1 /* Copyright (c) 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License. */
17 #include "ofproto/ofproto-dpif-xlate.h"
23 #include "byte-order.h"
28 #include "dynamic-string.h"
33 #include "mac-learning.h"
34 #include "meta-flow.h"
35 #include "multipath.h"
36 #include "netdev-vport.h"
39 #include "odp-execute.h"
40 #include "ofp-actions.h"
41 #include "ofproto/ofproto-dpif-ipfix.h"
42 #include "ofproto/ofproto-dpif-mirror.h"
43 #include "ofproto/ofproto-dpif-sflow.h"
44 #include "ofproto/ofproto-dpif.h"
48 COVERAGE_DEFINE(xlate_actions);
50 VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate);
52 /* Maximum depth of flow table recursion (due to resubmit actions) in a
53 * flow translation. */
54 #define MAX_RESUBMIT_RECURSION 64
57 struct hmap_node hmap_node; /* Node in global 'xbridges' map. */
58 struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */
60 struct list xbundles; /* Owned xbundles. */
61 struct hmap xports; /* Indexed by ofp_port. */
63 char *name; /* Name used in log messages. */
64 struct dpif *dpif; /* Datapath interface. */
65 struct mac_learning *ml; /* Mac learning handle. */
66 struct mbridge *mbridge; /* Mirroring. */
67 struct dpif_sflow *sflow; /* SFlow handle, or null. */
68 struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */
69 struct stp *stp; /* STP or null if disabled. */
71 enum ofp_config_flags frag; /* Fragmentation handling. */
72 bool has_netflow; /* Bridge runs netflow? */
73 bool has_in_band; /* Bridge has in band control? */
74 bool forward_bpdu; /* Bridge forwards STP BPDUs? */
78 struct hmap_node hmap_node; /* In global 'xbundles' map. */
79 struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */
81 struct list list_node; /* In parent 'xbridges' list. */
82 struct xbridge *xbridge; /* Parent xbridge. */
84 struct list xports; /* Contains "struct xport"s. */
86 char *name; /* Name used in log messages. */
87 struct bond *bond; /* Nonnull iff more than one port. */
88 struct lacp *lacp; /* LACP handle or null. */
90 enum port_vlan_mode vlan_mode; /* VLAN mode. */
91 int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */
92 unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1.
93 * NULL if all VLANs are trunked. */
94 bool use_priority_tags; /* Use 802.1p tag for frames in VLAN 0? */
95 bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */
99 struct hmap_node hmap_node; /* Node in global 'xports' map. */
100 struct ofport_dpif *ofport; /* Key in global 'xports map. */
102 struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */
103 ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */
105 odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */
107 struct list bundle_node; /* In parent xbundle (if it exists). */
108 struct xbundle *xbundle; /* Parent xbundle or null. */
110 struct netdev *netdev; /* 'ofport''s netdev. */
112 struct xbridge *xbridge; /* Parent bridge. */
113 struct xport *peer; /* Patch port peer or null. */
115 enum ofputil_port_config config; /* OpenFlow port configuration. */
116 int stp_port_no; /* STP port number or 0 if not in use. */
118 struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */
120 bool may_enable; /* May be enabled in bonds. */
121 bool is_tunnel; /* Is a tunnel port. */
123 struct cfm *cfm; /* CFM handle or null. */
124 struct bfd *bfd; /* BFD handle or null. */
128 struct xlate_in *xin;
129 struct xlate_out *xout;
131 const struct xbridge *xbridge;
133 /* Flow at the last commit. */
134 struct flow base_flow;
136 /* Tunnel IP destination address as received. This is stored separately
137 * as the base_flow.tunnel is cleared on init to reflect the datapath
138 * behavior. Used to make sure not to send tunneled output to ourselves,
139 * which might lead to an infinite loop. This could happen easily
140 * if a tunnel is marked as 'ip_remote=flow', and the flow does not
141 * actually set the tun_dst field. */
142 ovs_be32 orig_tunnel_ip_dst;
144 /* Stack for the push and pop actions. Each stack element is of type
145 * "union mf_subvalue". */
146 union mf_subvalue init_stack[1024 / sizeof(union mf_subvalue)];
149 /* The rule that we are currently translating, or NULL. */
150 struct rule_dpif *rule;
152 int recurse; /* Recursion level, via xlate_table_action. */
153 uint32_t orig_skb_priority; /* Priority when packet arrived. */
154 uint8_t table_id; /* OpenFlow table ID where flow was found. */
155 uint32_t sflow_n_outputs; /* Number of output ports. */
156 odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */
157 uint16_t user_cookie_offset;/* Used for user_action_cookie fixup. */
158 bool exit; /* No further actions should be processed. */
161 /* A controller may use OFPP_NONE as the ingress port to indicate that
162 * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for
163 * when an input bundle is needed for validation (e.g., mirroring or
164 * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have
165 * any 'port' structs, so care must be taken when dealing with it.
166 * The bundle's name and vlan mode are initialized in lookup_input_bundle() */
167 static struct xbundle ofpp_none_bundle;
169 /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from
170 * 'priority' (the datapath's term for QoS queue) to the dscp bits which all
171 * traffic egressing the 'ofport' with that priority should be marked with. */
172 struct skb_priority_to_dscp {
173 struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */
174 uint32_t skb_priority; /* Priority of this queue (see struct flow). */
176 uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */
179 static struct hmap xbridges = HMAP_INITIALIZER(&xbridges);
180 static struct hmap xbundles = HMAP_INITIALIZER(&xbundles);
181 static struct hmap xports = HMAP_INITIALIZER(&xports);
183 static bool may_receive(const struct xport *, struct xlate_ctx *);
184 static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len,
186 static void xlate_normal(struct xlate_ctx *);
187 static void xlate_report(struct xlate_ctx *, const char *);
188 static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port,
189 uint8_t table_id, bool may_packet_in);
190 static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn);
191 static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid);
192 static void output_normal(struct xlate_ctx *, const struct xbundle *,
194 static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port);
196 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
198 static struct xbridge *xbridge_lookup(const struct ofproto_dpif *);
199 static struct xbundle *xbundle_lookup(const struct ofbundle *);
200 static struct xport *xport_lookup(struct ofport_dpif *);
201 static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port);
202 static struct skb_priority_to_dscp *get_skb_priority(const struct xport *,
203 uint32_t skb_priority);
204 static void clear_skb_priorities(struct xport *);
205 static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority,
209 xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name,
210 struct dpif *dpif, const struct mac_learning *ml,
211 struct stp *stp, const struct mbridge *mbridge,
212 const struct dpif_sflow *sflow,
213 const struct dpif_ipfix *ipfix, enum ofp_config_flags frag,
214 bool forward_bpdu, bool has_in_band, bool has_netflow)
216 struct xbridge *xbridge = xbridge_lookup(ofproto);
219 xbridge = xzalloc(sizeof *xbridge);
220 xbridge->ofproto = ofproto;
222 hmap_insert(&xbridges, &xbridge->hmap_node, hash_pointer(ofproto, 0));
223 hmap_init(&xbridge->xports);
224 list_init(&xbridge->xbundles);
227 if (xbridge->ml != ml) {
228 mac_learning_unref(xbridge->ml);
229 xbridge->ml = mac_learning_ref(ml);
232 if (xbridge->mbridge != mbridge) {
233 mbridge_unref(xbridge->mbridge);
234 xbridge->mbridge = mbridge_ref(mbridge);
237 if (xbridge->sflow != sflow) {
238 dpif_sflow_unref(xbridge->sflow);
239 xbridge->sflow = dpif_sflow_ref(sflow);
242 if (xbridge->ipfix != ipfix) {
243 dpif_ipfix_unref(xbridge->ipfix);
244 xbridge->ipfix = dpif_ipfix_ref(ipfix);
247 if (xbridge->stp != stp) {
248 stp_unref(xbridge->stp);
249 xbridge->stp = stp_ref(stp);
253 xbridge->name = xstrdup(name);
255 xbridge->dpif = dpif;
256 xbridge->forward_bpdu = forward_bpdu;
257 xbridge->has_in_band = has_in_band;
258 xbridge->has_netflow = has_netflow;
259 xbridge->frag = frag;
263 xlate_remove_ofproto(struct ofproto_dpif *ofproto)
265 struct xbridge *xbridge = xbridge_lookup(ofproto);
266 struct xbundle *xbundle, *next_xbundle;
267 struct xport *xport, *next_xport;
