpacts_len, &ctx);
2518
2519 /* We've let OFPP_NORMAL and the learning action look at the
2520 * packet, so drop it now if forwarding is disabled. */
9d189a50 2521 if (in_port && !xport_stp_forward_state(in_port)) {
9583bc14
EJ
2522 ctx.xout->odp_actions.size = sample_actions_len;
2523 }
2524 }
2525
46c88433 2526 if (ctx.xbridge->has_in_band
ce4a6b76
BP
2527 && in_band_must_output_to_local_port(flow)
2528 && !actions_output_to_local_port(&ctx)) {
9583bc14
EJ
2529 compose_output_action(&ctx, OFPP_LOCAL);
2530 }
aaa0fbae
EJ
2531
2532 fix_sflow_action(&ctx);
2533
46c88433 2534 if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
9583bc14
EJ
2535 add_mirror_actions(&ctx, &orig_flow);
2536 }
9583bc14
EJ
2537 }
2538
2539 ofpbuf_uninit(&ctx.stack);
2540
2541 /* Clear the metadata and register wildcard masks, because we won't
2542 * use non-header fields as part of the cache. */
33bf9176
BP
2543 memset(&wc->masks.metadata, 0, sizeof wc->masks.metadata);
2544 memset(&wc->masks.regs, 0, sizeof wc->masks.regs);
9583bc14 2545}