]> git.proxmox.com Git - ovs.git/blame - ofproto/ofproto-dpif-xlate.c
bfd: Require bfd control packet received in forwarding_if_rx mode.
[ovs.git] / ofproto / ofproto-dpif-xlate.c
CommitLineData
0934ebba 1/* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014 Nicira, Inc.
9583bc14
EJ
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
8449c4d6
EJ
19#include <errno.h>
20
db7d4e46 21#include "bfd.h"
9583bc14
EJ
22#include "bitmap.h"
23#include "bond.h"
24#include "bundle.h"
25#include "byte-order.h"
db7d4e46 26#include "cfm.h"
9583bc14
EJ
27#include "connmgr.h"
28#include "coverage.h"
29#include "dpif.h"
30#include "dynamic-string.h"
f7f1ea29 31#include "in-band.h"
db7d4e46 32#include "lacp.h"
9583bc14 33#include "learn.h"
46c88433 34#include "list.h"
9583bc14
EJ
35#include "mac-learning.h"
36#include "meta-flow.h"
37#include "multipath.h"
38#include "netdev-vport.h"
39#include "netlink.h"
40#include "nx-match.h"
41#include "odp-execute.h"
42#include "ofp-actions.h"
43#include "ofproto/ofproto-dpif-ipfix.h"
ec7ceaed 44#include "ofproto/ofproto-dpif-mirror.h"
60d02c72 45#include "ofproto/ofproto-dpif-monitor.h"
9583bc14
EJ
46#include "ofproto/ofproto-dpif-sflow.h"
47#include "ofproto/ofproto-dpif.h"
6f00e29b 48#include "ofproto/ofproto-provider.h"
9583bc14
EJ
49#include "tunnel.h"
50#include "vlog.h"
51
46c88433 52COVERAGE_DEFINE(xlate_actions);
0f032e95 53COVERAGE_DEFINE(xlate_actions_oversize);
8bfd0fda 54COVERAGE_DEFINE(xlate_actions_mpls_overflow);
9583bc14
EJ
55
56VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate);
57
8a553e9a
EJ
58/* Maximum depth of flow table recursion (due to resubmit actions) in a
59 * flow translation. */
60#define MAX_RESUBMIT_RECURSION 64
adcf00ba
AZ
61#define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in
62 internal table. */
8a553e9a 63
98b07853
BP
64/* Maximum number of resubmit actions in a flow translation, whether they are
65 * recursive or not. */
66#define MAX_RESUBMITS (MAX_RESUBMIT_RECURSION * MAX_RESUBMIT_RECURSION)
67
dc24a00f
EJ
68struct ovs_rwlock xlate_rwlock = OVS_RWLOCK_INITIALIZER;
69
46c88433
EJ
70struct xbridge {
71 struct hmap_node hmap_node; /* Node in global 'xbridges' map. */
72 struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */
73
74 struct list xbundles; /* Owned xbundles. */
75 struct hmap xports; /* Indexed by ofp_port. */
76
77 char *name; /* Name used in log messages. */
89a8a7f0 78 struct dpif *dpif; /* Datapath interface. */
46c88433
EJ
79 struct mac_learning *ml; /* Mac learning handle. */
80 struct mbridge *mbridge; /* Mirroring. */
81 struct dpif_sflow *sflow; /* SFlow handle, or null. */
82 struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */
ce3955be 83 struct netflow *netflow; /* Netflow handle, or null. */
9d189a50 84 struct stp *stp; /* STP or null if disabled. */
46c88433 85
ec89fc6f
EJ
86 /* Special rules installed by ofproto-dpif. */
87 struct rule_dpif *miss_rule;
88 struct rule_dpif *no_packet_in_rule;
89
46c88433 90 enum ofp_config_flags frag; /* Fragmentation handling. */
46c88433
EJ
91 bool has_in_band; /* Bridge has in band control? */
92 bool forward_bpdu; /* Bridge forwards STP BPDUs? */
4b97b70d 93
adcf00ba
AZ
94 /* True if the datapath supports recirculation. */
95 bool enable_recirc;
96
4b97b70d
BP
97 /* True if the datapath supports variable-length
98 * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions.
99 * False if the datapath supports only 8-byte (or shorter) userdata. */
100 bool variable_length_userdata;
8bfd0fda
BP
101
102 /* Number of MPLS label stack entries that the datapath supports
103 * in matches. */
104 size_t max_mpls_depth;
46c88433
EJ
105};
106
107struct xbundle {
108 struct hmap_node hmap_node; /* In global 'xbundles' map. */
109 struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */
110
111 struct list list_node; /* In parent 'xbridges' list. */
112 struct xbridge *xbridge; /* Parent xbridge. */
113
114 struct list xports; /* Contains "struct xport"s. */
115
116 char *name; /* Name used in log messages. */
117 struct bond *bond; /* Nonnull iff more than one port. */
118 struct lacp *lacp; /* LACP handle or null. */
119
120 enum port_vlan_mode vlan_mode; /* VLAN mode. */
121 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
122 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
123 * NULL if all VLANs are trunked. */
124 bool use_priority_tags; /* Use 802.1p tag for frames in VLAN 0? */
125 bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */
126};
127
128struct xport {
129 struct hmap_node hmap_node; /* Node in global 'xports' map. */
130 struct ofport_dpif *ofport; /* Key in global 'xports map. */
131
132 struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */
133 ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */
134
135 odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */
136
137 struct list bundle_node; /* In parent xbundle (if it exists). */
138 struct xbundle *xbundle; /* Parent xbundle or null. */
139
140 struct netdev *netdev; /* 'ofport''s netdev. */
141
142 struct xbridge *xbridge; /* Parent bridge. */
143 struct xport *peer; /* Patch port peer or null. */
144
145 enum ofputil_port_config config; /* OpenFlow port configuration. */
dd8cd4b4 146 enum ofputil_port_state state; /* OpenFlow port state. */
92cf817b 147 int stp_port_no; /* STP port number or -1 if not in use. */
46c88433 148
55954f6e
EJ
149 struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */
150
46c88433
EJ
151 bool may_enable; /* May be enabled in bonds. */
152 bool is_tunnel; /* Is a tunnel port. */
153
154 struct cfm *cfm; /* CFM handle or null. */
155 struct bfd *bfd; /* BFD handle or null. */
156};
157
4d0acc70
EJ
158struct xlate_ctx {
159 struct xlate_in *xin;
160 struct xlate_out *xout;
161
46c88433 162 const struct xbridge *xbridge;
4d0acc70
EJ
163
164 /* Flow at the last commit. */
165 struct flow base_flow;
166
167 /* Tunnel IP destination address as received. This is stored separately
168 * as the base_flow.tunnel is cleared on init to reflect the datapath
169 * behavior. Used to make sure not to send tunneled output to ourselves,
170 * which might lead to an infinite loop. This could happen easily
171 * if a tunnel is marked as 'ip_remote=flow', and the flow does not
172 * actually set the tun_dst field. */
173 ovs_be32 orig_tunnel_ip_dst;
174
175 /* Stack for the push and pop actions. Each stack element is of type
176 * "union mf_subvalue". */
177 union mf_subvalue init_stack[1024 / sizeof(union mf_subvalue)];
178 struct ofpbuf stack;
179
180 /* The rule that we are currently translating, or NULL. */
181 struct rule_dpif *rule;
182
98b07853
BP
183 /* Resubmit statistics, via xlate_table_action(). */
184 int recurse; /* Current resubmit nesting depth. */
185 int resubmits; /* Total number of resubmits. */
5a070238 186 bool in_group; /* Currently translating ofgroup, if true. */
98b07853 187
4d0acc70
EJ
188 uint32_t orig_skb_priority; /* Priority when packet arrived. */
189 uint8_t table_id; /* OpenFlow table ID where flow was found. */
190 uint32_t sflow_n_outputs; /* Number of output ports. */
4e022ec0 191 odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */
4d0acc70
EJ
192 uint16_t user_cookie_offset;/* Used for user_action_cookie fixup. */
193 bool exit; /* No further actions should be processed. */
7fdb60a7
SH
194
195 /* OpenFlow 1.1+ action set.
196 *
197 * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS.
198 * When translation is otherwise complete, ofpacts_execute_action_set()
199 * converts it to a set of "struct ofpact"s that can be translated into
200 * datapath actions. */
201 struct ofpbuf action_set; /* Action set. */
202 uint64_t action_set_stub[1024 / 8];
4d0acc70
EJ
203};
204
9583bc14
EJ
205/* A controller may use OFPP_NONE as the ingress port to indicate that
206 * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for
207 * when an input bundle is needed for validation (e.g., mirroring or
208 * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have
3548d242
BP
209 * any 'port' structs, so care must be taken when dealing with it. */
210static struct xbundle ofpp_none_bundle = {
211 .name = "OFPP_NONE",
212 .vlan_mode = PORT_VLAN_TRUNK
213};
9583bc14 214
55954f6e
EJ
215/* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from
216 * 'priority' (the datapath's term for QoS queue) to the dscp bits which all
217 * traffic egressing the 'ofport' with that priority should be marked with. */
218struct skb_priority_to_dscp {
219 struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */
220 uint32_t skb_priority; /* Priority of this queue (see struct flow). */
221
222 uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */
223};
224
b256dc52
JS
225enum xc_type {
226 XC_RULE,
227 XC_BOND,
228 XC_NETDEV,
229 XC_NETFLOW,
230 XC_MIRROR,
231 XC_LEARN,
232 XC_NORMAL,
233 XC_FIN_TIMEOUT,
234};
235
236/* xlate_cache entries hold enough information to perform the side effects of
237 * xlate_actions() for a rule, without needing to perform rule translation
238 * from scratch. The primary usage of these is to submit statistics to objects
239 * that a flow relates to, although they may be used for other effects as well
240 * (for instance, refreshing hard timeouts for learned flows). */
241struct xc_entry {
242 enum xc_type type;
243 union {
244 struct rule_dpif *rule;
245 struct {
246 struct netdev *tx;
247 struct netdev *rx;
248 struct bfd *bfd;
249 } dev;
250 struct {
251 struct netflow *netflow;
252 struct flow *flow;
253 ofp_port_t iface;
254 } nf;
255 struct {
256 struct mbridge *mbridge;
257 mirror_mask_t mirrors;
258 } mirror;
259 struct {
260 struct bond *bond;
261 struct flow *flow;
262 uint16_t vid;
263 } bond;
264 struct {
265 struct ofproto_dpif *ofproto;
266 struct rule_dpif *rule;
267 } learn;
268 struct {
269 struct ofproto_dpif *ofproto;
270 struct flow *flow;
271 int vlan;
272 } normal;
273 struct {
274 struct rule_dpif *rule;
275 uint16_t idle;
276 uint16_t hard;
277 } fin;
278 } u;
279};
280
281#define XC_ENTRY_FOR_EACH(entry, entries, xcache) \
282 entries = xcache->entries; \
283 for (entry = ofpbuf_try_pull(&entries, sizeof *entry); \
284 entry; \
285 entry = ofpbuf_try_pull(&entries, sizeof *entry))
286
287struct xlate_cache {
288 struct ofpbuf entries;
289};
290
46c88433
EJ
291static struct hmap xbridges = HMAP_INITIALIZER(&xbridges);
292static struct hmap xbundles = HMAP_INITIALIZER(&xbundles);
293static struct hmap xports = HMAP_INITIALIZER(&xports);
294
295static bool may_receive(const struct xport *, struct xlate_ctx *);
9583bc14
EJ
296static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len,
297 struct xlate_ctx *);
91d6cd12
AW
298static void xlate_actions__(struct xlate_in *, struct xlate_out *)
299 OVS_REQ_RDLOCK(xlate_rwlock);
adcf00ba
AZ
300static void xlate_normal(struct xlate_ctx *);
301static void xlate_report(struct xlate_ctx *, const char *);
6d328fa2
SH
302static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port,
303 uint8_t table_id, bool may_packet_in,
304 bool honor_table_miss);
46c88433
EJ
305static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn);
306static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid);
307static void output_normal(struct xlate_ctx *, const struct xbundle *,
9583bc14 308 uint16_t vlan);
4e022ec0 309static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port);
9583bc14 310
46c88433
EJ
311static struct xbridge *xbridge_lookup(const struct ofproto_dpif *);
312static struct xbundle *xbundle_lookup(const struct ofbundle *);
5e6af486 313static struct xport *xport_lookup(const struct ofport_dpif *);
46c88433 314static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port);
55954f6e
EJ
315static struct skb_priority_to_dscp *get_skb_priority(const struct xport *,
316 uint32_t skb_priority);
317static void clear_skb_priorities(struct xport *);
318static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority,
319 uint8_t *dscp);
46c88433 320
b256dc52
JS
321static struct xc_entry *xlate_cache_add_entry(struct xlate_cache *xc,
322 enum xc_type type);
323
46c88433
EJ
324void
325xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name,
ec89fc6f
EJ
326 struct dpif *dpif, struct rule_dpif *miss_rule,
327 struct rule_dpif *no_packet_in_rule,
328 const struct mac_learning *ml, struct stp *stp,
329 const struct mbridge *mbridge,
46c88433 330 const struct dpif_sflow *sflow,
ce3955be
EJ
331 const struct dpif_ipfix *ipfix,
332 const struct netflow *netflow, enum ofp_config_flags frag,
4b97b70d 333 bool forward_bpdu, bool has_in_band,
adcf00ba 334 bool enable_recirc,
8bfd0fda
BP
335 bool variable_length_userdata,
336 size_t max_mpls_depth)
46c88433
EJ
337{
338 struct xbridge *xbridge = xbridge_lookup(ofproto);
339
340 if (!xbridge) {
341 xbridge = xzalloc(sizeof *xbridge);
342 xbridge->ofproto = ofproto;
343
344 hmap_insert(&xbridges, &xbridge->hmap_node, hash_pointer(ofproto, 0));
345 hmap_init(&xbridge->xports);
346 list_init(&xbridge->xbundles);
347 }
348
349 if (xbridge->ml != ml) {
350 mac_learning_unref(xbridge->ml);
351 xbridge->ml = mac_learning_ref(ml);
352 }
353
354 if (xbridge->mbridge != mbridge) {
355 mbridge_unref(xbridge->mbridge);
356 xbridge->mbridge = mbridge_ref(mbridge);
357 }
358
359 if (xbridge->sflow != sflow) {
360 dpif_sflow_unref(xbridge->sflow);
361 xbridge->sflow = dpif_sflow_ref(sflow);
362 }
363
364 if (xbridge->ipfix != ipfix) {
365 dpif_ipfix_unref(xbridge->ipfix);
366 xbridge->ipfix = dpif_ipfix_ref(ipfix);
367 }
368
9d189a50
EJ
369 if (xbridge->stp != stp) {
370 stp_unref(xbridge->stp);
371 xbridge->stp = stp_ref(stp);
372 }
373
ce3955be
EJ
374 if (xbridge->netflow != netflow) {
375 netflow_unref(xbridge->netflow);
376 xbridge->netflow = netflow_ref(netflow);
377 }
378
46c88433
EJ
379 free(xbridge->name);
380 xbridge->name = xstrdup(name);
381
89a8a7f0 382 xbridge->dpif = dpif;
46c88433
EJ
383 xbridge->forward_bpdu = forward_bpdu;
384 xbridge->has_in_band = has_in_band;
46c88433 385 xbridge->frag = frag;
ec89fc6f
EJ
386 xbridge->miss_rule = miss_rule;
387 xbridge->no_packet_in_rule = no_packet_in_rule;
adcf00ba 388 xbridge->enable_recirc = enable_recirc;
4b97b70d 389 xbridge->variable_length_userdata = variable_length_userdata;
8bfd0fda 390 xbridge->max_mpls_depth = max_mpls_depth;
46c88433
EJ
391}
392
393void
394xlate_remove_ofproto(struct ofproto_dpif *ofproto)
395{
396 struct xbridge *xbridge = xbridge_lookup(ofproto);
397 struct xbundle *xbundle, *next_xbundle;
398 struct xport *xport, *next_xport;
399
400 if (!xbridge) {
401 return;
402 }
403
404 HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) {
405 xlate_ofport_remove(xport->ofport);
406 }
407
408 LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) {
409 xlate_bundle_remove(xbundle->ofbundle);
410 }
411
412 hmap_remove(&xbridges, &xbridge->hmap_node);
795cc5c1
EJ
413 mac_learning_unref(xbridge->ml);
414 mbridge_unref(xbridge->mbridge);
415 dpif_sflow_unref(xbridge->sflow);
416 dpif_ipfix_unref(xbridge->ipfix);
417 stp_unref(xbridge->stp);
418 hmap_destroy(&xbridge->xports);
46c88433
EJ
419 free(xbridge->name);
420 free(xbridge);
421}
422
423void
424xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle,
425 const char *name, enum port_vlan_mode vlan_mode, int vlan,
426 unsigned long *trunks, bool use_priority_tags,
427 const struct bond *bond, const struct lacp *lacp,
428 bool floodable)
429{
430 struct xbundle *xbundle = xbundle_lookup(ofbundle);
431
432 if (!xbundle) {
433 xbundle = xzalloc(sizeof *xbundle);
434 xbundle->ofbundle = ofbundle;
435 xbundle->xbridge = xbridge_lookup(ofproto);
436
437 hmap_insert(&xbundles, &xbundle->hmap_node, hash_pointer(ofbundle, 0));
438 list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node);
439 list_init(&xbundle->xports);
440 }
441
442 ovs_assert(xbundle->xbridge);
443
444 free(xbundle->name);
445 xbundle->name = xstrdup(name);
446
447 xbundle->vlan_mode = vlan_mode;
448 xbundle->vlan = vlan;
449 xbundle->trunks = trunks;
450 xbundle->use_priority_tags = use_priority_tags;
451 xbundle->floodable = floodable;
452
453 if (xbundle->bond != bond) {
454 bond_unref(xbundle->bond);
455 xbundle->bond = bond_ref(bond);
456 }
457
458 if (xbundle->lacp != lacp) {
459 lacp_unref(xbundle->lacp);
460 xbundle->lacp = lacp_ref(lacp);
461 }
462}
463
464void
465xlate_bundle_remove(struct ofbundle *ofbundle)
466{
467 struct xbundle *xbundle = xbundle_lookup(ofbundle);
468 struct xport *xport, *next;
469
470 if (!xbundle) {
471 return;
472 }
473
474 LIST_FOR_EACH_SAFE (xport, next, bundle_node, &xbundle->xports) {
475 list_remove(&xport->bundle_node);
476 xport->xbundle = NULL;
477 }
478
479 hmap_remove(&xbundles, &xbundle->hmap_node);
480 list_remove(&xbundle->list_node);
481 bond_unref(xbundle->bond);
482 lacp_unref(xbundle->lacp);
483 free(xbundle->name);
484 free(xbundle);
485}
486
487void
488xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle,
489 struct ofport_dpif *ofport, ofp_port_t ofp_port,
490 odp_port_t odp_port, const struct netdev *netdev,
491 const struct cfm *cfm, const struct bfd *bfd,
9d189a50 492 struct ofport_dpif *peer, int stp_port_no,
55954f6e 493 const struct ofproto_port_queue *qdscp_list, size_t n_qdscp,
dd8cd4b4
SH
494 enum ofputil_port_config config,
495 enum ofputil_port_state state, bool is_tunnel,
9d189a50 496 bool may_enable)
46c88433
EJ
497{
498 struct xport *xport = xport_lookup(ofport);
55954f6e 499 size_t i;
46c88433
EJ
500
501 if (!xport) {
502 xport = xzalloc(sizeof *xport);
503 xport->ofport = ofport;
504 xport->xbridge = xbridge_lookup(ofproto);
505 xport->ofp_port = ofp_port;
506
55954f6e 507 hmap_init(&xport->skb_priorities);
46c88433
EJ
508 hmap_insert(&xports, &xport->hmap_node, hash_pointer(ofport, 0));
509 hmap_insert(&xport->xbridge->xports, &xport->ofp_node,
510 hash_ofp_port(xport->ofp_port));
511 }
512
513 ovs_assert(xport->ofp_port == ofp_port);
514
515 xport->config = config;
dd8cd4b4 516 xport->state = state;
9d189a50 517 xport->stp_port_no = stp_port_no;
46c88433
EJ
518 xport->is_tunnel = is_tunnel;
519 xport->may_enable = may_enable;
520 xport->odp_port = odp_port;
521
522 if (xport->netdev != netdev) {
523 netdev_close(xport->netdev);
524 xport->netdev = netdev_ref(netdev);
525 }
526
527 if (xport->cfm != cfm) {
528 cfm_unref(xport->cfm);
529 xport->cfm = cfm_ref(cfm);
530 }
531
532 if (xport->bfd != bfd) {
533 bfd_unref(xport->bfd);
534 xport->bfd = bfd_ref(bfd);
535 }
536
537 if (xport->peer) {
538 xport->peer->peer = NULL;
539 }
5e6af486 540 xport->peer = xport_lookup(peer);
46c88433
EJ
541 if (xport->peer) {
542 xport->peer->peer = xport;
543 }
544
545 if (xport->xbundle) {
546 list_remove(&xport->bundle_node);
547 }
5e6af486 548 xport->xbundle = xbundle_lookup(ofbundle);
46c88433
EJ
549 if (xport->xbundle) {
550 list_insert(&xport->xbundle->xports, &xport->bundle_node);
551 }
55954f6e
EJ
552
553 clear_skb_priorities(xport);
554 for (i = 0; i < n_qdscp; i++) {
555 struct skb_priority_to_dscp *pdscp;
556 uint32_t skb_priority;
557
89a8a7f0
EJ
558 if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue,
559 &skb_priority)) {
55954f6e
EJ
560 continue;
561 }
562
563 pdscp = xmalloc(sizeof *pdscp);
564 pdscp->skb_priority = skb_priority;
565 pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK;
566 hmap_insert(&xport->skb_priorities, &pdscp->hmap_node,
567 hash_int(pdscp->skb_priority, 0));
568 }
46c88433
EJ
569}
570
571void
572xlate_ofport_remove(struct ofport_dpif *ofport)
573{
574 struct xport *xport = xport_lookup(ofport);
575
576 if (!xport) {
577 return;
578 }
579
580 if (xport->peer) {
581 xport->peer->peer = NULL;
582 xport->peer = NULL;
583 }
584
e621a12d
EJ
585 if (xport->xbundle) {
586 list_remove(&xport->bundle_node);
587 }
588
55954f6e
EJ
589 clear_skb_priorities(xport);
590 hmap_destroy(&xport->skb_priorities);
591
46c88433
EJ
592 hmap_remove(&xports, &xport->hmap_node);
593 hmap_remove(&xport->xbridge->xports, &xport->ofp_node);
594
595 netdev_close(xport->netdev);
596 cfm_unref(xport->cfm);
597 bfd_unref(xport->bfd);
598 free(xport);
599}
600
8449c4d6
EJ
601/* Given a datpath, packet, and flow metadata ('backer', 'packet', and 'key'
602 * respectively), populates 'flow' with the result of odp_flow_key_to_flow().