273 HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) {
274 xlate_ofport_remove(xport->ofport);
277 LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) {
278 xlate_bundle_remove(xbundle->ofbundle);
281 hmap_remove(&xbridges, &xbridge->hmap_node);
287 xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle,
288 const char *name, enum port_vlan_mode vlan_mode, int vlan,
289 unsigned long *trunks, bool use_priority_tags,
290 const struct bond *bond, const struct lacp *lacp,
293 struct xbundle *xbundle = xbundle_lookup(ofbundle);
296 xbundle = xzalloc(sizeof *xbundle);
297 xbundle->ofbundle = ofbundle;
298 xbundle->xbridge = xbridge_lookup(ofproto);
300 hmap_insert(&xbundles, &xbundle->hmap_node, hash_pointer(ofbundle, 0));
301 list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node);
302 list_init(&xbundle->xports);
305 ovs_assert(xbundle->xbridge);
308 xbundle->name = xstrdup(name);
310 xbundle->vlan_mode = vlan_mode;
311 xbundle->vlan = vlan;
312 xbundle->trunks = trunks;
313 xbundle->use_priority_tags = use_priority_tags;
314 xbundle->floodable = floodable;
316 if (xbundle->bond != bond) {
317 bond_unref(xbundle->bond);
318 xbundle->bond = bond_ref(bond);
321 if (xbundle->lacp != lacp) {
322 lacp_unref(xbundle->lacp);
323 xbundle->lacp = lacp_ref(lacp);
328 xlate_bundle_remove(struct ofbundle *ofbundle)
330 struct xbundle *xbundle = xbundle_lookup(ofbundle);
331 struct xport *xport, *next;
337 LIST_FOR_EACH_SAFE (xport, next, bundle_node, &xbundle->xports) {
338 list_remove(&xport->bundle_node);
339 xport->xbundle = NULL;
342 hmap_remove(&xbundles, &xbundle->hmap_node);
343 list_remove(&xbundle->list_node);
344 bond_unref(xbundle->bond);
345 lacp_unref(xbundle->lacp);
351 xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle,
352 struct ofport_dpif *ofport, ofp_port_t ofp_port,
353 odp_port_t odp_port, const struct netdev *netdev,
354 const struct cfm *cfm, const struct bfd *bfd,
355 struct ofport_dpif *peer, int stp_port_no,
356 const struct ofproto_port_queue *qdscp_list, size_t n_qdscp,
357 enum ofputil_port_config config, bool is_tunnel,
360 struct xport *xport = xport_lookup(ofport);
364 xport = xzalloc(sizeof *xport);
365 xport->ofport = ofport;
366 xport->xbridge = xbridge_lookup(ofproto);
367 xport->ofp_port = ofp_port;
369 hmap_init(&xport->skb_priorities);
370 hmap_insert(&xports, &xport->hmap_node, hash_pointer(ofport, 0));
371 hmap_insert(&xport->xbridge->xports, &xport->ofp_node,
372 hash_ofp_port(xport->ofp_port));
375 ovs_assert(xport->ofp_port == ofp_port);
377 xport->config = config;
378 xport->stp_port_no = stp_port_no;
379 xport->is_tunnel = is_tunnel;
380 xport->may_enable = may_enable;
381 xport->odp_port = odp_port;
383 if (xport->netdev != netdev) {
384 netdev_close(xport->netdev);
385 xport->netdev = netdev_ref(netdev);
388 if (xport->cfm != cfm) {
389 cfm_unref(xport->cfm);
390 xport->cfm = cfm_ref(cfm);
393 if (xport->bfd != bfd) {
394 bfd_unref(xport->bfd);
395 xport->bfd = bfd_ref(bfd);
399 xport->peer->peer = NULL;
401 xport->peer = peer ? xport_lookup(peer) : NULL;
403 xport->peer->peer = xport;
406 if (xport->xbundle) {
407 list_remove(&xport->bundle_node);
409 xport->xbundle = ofbundle ? xbundle_lookup(ofbundle) : NULL;
410 if (xport->xbundle) {
411 list_insert(&xport->xbundle->xports, &xport->bundle_node);
414 clear_skb_priorities(xport);
415 for (i = 0; i < n_qdscp; i++) {
416 struct skb_priority_to_dscp *pdscp;
417 uint32_t skb_priority;
419 if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue,
424 pdscp = xmalloc(sizeof *pdscp);
425 pdscp->skb_priority = skb_priority;
426 pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK;
427 hmap_insert(&xport->skb_priorities, &pdscp->hmap_node,
428 hash_int(pdscp->skb_priority, 0));
433 xlate_ofport_remove(struct ofport_dpif *ofport)
435 struct xport *xport = xport_lookup(ofport);
442 xport->peer->peer = NULL;
446 if (xport->xbundle) {
447 list_remove(&xport->bundle_node);
450 clear_skb_priorities(xport);
451 hmap_destroy(&xport->skb_priorities);
453 hmap_remove(&xports, &xport->hmap_node);
454 hmap_remove(&xport->xbridge->xports, &xport->ofp_node);
456 netdev_close(xport->netdev);
457 cfm_unref(xport->cfm);
458 bfd_unref(xport->bfd);
462 static struct xbridge *
463 xbridge_lookup(const struct ofproto_dpif *ofproto)
465 struct xbridge *xbridge;
467 HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0),
469 if (xbridge->ofproto == ofproto) {
476 static struct xbundle *
477 xbundle_lookup(const struct ofbundle *ofbundle)
479 struct xbundle *xbundle;
481 HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0),
483 if (xbundle->ofbundle == ofbundle) {
490 static struct xport *
491 xport_lookup(struct ofport_dpif *ofport)
495 HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0),
497 if (xport->ofport == ofport) {
504 static struct stp_port *
505 xport_get_stp_port(const struct xport *xport)
507 return xport->xbridge->stp && xport->stp_port_no
508 ? stp_get_port(xport->xbridge->stp, xport->stp_port_no)
512 static enum stp_state
513 xport_stp_learn_state(const struct xport *xport)
515 struct stp_port *sp = xport_get_stp_port(xport);
516 return stp_learn_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
520 xport_stp_forward_state(const struct xport *xport)
522 struct stp_port *sp = xport_get_stp_port(xport);
523 return stp_forward_in_state(sp ? stp_port_get_state(sp) : STP_DISABLED);
526 /* Returns true if STP should process 'flow'. Sets fields in 'wc' that
527 * were used to make the determination.*/
529 stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc)
531 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
532 return eth_addr_equals(flow->dl_dst, eth_addr_stp);
536 stp_process_packet(const struct xport *xport, const struct ofpbuf *packet)
538 struct stp_port *sp = xport_get_stp_port(xport);
539 struct ofpbuf payload = *packet;
540 struct eth_header *eth = payload.data;
542 /* Sink packets on ports that have STP disabled when the bridge has
544 if (!sp || stp_port_get_state(sp) == STP_DISABLED) {
548 /* Trim off padding on payload. */
549 if (payload.size > ntohs(eth->eth_type) + ETH_HEADER_LEN) {
550 payload.size = ntohs(eth->eth_type) + ETH_HEADER_LEN;
553 if (ofpbuf_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) {
554 stp_received_bpdu(sp, payload.data, payload.size);
558 static struct xport *
559 get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port)
563 HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port),
565 if (xport->ofp_port == ofp_port) {
573 ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port)
575 const struct xport *xport = get_ofp_port(xbridge, ofp_port);
576 return xport ? xport->odp_port : ODPP_NONE;
580 xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan)
582 return (bundle->vlan_mode != PORT_VLAN_ACCESS
583 && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan)));
587 xbundle_includes_vlan(const struct xbundle *xbundle, uint16_t vlan)
589 return vlan == xbundle->vlan || xbundle_trunks_vlan(xbundle, vlan);
593 xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle)
595 return xbundle != &ofpp_none_bundle
596 ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle)
601 xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle)
603 return xbundle != &ofpp_none_bundle
604 ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle)
609 xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle)
611 return xbundle != &ofpp_none_bundle
612 ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle)
616 static struct xbundle *
617 lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port,
618 bool warn, struct xport **in_xportp)
622 /* Find the port and bundle for the received packet. */
623 xport = get_ofp_port(xbridge, in_port);
627 if (xport && xport->xbundle) {
628 return xport->xbundle;
631 /* Special-case OFPP_NONE, which a controller may use as the ingress
632 * port for traffic that it is sourcing. */
633 if (in_port == OFPP_NONE) {
634 ofpp_none_bundle.name = "OFPP_NONE";
635 ofpp_none_bundle.vlan_mode = PORT_VLAN_TRUNK;
636 return &ofpp_none_bundle;
639 /* Odd. A few possible reasons here:
641 * - We deleted a port but there are still a few packets queued up
644 * - Someone externally added a port (e.g. "ovs-dpctl add-if") that
645 * we don't know about.
647 * - The ofproto client didn't configure the port as part of a bundle.
648 * This is particularly likely to happen if a packet was received on the
649 * port after it was created, but before the client had a chance to
650 * configure its bundle.
653 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
655 VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown "
656 "port %"PRIu16, xbridge->name, in_port);
662 add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow)
664 const struct xbridge *xbridge = ctx->xbridge;
665 mirror_mask_t mirrors;
666 struct xbundle *in_xbundle;
670 mirrors = ctx->xout->mirrors;
671 ctx->xout->mirrors = 0;
673 in_xbundle = lookup_input_bundle(xbridge, orig_flow->in_port.ofp_port,
674 ctx->xin->packet != NULL, NULL);
678 mirrors |= xbundle_mirror_src(xbridge, in_xbundle);
680 /* Drop frames on bundles reserved for mirroring. */
681 if (xbundle_mirror_out(xbridge, in_xbundle)) {
682 if (ctx->xin->packet != NULL) {
683 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
684 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
685 "%s, which is reserved exclusively for mirroring",
686 ctx->xbridge->name, in_xbundle->name);
688 ofpbuf_clear(&ctx->xout->odp_actions);
693 vid = vlan_tci_to_vid(orig_flow->vlan_tci);
694 if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) {
697 vlan = input_vid_to_vlan(in_xbundle, vid);
703 /* Restore the original packet before adding the mirror actions. */
704 ctx->xin->flow = *orig_flow;
707 mirror_mask_t dup_mirrors;
708 struct ofbundle *out;
709 unsigned long *vlans;
714 has_mirror = mirror_get(xbridge->mbridge, mirror_mask_ffs(mirrors) - 1,
715 &vlans, &dup_mirrors, &out, &out_vlan);
716 ovs_assert(has_mirror);
719 ctx->xout->wc.masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK);
721 vlan_mirrored = !vlans || bitmap_is_set(vlans, vlan);
724 if (!vlan_mirrored) {
725 mirrors = zero_rightmost_1bit(mirrors);
729 mirrors &= ~dup_mirrors;
730 ctx->xout->mirrors |= dup_mirrors;
732 struct xbundle *out_xbundle = xbundle_lookup(out);
734 output_normal(ctx, out_xbundle, vlan);
736 } else if (vlan != out_vlan
737 && !eth_addr_is_reserved(orig_flow->dl_dst)) {
738 struct xbundle *xbundle;
740 LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) {
741 if (xbundle_includes_vlan(xbundle, out_vlan)
742 && !xbundle_mirror_out(xbridge, xbundle)) {
743 output_normal(ctx, xbundle, out_vlan);
750 /* Given 'vid', the VID obtained from the 802.1Q header that was received as
751 * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle',
752 * the bundle on which the packet was received, returns the VLAN to which the
755 * Both 'vid' and the return value are in the range 0...4095. */
757 input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid)
759 switch (in_xbundle->vlan_mode) {
760 case PORT_VLAN_ACCESS:
761 return in_xbundle->vlan;
764 case PORT_VLAN_TRUNK:
767 case PORT_VLAN_NATIVE_UNTAGGED:
768 case PORT_VLAN_NATIVE_TAGGED:
769 return vid ? vid : in_xbundle->vlan;
776 /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'.
777 * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs
780 * 'vid' should be the VID obtained from the 802.1Q header that was received as
781 * part of a packet (specify 0 if there was no 802.1Q header), in the range
784 input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn)
786 /* Allow any VID on the OFPP_NONE port. */
787 if (in_xbundle == &ofpp_none_bundle) {
791 switch (in_xbundle->vlan_mode) {
792 case PORT_VLAN_ACCESS:
795 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
796 VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" tagged "
797 "packet received on port %s configured as VLAN "
798 "%"PRIu16" access port", vid, in_xbundle->name,
805 case PORT_VLAN_NATIVE_UNTAGGED:
806 case PORT_VLAN_NATIVE_TAGGED:
808 /* Port must always carry its native VLAN. */
812 case PORT_VLAN_TRUNK:
813 if (!xbundle_includes_vlan(in_xbundle, vid)) {
815 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
816 VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" packet "
817 "received on port %s not configured for trunking "
818 "VLAN %"PRIu16, vid, in_xbundle->name, vid);
830 /* Given 'vlan', the VLAN that a packet belongs to, and
831 * 'out_xbundle', a bundle on which the packet is to be output, returns the VID
832 * that should be included in the 802.1Q header. (If the return value is 0,
833 * then the 802.1Q header should only be included in the packet if there is a
836 * Both 'vlan' and the return value are in the range 0...4095. */
838 output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan)
840 switch (out_xbundle->vlan_mode) {
841 case PORT_VLAN_ACCESS:
844 case PORT_VLAN_TRUNK:
845 case PORT_VLAN_NATIVE_TAGGED:
848 case PORT_VLAN_NATIVE_UNTAGGED:
849 return vlan == out_xbundle->vlan ? 0 : vlan;
857 output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle,
860 ovs_be16 *flow_tci = &ctx->xin->flow.vlan_tci;
862 ovs_be16 tci, old_tci;
865 vid = output_vlan_to_vid(out_xbundle, vlan);
866 if (list_is_empty(&out_xbundle->xports)) {
867 /* Partially configured bundle with no slaves. Drop the packet. */
869 } else if (!out_xbundle->bond) {
870 xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport,
873 struct ofport_dpif *ofport;
875 ofport = bond_choose_output_slave(out_xbundle->bond, &ctx->xin->flow,
876 &ctx->xout->wc, vid);
877 xport = ofport ? xport_lookup(ofport) : NULL;
880 /* No slaves enabled, so drop packet. */
887 if (tci || out_xbundle->use_priority_tags) {
888 tci |= *flow_tci & htons(VLAN_PCP_MASK);
890 tci |= htons(VLAN_CFI);
895 compose_output_action(ctx, xport->ofp_port);
899 /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after
900 * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to
901 * indicate this; newer upstream kernels use gratuitous ARP requests. */
903 is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc)
905 if (flow->dl_type != htons(ETH_TYPE_ARP)) {
909 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
910 if (!eth_addr_is_broadcast(flow->dl_dst)) {
914 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
915 if (flow->nw_proto == ARP_OP_REPLY) {
917 } else if (flow->nw_proto == ARP_OP_REQUEST) {
918 memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src);
919 memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst);
921 return flow->nw_src == flow->nw_dst;
928 update_learning_table(const struct xbridge *xbridge,
929 const struct flow *flow, struct flow_wildcards *wc,
930 int vlan, struct xbundle *in_xbundle)
932 struct mac_entry *mac;
934 /* Don't learn the OFPP_NONE port. */
935 if (in_xbundle == &ofpp_none_bundle) {
939 ovs_rwlock_wrlock(&xbridge->ml->rwlock);
940 if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) {
944 mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan);
945 if (is_gratuitous_arp(flow, wc)) {
946 /* We don't want to learn from gratuitous ARP packets that are
947 * reflected back over bond slaves so we lock the learning table. */
948 if (!in_xbundle->bond) {
949 mac_entry_set_grat_arp_lock(mac);
950 } else if (mac_entry_is_grat_arp_locked(mac)) {
955 if (mac->port.p != in_xbundle->ofbundle) {
956 /* The log messages here could actually be useful in debugging,
957 * so keep the rate limit relatively high. */
958 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300);
959 VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is "
960 "on port %s in VLAN %d",
961 xbridge->name, ETH_ADDR_ARGS(flow->dl_src),
962 in_xbundle->name, vlan);
964 mac->port.p = in_xbundle->ofbundle;
965 mac_learning_changed(xbridge->ml);
968 ovs_rwlock_unlock(&xbridge->ml->rwlock);
971 /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or
972 * dropped. Returns true if they may be forwarded, false if they should be
975 * 'in_port' must be the xport that corresponds to flow->in_port.
976 * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull).
978 * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as
979 * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as
980 * checked by input_vid_is_valid().
982 * May also add tags to '*tags', although the current implementation only does
983 * so in one special case.