836fbda7
SH
603 * Optionally populates 'ofproto' with the ofproto_dpif, 'odp_in_port' with
604 * the datapath in_port, that 'packet' ingressed, and 'ipfix', 'sflow', and
605 * 'netflow' with the appropriate handles for those protocols if they're
606 * enabled. Caller is responsible for unrefing them.
8449c4d6
EJ
607 *
608 * If 'ofproto' is nonnull, requires 'flow''s in_port to exist. Otherwise sets
609 * 'flow''s in_port to OFPP_NONE.
610 *
611 * This function does post-processing on data returned from
612 * odp_flow_key_to_flow() to help make VLAN splinters transparent to the rest
613 * of the upcall processing logic. In particular, if the extracted in_port is
614 * a VLAN splinter port, it replaces flow->in_port by the "real" port, sets
615 * flow->vlan_tci correctly for the VLAN of the VLAN splinter port, and pushes
616 * a VLAN header onto 'packet' (if it is nonnull).
617 *
618 * Similarly, this function also includes some logic to help with tunnels. It
619 * may modify 'flow' as necessary to make the tunneling implementation
620 * transparent to the upcall processing logic.
621 *
622 * Returns 0 if successful, ENODEV if the parsed flow has no associated ofport,
623 * or some other positive errno if there are other problems. */
624int
625xlate_receive(const struct dpif_backer *backer, struct ofpbuf *packet,
836fbda7 626 const struct nlattr *key, size_t key_len, struct flow *flow,
1dfdb9b3
EJ
627 struct ofproto_dpif **ofproto, struct dpif_ipfix **ipfix,
628 struct dpif_sflow **sflow, struct netflow **netflow,
629 odp_port_t *odp_in_port)
8449c4d6 630{
8449c4d6
EJ
631 const struct xport *xport;
632 int error = ENODEV;
633
dc24a00f 634 ovs_rwlock_rdlock(&xlate_rwlock);
836fbda7 635 if (odp_flow_key_to_flow(key, key_len, flow) == ODP_FIT_ERROR) {
8449c4d6
EJ
636 error = EINVAL;
637 goto exit;
638 }
639
640 if (odp_in_port) {
641 *odp_in_port = flow->in_port.odp_port;
642 }
643
644 xport = xport_lookup(tnl_port_should_receive(flow)
d6fc5f57
EJ
645 ? tnl_port_receive(flow)
646 : odp_port_to_ofport(backer, flow->in_port.odp_port));
8449c4d6
EJ
647
648 flow->in_port.ofp_port = xport ? xport->ofp_port : OFPP_NONE;
649 if (!xport) {
650 goto exit;
651 }
652
653 if (vsp_adjust_flow(xport->xbridge->ofproto, flow)) {
654 if (packet) {
655 /* Make the packet resemble the flow, so that it gets sent to
656 * an OpenFlow controller properly, so that it looks correct
657 * for sFlow, and so that flow_extract() will get the correct
da546e07 658 * vlan_tci if it is called on 'packet'. */
1bf02876 659 eth_push_vlan(packet, htons(ETH_TYPE_VLAN), flow->vlan_tci);
8449c4d6 660 }
8449c4d6
EJ
661 }
662 error = 0;
663
664 if (ofproto) {
665 *ofproto = xport->xbridge->ofproto;
666 }
667
1dfdb9b3
EJ
668 if (ipfix) {
669 *ipfix = dpif_ipfix_ref(xport->xbridge->ipfix);
670 }
671
672 if (sflow) {
673 *sflow = dpif_sflow_ref(xport->xbridge->sflow);
674 }
675
676 if (netflow) {
677 *netflow = netflow_ref(xport->xbridge->netflow);
678 }
679
8449c4d6 680exit:
dc24a00f 681 ovs_rwlock_unlock(&xlate_rwlock);
8449c4d6
EJ
682 return error;
683}
684
46c88433
EJ
685static struct xbridge *
686xbridge_lookup(const struct ofproto_dpif *ofproto)
687{
688 struct xbridge *xbridge;
689
5e6af486
EJ
690 if (!ofproto) {
691 return NULL;
692 }
693
46c88433
EJ
694 HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0),
695 &xbridges) {
696 if (xbridge->ofproto == ofproto) {
697 return xbridge;
698 }
699 }
700 return NULL;
701}
702
703static struct xbundle *
704xbundle_lookup(const struct ofbundle *ofbundle)
705{
706 struct xbundle *xbundle;
707
5e6af486
EJ
708 if (!ofbundle) {
709 return NULL;
710 }
711
46c88433
EJ
712 HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0),
713 &xbundles) {
714 if (xbundle->ofbundle == ofbundle) {
715 return xbundle;
716 }
717 }
718 return NULL;
719}
720
721static struct xport *
5e6af486 722xport_lookup(const struct ofport_dpif *ofport)
46c88433
EJ
723{
724 struct xport *xport;
725
5e6af486
EJ
726 if (!ofport) {
727 return NULL;
728 }
729
46c88433
EJ
730 HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0),
731 &xports) {
732 if (xport->ofport == ofport) {
733 return xport;
734 }
735 }
736 return NULL;
737}
738
40085e56
EJ
739static struct stp_port *
740xport_get_stp_port(const struct xport *xport)
741{
92cf817b 742 return xport->xbridge->stp && xport->stp_port_no != -1
40085e56
EJ
743 ? stp_get_port(xport->xbridge->stp, xport->stp_port_no)
744 : NULL;
745}
9d189a50 746
0d1cee12 747static bool
9d189a50
EJ
748xport_stp_learn_state(const struct xport *xport)
749{
40085e56
EJ
750 struct stp_port *sp = xport_get_stp_port(xport);
751 return stp_learn_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
9d189a50
EJ
752}
753
754static bool
755xport_stp_forward_state(const struct xport *xport)
756{
40085e56
EJ
757 struct stp_port *sp = xport_get_stp_port(xport);
758 return stp_forward_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
9d189a50
EJ
759}
760
0d1cee12
K
761static bool
762xport_stp_listen_state(const struct xport *xport)
763{
764 struct stp_port *sp = xport_get_stp_port(xport);
765 return stp_listen_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
766}
767
9d189a50
EJ
768/* Returns true if STP should process 'flow'. Sets fields in 'wc' that
769 * were used to make the determination.*/
770static bool
771stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc)
772{
bbbca389 773 /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */
9d189a50 774 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
bbbca389 775 return is_stp(flow);
9d189a50
EJ
776}
777
778static void
779stp_process_packet(const struct xport *xport, const struct ofpbuf *packet)
780{
40085e56 781 struct stp_port *sp = xport_get_stp_port(xport);
9d189a50 782 struct ofpbuf payload = *packet;
1f317cb5 783 struct eth_header *eth = ofpbuf_data(&payload);
9d189a50
EJ
784
785 /* Sink packets on ports that have STP disabled when the bridge has
786 * STP enabled. */
787 if (!sp || stp_port_get_state(sp) == STP_DISABLED) {
788 return;
789 }
790
791 /* Trim off padding on payload. */
1f317cb5
PS
792 if (ofpbuf_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) {
793 ofpbuf_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN);
9d189a50
EJ
794 }
795
796 if (ofpbuf_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) {
1f317cb5 797 stp_received_bpdu(sp, ofpbuf_data(&payload), ofpbuf_size(&payload));
9d189a50
EJ
798 }
799}
800
46c88433
EJ
801static struct xport *
802get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port)
803{
804 struct xport *xport;
805
806 HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port),
807 &xbridge->xports) {
808 if (xport->ofp_port == ofp_port) {
809 return xport;
810 }
811 }
812 return NULL;
813}
814
815static odp_port_t
816ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port)
817{
818 const struct xport *xport = get_ofp_port(xbridge, ofp_port);
819 return xport ? xport->odp_port : ODPP_NONE;
820}
821
dd8cd4b4
SH
822static bool
823odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port)
824{
825 struct xport *xport;
826
827 xport = get_ofp_port(ctx->xbridge, ofp_port);
828 if (!xport || xport->config & OFPUTIL_PC_PORT_DOWN ||
829 xport->state & OFPUTIL_PS_LINK_DOWN) {
830 return false;
831 }
832
833 return true;
834}
835
836static const struct ofputil_bucket *
837group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *,
838 int depth);
839
840static bool
841group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth)
842{
843 struct group_dpif *group;
844 bool hit;
845
846 hit = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group);
847 if (!hit) {
848 return false;
849 }
850
851 hit = group_first_live_bucket(ctx, group, depth) != NULL;
852
853 group_dpif_release(group);
854 return hit;
855}
856
857#define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */
858
859static bool
860bucket_is_alive(const struct xlate_ctx *ctx,
861 const struct ofputil_bucket *bucket, int depth)
862{
863 if (depth >= MAX_LIVENESS_RECURSION) {
864 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
865
866 VLOG_WARN_RL(&rl, "bucket chaining exceeded %d links",
867 MAX_LIVENESS_RECURSION);
868 return false;
869 }
870
fe7e5749 871 return !ofputil_bucket_has_liveness(bucket) ||
d6fc5f57
EJ
872 (bucket->watch_port != OFPP_ANY &&
873 odp_port_is_alive(ctx, bucket->watch_port)) ||
874 (bucket->watch_group != OFPG_ANY &&
875 group_is_alive(ctx, bucket->watch_group, depth + 1));
dd8cd4b4
SH
876}
877
878static const struct ofputil_bucket *
879group_first_live_bucket(const struct xlate_ctx *ctx,
880 const struct group_dpif *group, int depth)
881{
882 struct ofputil_bucket *bucket;
883 const struct list *buckets;
884
885 group_dpif_get_buckets(group, &buckets);
886 LIST_FOR_EACH (bucket, list_node, buckets) {
887 if (bucket_is_alive(ctx, bucket, depth)) {
888 return bucket;
889 }
890 }
891
892 return NULL;
893}
894
fe7e5749
SH
895static const struct ofputil_bucket *
896group_best_live_bucket(const struct xlate_ctx *ctx,
897 const struct group_dpif *group,
898 uint32_t basis)
899{
900 const struct ofputil_bucket *best_bucket = NULL;
901 uint32_t best_score = 0;
902 int i = 0;
903
904 const struct ofputil_bucket *bucket;
905 const struct list *buckets;
906
907 group_dpif_get_buckets(group, &buckets);
908 LIST_FOR_EACH (bucket, list_node, buckets) {
909 if (bucket_is_alive(ctx, bucket, 0)) {
7cb279c2 910 uint32_t score = (hash_int(i, basis) & 0xffff) * bucket->weight;
fe7e5749
SH
911 if (score >= best_score) {
912 best_bucket = bucket;
913 best_score = score;
914 }
915 }
916 i++;
917 }
918
919 return best_bucket;
920}
921
9583bc14 922static bool
46c88433 923xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan)
9583bc14
EJ
924{
925 return (bundle->vlan_mode != PORT_VLAN_ACCESS
926 && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan)));
927}
928
929static bool
46c88433
EJ
930xbundle_includes_vlan(const struct xbundle *xbundle, uint16_t vlan)
931{
932 return vlan == xbundle->vlan || xbundle_trunks_vlan(xbundle, vlan);
933}
934
935static mirror_mask_t
936xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle)
937{
938 return xbundle != &ofpp_none_bundle
939 ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle)
940 : 0;
941}
942
943static mirror_mask_t
944xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle)
9583bc14 945{
46c88433
EJ
946 return xbundle != &ofpp_none_bundle
947 ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle)
948 : 0;
9583bc14
EJ
949}
950
46c88433
EJ
951static mirror_mask_t
952xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle)
9583bc14 953{
46c88433
EJ
954 return xbundle != &ofpp_none_bundle
955 ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle)
956 : 0;
957}
958
959static struct xbundle *
960lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port,
961 bool warn, struct xport **in_xportp)
962{
963 struct xport *xport;
9583bc14
EJ
964
965 /* Find the port and bundle for the received packet. */
46c88433
EJ
966 xport = get_ofp_port(xbridge, in_port);
967 if (in_xportp) {
968 *in_xportp = xport;
9583bc14 969 }
46c88433
EJ
970 if (xport && xport->xbundle) {
971 return xport->xbundle;
9583bc14
EJ
972 }
973
974 /* Special-case OFPP_NONE, which a controller may use as the ingress
975 * port for traffic that it is sourcing. */
976 if (in_port == OFPP_NONE) {
977 return &ofpp_none_bundle;
978 }
979
980 /* Odd. A few possible reasons here:
981 *
982 * - We deleted a port but there are still a few packets queued up
983 * from it.
984 *
985 * - Someone externally added a port (e.g. "ovs-dpctl add-if") that
986 * we don't know about.
987 *
988 * - The ofproto client didn't configure the port as part of a bundle.
989 * This is particularly likely to happen if a packet was received on the
990 * port after it was created, but before the client had a chance to
991 * configure its bundle.