986 is_admissible(struct xlate_ctx *ctx, struct xport *in_port,
989 struct xbundle *in_xbundle = in_port->xbundle;
990 const struct xbridge *xbridge = ctx->xbridge;
991 struct flow *flow = &ctx->xin->flow;
993 /* Drop frames for reserved multicast addresses
994 * only if forward_bpdu option is absent. */
995 if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) {
996 xlate_report(ctx, "packet has reserved destination MAC, dropping");
1000 if (in_xbundle->bond) {
1001 struct mac_entry *mac;
1003 switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport,
1009 xlate_report(ctx, "bonding refused admissibility, dropping");
1012 case BV_DROP_IF_MOVED:
1013 ovs_rwlock_rdlock(&xbridge->ml->rwlock);
1014 mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan);
1015 if (mac && mac->port.p != in_xbundle->ofbundle &&
1016 (!is_gratuitous_arp(flow, &ctx->xout->wc)
1017 || mac_entry_is_grat_arp_locked(mac))) {
1018 ovs_rwlock_unlock(&xbridge->ml->rwlock);
1019 xlate_report(ctx, "SLB bond thinks this packet looped back, "
1023 ovs_rwlock_unlock(&xbridge->ml->rwlock);
1032 xlate_normal(struct xlate_ctx *ctx)
1034 struct flow_wildcards *wc = &ctx->xout->wc;
1035 struct flow *flow = &ctx->xin->flow;
1036 struct xbundle *in_xbundle;
1037 struct xport *in_port;
1038 struct mac_entry *mac;
1042 ctx->xout->has_normal = true;
1044 memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src);
1045 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
1046 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
1048 in_xbundle = lookup_input_bundle(ctx->xbridge, flow->in_port.ofp_port,
1049 ctx->xin->packet != NULL, &in_port);
1051 xlate_report(ctx, "no input bundle, dropping");
1055 /* Drop malformed frames. */
1056 if (flow->dl_type == htons(ETH_TYPE_VLAN) &&
1057 !(flow->vlan_tci & htons(VLAN_CFI))) {
1058 if (ctx->xin->packet != NULL) {
1059 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1060 VLOG_WARN_RL(&rl, "bridge %s: dropping packet with partial "
1061 "VLAN tag received on port %s",
1062 ctx->xbridge->name, in_xbundle->name);
1064 xlate_report(ctx, "partial VLAN tag, dropping");
1068 /* Drop frames on bundles reserved for mirroring. */
1069 if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) {
1070 if (ctx->xin->packet != NULL) {
1071 static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5);
1072 VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port "
1073 "%s, which is reserved exclusively for mirroring",
1074 ctx->xbridge->name, in_xbundle->name);
1076 xlate_report(ctx, "input port is mirror output port, dropping");
1081 vid = vlan_tci_to_vid(flow->vlan_tci);
1082 if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) {
1083 xlate_report(ctx, "disallowed VLAN VID for this input port, dropping");
1086 vlan = input_vid_to_vlan(in_xbundle, vid);
1088 /* Check other admissibility requirements. */
1089 if (in_port && !is_admissible(ctx, in_port, vlan)) {
1093 /* Learn source MAC. */
1094 if (ctx->xin->may_learn) {
1095 update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle);
1098 /* Determine output bundle. */
1099 ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock);
1100 mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan);
1102 struct xbundle *mac_xbundle = xbundle_lookup(mac->port.p);
1103 if (mac_xbundle && mac_xbundle != in_xbundle) {
1104 xlate_report(ctx, "forwarding to learned port");
1105 output_normal(ctx, mac_xbundle, vlan);
1106 } else if (!mac_xbundle) {
1107 xlate_report(ctx, "learned port is unknown, dropping");
1109 xlate_report(ctx, "learned port is input port, dropping");
1112 struct xbundle *xbundle;
1114 xlate_report(ctx, "no learned MAC for destination, flooding");
1115 LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) {
1116 if (xbundle != in_xbundle
1117 && xbundle_includes_vlan(xbundle, vlan)
1118 && xbundle->floodable
1119 && !xbundle_mirror_out(ctx->xbridge, xbundle)) {
1120 output_normal(ctx, xbundle, vlan);
1123 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
1125 ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock);
1128 /* Compose SAMPLE action for sFlow or IPFIX. The given probability is
1129 * the number of packets out of UINT32_MAX to sample. The given
1130 * cookie is passed back in the callback for each sampled packet.
1133 compose_sample_action(const struct xbridge *xbridge,
1134 struct ofpbuf *odp_actions,
1135 const struct flow *flow,
1136 const uint32_t probability,
1137 const union user_action_cookie *cookie,
1138 const size_t cookie_size)
1140 size_t sample_offset, actions_offset;
1141 odp_port_t odp_port;
1145 sample_offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SAMPLE);
1147 nl_msg_put_u32(odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability);
1149 actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS);
1151 odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port);
1152 pid = dpif_port_get_pid(xbridge->dpif, odp_port);
1153 cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, odp_actions);
1155 nl_msg_end_nested(odp_actions, actions_offset);
1156 nl_msg_end_nested(odp_actions, sample_offset);
1157 return cookie_offset;
1161 compose_sflow_cookie(const struct xbridge *xbridge, ovs_be16 vlan_tci,
1162 odp_port_t odp_port, unsigned int n_outputs,
1163 union user_action_cookie *cookie)
1167 cookie->type = USER_ACTION_COOKIE_SFLOW;
1168 cookie->sflow.vlan_tci = vlan_tci;
1170 /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output
1171 * port information") for the interpretation of cookie->output. */
1172 switch (n_outputs) {
1174 /* 0x40000000 | 256 means "packet dropped for unknown reason". */
1175 cookie->sflow.output = 0x40000000 | 256;
1179 ifindex = dpif_sflow_odp_port_to_ifindex(xbridge->sflow, odp_port);
1181 cookie->sflow.output = ifindex;
1186 /* 0x80000000 means "multiple output ports. */
1187 cookie->sflow.output = 0x80000000 | n_outputs;
1192 /* Compose SAMPLE action for sFlow bridge sampling. */
1194 compose_sflow_action(const struct xbridge *xbridge,
1195 struct ofpbuf *odp_actions,
1196 const struct flow *flow,
1197 odp_port_t odp_port)
1199 uint32_t probability;
1200 union user_action_cookie cookie;
1202 if (!xbridge->sflow || flow->in_port.ofp_port == OFPP_NONE) {
1206 probability = dpif_sflow_get_probability(xbridge->sflow);
1207 compose_sflow_cookie(xbridge, htons(0), odp_port,
1208 odp_port == ODPP_NONE ? 0 : 1, &cookie);
1210 return compose_sample_action(xbridge, odp_actions, flow, probability,
1211 &cookie, sizeof cookie.sflow);
1215 compose_flow_sample_cookie(uint16_t probability, uint32_t collector_set_id,
1216 uint32_t obs_domain_id, uint32_t obs_point_id,
1217 union user_action_cookie *cookie)
1219 cookie->type = USER_ACTION_COOKIE_FLOW_SAMPLE;
1220 cookie->flow_sample.probability = probability;
1221 cookie->flow_sample.collector_set_id = collector_set_id;
1222 cookie->flow_sample.obs_domain_id = obs_domain_id;
1223 cookie->flow_sample.obs_point_id = obs_point_id;
1227 compose_ipfix_cookie(union user_action_cookie *cookie)
1229 cookie->type = USER_ACTION_COOKIE_IPFIX;
1232 /* Compose SAMPLE action for IPFIX bridge sampling. */
1234 compose_ipfix_action(const struct xbridge *xbridge,
1235 struct ofpbuf *odp_actions,
1236 const struct flow *flow)
1238 uint32_t probability;
1239 union user_action_cookie cookie;
1241 if (!xbridge->ipfix || flow->in_port.ofp_port == OFPP_NONE) {
1245 probability = dpif_ipfix_get_bridge_exporter_probability(xbridge->ipfix);
1246 compose_ipfix_cookie(&cookie);
1248 compose_sample_action(xbridge, odp_actions, flow, probability,
1249 &cookie, sizeof cookie.ipfix);
1252 /* SAMPLE action for sFlow must be first action in any given list of
1253 * actions. At this point we do not have all information required to
1254 * build it. So try to build sample action as complete as possible. */
1256 add_sflow_action(struct xlate_ctx *ctx)
1258 ctx->user_cookie_offset = compose_sflow_action(ctx->xbridge,
1259 &ctx->xout->odp_actions,
1260 &ctx->xin->flow, ODPP_NONE);
1261 ctx->sflow_odp_port = 0;
1262 ctx->sflow_n_outputs = 0;
1265 /* SAMPLE action for IPFIX must be 1st or 2nd action in any given list
1266 * of actions, eventually after the SAMPLE action for sFlow. */
1268 add_ipfix_action(struct xlate_ctx *ctx)
1270 compose_ipfix_action(ctx->xbridge, &ctx->xout->odp_actions,
1274 /* Fix SAMPLE action according to data collected while composing ODP actions.