992 */
993 if (warn) {
994 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
995
996 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
46c88433 997 "port %"PRIu16, xbridge->name, in_port);
9583bc14
EJ
998 }
999 return NULL;
1000}
1001
1002static void
1003add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow)
1004{
46c88433 1005 const struct xbridge *xbridge = ctx->xbridge;
9583bc14 1006 mirror_mask_t mirrors;
46c88433 1007 struct xbundle *in_xbundle;
9583bc14
EJ
1008 uint16_t vlan;
1009 uint16_t vid;
cdf5d3a5
EJ
1010
1011 mirrors = ctx->xout->mirrors;
1012 ctx->xout->mirrors = 0;
9583bc14 1013
46c88433
EJ
1014 in_xbundle = lookup_input_bundle(xbridge, orig_flow->in_port.ofp_port,
1015 ctx->xin->packet != NULL, NULL);
1016 if (!in_xbundle) {
9583bc14
EJ
1017 return;
1018 }
46c88433 1019 mirrors |= xbundle_mirror_src(xbridge, in_xbundle);
9583bc14
EJ
1020
1021 /* Drop frames on bundles reserved for mirroring. */
46c88433 1022 if (xbundle_mirror_out(xbridge, in_xbundle)) {
9583bc14
EJ
1023 if (ctx->xin->packet != NULL) {
1024 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1025 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
1026 "%s, which is reserved exclusively for mirroring",
46c88433 1027 ctx->xbridge->name, in_xbundle->name);
9583bc14 1028 }
aaa0fbae 1029 ofpbuf_clear(&ctx->xout->odp_actions);
9583bc14
EJ
1030 return;
1031 }
1032
1033 /* Check VLAN. */
1034 vid = vlan_tci_to_vid(orig_flow->vlan_tci);
46c88433 1035 if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) {
9583bc14
EJ
1036 return;
1037 }
46c88433 1038 vlan = input_vid_to_vlan(in_xbundle, vid);
9583bc14 1039
9583bc14
EJ
1040 if (!mirrors) {
1041 return;
1042 }
1043
1044 /* Restore the original packet before adding the mirror actions. */
1045 ctx->xin->flow = *orig_flow;
1046
1047 while (mirrors) {
ec7ceaed
EJ
1048 mirror_mask_t dup_mirrors;
1049 struct ofbundle *out;
1050 unsigned long *vlans;
1051 bool vlan_mirrored;
1052 bool has_mirror;
1053 int out_vlan;
1054
4b6ab2b0 1055 has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors),
ec7ceaed
EJ
1056 &vlans, &dup_mirrors, &out, &out_vlan);
1057 ovs_assert(has_mirror);
1058
1059 if (vlans) {
9583bc14
EJ
1060 ctx->xout->wc.masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK);
1061 }
ec7ceaed
EJ
1062 vlan_mirrored = !vlans || bitmap_is_set(vlans, vlan);
1063 free(vlans);
9583bc14 1064
ec7ceaed 1065 if (!vlan_mirrored) {
9583bc14
EJ
1066 mirrors = zero_rightmost_1bit(mirrors);
1067 continue;
1068 }
1069
ec7ceaed
EJ
1070 mirrors &= ~dup_mirrors;
1071 ctx->xout->mirrors |= dup_mirrors;
1072 if (out) {
46c88433
EJ
1073 struct xbundle *out_xbundle = xbundle_lookup(out);
1074 if (out_xbundle) {
1075 output_normal(ctx, out_xbundle, vlan);
1076 }
ec7ceaed 1077 } else if (vlan != out_vlan
9583bc14 1078 && !eth_addr_is_reserved(orig_flow->dl_dst)) {
46c88433 1079 struct xbundle *xbundle;
9583bc14 1080
46c88433
EJ
1081 LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) {
1082 if (xbundle_includes_vlan(xbundle, out_vlan)
1083 && !xbundle_mirror_out(xbridge, xbundle)) {
1084 output_normal(ctx, xbundle, out_vlan);
9583bc14
EJ
1085 }
1086 }
1087 }
1088 }
1089}
1090
1091/* Given 'vid', the VID obtained from the 802.1Q header that was received as
46c88433 1092 * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle',
9583bc14
EJ
1093 * the bundle on which the packet was received, returns the VLAN to which the
1094 * packet belongs.
1095 *
1096 * Both 'vid' and the return value are in the range 0...4095. */
1097static uint16_t
46c88433 1098input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid)
9583bc14 1099{
46c88433 1100 switch (in_xbundle->vlan_mode) {
9583bc14 1101 case PORT_VLAN_ACCESS:
46c88433 1102 return in_xbundle->vlan;
9583bc14
EJ
1103 break;
1104
1105 case PORT_VLAN_TRUNK:
1106 return vid;
1107
1108 case PORT_VLAN_NATIVE_UNTAGGED:
1109 case PORT_VLAN_NATIVE_TAGGED:
46c88433 1110 return vid ? vid : in_xbundle->vlan;
9583bc14
EJ
1111
1112 default:
428b2edd 1113 OVS_NOT_REACHED();
9583bc14
EJ
1114 }
1115}
1116
46c88433 1117/* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'.
9583bc14
EJ
1118 * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs
1119 * a warning.
1120 *
1121 * 'vid' should be the VID obtained from the 802.1Q header that was received as
1122 * part of a packet (specify 0 if there was no 802.1Q header), in the range
1123 * 0...4095. */
1124static bool
46c88433 1125input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn)
9583bc14
EJ
1126{
1127 /* Allow any VID on the OFPP_NONE port. */
46c88433 1128 if (in_xbundle == &ofpp_none_bundle) {
9583bc14
EJ
1129 return true;
1130 }
1131
46c88433 1132 switch (in_xbundle->vlan_mode) {
9583bc14
EJ
1133 case PORT_VLAN_ACCESS:
1134 if (vid) {
1135 if (warn) {
1136 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
46c88433 1137 VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" tagged "
9583bc14 1138 "packet received on port %s configured as VLAN "
46c88433
EJ
1139 "%"PRIu16" access port", vid, in_xbundle->name,
1140 in_xbundle->vlan);
9583bc14
EJ
1141 }
1142 return false;
1143 }
1144 return true;
1145
1146 case PORT_VLAN_NATIVE_UNTAGGED:
1147 case PORT_VLAN_NATIVE_TAGGED:
1148 if (!vid) {
1149 /* Port must always carry its native VLAN. */
1150 return true;
1151 }
1152 /* Fall through. */
1153 case PORT_VLAN_TRUNK:
46c88433 1154 if (!xbundle_includes_vlan(in_xbundle, vid)) {
9583bc14
EJ
1155 if (warn) {
1156 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
46c88433 1157 VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" packet "
9583bc14 1158 "received on port %s not configured for trunking "
46c88433 1159 "VLAN %"PRIu16, vid, in_xbundle->name, vid);
9583bc14
EJ
1160 }
1161 return false;
1162 }
1163 return true;
1164
1165 default:
428b2edd 1166 OVS_NOT_REACHED();
9583bc14
EJ
1167 }
1168
1169}
1170
1171/* Given 'vlan', the VLAN that a packet belongs to, and
46c88433 1172 * 'out_xbundle', a bundle on which the packet is to be output, returns the VID
9583bc14
EJ
1173 * that should be included in the 802.1Q header. (If the return value is 0,
1174 * then the 802.1Q header should only be included in the packet if there is a
1175 * nonzero PCP.)
1176 *
1177 * Both 'vlan' and the return value are in the range 0...4095. */
1178static uint16_t
46c88433 1179output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan)
9583bc14 1180{
46c88433 1181 switch (out_xbundle->vlan_mode) {
9583bc14
EJ
1182 case PORT_VLAN_ACCESS:
1183 return 0;
1184
1185 case PORT_VLAN_TRUNK:
1186 case PORT_VLAN_NATIVE_TAGGED:
1187 return vlan;
1188
1189 case PORT_VLAN_NATIVE_UNTAGGED:
46c88433 1190 return vlan == out_xbundle->vlan ? 0 : vlan;
9583bc14
EJ
1191
1192 default:
428b2edd 1193 OVS_NOT_REACHED();
9583bc14
EJ
1194 }
1195}
1196
1197static void
46c88433 1198output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle,
9583bc14
EJ
1199 uint16_t vlan)
1200{
33bf9176 1201 ovs_be16 *flow_tci = &ctx->xin->flow.vlan_tci;
9583bc14
EJ
1202 uint16_t vid;
1203 ovs_be16 tci, old_tci;
46c88433 1204 struct xport *xport;
9583bc14 1205
46c88433
EJ
1206 vid = output_vlan_to_vid(out_xbundle, vlan);
1207 if (list_is_empty(&out_xbundle->xports)) {
1208 /* Partially configured bundle with no slaves. Drop the packet. */
1209 return;
1210 } else if (!out_xbundle->bond) {
adcf00ba 1211 ctx->xout->use_recirc = false;
46c88433
EJ
1212 xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport,
1213 bundle_node);
9583bc14 1214 } else {
46c88433 1215 struct ofport_dpif *ofport;
adcf00ba 1216 struct xlate_recirc *xr = &ctx->xout->recirc;
54ecb5a2 1217 struct flow_wildcards *wc = &ctx->xout->wc;
adcf00ba
AZ
1218
1219 if (ctx->xbridge->enable_recirc) {
1220 ctx->xout->use_recirc = bond_may_recirc(
62ac1f20 1221 out_xbundle->bond, &xr->recirc_id, &xr->hash_basis);
adcf00ba
AZ
1222
1223 if (ctx->xout->use_recirc) {
1224 /* Only TCP mode uses recirculation. */
347bf289 1225 xr->hash_alg = OVS_HASH_ALG_L4;
adcf00ba 1226 bond_update_post_recirc_rules(out_xbundle->bond, false);
54ecb5a2
AZ
1227
1228 /* Recirculation does not require unmasking hash fields. */
1229 wc = NULL;
adcf00ba
AZ
1230 }
1231 }
46c88433 1232
54ecb5a2
AZ
1233 ofport = bond_choose_output_slave(out_xbundle->bond,
1234 &ctx->xin->flow, wc, vid);
5e6af486 1235 xport = xport_lookup(ofport);
46c88433
EJ
1236
1237 if (!xport) {
9583bc14
EJ
1238 /* No slaves enabled, so drop packet. */
1239 return;
1240 }
d6fc5f57 1241
b256dc52
JS
1242 /* If ctx->xout->use_recirc is set, the main thread will handle stats
1243 * accounting for this bond. */
1244 if (!ctx->xout->use_recirc) {
1245 if (ctx->xin->resubmit_stats) {
1246 bond_account(out_xbundle->bond, &ctx->xin->flow, vid,
1247 ctx->xin->resubmit_stats->n_bytes);
1248 }
1249 if (ctx->xin->xcache) {
1250 struct xc_entry *entry;
1251 struct flow *flow;
1252
1253 flow = &ctx->xin->flow;
1254 entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND);
1255 entry->u.bond.bond = bond_ref(out_xbundle->bond);
1256 entry->u.bond.flow = xmemdup(flow, sizeof *flow);
1257 entry->u.bond.vid = vid;
1258 }
d6fc5f57 1259 }
9583bc14
EJ
1260 }
1261
33bf9176 1262 old_tci = *flow_tci;
9583bc14 1263 tci = htons(vid);
46c88433 1264 if (tci || out_xbundle->use_priority_tags) {
33bf9176 1265 tci |= *flow_tci & htons(VLAN_PCP_MASK);
9583bc14
EJ
1266 if (tci) {
1267 tci |= htons(VLAN_CFI);
1268 }
1269 }
33bf9176 1270 *flow_tci = tci;
9583bc14 1271
46c88433 1272 compose_output_action(ctx, xport->ofp_port);
33bf9176 1273 *flow_tci = old_tci;
9583bc14
EJ
1274}
1275
1276/* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
1277 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
1278 * indicate this; newer upstream kernels use gratuitous ARP requests. */
1279static bool
1280is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc)
1281{
1282 if (flow->dl_type != htons(ETH_TYPE_ARP)) {
1283 return false;
1284 }
1285
1286 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
1287 if (!eth_addr_is_broadcast(flow->dl_dst)) {
1288 return false;
1289 }
1290
1291 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
1292 if (flow->nw_proto == ARP_OP_REPLY) {
1293 return true;
1294 } else if (flow->nw_proto == ARP_OP_REQUEST) {
1295 memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src);
1296 memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst);
1297
1298 return flow->nw_src == flow->nw_dst;
1299 } else {
1300 return false;
1301 }
1302}
1303
ee047520
BP
1304/* Checks whether a MAC learning update is necessary for MAC learning table
1305 * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in
1306 * 'vlan'.
1307 *
1308 * Most packets processed through the MAC learning table do not actually
1309 * change it in any way. This function requires only a read lock on the MAC
1310 * learning table, so it is much cheaper in this common case.
1311 *
1312 * Keep the code here synchronized with that in update_learning_table__()
1313 * below. */
1314static bool
1315is_mac_learning_update_needed(const struct mac_learning *ml,
1316 const struct flow *flow,
1317 struct flow_wildcards *wc,
1318 int vlan, struct xbundle *in_xbundle)
d6fc5f57 1319OVS_REQ_RDLOCK(ml->rwlock)
9583bc14
EJ
1320{
1321 struct mac_entry *mac;
1322
ee047520
BP
1323 if (!mac_learning_may_learn(ml, flow->dl_src, vlan)) {
1324 return false;
1325 }
1326
1327 mac = mac_learning_lookup(ml, flow->dl_src, vlan);
1328 if (!mac || mac_entry_age(ml, mac)) {
1329 return true;
9583bc14
EJ
1330 }
1331
ee047520
BP
1332 if (is_gratuitous_arp(flow, wc)) {
1333 /* We don't want to learn from gratuitous ARP packets that are
1334 * reflected back over bond slaves so we lock the learning table. */
1335 if (!in_xbundle->bond) {
1336 return true;
1337 } else if (mac_entry_is_grat_arp_locked(mac)) {
1338 return false;
1339 }
1340 }
1341
1342 return mac->port.p != in_xbundle->ofbundle;
1343}
1344
1345
1346/* Updates MAC learning table 'ml' given that a packet matching 'flow' was
1347 * received on 'in_xbundle' in 'vlan'.
1348 *
1349 * This code repeats all the checks in is_mac_learning_update_needed() because
1350 * the lock was released between there and here and thus the MAC learning state
1351 * could have changed.
1352 *
1353 * Keep the code here synchronized with that in is_mac_learning_update_needed()
1354 * above. */
1355static void
1356update_learning_table__(const struct xbridge *xbridge,
1357 const struct flow *flow, struct flow_wildcards *wc,
1358 int vlan, struct xbundle *in_xbundle)
d6fc5f57 1359OVS_REQ_WRLOCK(xbridge->ml->rwlock)
ee047520
BP
1360{
1361 struct mac_entry *mac;
1362
46c88433 1363 if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) {
ee047520 1364 return;
9583bc14
EJ
1365 }
1366
46c88433 1367 mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan);
9583bc14
EJ
1368 if (is_gratuitous_arp(flow, wc)) {
1369 /* We don't want to learn from gratuitous ARP packets that are
1370 * reflected back over bond slaves so we lock the learning table. */
46c88433 1371 if (!in_xbundle->bond) {
9583bc14
EJ
1372 mac_entry_set_grat_arp_lock(mac);
1373 } else if (mac_entry_is_grat_arp_locked(mac)) {
ee047520 1374 return;
9583bc14
EJ
1375 }
1376 }
1377
30618594 1378 if (mac->port.p != in_xbundle->ofbundle) {
9583bc14
EJ
1379 /* The log messages here could actually be useful in debugging,
1380 * so keep the rate limit relatively high. */
1381 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
ee047520 1382
9583bc14
EJ
1383 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
1384 "on port %s in VLAN %d",
46c88433
EJ
1385 xbridge->name, ETH_ADDR_ARGS(flow->dl_src),
1386 in_xbundle->name, vlan);
9583bc14 1387
46c88433 1388 mac->port.p = in_xbundle->ofbundle;
30618594 1389 mac_learning_changed(xbridge->ml);
9583bc14 1390 }
ee047520
BP
1391}
1392
1393static void
1394update_learning_table(const struct xbridge *xbridge,
1395 const struct flow *flow, struct flow_wildcards *wc,
1396 int vlan, struct xbundle *in_xbundle)
1397{
1398 bool need_update;
1399
1400 /* Don't learn the OFPP_NONE port. */
1401 if (in_xbundle == &ofpp_none_bundle) {
1402 return;
1403 }
1404
1405 /* First try the common case: no change to MAC learning table. */
1406 ovs_rwlock_rdlock(&xbridge->ml->rwlock);
1407 need_update = is_mac_learning_update_needed(xbridge->ml, flow, wc, vlan,
1408 in_xbundle);
509c0149 1409 ovs_rwlock_unlock(&xbridge->ml->rwlock);
ee047520
BP
1410
1411 if (need_update) {
1412 /* Slow path: MAC learning table might need an update. */
1413 ovs_rwlock_wrlock(&xbridge->ml->rwlock);
1414 update_learning_table__(xbridge, flow, wc, vlan, in_xbundle);
1415 ovs_rwlock_unlock(&xbridge->ml->rwlock);
1416 }
9583bc14
EJ
1417}
1418
46c88433 1419/* Determines whether packets in 'flow' within 'xbridge' should be forwarded or
9583bc14
EJ
1420 * dropped. Returns true if they may be forwarded, false if they should be
1421 * dropped.
1422 *
46c88433 1423 * 'in_port' must be the xport that corresponds to flow->in_port.
9583bc14
EJ
1424 * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull).
1425 *
1426 * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as
1427 * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as
1428 * checked by input_vid_is_valid().
1429 *
1430 * May also add tags to '*tags', although the current implementation only does
1431 * so in one special case.