1275 * We need to fix SAMPLE actions OVS_SAMPLE_ATTR_ACTIONS attribute, i.e. nested
1276 * USERSPACE action's user-cookie which is required for sflow. */
1278 fix_sflow_action(struct xlate_ctx *ctx)
1280 const struct flow *base = &ctx->base_flow;
1281 union user_action_cookie *cookie;
1283 if (!ctx->user_cookie_offset) {
1287 cookie = ofpbuf_at(&ctx->xout->odp_actions, ctx->user_cookie_offset,
1288 sizeof cookie->sflow);
1289 ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW);
1291 compose_sflow_cookie(ctx->xbridge, base->vlan_tci,
1292 ctx->sflow_odp_port, ctx->sflow_n_outputs, cookie);
1295 static enum slow_path_reason
1296 process_special(struct xlate_ctx *ctx, const struct flow *flow,
1297 const struct xport *xport, const struct ofpbuf *packet)
1299 struct flow_wildcards *wc = &ctx->xout->wc;
1300 const struct xbridge *xbridge = ctx->xbridge;
1304 } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) {
1306 cfm_process_heartbeat(xport->cfm, packet);
1309 } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) {
1311 bfd_process_packet(xport->bfd, flow, packet);
1314 } else if (xport->xbundle && xport->xbundle->lacp
1315 && flow->dl_type == htons(ETH_TYPE_LACP)) {
1317 lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet);
1320 } else if (xbridge->stp && stp_should_process_flow(flow, wc)) {
1322 stp_process_packet(xport, packet);
1331 compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port,
1334 const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port);
1335 struct flow_wildcards *wc = &ctx->xout->wc;
1336 struct flow *flow = &ctx->xin->flow;
1337 ovs_be16 flow_vlan_tci;
1338 uint32_t flow_skb_mark;
1339 uint8_t flow_nw_tos;
1340 odp_port_t out_port, odp_port;
1343 /* If 'struct flow' gets additional metadata, we'll need to zero it out
1344 * before traversing a patch port. */
1345 BUILD_ASSERT_DECL(FLOW_WC_SEQ == 20);
1348 xlate_report(ctx, "Nonexistent output port");
1350 } else if (xport->config & OFPUTIL_PC_NO_FWD) {
1351 xlate_report(ctx, "OFPPC_NO_FWD set, skipping output");
1353 } else if (check_stp && !xport_stp_forward_state(xport)) {
1354 xlate_report(ctx, "STP not in forwarding state, skipping output");
1358 if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) {
1359 ctx->xout->mirrors |= xbundle_mirror_dst(xport->xbundle->xbridge,
1364 const struct xport *peer = xport->peer;
1365 struct flow old_flow = ctx->xin->flow;
1366 enum slow_path_reason special;
1368 ctx->xbridge = peer->xbridge;
1369 flow->in_port.ofp_port = peer->ofp_port;
1370 flow->metadata = htonll(0);
1371 memset(&flow->tunnel, 0, sizeof flow->tunnel);
1372 memset(flow->regs, 0, sizeof flow->regs);
1374 special = process_special(ctx, &ctx->xin->flow, peer,
1377 ctx->xout->slow = special;
1378 } else if (may_receive(peer, ctx)) {
1379 if (xport_stp_forward_state(peer)) {
1380 xlate_table_action(ctx, flow->in_port.ofp_port, 0, true);
1382 /* Forwarding is disabled by STP. Let OFPP_NORMAL and the
1383 * learning action look at the packet, then drop it. */
1384 struct flow old_base_flow = ctx->base_flow;
1385 size_t old_size = ctx->xout->odp_actions.size;
1386 mirror_mask_t old_mirrors = ctx->xout->mirrors;
1387 xlate_table_action(ctx, flow->in_port.ofp_port, 0, true);
1388 ctx->xout->mirrors = old_mirrors;
1389 ctx->base_flow = old_base_flow;
1390 ctx->xout->odp_actions.size = old_size;
1394 ctx->xin->flow = old_flow;
1395 ctx->xbridge = xport->xbundle->xbridge;
1397 if (ctx->xin->resubmit_stats) {
1398 netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats);
1399 netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats);
1405 flow_vlan_tci = flow->vlan_tci;
1406 flow_skb_mark = flow->skb_mark;
1407 flow_nw_tos = flow->nw_tos;
1409 if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) {
1410 wc->masks.nw_tos |= IP_ECN_MASK;
1411 flow->nw_tos &= ~IP_DSCP_MASK;
1412 flow->nw_tos |= dscp;
1415 if (xport->is_tunnel) {
1416 /* Save tunnel metadata so that changes made due to
1417 * the Logical (tunnel) Port are not visible for any further
1418 * matches, while explicit set actions on tunnel metadata are.
1420 struct flow_tnl flow_tnl = flow->tunnel;
1421 odp_port = tnl_port_send(xport->ofport, flow, &ctx->xout->wc);
1422 if (odp_port == ODPP_NONE) {
1423 xlate_report(ctx, "Tunneling decided against output");
1424 goto out; /* restore flow_nw_tos */
1426 if (flow->tunnel.ip_dst == ctx->orig_tunnel_ip_dst) {
1427 xlate_report(ctx, "Not tunneling to our own address");
1428 goto out; /* restore flow_nw_tos */
1430 if (ctx->xin->resubmit_stats) {
1431 netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats);
1433 out_port = odp_port;
1434 commit_odp_tunnel_action(flow, &ctx->base_flow,
1435 &ctx->xout->odp_actions);
1436 flow->tunnel = flow_tnl; /* Restore tunnel metadata */
1438 ofp_port_t vlandev_port;
1440 odp_port = xport->odp_port;
1441 if (ofproto_has_vlan_splinters(ctx->xbridge->ofproto)) {
1442 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
1444 vlandev_port = vsp_realdev_to_vlandev(ctx->xbridge->ofproto, ofp_port,
1446 if (vlandev_port == ofp_port) {
1447 out_port = odp_port;
1449 out_port = ofp_port_to_odp_port(ctx->xbridge, vlandev_port);
1450 flow->vlan_tci = htons(0);
1452 flow->skb_mark &= ~IPSEC_MARK;
1455 if (out_port != ODPP_NONE) {
1456 commit_odp_actions(flow, &ctx->base_flow,
1457 &ctx->xout->odp_actions, &ctx->xout->wc);
1458 nl_msg_put_odp_port(&ctx->xout->odp_actions, OVS_ACTION_ATTR_OUTPUT,
1461 ctx->sflow_odp_port = odp_port;
1462 ctx->sflow_n_outputs++;
1463 ctx->xout->nf_output_iface = ofp_port;
1468 flow->vlan_tci = flow_vlan_tci;
1469 flow->skb_mark = flow_skb_mark;
1470 flow->nw_tos = flow_nw_tos;
1474 compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port)
1476 compose_output_action__(ctx, ofp_port, true);
1479 /* Common rule processing in one place to avoid duplicating code. */
1480 static struct rule_dpif *
1481 ctx_rule_hooks(struct xlate_ctx *ctx, struct rule_dpif *rule,
1484 if (ctx->xin->resubmit_hook) {
1485 ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse);
1487 if (rule == NULL && may_packet_in) {
1489 * check if table configuration flags
1490 * OFPTC_TABLE_MISS_CONTROLLER, default.
1491 * OFPTC_TABLE_MISS_CONTINUE,
1492 * OFPTC_TABLE_MISS_DROP
1493 * When OF1.0, OFPTC_TABLE_MISS_CONTINUE is used. What to do?
1495 rule = rule_dpif_miss_rule(ctx->xbridge->ofproto, &ctx->xin->flow);
1497 if (rule && ctx->xin->resubmit_stats) {
1498 rule_credit_stats(rule, ctx->xin->resubmit_stats);
1504 xlate_table_action(struct xlate_ctx *ctx,
1505 ofp_port_t in_port, uint8_t table_id, bool may_packet_in)
1507 if (ctx->recurse < MAX_RESUBMIT_RECURSION) {
1508 struct rule_dpif *rule;
1509 ofp_port_t old_in_port = ctx->xin->flow.in_port.ofp_port;
1510 uint8_t old_table_id = ctx->table_id;
1512 ctx->table_id = table_id;
1514 /* Look up a flow with 'in_port' as the input port. */
1515 ctx->xin->flow.in_port.ofp_port = in_port;
1516 rule = rule_dpif_lookup_in_table(ctx->xbridge->ofproto,
1517 &ctx->xin->flow, &ctx->xout->wc,
1520 /* Restore the original input port. Otherwise OFPP_NORMAL and
1521 * OFPP_IN_PORT will have surprising behavior. */
1522 ctx->xin->flow.in_port.ofp_port = old_in_port;
1524 rule = ctx_rule_hooks(ctx, rule, may_packet_in);
1527 struct rule_dpif *old_rule = ctx->rule;
1531 do_xlate_actions(rule->up.ofpacts, rule->up.ofpacts_len, ctx);
1532 ctx->rule = old_rule;
1536 ctx->table_id = old_table_id;
1538 static struct vlog_rate_limit recurse_rl = VLOG_RATE_LIMIT_INIT(1, 1);
1540 VLOG_ERR_RL(&recurse_rl, "resubmit actions recursed over %d times",
1541 MAX_RESUBMIT_RECURSION);
1546 xlate_ofpact_resubmit(struct xlate_ctx *ctx,
1547 const struct ofpact_resubmit *resubmit)
1552 in_port = resubmit->in_port;
1553 if (in_port == OFPP_IN_PORT) {
1554 in_port = ctx->xin->flow.in_port.ofp_port;
1557 table_id = resubmit->table_id;
1558 if (table_id == 255) {
1559 table_id = ctx->table_id;
1562 xlate_table_action(ctx, in_port, table_id, false);
1566 flood_packets(struct xlate_ctx *ctx, bool all)
1568 const struct xport *xport;
1570 HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) {
1571 if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) {
1576 compose_output_action__(ctx, xport->ofp_port, false);
1577 } else if (!(xport->config & OFPUTIL_PC_NO_FLOOD)) {
1578 compose_output_action(ctx, xport->ofp_port);
1582 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
1586 execute_controller_action(struct xlate_ctx *ctx, int len,
1587 enum ofp_packet_in_reason reason,
1588 uint16_t controller_id)
1590 struct ofputil_packet_in pin;
1591 struct ofpbuf *packet;
1594 ovs_assert(!ctx->xout->slow || ctx->xout->slow == SLOW_CONTROLLER);
1595 ctx->xout->slow = SLOW_CONTROLLER;
1596 if (!ctx->xin->packet) {
1600 packet = ofpbuf_clone(ctx->xin->packet);
1602 key.skb_priority = 0;
1604 memset(&key.tunnel, 0, sizeof key.tunnel);
1606 commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
1607 &ctx->xout->odp_actions, &ctx->xout->wc);
1609 odp_execute_actions(NULL, packet, &key, ctx->xout->odp_actions.data,
1610 ctx->xout->odp_actions.size, NULL, NULL);
1612 pin.packet = packet->data;
1613 pin.packet_len = packet->size;
1614 pin.reason = reason;
1615 pin.controller_id = controller_id;
1616 pin.table_id = ctx->table_id;
1617 pin.cookie = ctx->rule ? ctx->rule->up.flow_cookie : 0;
1620 flow_get_metadata(&ctx->xin->flow, &pin.fmd);
1622 ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, &pin);
1623 ofpbuf_delete(packet);
1627 compose_mpls_push_action(struct xlate_ctx *ctx, ovs_be16 eth_type)
1629 struct flow_wildcards *wc = &ctx->xout->wc;
1630 struct flow *flow = &ctx->xin->flow;
1632 ovs_assert(eth_type_mpls(eth_type));
1634 memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
1635 memset(&wc->masks.mpls_depth, 0xff, sizeof wc->masks.mpls_depth);
1637 if (flow->mpls_depth) {
1638 flow->mpls_lse &= ~htonl(MPLS_BOS_MASK);
1644 if (flow->dl_type == htons(ETH_TYPE_IPV6)) {
1645 label = htonl(0x2); /* IPV6 Explicit Null. */
1647 label = htonl(0x0); /* IPV4 Explicit Null. */
1649 wc->masks.nw_tos |= IP_DSCP_MASK;
1650 wc->masks.nw_ttl = 0xff;
1651 tc = (flow->nw_tos & IP_DSCP_MASK) >> 2;
1652 ttl = flow->nw_ttl ? flow->nw_ttl : 0x40;
1653 flow->mpls_lse = set_mpls_lse_values(ttl, tc, 1, label);
1654 flow->mpls_depth = 1;
1656 flow->dl_type = eth_type;
1660 compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type)
1662 struct flow_wildcards *wc = &ctx->xout->wc;
1663 struct flow *flow = &ctx->xin->flow;
1665 ovs_assert(eth_type_mpls(ctx->xin->flow.dl_type));
1666 ovs_assert(!eth_type_mpls(eth_type));
1668 memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
1669 memset(&wc->masks.mpls_depth, 0xff, sizeof wc->masks.mpls_depth);
1671 if (flow->mpls_depth) {
1673 flow->mpls_lse = htonl(0);
1674 if (!flow->mpls_depth) {
1675 flow->dl_type = eth_type;
1681 compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids)
1683 struct flow *flow = &ctx->xin->flow;
1685 if (!is_ip_any(flow)) {
1689 ctx->xout->wc.masks.nw_ttl = 0xff;
1690 if (flow->nw_ttl > 1) {
1696 for (i = 0; i < ids->n_controllers; i++) {
1697 execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL,
1701 /* Stop processing for current table. */
1707 compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl)
1709 if (!eth_type_mpls(ctx->xin->flow.dl_type)) {
1713 ctx->xout->wc.masks.mpls_lse |= htonl(MPLS_TTL_MASK);
1714 set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse, ttl);
1719 compose_dec_mpls_ttl_action(struct xlate_ctx *ctx)
1721 struct flow *flow = &ctx->xin->flow;
1722 uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse);
1723 struct flow_wildcards *wc = &ctx->xout->wc;
1725 memset(&wc->masks.mpls_lse, 0xff, sizeof wc->masks.mpls_lse);
1727 if (!eth_type_mpls(flow->dl_type)) {
1733 set_mpls_lse_ttl(&flow->mpls_lse, ttl);
1736 execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0);
1738 /* Stop processing for current table. */
1744 xlate_output_action(struct xlate_ctx *ctx,
1745 ofp_port_t port, uint16_t max_len, bool may_packet_in)
1747 ofp_port_t prev_nf_output_iface = ctx->xout->nf_output_iface;
1749 ctx->xout->nf_output_iface = NF_OUT_DROP;
1753 compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port);
1756 xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
1763 flood_packets(ctx, false);
1766 flood_packets(ctx, true);
1768 case OFPP_CONTROLLER:
1769 execute_controller_action(ctx, max_len, OFPR_ACTION, 0);
1775 if (port != ctx->xin->flow.in_port.ofp_port) {
1776 compose_output_action(ctx, port);
1778 xlate_report(ctx, "skipping output to input port");
1783 if (prev_nf_output_iface == NF_OUT_FLOOD) {
1784 ctx->xout->nf_output_iface = NF_OUT_FLOOD;
1785 } else if (ctx->xout->nf_output_iface == NF_OUT_DROP) {
1786 ctx->xout->nf_output_iface = prev_nf_output_iface;
1787 } else if (prev_nf_output_iface != NF_OUT_DROP &&
1788 ctx->xout->nf_output_iface != NF_OUT_FLOOD) {
1789 ctx->xout->nf_output_iface = NF_OUT_MULTI;
1794 xlate_output_reg_action(struct xlate_ctx *ctx,
1795 const struct ofpact_output_reg *or)
1797 uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow);
1798 if (port <= UINT16_MAX) {
1799 union mf_subvalue value;
1801 memset(&value, 0xff, sizeof value);
1802 mf_write_subfield_flow(&or->src, &value, &ctx->xout->wc.masks);
1803 xlate_output_action(ctx, u16_to_ofp(port),
1804 or->max_len, false);
1809 xlate_enqueue_action(struct xlate_ctx *ctx,
1810 const struct ofpact_enqueue *enqueue)
1812 ofp_port_t ofp_port = enqueue->port;
1813 uint32_t queue_id = enqueue->queue;
1814 uint32_t flow_priority, priority;
1817 /* Translate queue to priority. */
1818 error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority);
1820 /* Fall back to ordinary output action. */
1821 xlate_output_action(ctx, enqueue->port, 0, false);
1825 /* Check output port. */
1826 if (ofp_port == OFPP_IN_PORT) {
1827 ofp_port = ctx->xin->flow.in_port.ofp_port;
1828 } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) {
1832 /* Add datapath actions. */
1833 flow_priority = ctx->xin->flow.skb_priority;
1834 ctx->xin->flow.skb_priority = priority;
1835 compose_output_action(ctx, ofp_port);
1836 ctx->xin->flow.skb_priority = flow_priority;
1838 /* Update NetFlow output port. */
1839 if (ctx->xout->nf_output_iface == NF_OUT_DROP) {
1840 ctx->xout->nf_output_iface = ofp_port;
1841 } else if (ctx->xout->nf_output_iface != NF_OUT_FLOOD) {
1842 ctx->xout->nf_output_iface = NF_OUT_MULTI;
1847 xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id)
1849 uint32_t skb_priority;
1851 if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) {
1852 ctx->xin->flow.skb_priority = skb_priority;
1854 /* Couldn't translate queue to a priority. Nothing to do. A warning