1432 */
1433static bool
46c88433 1434is_admissible(struct xlate_ctx *ctx, struct xport *in_port,
9583bc14
EJ
1435 uint16_t vlan)
1436{
46c88433
EJ
1437 struct xbundle *in_xbundle = in_port->xbundle;
1438 const struct xbridge *xbridge = ctx->xbridge;
9583bc14 1439 struct flow *flow = &ctx->xin->flow;
9583bc14
EJ
1440
1441 /* Drop frames for reserved multicast addresses
1442 * only if forward_bpdu option is absent. */
46c88433 1443 if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) {
9583bc14
EJ
1444 xlate_report(ctx, "packet has reserved destination MAC, dropping");
1445 return false;
1446 }
1447
46c88433 1448 if (in_xbundle->bond) {
9583bc14
EJ
1449 struct mac_entry *mac;
1450
46c88433 1451 switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport,
4a1b8f30 1452 flow->dl_dst)) {
9583bc14
EJ
1453 case BV_ACCEPT:
1454 break;
1455
1456 case BV_DROP:
1457 xlate_report(ctx, "bonding refused admissibility, dropping");
1458 return false;
1459
1460 case BV_DROP_IF_MOVED:
509c0149 1461 ovs_rwlock_rdlock(&xbridge->ml->rwlock);
30618594 1462 mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan);
46c88433 1463 if (mac && mac->port.p != in_xbundle->ofbundle &&
9583bc14
EJ
1464 (!is_gratuitous_arp(flow, &ctx->xout->wc)
1465 || mac_entry_is_grat_arp_locked(mac))) {
509c0149 1466 ovs_rwlock_unlock(&xbridge->ml->rwlock);
9583bc14 1467 xlate_report(ctx, "SLB bond thinks this packet looped back, "
d6fc5f57 1468 "dropping");
9583bc14
EJ
1469 return false;
1470 }
509c0149 1471 ovs_rwlock_unlock(&xbridge->ml->rwlock);
9583bc14
EJ
1472 break;
1473 }
1474 }
1475
1476 return true;
1477}
1478
1479static void
1480xlate_normal(struct xlate_ctx *ctx)
1481{
33bf9176
BP
1482 struct flow_wildcards *wc = &ctx->xout->wc;
1483 struct flow *flow = &ctx->xin->flow;
46c88433
EJ
1484 struct xbundle *in_xbundle;
1485 struct xport *in_port;
9583bc14 1486 struct mac_entry *mac;
d6d5bbc9 1487 void *mac_port;
9583bc14
EJ
1488 uint16_t vlan;
1489 uint16_t vid;
1490
1491 ctx->xout->has_normal = true;
1492
33bf9176
BP
1493 memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src);
1494 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
1dd35f8a 1495 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
9583bc14 1496
46c88433
EJ
1497 in_xbundle = lookup_input_bundle(ctx->xbridge, flow->in_port.ofp_port,
1498 ctx->xin->packet != NULL, &in_port);
1499 if (!in_xbundle) {
9583bc14
EJ
1500 xlate_report(ctx, "no input bundle, dropping");
1501 return;
1502 }
1503
1504 /* Drop malformed frames. */
33bf9176
BP
1505 if (flow->dl_type == htons(ETH_TYPE_VLAN) &&
1506 !(flow->vlan_tci & htons(VLAN_CFI))) {
9583bc14
EJ
1507 if (ctx->xin->packet != NULL) {
1508 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1509 VLOG_WARN_RL(&rl, "bridge %s: dropping packet with partial "
1510 "VLAN tag received on port %s",
46c88433 1511 ctx->xbridge->name, in_xbundle->name);
9583bc14
EJ
1512 }
1513 xlate_report(ctx, "partial VLAN tag, dropping");
1514 return;
1515 }
1516
1517 /* Drop frames on bundles reserved for mirroring. */
46c88433 1518 if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) {
9583bc14
EJ
1519 if (ctx->xin->packet != NULL) {
1520 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1521 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
1522 "%s, which is reserved exclusively for mirroring",
46c88433 1523 ctx->xbridge->name, in_xbundle->name);
9583bc14
EJ
1524 }
1525 xlate_report(ctx, "input port is mirror output port, dropping");
1526 return;
1527 }
1528
1529 /* Check VLAN. */
33bf9176 1530 vid = vlan_tci_to_vid(flow->vlan_tci);
46c88433 1531 if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) {
9583bc14
EJ
1532 xlate_report(ctx, "disallowed VLAN VID for this input port, dropping");
1533 return;
1534 }
46c88433 1535 vlan = input_vid_to_vlan(in_xbundle, vid);
9583bc14
EJ
1536
1537 /* Check other admissibility requirements. */
1538 if (in_port && !is_admissible(ctx, in_port, vlan)) {
1539 return;
1540 }
1541
1542 /* Learn source MAC. */
1543 if (ctx->xin->may_learn) {
46c88433 1544 update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle);
9583bc14 1545 }
b256dc52
JS
1546 if (ctx->xin->xcache) {
1547 struct xc_entry *entry;
1548
1549 /* Save enough info to update mac learning table later. */
1550 entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL);
1551 entry->u.normal.ofproto = ctx->xin->ofproto;
1552 entry->u.normal.flow = xmemdup(flow, sizeof *flow);
1553 entry->u.normal.vlan = vlan;
1554 }
9583bc14
EJ
1555
1556 /* Determine output bundle. */
509c0149 1557 ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock);
30618594 1558 mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan);
d6d5bbc9
EJ
1559 mac_port = mac ? mac->port.p : NULL;
1560 ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock);
1561
1562 if (mac_port) {
1563 struct xbundle *mac_xbundle = xbundle_lookup(mac_port);
46c88433 1564 if (mac_xbundle && mac_xbundle != in_xbundle) {
9583bc14 1565 xlate_report(ctx, "forwarding to learned port");
46c88433
EJ
1566 output_normal(ctx, mac_xbundle, vlan);
1567 } else if (!mac_xbundle) {
1568 xlate_report(ctx, "learned port is unknown, dropping");
9583bc14
EJ
1569 } else {
1570 xlate_report(ctx, "learned port is input port, dropping");
1571 }
1572 } else {
46c88433 1573 struct xbundle *xbundle;
9583bc14
EJ
1574
1575 xlate_report(ctx, "no learned MAC for destination, flooding");
46c88433
EJ
1576 LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) {
1577 if (xbundle != in_xbundle
1578 && xbundle_includes_vlan(xbundle, vlan)
1579 && xbundle->floodable
1580 && !xbundle_mirror_out(ctx->xbridge, xbundle)) {
1581 output_normal(ctx, xbundle, vlan);
9583bc14
EJ
1582 }
1583 }
1584 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
1585 }
1586}
1587
1588/* Compose SAMPLE action for sFlow or IPFIX. The given probability is
1589 * the number of packets out of UINT32_MAX to sample. The given
1590 * cookie is passed back in the callback for each sampled packet.
1591 */
1592static size_t
46c88433 1593compose_sample_action(const struct xbridge *xbridge,
9583bc14
EJ
1594 struct ofpbuf *odp_actions,
1595 const struct flow *flow,
1596 const uint32_t probability,
1597 const union user_action_cookie *cookie,
1598 const size_t cookie_size)
1599{
1600 size_t sample_offset, actions_offset;
89a8a7f0 1601 odp_port_t odp_port;
9583bc14 1602 int cookie_offset;
89a8a7f0 1603 uint32_t pid;
9583bc14
EJ
1604
1605 sample_offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SAMPLE);
1606
1607 nl_msg_put_u32(odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability);
1608
1609 actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS);
89a8a7f0
EJ
1610
1611 odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port);
9a159f74
AW
1612 pid = dpif_port_get_pid(xbridge->dpif, odp_port,
1613 flow_hash_5tuple(flow, 0));
1614 cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size,
1615 odp_actions);
9583bc14
EJ
1616
1617 nl_msg_end_nested(odp_actions, actions_offset);
1618 nl_msg_end_nested(odp_actions, sample_offset);
1619 return cookie_offset;
1620}
1621
1622static void
46c88433
EJ
1623compose_sflow_cookie(const struct xbridge *xbridge, ovs_be16 vlan_tci,
1624 odp_port_t odp_port, unsigned int n_outputs,
1625 union user_action_cookie *cookie)
9583bc14
EJ
1626{
1627 int ifindex;
1628
1629 cookie->type = USER_ACTION_COOKIE_SFLOW;
1630 cookie->sflow.vlan_tci = vlan_tci;
1631
1632 /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output
1633 * port information") for the interpretation of cookie->output. */
1634 switch (n_outputs) {
1635 case 0:
1636 /* 0x40000000 | 256 means "packet dropped for unknown reason". */
1637 cookie->sflow.output = 0x40000000 | 256;
1638 break;
1639
1640 case 1:
46c88433 1641 ifindex = dpif_sflow_odp_port_to_ifindex(xbridge->sflow, odp_port);
9583bc14
EJ
1642 if (ifindex) {
1643 cookie->sflow.output = ifindex;
1644 break;
1645 }
1646 /* Fall through. */
1647 default:
1648 /* 0x80000000 means "multiple output ports. */
1649 cookie->sflow.output = 0x80000000 | n_outputs;
1650 break;
1651 }
1652}
1653
1654/* Compose SAMPLE action for sFlow bridge sampling. */
1655static size_t
46c88433 1656compose_sflow_action(const struct xbridge *xbridge,
9583bc14
EJ
1657 struct ofpbuf *odp_actions,
1658 const struct flow *flow,
4e022ec0 1659 odp_port_t odp_port)
9583bc14
EJ
1660{
1661 uint32_t probability;
1662 union user_action_cookie cookie;
1663
46c88433 1664 if (!xbridge->sflow || flow->in_port.ofp_port == OFPP_NONE) {
9583bc14
EJ
1665 return 0;
1666 }
1667
46c88433
EJ
1668 probability = dpif_sflow_get_probability(xbridge->sflow);
1669 compose_sflow_cookie(xbridge, htons(0), odp_port,
4e022ec0 1670 odp_port == ODPP_NONE ? 0 : 1, &cookie);
9583bc14 1671
46c88433 1672 return compose_sample_action(xbridge, odp_actions, flow, probability,
9583bc14
EJ
1673 &cookie, sizeof cookie.sflow);
1674}
1675
1676static void
1677compose_flow_sample_cookie(uint16_t probability, uint32_t collector_set_id,
1678 uint32_t obs_domain_id, uint32_t obs_point_id,
1679 union user_action_cookie *cookie)
1680{
1681 cookie->type = USER_ACTION_COOKIE_FLOW_SAMPLE;
1682 cookie->flow_sample.probability = probability;
1683 cookie->flow_sample.collector_set_id = collector_set_id;
1684 cookie->flow_sample.obs_domain_id = obs_domain_id;
1685 cookie->flow_sample.obs_point_id = obs_point_id;
1686}
1687
1688static void
1689compose_ipfix_cookie(union user_action_cookie *cookie)
1690{
1691 cookie->type = USER_ACTION_COOKIE_IPFIX;
1692}
1693
1694/* Compose SAMPLE action for IPFIX bridge sampling. */
1695static void
46c88433 1696compose_ipfix_action(const struct xbridge *xbridge,
9583bc14
EJ
1697 struct ofpbuf *odp_actions,
1698 const struct flow *flow)
1699{
1700 uint32_t probability;
1701 union user_action_cookie cookie;
1702
46c88433 1703 if (!xbridge->ipfix || flow->in_port.ofp_port == OFPP_NONE) {
9583bc14
EJ
1704 return;
1705 }
1706
46c88433 1707 probability = dpif_ipfix_get_bridge_exporter_probability(xbridge->ipfix);
9583bc14
EJ
1708 compose_ipfix_cookie(&cookie);
1709
46c88433 1710 compose_sample_action(xbridge, odp_actions, flow, probability,
9583bc14
EJ
1711 &cookie, sizeof cookie.ipfix);
1712}
1713
1714/* SAMPLE action for sFlow must be first action in any given list of
1715 * actions. At this point we do not have all information required to
1716 * build it. So try to build sample action as complete as possible. */
1717static void
1718add_sflow_action(struct xlate_ctx *ctx)
1719{
46c88433 1720 ctx->user_cookie_offset = compose_sflow_action(ctx->xbridge,
9583bc14 1721 &ctx->xout->odp_actions,
4e022ec0 1722 &ctx->xin->flow, ODPP_NONE);
9583bc14
EJ
1723 ctx->sflow_odp_port = 0;
1724 ctx->sflow_n_outputs = 0;
1725}
1726
1727/* SAMPLE action for IPFIX must be 1st or 2nd action in any given list
1728 * of actions, eventually after the SAMPLE action for sFlow. */
1729static void
1730add_ipfix_action(struct xlate_ctx *ctx)
1731{
46c88433 1732 compose_ipfix_action(ctx->xbridge, &ctx->xout->odp_actions,
9583bc14
EJ
1733 &ctx->xin->flow);
1734}
1735
1736/* Fix SAMPLE action according to data collected while composing ODP actions.
1737 * We need to fix SAMPLE actions OVS_SAMPLE_ATTR_ACTIONS attribute, i.e. nested
1738 * USERSPACE action's user-cookie which is required for sflow. */
1739static void
1740fix_sflow_action(struct xlate_ctx *ctx)
1741{
1742 const struct flow *base = &ctx->base_flow;
1743 union user_action_cookie *cookie;
1744
1745 if (!ctx->user_cookie_offset) {
1746 return;
1747 }
1748
1749 cookie = ofpbuf_at(&ctx->xout->odp_actions, ctx->user_cookie_offset,
1750 sizeof cookie->sflow);
1751 ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW);
1752
46c88433 1753 compose_sflow_cookie(ctx->xbridge, base->vlan_tci,
9583bc14
EJ
1754 ctx->sflow_odp_port, ctx->sflow_n_outputs, cookie);
1755}
1756
db7d4e46 1757static enum slow_path_reason
642dc74d 1758process_special(struct xlate_ctx *ctx, const struct flow *flow,
46c88433 1759 const struct xport *xport, const struct ofpbuf *packet)
db7d4e46 1760{
642dc74d 1761 struct flow_wildcards *wc = &ctx->xout->wc;
46c88433 1762 const struct xbridge *xbridge = ctx->xbridge;
642dc74d 1763
46c88433 1764 if (!xport) {
db7d4e46 1765 return 0;
46c88433 1766 } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) {
db7d4e46 1767 if (packet) {
46c88433 1768 cfm_process_heartbeat(xport->cfm, packet);
db7d4e46
JP
1769 }
1770 return SLOW_CFM;
fab52e16 1771 } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) {
db7d4e46 1772 if (packet) {
46c88433 1773 bfd_process_packet(xport->bfd, flow, packet);
60d02c72
AW
1774 /* If POLL received, immediately sends FINAL back. */
1775 if (bfd_should_send_packet(xport->bfd)) {
1776 if (xport->peer) {
1777 ofproto_dpif_monitor_port_send_soon(xport->ofport);
1778 } else {
1779 ofproto_dpif_monitor_port_send_soon_safe(xport->ofport);
1780 }
1781 }
db7d4e46
JP
1782 }
1783 return SLOW_BFD;
46c88433 1784 } else if (xport->xbundle && xport->xbundle->lacp
db7d4e46
JP
1785 && flow->dl_type == htons(ETH_TYPE_LACP)) {
1786 if (packet) {
46c88433 1787 lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet);
db7d4e46
JP
1788 }
1789 return SLOW_LACP;
9d189a50 1790 } else if (xbridge->stp && stp_should_process_flow(flow, wc)) {
db7d4e46 1791 if (packet) {
9d189a50 1792 stp_process_packet(xport, packet);
db7d4e46
JP
1793 }
1794 return SLOW_STP;
1795 } else {
1796 return 0;
1797 }
1798}
1799
9583bc14 1800static void
4e022ec0 1801compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port,
9583bc14
EJ
1802 bool check_stp)
1803{
46c88433 1804 const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port);
1dd35f8a 1805 struct flow_wildcards *wc = &ctx->xout->wc;
33bf9176 1806 struct flow *flow = &ctx->xin->flow;
9583bc14 1807 ovs_be16 flow_vlan_tci;
1362e248 1808 uint32_t flow_pkt_mark;
9583bc14 1809 uint8_t flow_nw_tos;
4e022ec0 1810 odp_port_t out_port, odp_port;
ca077186 1811 uint8_t dscp;
9583bc14
EJ
1812
1813 /* If 'struct flow' gets additional metadata, we'll need to zero it out
1814 * before traversing a patch port. */
419681da 1815 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 26);
9583bc14 1816
46c88433 1817 if (!xport) {
9583bc14
EJ
1818 xlate_report(ctx, "Nonexistent output port");
1819 return;
46c88433 1820 } else if (xport->config & OFPUTIL_PC_NO_FWD) {
9583bc14
EJ
1821 xlate_report(ctx, "OFPPC_NO_FWD set, skipping output");
1822 return;
0d1cee12 1823 } else if (check_stp) {
bbbca389 1824 if (is_stp(&ctx->base_flow)) {
0d1cee12
K
1825 if (!xport_stp_listen_state(xport)) {
1826 xlate_report(ctx, "STP not in listening state, "
1827 "skipping bpdu output");
1828 return;
1829 }
1830 } else if (!xport_stp_forward_state(xport)) {
1831 xlate_report(ctx, "STP not in forwarding state, "
1832 "skipping output");
1833 return;
1834 }
9583bc14
EJ
1835 }
1836
46c88433
EJ
1837 if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) {
1838 ctx->xout->mirrors |= xbundle_mirror_dst(xport->xbundle->xbridge,
1839 xport->xbundle);
cdf5d3a5
EJ
1840 }
1841
46c88433
EJ
1842 if (xport->peer) {
1843 const struct xport *peer = xport->peer;
9583bc14 1844 struct flow old_flow = ctx->xin->flow;
9583bc14 1845 enum slow_path_reason special;
9583bc14 1846
46c88433
EJ
1847 ctx->xbridge = peer->xbridge;
1848 flow->in_port.ofp_port = peer->ofp_port;
33bf9176
BP
1849 flow->metadata = htonll(0);
1850 memset(&flow->tunnel, 0, sizeof flow->tunnel);
1851 memset(flow->regs, 0, sizeof flow->regs);
9583bc14 1852
642dc74d 1853 special = process_special(ctx, &ctx->xin->flow, peer,
9583bc14
EJ
1854 ctx->xin->packet);
1855 if (special) {
04594cd5 1856 ctx->xout->slow |= special;
ddd3c975 1857 } else if (may_receive(peer, ctx)) {
9d189a50 1858 if (xport_stp_forward_state(peer)) {
6d328fa2 1859 xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true);
9583bc14
EJ
1860 } else {
1861 /* Forwarding is disabled by STP. Let OFPP_NORMAL and the
1862 * learning action look at the packet, then drop it. */
1863 struct flow old_base_flow = ctx->base_flow;
1f317cb5 1864 size_t old_size = ofpbuf_size(&ctx->xout->odp_actions);
cdf5d3a5 1865 mirror_mask_t old_mirrors = ctx->xout->mirrors;
6d328fa2 1866 xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true);
cdf5d3a5 1867 ctx->xout->mirrors = old_mirrors;
9583bc14 1868 ctx->base_flow = old_base_flow;
1f317cb5 1869 ofpbuf_set_size(&ctx->xout->odp_actions, old_size);
9583bc14
EJ
1870 }
1871 }
1872
1873 ctx->xin->flow = old_flow;
832554e3 1874 ctx->xbridge = xport->xbridge;
9583bc14
EJ
1875
1876 if (ctx->xin->resubmit_stats) {
46c88433
EJ
1877 netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats);
1878 netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats);
a1aeea86
AW
1879 if (peer->bfd) {
1880 bfd_account_rx(peer->bfd, ctx->xin->resubmit_stats);
1881 }
9583bc14 1882 }
b256dc52
JS
1883 if (ctx->xin->xcache) {
1884 struct xc_entry *entry;
1885
1886 entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV);
1887 entry->u.dev.tx = netdev_ref(xport->netdev);
1888 entry->u.dev.rx = netdev_ref(peer->netdev);
1889 entry->u.dev.bfd = bfd_ref(peer->bfd);
1890 }
9583bc14
EJ
1891
1892 return;
1893 }
1894
33bf9176 1895 flow_vlan_tci = flow->vlan_tci;
1362e248 1896 flow_pkt_mark = flow->pkt_mark;
33bf9176 1897 flow_nw_tos = flow->nw_tos;
9583bc14 1898
55954f6e 1899 if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) {
1ac9326d 1900 wc->masks.nw_tos |= IP_DSCP_MASK;
33bf9176 1901 flow->nw_tos &= ~IP_DSCP_MASK;
ca077186 1902 flow->nw_tos |= dscp;
9583bc14
EJ
1903 }
1904
46c88433 1905 if (xport->is_tunnel) {
9583bc14
EJ
1906 /* Save tunnel metadata so that changes made due to
1907 * the Logical (tunnel) Port are not visible for any further
1908 * matches, while explicit set actions on tunnel metadata are.
1909 */
33bf9176 1910 struct flow_tnl flow_tnl = flow->tunnel;
46c88433 1911 odp_port = tnl_port_send(xport->ofport, flow, &ctx->xout->wc);
4e022ec0 1912 if (odp_port == ODPP_NONE) {
9583bc14
EJ
1913 xlate_report(ctx, "Tunneling decided against output");
1914 goto out; /* restore flow_nw_tos */
1915 }
33bf9176 1916 if (flow->tunnel.ip_dst == ctx->orig_tunnel_ip_dst) {
9583bc14
EJ
1917 xlate_report(ctx, "Not tunneling to our own address");
1918 goto out; /* restore flow_nw_tos */
1919 }
1920 if (ctx->xin->resubmit_stats) {
46c88433 1921 netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats);
9583bc14 1922 }
b256dc52
JS
1923 if (ctx->xin->xcache) {
1924 struct xc_entry *entry;
1925
1926 entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV);
1927 entry->u.dev.tx = netdev_ref(xport->netdev);
1928 }
9583bc14 1929 out_port = odp_port;
33bf9176 1930 commit_odp_tunnel_action(flow, &ctx->base_flow,
9583bc14 1931 &ctx->xout->odp_actions);
33bf9176 1932 flow->tunnel = flow_tnl; /* Restore tunnel metadata */
9583bc14 1933 } else {
46c88433 1934 odp_port = xport->odp_port;
7614e5d0 1935 out_port = odp_port;
46c88433 1936 if (ofproto_has_vlan_splinters(ctx->xbridge->ofproto)) {
7614e5d0
JR
1937 ofp_port_t vlandev_port;
1938
1dd35f8a 1939 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
7614e5d0
JR
1940 vlandev_port = vsp_realdev_to_vlandev(ctx->xbridge->ofproto,
1941 ofp_port, flow->vlan_tci);
1942 if (vlandev_port != ofp_port) {
1943 out_port = ofp_port_to_odp_port(ctx->xbridge, vlandev_port);
1944 flow->vlan_tci = htons(0);
1945 }
9583bc14 1946 }
9583bc14 1947 }
9583bc14 1948
4e022ec0 1949 if (out_port != ODPP_NONE) {
7fd91025
BP
1950 ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow,
1951 &ctx->xout->odp_actions,
8bfd0fda 1952 &ctx->xout->wc);
adcf00ba
AZ
1953
1954 if (ctx->xout->use_recirc) {
347bf289 1955 struct ovs_action_hash *act_hash;
adcf00ba
AZ
1956 struct xlate_recirc *xr = &ctx->xout->recirc;
1957
347bf289
AZ
1958 /* Hash action. */
1959 act_hash = nl_msg_put_unspec_uninit(&ctx->xout->odp_actions,
1960 OVS_ACTION_ATTR_HASH,
1961 sizeof *act_hash);
1962 act_hash->hash_alg = xr->hash_alg;
62ac1f20 1963 act_hash->hash_basis = xr->hash_basis;
347bf289
AZ
1964
1965 /* Recirc action. */
1966 nl_msg_put_u32(&ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC,
1967 xr->recirc_id);
adcf00ba
AZ
1968 } else {
1969 nl_msg_put_odp_port(&ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
1970 out_port);
1971 }
9583bc14 1972
6cbbf4fa
EJ
1973 ctx->sflow_odp_port = odp_port;
1974 ctx->sflow_n_outputs++;
1975 ctx->xout->nf_output_iface = ofp_port;
1976 }
1977
1978 out:
9583bc14 1979 /* Restore flow */
33bf9176 1980 flow->vlan_tci = flow_vlan_tci;
1362e248 1981 flow->pkt_mark = flow_pkt_mark;
33bf9176 1982 flow->nw_tos = flow_nw_tos;
9583bc14
EJ
1983}
1984
1985static void
4e022ec0 1986compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port)
9583bc14
EJ
1987{
1988 compose_output_action__(ctx, ofp_port, true);
1989}
1990
bb61b33d
BP
1991static void
1992xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule)
bb61b33d
BP
1993{
1994 struct rule_dpif *old_rule = ctx->rule;
dc723c44 1995 const struct rule_actions *actions;
bb61b33d
BP
1996
1997 if (ctx->xin->resubmit_stats) {
70742c7f 1998 rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats);
bb61b33d
BP
1999 }
2000
98b07853 2001 ctx->resubmits++;
bb61b33d
BP
2002 ctx->recurse++;
2003 ctx->rule = rule;
6f00e29b
BP
2004 actions = rule_dpif_get_actions(rule);
2005 do_xlate_actions(actions->ofpacts, actions->ofpacts_len, ctx);
bb61b33d
BP
2006 ctx->rule = old_rule;
2007 ctx->recurse--;
bb61b33d
BP
2008}
2009
bd3240ba
SH
2010static bool
2011xlate_resubmit_resource_check(struct xlate_ctx *ctx)
9583bc14 2012{
98b07853
BP
2013 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2014
adcf00ba 2015 if (ctx->recurse >= MAX_RESUBMIT_RECURSION + MAX_INTERNAL_RESUBMITS) {
98b07853
BP
2016 VLOG_ERR_RL(&rl, "resubmit actions recursed over %d times",
2017 MAX_RESUBMIT_RECURSION);
adcf00ba 2018 } else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) {
98b07853 2019 VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS);
1f317cb5 2020 } else if (ofpbuf_size(&ctx->xout->odp_actions) > UINT16_MAX) {
98b07853 2021 VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of actions");
1f317cb5 2022 } else if (ofpbuf_size(&ctx->stack) >= 65536) {
98b07853
BP
2023 VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of stack");
2024 } else {
bd3240ba
SH
2025 return true;
2026 }
2027
2028 return false;
2029}
2030
2031static void
6d328fa2
SH
2032xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id,
2033 bool may_packet_in, bool honor_table_miss)
bd3240ba
SH
2034{
2035 if (xlate_resubmit_resource_check(ctx)) {
4e022ec0 2036 ofp_port_t old_in_port = ctx->xin->flow.in_port.ofp_port;
3f207910 2037 bool skip_wildcards = ctx->xin->skip_wildcards;
9583bc14 2038 uint8_t old_table_id = ctx->table_id;
3f207910 2039 struct rule_dpif *rule;
6d328fa2
SH
2040 enum rule_dpif_lookup_verdict verdict;
2041 enum ofputil_port_config config = 0;
9583bc14
EJ
2042
2043 ctx->table_id = table_id;
2044
bb61b33d
BP
2045 /* Look up a flow with 'in_port' as the input port. Then restore the
2046 * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will
2047 * have surprising behavior). */
4e022ec0 2048 ctx->xin->flow.in_port.ofp_port = in_port;
6d328fa2
SH
2049 verdict = rule_dpif_lookup_from_table(ctx->xbridge->ofproto,
2050 &ctx->xin->flow,
2051 !skip_wildcards
2052 ? &ctx->xout->wc : NULL,
2053 honor_table_miss,
83709dfa
JR
2054 &ctx->table_id, &rule,
2055 ctx->xin->xcache != NULL);
4e022ec0 2056 ctx->xin->flow.in_port.ofp_port = old_in_port;
9583bc14 2057
ad3efdcb
EJ
2058 if (ctx->xin->resubmit_hook) {
2059 ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse);
2060 }
2061
6d328fa2
SH
2062 switch (verdict) {
2063 case RULE_DPIF_LOOKUP_VERDICT_MATCH:
2064 goto match;
2065 case RULE_DPIF_LOOKUP_VERDICT_CONTROLLER:
2066 if (may_packet_in) {
2067 struct xport *xport;
2068
2069 xport = get_ofp_port(ctx->xbridge,
2070 ctx->xin->flow.in_port.ofp_port);
2071 config = xport ? xport->config : 0;
2072 break;
2073 }
2074 /* Fall through to drop */
2075 case RULE_DPIF_LOOKUP_VERDICT_DROP:
2076 config = OFPUTIL_PC_NO_PACKET_IN;
2077 break;
6b83a3c5
SH
2078 case RULE_DPIF_LOOKUP_VERDICT_DEFAULT:
2079 if (!ofproto_dpif_wants_packet_in_on_miss(ctx->xbridge->ofproto)) {
2080 config = OFPUTIL_PC_NO_PACKET_IN;
2081 }
2082 break;
6d328fa2
SH
2083 default:
2084 OVS_NOT_REACHED();
a2143702 2085 }
6d328fa2
SH
2086
2087 choose_miss_rule(config, ctx->xbridge->miss_rule,
83709dfa
JR
2088 ctx->xbridge->no_packet_in_rule, &rule,
2089 ctx->xin->xcache != NULL);
6d328fa2
SH
2090
2091match:
a2143702 2092 if (rule) {
83709dfa
JR
2093 /* Fill in the cache entry here instead of xlate_recursively
2094 * to make the reference counting more explicit. We take a
2095 * reference in the lookups above if we are going to cache the
2096 * rule. */
2097 if (ctx->xin->xcache) {
2098 struct xc_entry *entry;
2099
2100 entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE);
2101 entry->u.rule = rule;
2102 }
bb61b33d 2103 xlate_recursively(ctx, rule);
ad3efdcb
EJ
2104 }
2105
9583bc14 2106 ctx->table_id = old_table_id;
98b07853 2107 return;
9583bc14 2108 }
98b07853
BP
2109
2110 ctx->exit = true;
9583bc14
EJ
2111}
2112
f4fb341b
SH
2113static void
2114xlate_group_bucket(struct xlate_ctx *ctx, const struct ofputil_bucket *bucket)
2115{
2116 uint64_t action_list_stub[1024 / 8];
2117 struct ofpbuf action_list, action_set;
2118
2119 ofpbuf_use_const(&action_set, bucket->ofpacts, bucket->ofpacts_len);
2120 ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub);
2121
2122 ofpacts_execute_action_set(&action_list, &action_set);
2123 ctx->recurse++;
1f317cb5 2124 do_xlate_actions(ofpbuf_data(&action_list), ofpbuf_size(&action_list), ctx);
f4fb341b
SH
2125 ctx->recurse--;
2126
2127 ofpbuf_uninit(&action_set);
2128 ofpbuf_uninit(&action_list);
2129}
2130
2131static void
2132xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group)
2133{
2134 const struct ofputil_bucket *bucket;
2135 const struct list *buckets;
2136 struct flow old_flow = ctx->xin->flow;
2137
2138 group_dpif_get_buckets(group, &buckets);
2139
2140 LIST_FOR_EACH (bucket, list_node, buckets) {
2141 xlate_group_bucket(ctx, bucket);
2142 /* Roll back flow to previous state.
2143 * This is equivalent to cloning the packet for each bucket.
2144 *
2145 * As a side effect any subsequently applied actions will
2146 * also effectively be applied to a clone of the packet taken
2147 * just before applying the all or indirect group. */
2148 ctx->xin->flow = old_flow;
2149 }
2150}
2151
dd8cd4b4
SH
2152static void
2153xlate_ff_group(struct xlate_ctx *ctx, struct group_dpif *group)
2154{
2155 const struct ofputil_bucket *bucket;
2156
2157 bucket = group_first_live_bucket(ctx, group, 0);
2158 if (bucket) {
2159 xlate_group_bucket(ctx, bucket);
2160 }
2161}
2162
fe7e5749
SH
2163static void
2164xlate_select_group(struct xlate_ctx *ctx, struct group_dpif *group)
2165{
2166 struct flow_wildcards *wc = &ctx->xout->wc;
2167 const struct ofputil_bucket *bucket;
2168 uint32_t basis;
2169
7e36ac42 2170 basis = hash_mac(ctx->xin->flow.dl_dst, 0, 0);
fe7e5749
SH
2171 bucket = group_best_live_bucket(ctx, group, basis);
2172 if (bucket) {
2173 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
2174 xlate_group_bucket(ctx, bucket);
2175 }
2176}
2177
f4fb341b
SH
2178static void
2179xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group)
2180{
5a070238
BP
2181 ctx->in_group = true;
2182
f4fb341b
SH
2183 switch (group_dpif_get_type(group)) {
2184 case OFPGT11_ALL:
2185 case OFPGT11_INDIRECT:
2186 xlate_all_group(ctx, group);
2187 break;
2188 case OFPGT11_SELECT:
fe7e5749 2189 xlate_select_group(ctx, group);
f4fb341b 2190 break;
dd8cd4b4
SH
2191 case OFPGT11_FF:
2192 xlate_ff_group(ctx, group);
2193 break;
f4fb341b 2194 default:
428b2edd 2195 OVS_NOT_REACHED();
f4fb341b
SH
2196 }
2197 group_dpif_release(group);
5a070238
BP
2198
2199 ctx->in_group = false;
2200}
2201
2202static bool
2203xlate_group_resource_check(struct xlate_ctx *ctx)
2204{
2205 if (!xlate_resubmit_resource_check(ctx)) {
2206 return false;
2207 } else if (ctx->in_group) {
2208 /* Prevent nested translation of OpenFlow groups.
2209 *
2210 * OpenFlow allows this restriction. We enforce this restriction only
2211 * because, with the current architecture, we would otherwise have to
2212 * take a possibly recursive read lock on the ofgroup rwlock, which is
2213 * unsafe given that POSIX allows taking a read lock to block if there
2214 * is a thread blocked on taking the write lock. Other solutions
2215 * without this restriction are also possible, but seem unwarranted
2216 * given the current limited use of groups. */
2217 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2218
2219 VLOG_ERR_RL(&rl, "cannot recursively translate OpenFlow group");
2220 return false;
2221 } else {
2222 return true;
2223 }
f4fb341b
SH
2224}
2225
2226static bool
2227xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id)
2228{
5a070238 2229 if (xlate_group_resource_check(ctx)) {
f4fb341b
SH
2230 struct group_dpif *group;
2231 bool got_group;
2232
2233 got_group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group);
2234 if (got_group) {
2235 xlate_group_action__(ctx, group);
2236 } else {
2237 return true;
2238 }
2239 }
2240
2241 return false;
2242}
2243
9583bc14
EJ
2244static void
2245xlate_ofpact_resubmit(struct xlate_ctx *ctx,
2246 const struct ofpact_resubmit *resubmit)
2247{
4e022ec0 2248 ofp_port_t in_port;
9583bc14 2249 uint8_t table_id;
adcf00ba
AZ
2250 bool may_packet_in = false;
2251 bool honor_table_miss = false;
2252
2253 if (ctx->rule && rule_dpif_is_internal(ctx->rule)) {
2254 /* Still allow missed packets to be sent to the controller
2255 * if resubmitting from an internal table. */
2256 may_packet_in = true;
2257 honor_table_miss = true;
2258 }
9583bc14
EJ
2259
2260 in_port = resubmit->in_port;
2261 if (in_port == OFPP_IN_PORT) {
4e022ec0 2262 in_port = ctx->xin->flow.in_port.ofp_port;
9583bc14
EJ
2263 }
2264
2265 table_id = resubmit->table_id;
2266 if (table_id == 255) {
2267 table_id = ctx->table_id;
2268 }
2269
adcf00ba
AZ
2270 xlate_table_action(ctx, in_port, table_id, may_packet_in,
2271 honor_table_miss);
9583bc14
EJ
2272}
2273
2274static void
2275flood_packets(struct xlate_ctx *ctx, bool all)
2276{
46c88433 2277 const struct xport *xport;
9583bc14 2278
46c88433
EJ
2279 HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) {
2280 if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) {
9583bc14
EJ
2281 continue;
2282 }
2283
2284 if (all) {
46c88433
EJ
2285 compose_output_action__(ctx, xport->ofp_port, false);
2286 } else if (!(xport->config & OFPUTIL_PC_NO_FLOOD)) {
2287 compose_output_action(ctx, xport->ofp_port);
9583bc14
EJ
2288 }
2289 }
2290
2291 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
2292}
2293
2294static void
2295execute_controller_action(struct xlate_ctx *ctx, int len,
2296 enum ofp_packet_in_reason reason,
2297 uint16_t controller_id)
2298{
0fb7792a 2299 struct ofproto_packet_in *pin;
9583bc14 2300 struct ofpbuf *packet;
758c456d 2301 struct pkt_metadata md = PKT_METADATA_INITIALIZER(0);
9583bc14 2302
04594cd5 2303 ctx->xout->slow |= SLOW_CONTROLLER;
9583bc14
EJ
2304 if (!ctx->xin->packet) {
2305 return;
2306 }
2307
2308 packet = ofpbuf_clone(ctx->xin->packet);
2309
7fd91025
BP
2310 ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
2311 &ctx->xout->odp_actions,
8bfd0fda 2312 &ctx->xout->wc);
9583bc14 2313
1f317cb5
PS
2314 odp_execute_actions(NULL, packet, false, &md,
2315 ofpbuf_data(&ctx->xout->odp_actions),
2316 ofpbuf_size(&ctx->xout->odp_actions), NULL);
9583bc14 2317
ada3a58d 2318 pin = xmalloc(sizeof *pin);
1f317cb5 2319 pin->up.packet_len = ofpbuf_size(packet);
0fb7792a
BP
2320 pin->up.packet = ofpbuf_steal_data(packet);
2321 pin->up.reason = reason;
0fb7792a 2322 pin->up.table_id = ctx->table_id;
d4fa4e79
BP
2323 pin->up.cookie = (ctx->rule
2324 ? rule_dpif_get_flow_cookie(ctx->rule)
2325 : OVS_BE64_MAX);
0fb7792a 2326
0fb7792a 2327 flow_get_metadata(&ctx->xin->flow, &pin->up.fmd);
9583bc14 2328
f11c7538 2329 pin->controller_id = controller_id;
d38a3c7b 2330 pin->send_len = len;
32260212
SH
2331 /* If a rule is a table-miss rule then this is
2332 * a table-miss handled by a table-miss rule.