1855 * has already been logged. */
1860 slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_)
1862 const struct xbridge *xbridge = xbridge_;
1873 case OFPP_CONTROLLER: /* Not supported by the bundle action. */
1876 port = get_ofp_port(xbridge, ofp_port);
1877 return port ? port->may_enable : false;
1882 xlate_bundle_action(struct xlate_ctx *ctx,
1883 const struct ofpact_bundle *bundle)
1887 port = bundle_execute(bundle, &ctx->xin->flow, &ctx->xout->wc,
1889 CONST_CAST(struct xbridge *, ctx->xbridge));
1890 if (bundle->dst.field) {
1891 nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow,
1894 xlate_output_action(ctx, port, 0, false);
1899 xlate_learn_action(struct xlate_ctx *ctx,
1900 const struct ofpact_learn *learn)
1902 struct ofputil_flow_mod *fm;
1903 struct ofpbuf ofpacts;
1905 ctx->xout->has_learn = true;
1907 learn_mask(learn, &ctx->xout->wc);
1909 if (!ctx->xin->may_learn) {
1913 fm = xmalloc(sizeof *fm);
1914 ofpbuf_init(&ofpacts, 0);
1915 learn_execute(learn, &ctx->xin->flow, fm, &ofpacts);
1917 ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm);
1920 /* Reduces '*timeout' to no more than 'max'. A value of zero in either case
1921 * means "infinite". */
1923 reduce_timeout(uint16_t max, uint16_t *timeout)
1925 if (max && (!*timeout || *timeout > max)) {
1931 xlate_fin_timeout(struct xlate_ctx *ctx,
1932 const struct ofpact_fin_timeout *oft)
1934 if (ctx->xin->tcp_flags & (TCP_FIN | TCP_RST) && ctx->rule) {
1935 struct rule_dpif *rule = ctx->rule;
1937 if (list_is_empty(&rule->up.expirable)) {
1938 list_insert(&rule->up.ofproto->expirable, &rule->up.expirable);
1941 reduce_timeout(oft->fin_idle_timeout, &rule->up.idle_timeout);
1942 reduce_timeout(oft->fin_hard_timeout, &rule->up.hard_timeout);
1947 xlate_sample_action(struct xlate_ctx *ctx,
1948 const struct ofpact_sample *os)
1950 union user_action_cookie cookie;
1951 /* Scale the probability from 16-bit to 32-bit while representing
1952 * the same percentage. */
1953 uint32_t probability = (os->probability << 16) | os->probability;
1955 commit_odp_actions(&ctx->xin->flow, &ctx->base_flow,
1956 &ctx->xout->odp_actions, &ctx->xout->wc);
1958 compose_flow_sample_cookie(os->probability, os->collector_set_id,
1959 os->obs_domain_id, os->obs_point_id, &cookie);
1960 compose_sample_action(ctx->xbridge, &ctx->xout->odp_actions, &ctx->xin->flow,
1961 probability, &cookie, sizeof cookie.flow_sample);
1965 may_receive(const struct xport *xport, struct xlate_ctx *ctx)
1967 if (xport->config & (eth_addr_equals(ctx->xin->flow.dl_dst, eth_addr_stp)
1968 ? OFPUTIL_PC_NO_RECV_STP
1969 : OFPUTIL_PC_NO_RECV)) {
1973 /* Only drop packets here if both forwarding and learning are
1974 * disabled. If just learning is enabled, we need to have
1975 * OFPP_NORMAL and the learning action have a look at the packet
1976 * before we can drop it. */
1977 if (!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) {
1985 tunnel_ecn_ok(struct xlate_ctx *ctx)
1987 if (is_ip_any(&ctx->base_flow)
1988 && (ctx->xin->flow.tunnel.ip_tos & IP_ECN_MASK) == IP_ECN_CE) {
1989 if ((ctx->base_flow.nw_tos & IP_ECN_MASK) == IP_ECN_NOT_ECT) {
1990 VLOG_WARN_RL(&rl, "dropping tunnel packet marked ECN CE"
1991 " but is not ECN capable");
1994 /* Set the ECN CE value in the tunneled packet. */
1995 ctx->xin->flow.nw_tos |= IP_ECN_CE;
2003 do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len,
2004 struct xlate_ctx *ctx)
2006 struct flow_wildcards *wc = &ctx->xout->wc;
2007 struct flow *flow = &ctx->xin->flow;
2008 bool was_evictable = true;
2009 const struct ofpact *a;
2012 /* Don't let the rule we're working on get evicted underneath us. */
2013 was_evictable = ctx->rule->up.evictable;
2014 ctx->rule->up.evictable = false;
2017 OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) {
2018 struct ofpact_controller *controller;
2019 const struct ofpact_metadata *metadata;
2027 xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port,
2028 ofpact_get_OUTPUT(a)->max_len, true);
2031 case OFPACT_CONTROLLER:
2032 controller = ofpact_get_CONTROLLER(a);
2033 execute_controller_action(ctx, controller->max_len,
2035 controller->controller_id);
2038 case OFPACT_ENQUEUE:
2039 xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a));
2042 case OFPACT_SET_VLAN_VID:
2043 wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI);
2044 flow->vlan_tci &= ~htons(VLAN_VID_MASK);
2045 flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid)
2049 case OFPACT_SET_VLAN_PCP:
2050 wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI);
2051 flow->vlan_tci &= ~htons(VLAN_PCP_MASK);
2053 htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp << VLAN_PCP_SHIFT)
2057 case OFPACT_STRIP_VLAN:
2058 memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
2059 flow->vlan_tci = htons(0);
2062 case OFPACT_PUSH_VLAN:
2063 /* XXX 802.1AD(QinQ) */
2064 memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci);
2065 flow->vlan_tci = htons(VLAN_CFI);
2068 case OFPACT_SET_ETH_SRC:
2069 memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src);
2070 memcpy(flow->dl_src, ofpact_get_SET_ETH_SRC(a)->mac, ETH_ADDR_LEN);
2073 case OFPACT_SET_ETH_DST:
2074 memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst);
2075 memcpy(flow->dl_dst, ofpact_get_SET_ETH_DST(a)->mac, ETH_ADDR_LEN);
2078 case OFPACT_SET_IPV4_SRC:
2079 memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src);
2080 if (flow->dl_type == htons(ETH_TYPE_IP)) {
2081 flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4;
2085 case OFPACT_SET_IPV4_DST:
2086 memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst);
2087 if (flow->dl_type == htons(ETH_TYPE_IP)) {
2088 flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4;
2092 case OFPACT_SET_IPV4_DSCP:
2093 wc->masks.nw_tos |= IP_DSCP_MASK;
2094 /* OpenFlow 1.0 only supports IPv4. */
2095 if (flow->dl_type == htons(ETH_TYPE_IP)) {
2096 flow->nw_tos &= ~IP_DSCP_MASK;
2097 flow->nw_tos |= ofpact_get_SET_IPV4_DSCP(a)->dscp;
2101 case OFPACT_SET_L4_SRC_PORT:
2102 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
2103 memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src);
2104 if (is_ip_any(flow)) {
2105 flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port);
2109 case OFPACT_SET_L4_DST_PORT:
2110 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
2111 memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst);
2112 if (is_ip_any(flow)) {
2113 flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port);
2117 case OFPACT_RESUBMIT:
2118 xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a));
2121 case OFPACT_SET_TUNNEL:
2122 flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id);
2125 case OFPACT_SET_QUEUE:
2126 xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id);
2129 case OFPACT_POP_QUEUE:
2130 flow->skb_priority = ctx->orig_skb_priority;
2133 case OFPACT_REG_MOVE:
2134 nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc);
2137 case OFPACT_REG_LOAD:
2138 nxm_execute_reg_load(ofpact_get_REG_LOAD(a), flow);
2141 case OFPACT_STACK_PUSH:
2142 nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc,
2146 case OFPACT_STACK_POP:
2147 nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc,
2151 case OFPACT_PUSH_MPLS:
2152 compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)->ethertype);
2155 case OFPACT_POP_MPLS:
2156 compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype);
2159 case OFPACT_SET_MPLS_TTL:
2160 if (compose_set_mpls_ttl_action(ctx,
2161 ofpact_get_SET_MPLS_TTL(a)->ttl)) {
2166 case OFPACT_DEC_MPLS_TTL:
2167 if (compose_dec_mpls_ttl_action(ctx)) {
2172 case OFPACT_DEC_TTL:
2173 wc->masks.nw_ttl = 0xff;
2174 if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) {
2180 /* Nothing to do. */
2183 case OFPACT_MULTIPATH:
2184 multipath_execute(ofpact_get_MULTIPATH(a), flow, wc);
2188 xlate_bundle_action(ctx, ofpact_get_BUNDLE(a));
2191 case OFPACT_OUTPUT_REG:
2192 xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a));
2196 xlate_learn_action(ctx, ofpact_get_LEARN(a));
2203 case OFPACT_FIN_TIMEOUT:
2204 memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto);
2205 ctx->xout->has_fin_timeout = true;
2206 xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a));
2209 case OFPACT_CLEAR_ACTIONS:
2211 * Nothing to do because writa-actions is not supported for now.
2212 * When writa-actions is supported, clear-actions also must
2213 * be supported at the same time.