2333 *
2334 * Else, if rule is internal and has a controller action,
2335 * the later being implied by the rule being processed here,
2336 * then this is a table-miss handled without a table-miss rule.
2337 *
2338 * Otherwise this is not a table-miss. */
2339 pin->miss_type = OFPROTO_PACKET_IN_NO_MISS;
2340 if (ctx->rule) {
2341 if (rule_dpif_is_table_miss(ctx->rule)) {
2342 pin->miss_type = OFPROTO_PACKET_IN_MISS_FLOW;
2343 } else if (rule_dpif_is_internal(ctx->rule)) {
2344 pin->miss_type = OFPROTO_PACKET_IN_MISS_WITHOUT_FLOW;
2345 }
2346 }
ada3a58d 2347 ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, pin);
9583bc14
EJ
2348 ofpbuf_delete(packet);
2349}
2350
8bfd0fda
BP
2351static void
2352compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls)
9583bc14 2353{
33bf9176
BP
2354 struct flow_wildcards *wc = &ctx->xout->wc;
2355 struct flow *flow = &ctx->xin->flow;
8bfd0fda 2356 int n;
33bf9176 2357
8bfd0fda 2358 ovs_assert(eth_type_mpls(mpls->ethertype));
b0a17866 2359
8bfd0fda
BP
2360 n = flow_count_mpls_labels(flow, wc);
2361 if (!n) {
8bfd0fda
BP
2362 ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow,
2363 &ctx->xout->odp_actions,
2364 &ctx->xout->wc);
2365 } else if (n >= FLOW_MAX_MPLS_LABELS) {
2366 if (ctx->xin->packet != NULL) {
2367 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2368 VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an "
2369 "MPLS push action can't be performed as it would "
2370 "have more MPLS LSEs than the %d supported.",
2371 ctx->xbridge->name, FLOW_MAX_MPLS_LABELS);
9583bc14 2372 }
8bfd0fda
BP
2373 ctx->exit = true;
2374 return;
2375 } else if (n >= ctx->xbridge->max_mpls_depth) {
2376 COVERAGE_INC(xlate_actions_mpls_overflow);
2377 ctx->xout->slow |= SLOW_ACTION;
9583bc14 2378 }
b0a17866 2379
8bfd0fda 2380 flow_push_mpls(flow, n, mpls->ethertype, wc);
9583bc14
EJ
2381}
2382
8bfd0fda 2383static void
9cfef3d0 2384compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type)
9583bc14 2385{
33bf9176 2386 struct flow_wildcards *wc = &ctx->xout->wc;
8bfd0fda
BP
2387 struct flow *flow = &ctx->xin->flow;
2388 int n = flow_count_mpls_labels(flow, wc);
33bf9176 2389
8bfd0fda
BP
2390 if (!flow_pop_mpls(flow, n, eth_type, wc) && n >= FLOW_MAX_MPLS_LABELS) {
2391 if (ctx->xin->packet != NULL) {
2392 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
2393 VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an "
2394 "MPLS pop action can't be performed as it has "
2395 "more MPLS LSEs than the %d supported.",
2396 ctx->xbridge->name, FLOW_MAX_MPLS_LABELS);
2397 }
2398 ctx->exit = true;
2399 ofpbuf_clear(&ctx->xout->odp_actions);
9583bc14
EJ
2400 }
2401}
2402
2403static bool
2404compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids)
2405{
33bf9176
BP
2406 struct flow *flow = &ctx->xin->flow;
2407
2408 if (!is_ip_any(flow)) {
9583bc14
EJ
2409 return false;
2410 }
2411
1dd35f8a 2412 ctx->xout->wc.masks.nw_ttl = 0xff;
33bf9176
BP
2413 if (flow->nw_ttl > 1) {
2414 flow->nw_ttl--;
9583bc14
EJ
2415 return false;
2416 } else {
2417 size_t i;
2418
2419 for (i = 0; i < ids->n_controllers; i++) {
2420 execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL,
2421 ids->cnt_ids[i]);
2422 }
2423
2424 /* Stop processing for current table. */
2425 return true;
2426 }
2427}
2428
8bfd0fda 2429static void
097d4939
JR
2430compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label)
2431{
8bfd0fda
BP
2432 if (eth_type_mpls(ctx->xin->flow.dl_type)) {
2433 ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK);
2434 set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label);
097d4939 2435 }
097d4939
JR
2436}
2437
8bfd0fda 2438static void
097d4939
JR
2439compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc)
2440{
8bfd0fda
BP
2441 if (eth_type_mpls(ctx->xin->flow.dl_type)) {
2442 ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_TC_MASK);
2443 set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc);
097d4939 2444 }
097d4939
JR
2445}
2446
8bfd0fda 2447static void
9cfef3d0 2448compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl)
9583bc14 2449{
8bfd0fda
BP
2450 if (eth_type_mpls(ctx->xin->flow.dl_type)) {
2451 ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK);
2452 set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl);
b0a17866 2453 }
9583bc14
EJ
2454}
2455
2456static bool
9cfef3d0 2457compose_dec_mpls_ttl_action(struct xlate_ctx *ctx)
9583bc14 2458{
33bf9176 2459 struct flow *flow = &ctx->xin->flow;
8bfd0fda 2460 uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]);
1dd35f8a
JP
2461 struct flow_wildcards *wc = &ctx->xout->wc;
2462
1dd35f8a 2463 memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
8bfd0fda
BP
2464 if (eth_type_mpls(flow->dl_type)) {
2465 if (ttl > 1) {
2466 ttl--;
2467 set_mpls_lse_ttl(&flow->mpls_lse[0], ttl);
2468 return false;
2469 } else {
2470 execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0);
9583bc14 2471
8bfd0fda
BP
2472 /* Stop processing for current table. */
2473 return true;
2474 }
9583bc14 2475 } else {
9583bc14
EJ
2476 return true;
2477 }
2478}
2479
2480static void
2481xlate_output_action(struct xlate_ctx *ctx,
4e022ec0 2482 ofp_port_t port, uint16_t max_len, bool may_packet_in)
9583bc14 2483{
4e022ec0 2484 ofp_port_t prev_nf_output_iface = ctx->xout->nf_output_iface;
9583bc14
EJ
2485
2486 ctx->xout->nf_output_iface = NF_OUT_DROP;
2487
2488 switch (port) {
2489 case OFPP_IN_PORT:
4e022ec0 2490 compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port);
9583bc14
EJ
2491 break;
2492 case OFPP_TABLE:
4e022ec0 2493 xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
6d328fa2 2494 0, may_packet_in, true);
9583bc14
EJ
2495 break;
2496 case OFPP_NORMAL:
2497 xlate_normal(ctx);
2498 break;
2499 case OFPP_FLOOD:
2500 flood_packets(ctx, false);
2501 break;
2502 case OFPP_ALL:
2503 flood_packets(ctx, true);
2504 break;
2505 case OFPP_CONTROLLER:
2506 execute_controller_action(ctx, max_len, OFPR_ACTION, 0);
2507 break;
2508 case OFPP_NONE:
2509 break;
2510 case OFPP_LOCAL:
2511 default:
4e022ec0 2512 if (port != ctx->xin->flow.in_port.ofp_port) {
9583bc14
EJ
2513 compose_output_action(ctx, port);
2514 } else {
2515 xlate_report(ctx, "skipping output to input port");
2516 }
2517 break;
2518 }
2519
2520 if (prev_nf_output_iface == NF_OUT_FLOOD) {
2521 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
2522 } else if (ctx->xout->nf_output_iface == NF_OUT_DROP) {
2523 ctx->xout->nf_output_iface = prev_nf_output_iface;
2524 } else if (prev_nf_output_iface != NF_OUT_DROP &&
2525 ctx->xout->nf_output_iface != NF_OUT_FLOOD) {
2526 ctx->xout->nf_output_iface = NF_OUT_MULTI;
2527 }
2528}
2529
2530static void
2531xlate_output_reg_action(struct xlate_ctx *ctx,
2532 const struct ofpact_output_reg *or)
2533{
2534 uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow);
2535 if (port <= UINT16_MAX) {
2536 union mf_subvalue value;
2537
2538 memset(&value, 0xff, sizeof value);
2539 mf_write_subfield_flow(&or->src, &value, &ctx->xout->wc.masks);
4e022ec0
AW
2540 xlate_output_action(ctx, u16_to_ofp(port),
2541 or->max_len, false);
9583bc14
EJ
2542 }
2543}
2544
2545static void
2546xlate_enqueue_action(struct xlate_ctx *ctx,
2547 const struct ofpact_enqueue *enqueue)
2548{
4e022ec0 2549 ofp_port_t ofp_port = enqueue->port;
9583bc14
EJ
2550 uint32_t queue_id = enqueue->queue;
2551 uint32_t flow_priority, priority;
2552 int error;
2553
2554 /* Translate queue to priority. */
89a8a7f0 2555 error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority);
9583bc14
EJ
2556 if (error) {
2557 /* Fall back to ordinary output action. */
2558 xlate_output_action(ctx, enqueue->port, 0, false);
2559 return;
2560 }
2561
2562 /* Check output port. */
2563 if (ofp_port == OFPP_IN_PORT) {
4e022ec0
AW
2564 ofp_port = ctx->xin->flow.in_port.ofp_port;
2565 } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) {
9583bc14
EJ
2566 return;
2567 }
2568
2569 /* Add datapath actions. */
2570 flow_priority = ctx->xin->flow.skb_priority;
2571 ctx->xin->flow.skb_priority = priority;
2572 compose_output_action(ctx, ofp_port);
2573 ctx->xin->flow.skb_priority = flow_priority;
2574
2575 /* Update NetFlow output port. */
2576 if (ctx->xout->nf_output_iface == NF_OUT_DROP) {
2577 ctx->xout->nf_output_iface = ofp_port;
2578 } else if (ctx->xout->nf_output_iface != NF_OUT_FLOOD) {
2579 ctx->xout->nf_output_iface = NF_OUT_MULTI;
2580 }
2581}
2582
2583static void
2584xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id)
2585{
2586 uint32_t skb_priority;
2587
89a8a7f0 2588 if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) {
9583bc14
EJ
2589 ctx->xin->flow.skb_priority = skb_priority;
2590 } else {
2591 /* Couldn't translate queue to a priority. Nothing to do. A warning
2592 * has already been logged. */
2593 }
2594}
2595
2596static bool
46c88433 2597slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_)
9583bc14 2598{
46c88433
EJ
2599 const struct xbridge *xbridge = xbridge_;
2600 struct xport *port;
9583bc14
EJ
2601
2602 switch (ofp_port) {
2603 case OFPP_IN_PORT:
2604 case OFPP_TABLE:
2605 case OFPP_NORMAL:
2606 case OFPP_FLOOD:
2607 case OFPP_ALL:
2608 case OFPP_NONE:
2609 return true;
2610 case OFPP_CONTROLLER: /* Not supported by the bundle action. */
2611 return false;
2612 default:
46c88433 2613 port = get_ofp_port(xbridge, ofp_port);
9583bc14
EJ
2614 return port ? port->may_enable : false;
2615 }
2616}
2617
2618static void
2619xlate_bundle_action(struct xlate_ctx *ctx,
2620 const struct ofpact_bundle *bundle)
2621{
4e022ec0 2622 ofp_port_t port;
9583bc14
EJ
2623
2624 port = bundle_execute(bundle, &ctx->xin->flow, &ctx->xout->wc,
46c88433
EJ
2625 slave_enabled_cb,
2626 CONST_CAST(struct xbridge *, ctx->xbridge));
9583bc14 2627 if (bundle->dst.field) {
f74e7df7
JP
2628 nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow,
2629 &ctx->xout->wc);
9583bc14
EJ
2630 } else {
2631 xlate_output_action(ctx, port, 0, false);
2632 }
2633}
2634
2635static void
2636xlate_learn_action(struct xlate_ctx *ctx,
2637 const struct ofpact_learn *learn)
2638{
1b63b91e
BP
2639 uint64_t ofpacts_stub[1024 / 8];
2640 struct ofputil_flow_mod fm;
9583bc14 2641 struct ofpbuf ofpacts;
9583bc14
EJ
2642
2643 ctx->xout->has_learn = true;
2644
2645 learn_mask(learn, &ctx->xout->wc);
2646
2647 if (!ctx->xin->may_learn) {
2648 return;
2649 }
2650
1b63b91e
BP
2651 ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub);
2652 learn_execute(learn, &ctx->xin->flow, &fm, &ofpacts);
2653 ofproto_dpif_flow_mod(ctx->xbridge->ofproto, &fm);
2654 ofpbuf_uninit(&ofpacts);
b256dc52
JS
2655
2656 if (ctx->xin->xcache) {
2657 struct xc_entry *entry;
2658
2659 entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN);
2660 entry->u.learn.ofproto = ctx->xin->ofproto;
83709dfa
JR
2661 /* Lookup the learned rule, taking a reference on it. The reference
2662 * is released when this cache entry is deleted. */
b256dc52 2663 rule_dpif_lookup(ctx->xbridge->ofproto, &ctx->xin->flow, NULL,
1a7c0cd7 2664 &entry->u.learn.rule, true);
b256dc52
JS
2665 }
2666}
2667
2668static void
2669xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags,
2670 uint16_t idle_timeout, uint16_t hard_timeout)
2671{
2672 if (tcp_flags & (TCP_FIN | TCP_RST)) {
2673 rule_dpif_reduce_timeouts(rule, idle_timeout, hard_timeout);
2674 }
9583bc14
EJ
2675}
2676
9583bc14
EJ
2677static void
2678xlate_fin_timeout(struct xlate_ctx *ctx,
2679 const struct ofpact_fin_timeout *oft)
2680{
b256dc52
JS
2681 if (ctx->rule) {
2682 xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags,
2683 oft->fin_idle_timeout, oft->fin_hard_timeout);
2684 if (ctx->xin->xcache) {
2685 struct xc_entry *entry;
2686
2687 entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT);
83709dfa
JR
2688 /* XC_RULE already holds a reference on the rule, none is taken
2689 * here. */
b256dc52
JS
2690 entry->u.fin.rule = ctx->rule;
2691 entry->u.fin.idle = oft->fin_idle_timeout;
2692 entry->u.fin.hard = oft->fin_hard_timeout;
b256dc52 2693 }
9583bc14
EJ
2694 }
2695}
2696
2697static void
2698xlate_sample_action(struct xlate_ctx *ctx,
2699 const struct ofpact_sample *os)
2700{
2701 union user_action_cookie cookie;
2702 /* Scale the probability from 16-bit to 32-bit while representing
2703 * the same percentage. */
2704 uint32_t probability = (os->probability << 16) | os->probability;
2705
4b97b70d
BP
2706 if (!ctx->xbridge->variable_length_userdata) {
2707 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
2708
2709 VLOG_ERR_RL(&rl, "ignoring NXAST_SAMPLE action because datapath "
2710 "lacks support (needs Linux 3.10+ or kernel module from "
2711 "OVS 1.11+)");
2712 return;
2713 }
2714
7fd91025
BP
2715 ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
2716 &ctx->xout->odp_actions,
8bfd0fda 2717 &ctx->xout->wc);
9583bc14
EJ
2718
2719 compose_flow_sample_cookie(os->probability, os->collector_set_id,
2720 os->obs_domain_id, os->obs_point_id, &cookie);
46c88433 2721 compose_sample_action(ctx->xbridge, &ctx->xout->odp_actions, &ctx->xin->flow,
9583bc14
EJ
2722 probability, &cookie, sizeof cookie.flow_sample);
2723}
2724
2725static bool
46c88433 2726may_receive(const struct xport *xport, struct xlate_ctx *ctx)
9583bc14 2727{
bbbca389 2728 if (xport->config & (is_stp(&ctx->xin->flow)
46c88433
EJ
2729 ? OFPUTIL_PC_NO_RECV_STP
2730 : OFPUTIL_PC_NO_RECV)) {
9583bc14
EJ
2731 return false;
2732 }
2733
2734 /* Only drop packets here if both forwarding and learning are
2735 * disabled. If just learning is enabled, we need to have
2736 * OFPP_NORMAL and the learning action have a look at the packet
2737 * before we can drop it. */
9d189a50 2738 if (!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) {
9583bc14
EJ
2739 return false;
2740 }
2741
2742 return true;
2743}
2744
7fdb60a7
SH
2745static void
2746xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact *a)
2747{
2748 struct ofpact_nest *on = ofpact_get_WRITE_ACTIONS(a);
2749 ofpbuf_put(&ctx->action_set, on->actions, ofpact_nest_get_action_len(on));
2750 ofpact_pad(&ctx->action_set);
2751}
2752
2753static void
2754xlate_action_set(struct xlate_ctx *ctx)
2755{
2756 uint64_t action_list_stub[1024 / 64];
2757 struct ofpbuf action_list;
2758
2759 ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub);
2760 ofpacts_execute_action_set(&action_list, &ctx->action_set);
1f317cb5 2761 do_xlate_actions(ofpbuf_data(&action_list), ofpbuf_size(&action_list), ctx);
7fdb60a7
SH
2762 ofpbuf_uninit(&action_list);
2763}
2764
9583bc14
EJ
2765static void
2766do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len,
2767 struct xlate_ctx *ctx)
2768{
33bf9176
BP
2769 struct flow_wildcards *wc = &ctx->xout->wc;
2770 struct flow *flow = &ctx->xin->flow;
9583bc14
EJ
2771 const struct ofpact *a;
2772
f47ea021
JR
2773 /* dl_type already in the mask, not set below. */
2774
9583bc14
EJ
2775 OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) {
2776 struct ofpact_controller *controller;
2777 const struct ofpact_metadata *metadata;
b2dd70be
JR
2778 const struct ofpact_set_field *set_field;
2779 const struct mf_field *mf;
9583bc14
EJ
2780
2781 if (ctx->exit) {
2782 break;
2783 }
2784
2785 switch (a->type) {
2786 case OFPACT_OUTPUT:
2787 xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port,
2788 ofpact_get_OUTPUT(a)->max_len, true);
2789 break;
2790
7395c052 2791 case OFPACT_GROUP:
f4fb341b
SH
2792 if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id)) {
2793 return;
2794 }
7395c052
NZ
2795 break;
2796
9583bc14
EJ
2797 case OFPACT_CONTROLLER:
2798 controller = ofpact_get_CONTROLLER(a);
2799 execute_controller_action(ctx, controller->max_len,
2800 controller->reason,
2801 controller->controller_id);
2802 break;
2803
2804 case OFPACT_ENQUEUE:
2805 xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a));
2806 break;
2807
2808 case OFPACT_SET_VLAN_VID:
f74e7df7 2809 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
ca287d20
JR
2810 if (flow->vlan_tci & htons(VLAN_CFI) ||
2811 ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) {
2812 flow->vlan_tci &= ~htons(VLAN_VID_MASK);
2813 flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid)
2814 | htons(VLAN_CFI));
2815 }
9583bc14
EJ
2816 break;
2817
2818 case OFPACT_SET_VLAN_PCP:
f74e7df7 2819 wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI);
ca287d20
JR
2820 if (flow->vlan_tci & htons(VLAN_CFI) ||
2821 ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) {
2822 flow->vlan_tci &= ~htons(VLAN_PCP_MASK);
2823 flow->vlan_tci |= htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp
2824 << VLAN_PCP_SHIFT) | VLAN_CFI);
2825 }
9583bc14
EJ
2826 break;
2827
2828 case OFPACT_STRIP_VLAN:
f74e7df7 2829 memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
33bf9176 2830 flow->vlan_tci = htons(0);
9583bc14
EJ
2831 break;
2832
2833 case OFPACT_PUSH_VLAN:
2834 /* XXX 802.1AD(QinQ) */
f74e7df7 2835 memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
33bf9176 2836 flow->vlan_tci = htons(VLAN_CFI);
9583bc14
EJ
2837 break;
2838
2839 case OFPACT_SET_ETH_SRC:
f74e7df7 2840 memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src);
33bf9176 2841 memcpy(flow->dl_src, ofpact_get_SET_ETH_SRC(a)->mac, ETH_ADDR_LEN);
9583bc14
EJ
2842 break;
2843
2844 case OFPACT_SET_ETH_DST:
f74e7df7 2845 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
33bf9176 2846 memcpy(flow->dl_dst, ofpact_get_SET_ETH_DST(a)->mac, ETH_ADDR_LEN);
9583bc14
EJ
2847 break;
2848
2849 case OFPACT_SET_IPV4_SRC:
33bf9176 2850 if (flow->dl_type == htons(ETH_TYPE_IP)) {
f47ea021 2851 memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src);
33bf9176 2852 flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4;
9583bc14
EJ
2853 }
2854 break;
2855
2856 case OFPACT_SET_IPV4_DST:
33bf9176 2857 if (flow->dl_type == htons(ETH_TYPE_IP)) {
f47ea021 2858 memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst);
33bf9176 2859 flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4;
9583bc14
EJ
2860 }
2861 break;
2862
04f01c24
BP
2863 case OFPACT_SET_IP_DSCP:
2864 if (is_ip_any(flow)) {
f47ea021 2865 wc->masks.nw_tos |= IP_DSCP_MASK;
33bf9176 2866 flow->nw_tos &= ~IP_DSCP_MASK;
04f01c24 2867 flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp;
9583bc14
EJ
2868 }
2869 break;
2870
ff14eb7a
JR
2871 case OFPACT_SET_IP_ECN:
2872 if (is_ip_any(flow)) {
2873 wc->masks.nw_tos |= IP_ECN_MASK;
2874 flow->nw_tos &= ~IP_ECN_MASK;
2875 flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn;
2876 }
2877 break;
2878
0c20dbe4
JR
2879 case OFPACT_SET_IP_TTL:
2880 if (is_ip_any(flow)) {
2881 wc->masks.nw_ttl = 0xff;
2882 flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl;
2883 }
2884 break;
2885
9583bc14 2886 case OFPACT_SET_L4_SRC_PORT:
33bf9176 2887 if (is_ip_any(flow)) {
f47ea021
JR
2888 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
2889 memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src);
33bf9176 2890 flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port);
9583bc14
EJ
2891 }
2892 break;
2893
2894 case OFPACT_SET_L4_DST_PORT:
33bf9176 2895 if (is_ip_any(flow)) {
f47ea021
JR
2896 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
2897 memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
33bf9176 2898 flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port);
9583bc14
EJ
2899 }
2900 break;
2901
2902 case OFPACT_RESUBMIT:
2903 xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a));
2904 break;
2905
2906 case OFPACT_SET_TUNNEL:
33bf9176 2907 flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id);
9583bc14
EJ
2908 break;
2909
2910 case OFPACT_SET_QUEUE:
2911 xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id);
2912 break;
2913
2914 case OFPACT_POP_QUEUE:
33bf9176 2915 flow->skb_priority = ctx->orig_skb_priority;
9583bc14
EJ
2916 break;
2917
2918 case OFPACT_REG_MOVE:
33bf9176 2919 nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc);
9583bc14
EJ
2920 break;
2921
2922 case OFPACT_REG_LOAD:
f47ea021 2923 nxm_execute_reg_load(ofpact_get_REG_LOAD(a), flow, wc);
9583bc14
EJ
2924 break;
2925
b2dd70be
JR
2926 case OFPACT_SET_FIELD:
2927 set_field = ofpact_get_SET_FIELD(a);
2928 mf = set_field->field;
b2dd70be
JR
2929
2930 /* Set field action only ever overwrites packet's outermost
2931 * applicable header fields. Do nothing if no header exists. */
4479846e
JR
2932 if (mf->id == MFF_VLAN_VID) {
2933 wc->masks.vlan_tci |= htons(VLAN_CFI);
2934 if (!(flow->vlan_tci & htons(VLAN_CFI))) {
2935 break;
2936 }
2937 } else if ((mf->id == MFF_MPLS_LABEL || mf->id == MFF_MPLS_TC)
2938 /* 'dl_type' is already unwildcarded. */
2939 && !eth_type_mpls(flow->dl_type)) {
2940 break;
b2dd70be 2941 }
4479846e
JR
2942
2943 mf_mask_field_and_prereqs(mf, &wc->masks);
2944 mf_set_flow_value(mf, &set_field->value, flow);
b2dd70be
JR
2945 break;
2946
9583bc14 2947 case OFPACT_STACK_PUSH:
33bf9176
BP
2948 nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc,
2949 &ctx->stack);
9583bc14
EJ
2950 break;
2951
2952 case OFPACT_STACK_POP:
f74e7df7
JP
2953 nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc,
2954 &ctx->stack);
9583bc14
EJ
2955 break;
2956
2957 case OFPACT_PUSH_MPLS:
8bfd0fda 2958 compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a));
9583bc14
EJ
2959 break;
2960
2961 case OFPACT_POP_MPLS:
8bfd0fda 2962 compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype);
9583bc14
EJ
2963 break;
2964
097d4939 2965 case OFPACT_SET_MPLS_LABEL:
8bfd0fda
BP
2966 compose_set_mpls_label_action(
2967 ctx, ofpact_get_SET_MPLS_LABEL(a)->label);
2968 break;
097d4939
JR
2969
2970 case OFPACT_SET_MPLS_TC:
8bfd0fda 2971 compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc);
097d4939
JR
2972 break;
2973
9583bc14 2974 case OFPACT_SET_MPLS_TTL:
8bfd0fda 2975 compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl);
9583bc14
EJ
2976 break;
2977
2978 case OFPACT_DEC_MPLS_TTL:
9cfef3d0 2979 if (compose_dec_mpls_ttl_action(ctx)) {
ad3efdcb 2980 return;
9583bc14
EJ
2981 }
2982 break;
2983
2984 case OFPACT_DEC_TTL:
f74e7df7 2985 wc->masks.nw_ttl = 0xff;
9583bc14 2986 if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) {
ad3efdcb 2987 return;
9583bc14
EJ
2988 }
2989 break;
2990
2991 case OFPACT_NOTE:
2992 /* Nothing to do. */
2993 break;
2994
2995 case OFPACT_MULTIPATH:
33bf9176 2996 multipath_execute(ofpact_get_MULTIPATH(a), flow, wc);
9583bc14
EJ
2997 break;
2998
2999 case OFPACT_BUNDLE:
9583bc14
EJ
3000 xlate_bundle_action(ctx, ofpact_get_BUNDLE(a));
3001 break;
3002
3003 case OFPACT_OUTPUT_REG:
3004 xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a));
3005 break;
3006
3007 case OFPACT_LEARN:
3008 xlate_learn_action(ctx, ofpact_get_LEARN(a));
3009 break;
3010
3011 case OFPACT_EXIT:
3012 ctx->exit = true;
3013 break;
3014
3015 case OFPACT_FIN_TIMEOUT:
33bf9176 3016 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
9583bc14
EJ
3017 ctx->xout->has_fin_timeout = true;
3018 xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a));
3019 break;
3020
3021 case OFPACT_CLEAR_ACTIONS:
7fdb60a7
SH
3022 ofpbuf_clear(&ctx->action_set);
3023 break;
3024
3025 case OFPACT_WRITE_ACTIONS:
3026 xlate_write_actions(ctx, a);
9583bc14
EJ
3027 break;
3028
3029 case OFPACT_WRITE_METADATA:
3030 metadata = ofpact_get_WRITE_METADATA(a);
33bf9176
BP
3031 flow->metadata &= ~metadata->mask;
3032 flow->metadata |= metadata->metadata & metadata->mask;
9583bc14
EJ
3033 break;
3034
638a19b0
JR
3035 case OFPACT_METER:
3036 /* Not implemented yet. */
3037 break;
3038
9583bc14 3039 case OFPACT_GOTO_TABLE: {
9583bc14 3040 struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a);
9583bc14
EJ
3041
3042 ovs_assert(ctx->table_id < ogt->table_id);
4468099e 3043 xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
6d328fa2 3044 ogt->table_id, true, true);
9583bc14
EJ
3045 break;
3046 }
3047
3048 case OFPACT_SAMPLE:
3049 xlate_sample_action(ctx, ofpact_get_SAMPLE(a));
3050 break;
3051 }
3052 }
9583bc14
EJ
3053}
3054
3055void
3056xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto,
3057 const struct flow *flow, struct rule_dpif *rule,
a66733a8 3058 uint16_t tcp_flags, const struct ofpbuf *packet)
9583bc14
EJ
3059{
3060 xin->ofproto = ofproto;
3061 xin->flow = *flow;
3062 xin->packet = packet;
3063 xin->may_learn = packet != NULL;
3064 xin->rule = rule;
b256dc52 3065 xin->xcache = NULL;
9583bc14
EJ
3066 xin->ofpacts = NULL;
3067 xin->ofpacts_len = 0;
3068 xin->tcp_flags = tcp_flags;
3069 xin->resubmit_hook = NULL;
3070 xin->report_hook = NULL;
3071 xin->resubmit_stats = NULL;
3f207910 3072 xin->skip_wildcards = false;
9583bc14
EJ
3073}
3074
3075void
3076xlate_out_uninit(struct xlate_out *xout)
3077{
3078 if (xout) {
3079 ofpbuf_uninit(&xout->odp_actions);
3080 }
3081}
3082
3083/* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts'
3084 * into datapath actions, using 'ctx', and discards the datapath actions. */
3085void
3086xlate_actions_for_side_effects(struct xlate_in *xin)
3087{
3088 struct xlate_out xout;
3089
3090 xlate_actions(xin, &xout);
3091 xlate_out_uninit(&xout);
3092}
3093
3094static void
3095xlate_report(struct xlate_ctx *ctx, const char *s)
3096{
3097 if (ctx->xin->report_hook) {
4d0acc70 3098 ctx->xin->report_hook(ctx->xin, s, ctx->recurse);
9583bc14
EJ
3099 }
3100}
3101
3102void
3103xlate_out_copy(struct xlate_out *dst, const struct xlate_out *src)
3104{
3105 dst->wc = src->wc;
9583bc14
EJ
3106 dst->slow = src->slow;
3107 dst->has_learn = src->has_learn;
3108 dst->has_normal = src->has_normal;
3109 dst->has_fin_timeout = src->has_fin_timeout;
3110 dst->nf_output_iface = src->nf_output_iface;
3111 dst->mirrors = src->mirrors;
3112
3113 ofpbuf_use_stub(&dst->odp_actions, dst->odp_actions_stub,
3114 sizeof dst->odp_actions_stub);
1f317cb5
PS
3115 ofpbuf_put(&dst->odp_actions, ofpbuf_data(&src->odp_actions),
3116 ofpbuf_size(&src->odp_actions));
9583bc14
EJ
3117}
3118\f
55954f6e
EJ
3119static struct skb_priority_to_dscp *
3120get_skb_priority(const struct xport *xport, uint32_t skb_priority)
3121{
3122 struct skb_priority_to_dscp *pdscp;
3123 uint32_t hash;
3124
3125 hash = hash_int(skb_priority, 0);
3126 HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) {
3127 if (pdscp->skb_priority == skb_priority) {
3128 return pdscp;
3129 }
3130 }
3131 return NULL;
3132}
3133
3134static bool
3135dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority,
3136 uint8_t *dscp)
3137{
3138 struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority);
3139 *dscp = pdscp ? pdscp->dscp : 0;
3140 return pdscp != NULL;
3141}
3142
3143static void
3144clear_skb_priorities(struct xport *xport)
3145{
3146 struct skb_priority_to_dscp *pdscp, *next;
3147
3148 HMAP_FOR_EACH_SAFE (pdscp, next, hmap_node, &xport->skb_priorities) {
3149 hmap_remove(&xport->skb_priorities, &pdscp->hmap_node);
3150 free(pdscp);
3151 }
3152}
3153
ce4a6b76
BP
3154static bool
3155actions_output_to_local_port(const struct xlate_ctx *ctx)
3156{
46c88433 3157 odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL);
ce4a6b76
BP
3158 const struct nlattr *a;
3159 unsigned int left;
3160
1f317cb5
PS
3161 NL_ATTR_FOR_EACH_UNSAFE (a, left, ofpbuf_data(&ctx->xout->odp_actions),
3162 ofpbuf_size(&ctx->xout->odp_actions)) {
ce4a6b76
BP
3163 if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT
3164 && nl_attr_get_odp_port(a) == local_odp_port) {
3165 return true;
3166 }
3167 }
3168 return false;
3169}
9583bc14 3170
91d6cd12
AW
3171/* Thread safe call to xlate_actions__(). */
3172void
3173xlate_actions(struct xlate_in *xin, struct xlate_out *xout)
6a4d3ab5 3174 OVS_EXCLUDED(xlate_rwlock)
91d6cd12
AW
3175{
3176 ovs_rwlock_rdlock(&xlate_rwlock);
3177 xlate_actions__(xin, xout);
3178 ovs_rwlock_unlock(&xlate_rwlock);
3179}
3180
9583bc14 3181/* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts'
56450a41
BP
3182 * into datapath actions in 'odp_actions', using 'ctx'.
3183 *
3184 * The caller must take responsibility for eventually freeing 'xout', with
3185 * xlate_out_uninit(). */
91d6cd12
AW
3186static void
3187xlate_actions__(struct xlate_in *xin, struct xlate_out *xout)
3188 OVS_REQ_RDLOCK(xlate_rwlock)
9583bc14 3189{
33bf9176
BP
3190 struct flow_wildcards *wc = &xout->wc;
3191 struct flow *flow = &xin->flow;
10c44245 3192 struct rule_dpif *rule = NULL;
33bf9176 3193
dc723c44 3194 const struct rule_actions *actions = NULL;
9583bc14
EJ
3195 enum slow_path_reason special;
3196 const struct ofpact *ofpacts;
46c88433 3197 struct xport *in_port;
9583bc14
EJ
3198 struct flow orig_flow;
3199 struct xlate_ctx ctx;
3200 size_t ofpacts_len;
62a7cc71 3201 bool tnl_may_send;
c56fac1b 3202 bool is_icmp;
9583bc14 3203
46c88433 3204 COVERAGE_INC(xlate_actions);
9583bc14
EJ
3205
3206 /* Flow initialization rules:
3207 * - 'base_flow' must match the kernel's view of the packet at the
3208 * time that action processing starts. 'flow' represents any
3209 * transformations we wish to make through actions.
3210 * - By default 'base_flow' and 'flow' are the same since the input
3211 * packet matches the output before any actions are applied.
3212 * - When using VLAN splinters, 'base_flow''s VLAN is set to the value
3213 * of the received packet as seen by the kernel. If we later output
3214 * to another device without any modifications this will cause us to
3215 * insert a new tag since the original one was stripped off by the
3216 * VLAN device.
3217 * - Tunnel metadata as received is retained in 'flow'. This allows
3218 * tunnel metadata matching also in later tables.
3219 * Since a kernel action for setting the tunnel metadata will only be
3220 * generated with actual tunnel output, changing the tunnel metadata
3221 * values in 'flow' (such as tun_id) will only have effect with a later
3222 * tunnel output action.