2217 case OFPACT_WRITE_METADATA:
2218 metadata = ofpact_get_WRITE_METADATA(a);
2219 flow->metadata &= ~metadata->mask;
2220 flow->metadata |= metadata->metadata & metadata->mask;
2224 /* Not implemented yet. */
2227 case OFPACT_GOTO_TABLE: {
2228 /* It is assumed that goto-table is the last action. */
2229 struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a);
2231 ovs_assert(ctx->table_id < ogt->table_id);
2232 xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port,
2233 ogt->table_id, true);
2238 xlate_sample_action(ctx, ofpact_get_SAMPLE(a));
2245 ctx->rule->up.evictable = was_evictable;
2250 xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto,
2251 const struct flow *flow, struct rule_dpif *rule,
2252 uint8_t tcp_flags, const struct ofpbuf *packet)
2254 xin->ofproto = ofproto;
2256 xin->packet = packet;
2257 xin->may_learn = packet != NULL;
2259 xin->ofpacts = NULL;
2260 xin->ofpacts_len = 0;
2261 xin->tcp_flags = tcp_flags;
2262 xin->resubmit_hook = NULL;
2263 xin->report_hook = NULL;
2264 xin->resubmit_stats = NULL;
2268 xlate_out_uninit(struct xlate_out *xout)
2271 ofpbuf_uninit(&xout->odp_actions);
2275 /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts'
2276 * into datapath actions, using 'ctx', and discards the datapath actions. */
2278 xlate_actions_for_side_effects(struct xlate_in *xin)
2280 struct xlate_out xout;
2282 xlate_actions(xin, &xout);
2283 xlate_out_uninit(&xout);
2287 xlate_report(struct xlate_ctx *ctx, const char *s)
2289 if (ctx->xin->report_hook) {
2290 ctx->xin->report_hook(ctx->xin, s, ctx->recurse);
2295 xlate_out_copy(struct xlate_out *dst, const struct xlate_out *src)
2298 dst->slow = src->slow;
2299 dst->has_learn = src->has_learn;
2300 dst->has_normal = src->has_normal;
2301 dst->has_fin_timeout = src->has_fin_timeout;
2302 dst->nf_output_iface = src->nf_output_iface;
2303 dst->mirrors = src->mirrors;
2305 ofpbuf_use_stub(&dst->odp_actions, dst->odp_actions_stub,
2306 sizeof dst->odp_actions_stub);
2307 ofpbuf_put(&dst->odp_actions, src->odp_actions.data,
2308 src->odp_actions.size);
2311 static struct skb_priority_to_dscp *
2312 get_skb_priority(const struct xport *xport, uint32_t skb_priority)
2314 struct skb_priority_to_dscp *pdscp;
2317 hash = hash_int(skb_priority, 0);
2318 HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) {
2319 if (pdscp->skb_priority == skb_priority) {
2327 dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority,
2330 struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority);
2331 *dscp = pdscp ? pdscp->dscp : 0;
2332 return pdscp != NULL;
2336 clear_skb_priorities(struct xport *xport)
2338 struct skb_priority_to_dscp *pdscp, *next;
2340 HMAP_FOR_EACH_SAFE (pdscp, next, hmap_node, &xport->skb_priorities) {
2341 hmap_remove(&xport->skb_priorities, &pdscp->hmap_node);
2347 actions_output_to_local_port(const struct xlate_ctx *ctx)
2349 odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL);
2350 const struct nlattr *a;
2353 NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->xout->odp_actions.data,
2354 ctx->xout->odp_actions.size) {
2355 if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT
2356 && nl_attr_get_odp_port(a) == local_odp_port) {
2363 /* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts'
2364 * into datapath actions in 'odp_actions', using 'ctx'. */
2366 xlate_actions(struct xlate_in *xin, struct xlate_out *xout)
2368 struct flow_wildcards *wc = &xout->wc;
2369 struct flow *flow = &xin->flow;
2371 enum slow_path_reason special;
2372 const struct ofpact *ofpacts;
2373 struct xport *in_port;
2374 struct flow orig_flow;
2375 struct xlate_ctx ctx;
2378 COVERAGE_INC(xlate_actions);
2380 /* Flow initialization rules:
2381 * - 'base_flow' must match the kernel's view of the packet at the
2382 * time that action processing starts. 'flow' represents any
2383 * transformations we wish to make through actions.
2384 * - By default 'base_flow' and 'flow' are the same since the input
2385 * packet matches the output before any actions are applied.
2386 * - When using VLAN splinters, 'base_flow''s VLAN is set to the value
2387 * of the received packet as seen by the kernel. If we later output
2388 * to another device without any modifications this will cause us to
2389 * insert a new tag since the original one was stripped off by the
2391 * - Tunnel metadata as received is retained in 'flow'. This allows
2392 * tunnel metadata matching also in later tables.
2393 * Since a kernel action for setting the tunnel metadata will only be
2394 * generated with actual tunnel output, changing the tunnel metadata
2395 * values in 'flow' (such as tun_id) will only have effect with a later
2396 * tunnel output action.
2397 * - Tunnel 'base_flow' is completely cleared since that is what the
2398 * kernel does. If we wish to maintain the original values an action
2399 * needs to be generated. */
2404 ctx.xout->has_learn = false;
2405 ctx.xout->has_normal = false;
2406 ctx.xout->has_fin_timeout = false;
2407 ctx.xout->nf_output_iface = NF_OUT_DROP;
2408 ctx.xout->mirrors = 0;
2409 ofpbuf_use_stub(&ctx.xout->odp_actions, ctx.xout->odp_actions_stub,
2410 sizeof ctx.xout->odp_actions_stub);
2411 ofpbuf_reserve(&ctx.xout->odp_actions, NL_A_U32_SIZE);
2413 ctx.xbridge = xbridge_lookup(xin->ofproto);
2418 ctx.rule = xin->rule;
2420 ctx.base_flow = *flow;
2421 memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel);
2422 ctx.orig_tunnel_ip_dst = flow->tunnel.ip_dst;
2424 flow_wildcards_init_catchall(wc);
2425 memset(&wc->masks.in_port, 0xff, sizeof wc->masks.in_port);
2426 memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority);
2427 memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type);
2428 wc->masks.nw_frag |= FLOW_NW_FRAG_MASK;
2430 if (tnl_port_should_receive(&ctx.xin->flow)) {
2431 memset(&wc->masks.tunnel, 0xff, sizeof wc->masks.tunnel);
2432 /* skb_mark is currently used only by tunnels but that will likely
2433 * change in the future. */
2434 memset(&wc->masks.skb_mark, 0xff, sizeof wc->masks.skb_mark);
2436 if (ctx.xbridge->has_netflow) {
2437 netflow_mask_wc(flow, wc);
2441 ctx.orig_skb_priority = flow->skb_priority;
2446 ofpacts = xin->ofpacts;
2447 ofpacts_len = xin->ofpacts_len;
2448 } else if (xin->rule) {
2449 ofpacts = xin->rule->up.ofpacts;
2450 ofpacts_len = xin->rule->up.ofpacts_len;
2455 ofpbuf_use_stub(&ctx.stack, ctx.init_stack, sizeof ctx.init_stack);
2457 if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
2458 /* Do this conditionally because the copy is expensive enough that it
2459 * shows up in profiles. */
2463 if (flow->nw_frag & FLOW_NW_FRAG_ANY) {
2464 switch (ctx.xbridge->frag) {
2465 case OFPC_FRAG_NORMAL:
2466 /* We must pretend that transport ports are unavailable. */
2467 flow->tp_src = ctx.base_flow.tp_src = htons(0);
2468 flow->tp_dst = ctx.base_flow.tp_dst = htons(0);
2471 case OFPC_FRAG_DROP:
2474 case OFPC_FRAG_REASM:
2477 case OFPC_FRAG_NX_MATCH:
2478 /* Nothing to do. */
2481 case OFPC_INVALID_TTL_TO_CONTROLLER:
2486 in_port = get_ofp_port(ctx.xbridge, flow->in_port.ofp_port);
2487 special = process_special(&ctx, flow, in_port, ctx.xin->packet);
2489 ctx.xout->slow = special;
2491 size_t sample_actions_len;
2493 if (flow->in_port.ofp_port
2494 != vsp_realdev_to_vlandev(ctx.xbridge->ofproto,
2495 flow->in_port.ofp_port,
2497 ctx.base_flow.vlan_tci = 0;
2500 add_sflow_action(&ctx);
2501 add_ipfix_action(&ctx);
2502 sample_actions_len = ctx.xout->odp_actions.size;
2504 if (tunnel_ecn_ok(&ctx) && (!in_port || may_receive(in_port, &ctx))) {
2505 do_xlate_actions(ofpacts, ofpacts_len, &ctx);
2507 /* We've let OFPP_NORMAL and the learning action look at the
2508 * packet, so drop it now if forwarding is disabled. */
2509 if (in_port && !xport_stp_forward_state(in_port)) {
2510 ctx.xout->odp_actions.size = sample_actions_len;
2514 if (ctx.xbridge->has_in_band
2515 && in_band_must_output_to_local_port(flow)
2516 && !actions_output_to_local_port(&ctx)) {
2517 compose_output_action(&ctx, OFPP_LOCAL);
2520 fix_sflow_action(&ctx);
2522 if (mbridge_has_mirrors(ctx.xbridge->mbridge)) {
2523 add_mirror_actions(&ctx, &orig_flow);
2527 ofpbuf_uninit(&ctx.stack);
2529 /* Clear the metadata and register wildcard masks, because we won't
2530 * use non-header fields as part of the cache. */
2531 memset(&wc->masks.metadata, 0, sizeof wc->masks.metadata);
2532 memset(&wc->masks.regs, 0, sizeof wc->masks.regs);