3223 * - Tunnel 'base_flow' is completely cleared since that is what the
3224 * kernel does. If we wish to maintain the original values an action
3225 * needs to be generated. */
3226
3227 ctx.xin = xin;
3228 ctx.xout = xout;
46c88433
EJ
3229 ctx.xout->slow = 0;
3230 ctx.xout->has_learn = false;
3231 ctx.xout->has_normal = false;
3232 ctx.xout->has_fin_timeout = false;
3233 ctx.xout->nf_output_iface = NF_OUT_DROP;
3234 ctx.xout->mirrors = 0;
3235 ofpbuf_use_stub(&ctx.xout->odp_actions, ctx.xout->odp_actions_stub,
3236 sizeof ctx.xout->odp_actions_stub);
3237 ofpbuf_reserve(&ctx.xout->odp_actions, NL_A_U32_SIZE);
3238
3239 ctx.xbridge = xbridge_lookup(xin->ofproto);
3240 if (!ctx.xbridge) {
83709dfa 3241 return;
46c88433 3242 }
9583bc14 3243
9583bc14
EJ
3244 ctx.rule = xin->rule;
3245
33bf9176 3246 ctx.base_flow = *flow;
9583bc14 3247 memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel);
33bf9176 3248 ctx.orig_tunnel_ip_dst = flow->tunnel.ip_dst;
9583bc14 3249
33bf9176
BP
3250 flow_wildcards_init_catchall(wc);
3251 memset(&wc->masks.in_port, 0xff, sizeof wc->masks.in_port);
1dd35f8a 3252 memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority);
7431e171 3253 memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type);
0934ebba
BP
3254 if (is_ip_any(flow)) {
3255 wc->masks.nw_frag |= FLOW_NW_FRAG_MASK;
3256 }
c56fac1b 3257 is_icmp = is_icmpv4(flow) || is_icmpv6(flow);
9583bc14 3258
62a7cc71 3259 tnl_may_send = tnl_xlate_init(&ctx.base_flow, flow, wc);
ce3955be 3260 if (ctx.xbridge->netflow) {
9b658910 3261 netflow_mask_wc(flow, wc);
9583bc14
EJ
3262 }
3263
9583bc14 3264 ctx.recurse = 0;
98b07853 3265 ctx.resubmits = 0;
5a070238 3266 ctx.in_group = false;
33bf9176 3267 ctx.orig_skb_priority = flow->skb_priority;
9583bc14
EJ
3268 ctx.table_id = 0;
3269 ctx.exit = false;
3270
10c44245 3271 if (!xin->ofpacts && !ctx.rule) {
6d328fa2
SH
3272 ctx.table_id = rule_dpif_lookup(ctx.xbridge->ofproto, flow,
3273 !xin->skip_wildcards ? wc : NULL,
83709dfa 3274 &rule, ctx.xin->xcache != NULL);
10c44245
EJ
3275 if (ctx.xin->resubmit_stats) {
3276 rule_dpif_credit_stats(rule, ctx.xin->resubmit_stats);
3277 }
b256dc52
JS
3278 if (ctx.xin->xcache) {
3279 struct xc_entry *entry;
3280
3281 entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE);
b256dc52
JS
3282 entry->u.rule = rule;
3283 }
10c44245
EJ
3284 ctx.rule = rule;
3285 }
ee1afdd5 3286 xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule);
adcf00ba 3287 xout->use_recirc = false;
10c44245 3288
9583bc14
EJ
3289 if (xin->ofpacts) {
3290 ofpacts = xin->ofpacts;
3291 ofpacts_len = xin->ofpacts_len;
10c44245
EJ
3292 } else if (ctx.rule) {
3293 actions = rule_dpif_get_actions(ctx.rule);
6f00e29b
BP
3294 ofpacts = actions->ofpacts;
3295 ofpacts_len = actions->ofpacts_len;
9583bc14 3296 } else {
428b2edd 3297 OVS_NOT_REACHED();
9583bc14
EJ
3298 }
3299
3300 ofpbuf_use_stub(&ctx.stack, ctx.init_stack, sizeof ctx.init_stack);
7fdb60a7
SH
3301 ofpbuf_use_stub(&ctx.action_set,
3302 ctx.action_set_stub, sizeof ctx.action_set_stub);
9583bc14 3303
ade6ad9c 3304 if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
9583bc14
EJ
3305 /* Do this conditionally because the copy is expensive enough that it
3306 * shows up in profiles. */
33bf9176 3307 orig_flow = *flow;
9583bc14
EJ
3308 }
3309
33bf9176 3310 if (flow->nw_frag & FLOW_NW_FRAG_ANY) {
46c88433 3311 switch (ctx.xbridge->frag) {
9583bc14
EJ
3312 case OFPC_FRAG_NORMAL:
3313 /* We must pretend that transport ports are unavailable. */
33bf9176
BP
3314 flow->tp_src = ctx.base_flow.tp_src = htons(0);
3315 flow->tp_dst = ctx.base_flow.tp_dst = htons(0);
9583bc14
EJ
3316 break;
3317
3318 case OFPC_FRAG_DROP:
83709dfa 3319 return;
9583bc14
EJ
3320
3321 case OFPC_FRAG_REASM:
428b2edd 3322 OVS_NOT_REACHED();
9583bc14
EJ
3323
3324 case OFPC_FRAG_NX_MATCH:
3325 /* Nothing to do. */
3326 break;
3327
3328 case OFPC_INVALID_TTL_TO_CONTROLLER:
428b2edd 3329 OVS_NOT_REACHED();
9583bc14
EJ
3330 }
3331 }
3332
46c88433 3333 in_port = get_ofp_port(ctx.xbridge, flow->in_port.ofp_port);
b256dc52
JS
3334 if (in_port && in_port->is_tunnel) {
3335 if (ctx.xin->resubmit_stats) {
3336 netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats);
3337 if (in_port->bfd) {
3338 bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats);
3339 }
3340 }
3341 if (ctx.xin->xcache) {
3342 struct xc_entry *entry;
3343
3344 entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV);
3345 entry->u.dev.rx = netdev_ref(in_port->netdev);
3346 entry->u.dev.bfd = bfd_ref(in_port->bfd);
d6fc5f57
EJ
3347 }
3348 }
3349
642dc74d 3350 special = process_special(&ctx, flow, in_port, ctx.xin->packet);
9583bc14 3351 if (special) {
04594cd5 3352 ctx.xout->slow |= special;
9583bc14 3353 } else {
9583bc14 3354 size_t sample_actions_len;
9583bc14 3355
4e022ec0 3356 if (flow->in_port.ofp_port
46c88433
EJ
3357 != vsp_realdev_to_vlandev(ctx.xbridge->ofproto,
3358 flow->in_port.ofp_port,
33bf9176 3359 flow->vlan_tci)) {
9583bc14
EJ
3360 ctx.base_flow.vlan_tci = 0;
3361 }
3362
3363 add_sflow_action(&ctx);
3364 add_ipfix_action(&ctx);
1f317cb5 3365 sample_actions_len = ofpbuf_size(&ctx.xout->odp_actions);
9583bc14 3366
62a7cc71 3367 if (tnl_may_send && (!in_port || may_receive(in_port, &ctx))) {
9583bc14
EJ
3368 do_xlate_actions(ofpacts, ofpacts_len, &ctx);
3369
3370 /* We've let OFPP_NORMAL and the learning action look at the
3371 * packet, so drop it now if forwarding is disabled. */
9d189a50 3372 if (in_port && !xport_stp_forward_state(in_port)) {
1f317cb5 3373 ofpbuf_set_size(&ctx.xout->odp_actions, sample_actions_len);
9583bc14
EJ
3374 }
3375 }
3376
1f317cb5 3377 if (ofpbuf_size(&ctx.action_set)) {
7fdb60a7
SH
3378 xlate_action_set(&ctx);
3379 }
3380
46c88433 3381 if (ctx.xbridge->has_in_band
ce4a6b76
BP
3382 && in_band_must_output_to_local_port(flow)
3383 && !actions_output_to_local_port(&ctx)) {
9583bc14
EJ
3384 compose_output_action(&ctx, OFPP_LOCAL);
3385 }
aaa0fbae
EJ
3386
3387 fix_sflow_action(&ctx);
3388
46c88433 3389 if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
9583bc14
EJ
3390 add_mirror_actions(&ctx, &orig_flow);
3391 }
9583bc14
EJ
3392 }
3393
1f317cb5 3394 if (nl_attr_oversized(ofpbuf_size(&ctx.xout->odp_actions))) {
542024c4 3395 /* These datapath actions are too big for a Netlink attribute, so we
0f032e95
BP
3396 * can't hand them to the kernel directly. dpif_execute() can execute
3397 * them one by one with help, so just mark the result as SLOW_ACTION to
3398 * prevent the flow from being installed. */
3399 COVERAGE_INC(xlate_actions_oversize);
3400 ctx.xout->slow |= SLOW_ACTION;
542024c4
BP
3401 }
3402
b256dc52
JS
3403 if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
3404 if (ctx.xin->resubmit_stats) {
3405 mirror_update_stats(ctx.xbridge->mbridge, xout->mirrors,
3406 ctx.xin->resubmit_stats->n_packets,
3407 ctx.xin->resubmit_stats->n_bytes);
3408 }
3409 if (ctx.xin->xcache) {
3410 struct xc_entry *entry;
3411
3412 entry = xlate_cache_add_entry(ctx.xin->xcache, XC_MIRROR);
3413 entry->u.mirror.mbridge = mbridge_ref(ctx.xbridge->mbridge);
3414 entry->u.mirror.mirrors = xout->mirrors;
3415 }
3416 }
3417
3418 if (ctx.xbridge->netflow) {
3419 const struct ofpact *ofpacts = actions->ofpacts;
3420 size_t ofpacts_len = actions->ofpacts_len;
3421
3422 /* Only update netflow if we don't have controller flow. We don't
3423 * report NetFlow expiration messages for such facets because they
3424 * are just part of the control logic for the network, not real
3425 * traffic. */
3426 if (ofpacts_len == 0
3427 || ofpacts->type != OFPACT_CONTROLLER
3428 || ofpact_next(ofpacts) < ofpact_end(ofpacts, ofpacts_len)) {
3429 if (ctx.xin->resubmit_stats) {
d6fc5f57
EJ
3430 netflow_flow_update(ctx.xbridge->netflow, flow,
3431 xout->nf_output_iface,
3432 ctx.xin->resubmit_stats);
3433 }
b256dc52
JS
3434 if (ctx.xin->xcache) {
3435 struct xc_entry *entry;
3436
3437 entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW);
3438 entry->u.nf.netflow = netflow_ref(ctx.xbridge->netflow);
3439 entry->u.nf.flow = xmemdup(flow, sizeof *flow);
3440 entry->u.nf.iface = xout->nf_output_iface;
3441 }
d6fc5f57
EJ
3442 }
3443 }
3444
9583bc14 3445 ofpbuf_uninit(&ctx.stack);
7fdb60a7 3446 ofpbuf_uninit(&ctx.action_set);
9583bc14
EJ
3447
3448 /* Clear the metadata and register wildcard masks, because we won't
3449 * use non-header fields as part of the cache. */
c11c6faa 3450 flow_wildcards_clear_non_packet_fields(wc);
dc24a00f 3451
c56fac1b
BP
3452 /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow uses
3453 * the low 8 bits of the 16-bit tp_src and tp_dst members to represent
3454 * these fields. The datapath interface, on the other hand, represents
3455 * them with just 8 bits each. This means that if the high 8 bits of the
3456 * masks for these fields somehow become set, then they will get chopped
3457 * off by a round trip through the datapath, and revalidation will spot
3458 * that as an inconsistency and delete the flow. Avoid the problem here by
3459 * making sure that only the low 8 bits of either field can be unwildcarded
3460 * for ICMP.
3461 */
3462 if (is_icmp) {
3463 wc->masks.tp_src &= htons(UINT8_MAX);
3464 wc->masks.tp_dst &= htons(UINT8_MAX);
3465 }
91d6cd12
AW
3466}
3467
3468/* Sends 'packet' out 'ofport'.
3469 * May modify 'packet'.
3470 * Returns 0 if successful, otherwise a positive errno value. */
3471int
3472xlate_send_packet(const struct ofport_dpif *ofport, struct ofpbuf *packet)
3473{
91d6cd12 3474 struct xport *xport;
91d6cd12 3475 struct ofpact_output output;
91d6cd12 3476 struct flow flow;
91d6cd12 3477
91d6cd12
AW
3478 ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output);
3479 /* Use OFPP_NONE as the in_port to avoid special packet processing. */
b5e7e61a
AZ
3480 flow_extract(packet, NULL, &flow);
3481 flow.in_port.ofp_port = OFPP_NONE;
91d6cd12
AW
3482
3483 ovs_rwlock_rdlock(&xlate_rwlock);
3484 xport = xport_lookup(ofport);
3485 if (!xport) {
91d6cd12 3486 ovs_rwlock_unlock(&xlate_rwlock);
02ea2703 3487 return EINVAL;
91d6cd12 3488 }
91d6cd12
AW
3489 output.port = xport->ofp_port;
3490 output.max_len = 0;
dc24a00f 3491 ovs_rwlock_unlock(&xlate_rwlock);
e491a67a
YT
3492
3493 return ofproto_dpif_execute_actions(xport->xbridge->ofproto, &flow, NULL,
3494 &output.ofpact, sizeof output,
3495 packet);
9583bc14 3496}
b256dc52
JS
3497
3498struct xlate_cache *
3499xlate_cache_new(void)
3500{
3501 struct xlate_cache *xcache = xmalloc(sizeof *xcache);
3502
3503 ofpbuf_init(&xcache->entries, 512);
3504 return xcache;
3505}
3506
3507static struct xc_entry *
3508xlate_cache_add_entry(struct xlate_cache *xcache, enum xc_type type)
3509{
3510 struct xc_entry *entry;
3511
3512 entry = ofpbuf_put_zeros(&xcache->entries, sizeof *entry);
3513 entry->type = type;
3514
3515 return entry;
3516}
3517
3518static void
3519xlate_cache_netdev(struct xc_entry *entry, const struct dpif_flow_stats *stats)
3520{
3521 if (entry->u.dev.tx) {
3522 netdev_vport_inc_tx(entry->u.dev.tx, stats);
3523 }
3524 if (entry->u.dev.rx) {
3525 netdev_vport_inc_rx(entry->u.dev.rx, stats);
3526 }
3527 if (entry->u.dev.bfd) {
3528 bfd_account_rx(entry->u.dev.bfd, stats);
3529 }
3530}
3531
3532static void
3533xlate_cache_normal(struct ofproto_dpif *ofproto, struct flow *flow, int vlan)
3534{
3535 struct xbridge *xbridge;
3536 struct xbundle *xbundle;
3537 struct flow_wildcards wc;
3538
3539 xbridge = xbridge_lookup(ofproto);
3540 if (!xbridge) {
3541 return;
3542 }
3543
3544 xbundle = lookup_input_bundle(xbridge, flow->in_port.ofp_port, false,
3545 NULL);
3546 if (!xbundle) {
3547 return;
3548 }
3549
3550 update_learning_table(xbridge, flow, &wc, vlan, xbundle);
3551}
3552
3553/* Push stats and perform side effects of flow translation. */
3554void
3555xlate_push_stats(struct xlate_cache *xcache, bool may_learn,
3556 const struct dpif_flow_stats *stats)
3557{
3558 struct xc_entry *entry;
3559 struct ofpbuf entries = xcache->entries;
3560
3561 XC_ENTRY_FOR_EACH (entry, entries, xcache) {
3562 switch (entry->type) {
3563 case XC_RULE:
3564 rule_dpif_credit_stats(entry->u.rule, stats);
3565 break;
3566 case XC_BOND:
3567 bond_account(entry->u.bond.bond, entry->u.bond.flow,
3568 entry->u.bond.vid, stats->n_bytes);
3569 break;
3570 case XC_NETDEV:
3571 xlate_cache_netdev(entry, stats);
3572 break;
3573 case XC_NETFLOW:
3574 netflow_flow_update(entry->u.nf.netflow, entry->u.nf.flow,
3575 entry->u.nf.iface, stats);
3576 break;
3577 case XC_MIRROR:
3578 mirror_update_stats(entry->u.mirror.mbridge,
3579 entry->u.mirror.mirrors,
3580 stats->n_packets, stats->n_bytes);
3581 break;
3582 case XC_LEARN:
3583 if (may_learn) {
3584 struct rule_dpif *rule = entry->u.learn.rule;
3585
3586 /* Reset the modified time for a rule that is equivalent to
83709dfa
JR
3587 * the currently cached rule. If the rule is not the exact
3588 * rule we have cached, update the reference that we have. */
3589 entry->u.learn.rule = ofproto_dpif_refresh_rule(rule);
b256dc52
JS
3590 }
3591 break;
3592 case XC_NORMAL:
3593 xlate_cache_normal(entry->u.normal.ofproto, entry->u.normal.flow,
3594 entry->u.normal.vlan);
3595 break;
3596 case XC_FIN_TIMEOUT:
3597 xlate_fin_timeout__(entry->u.fin.rule, stats->tcp_flags,
3598 entry->u.fin.idle, entry->u.fin.hard);
3599 break;
3600 default:
3601 OVS_NOT_REACHED();
3602 }
3603 }
3604}
3605
3606static void
3607xlate_dev_unref(struct xc_entry *entry)
3608{
3609 if (entry->u.dev.tx) {
3610 netdev_close(entry->u.dev.tx);
3611 }
3612 if (entry->u.dev.rx) {
3613 netdev_close(entry->u.dev.rx);
3614 }
3615 if (entry->u.dev.bfd) {
3616 bfd_unref(entry->u.dev.bfd);
3617 }
3618}
3619
3620static void
3621xlate_cache_clear_netflow(struct netflow *netflow, struct flow *flow)
3622{
3623 netflow_expire(netflow, flow);
3624 netflow_flow_clear(netflow, flow);
3625 netflow_unref(netflow);
3626 free(flow);
3627}
3628
3629void
3630xlate_cache_clear(struct xlate_cache *xcache)
3631{
3632 struct xc_entry *entry;
3633 struct ofpbuf entries;
3634
3635 if (!xcache) {
3636 return;
3637 }
3638
3639 XC_ENTRY_FOR_EACH (entry, entries, xcache) {
3640 switch (entry->type) {
3641 case XC_RULE:
3642 rule_dpif_unref(entry->u.rule);
3643 break;
3644 case XC_BOND:
3645 free(entry->u.bond.flow);
3646 bond_unref(entry->u.bond.bond);
3647 break;
3648 case XC_NETDEV:
3649 xlate_dev_unref(entry);
3650 break;
3651 case XC_NETFLOW:
3652 xlate_cache_clear_netflow(entry->u.nf.netflow, entry->u.nf.flow);
3653 break;
3654 case XC_MIRROR:
3655 mbridge_unref(entry->u.mirror.mbridge);
3656 break;
3657 case XC_LEARN:
83709dfa 3658 /* 'u.learn.rule' is the learned rule. */
b256dc52
JS
3659 rule_dpif_unref(entry->u.learn.rule);
3660 break;
3661 case XC_NORMAL:
3662 free(entry->u.normal.flow);
3663 break;
3664 case XC_FIN_TIMEOUT:
83709dfa
JR
3665 /* 'u.fin.rule' is always already held as a XC_RULE, which
3666 * has already released it's reference above. */
b256dc52
JS
3667 break;
3668 default:
3669 OVS_NOT_REACHED();
3670 }
3671 }
3672
3673 ofpbuf_clear(&xcache->entries);
3674}
3675
3676void
3677xlate_cache_delete(struct xlate_cache *xcache)
3678{
3679 xlate_cache_clear(xcache);
3680 ofpbuf_uninit(&xcache->entries);
3681 free(xcache);
3682}