]>
Commit | Line | Data |
---|---|---|
81f71381 | 1 | /* Copyright (c) 2009-2017, 2019-2020 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 | 19 | #include <errno.h> |
b2befd5b BP |
20 | #include <sys/types.h> |
21 | #include <netinet/in.h> | |
a36de779 PS |
22 | #include <arpa/inet.h> |
23 | #include <net/if.h> | |
24 | #include <sys/socket.h> | |
8449c4d6 | 25 | |
db7d4e46 | 26 | #include "bfd.h" |
9583bc14 EJ |
27 | #include "bitmap.h" |
28 | #include "bond.h" | |
29 | #include "bundle.h" | |
30 | #include "byte-order.h" | |
db7d4e46 | 31 | #include "cfm.h" |
9583bc14 EJ |
32 | #include "connmgr.h" |
33 | #include "coverage.h" | |
46445c63 | 34 | #include "csum.h" |
e14deea0 | 35 | #include "dp-packet.h" |
9583bc14 | 36 | #include "dpif.h" |
f7f1ea29 | 37 | #include "in-band.h" |
db7d4e46 | 38 | #include "lacp.h" |
9583bc14 EJ |
39 | #include "learn.h" |
40 | #include "mac-learning.h" | |
6d95c4e8 | 41 | #include "mcast-snooping.h" |
9583bc14 EJ |
42 | #include "multipath.h" |
43 | #include "netdev-vport.h" | |
44 | #include "netlink.h" | |
45 | #include "nx-match.h" | |
46 | #include "odp-execute.h" | |
9583bc14 | 47 | #include "ofproto/ofproto-dpif-ipfix.h" |
ec7ceaed | 48 | #include "ofproto/ofproto-dpif-mirror.h" |
60d02c72 | 49 | #include "ofproto/ofproto-dpif-monitor.h" |
9583bc14 | 50 | #include "ofproto/ofproto-dpif-sflow.h" |
2d9b49dd | 51 | #include "ofproto/ofproto-dpif-trace.h" |
901a517e | 52 | #include "ofproto/ofproto-dpif-xlate-cache.h" |
9583bc14 | 53 | #include "ofproto/ofproto-dpif.h" |
6f00e29b | 54 | #include "ofproto/ofproto-provider.h" |
b598f214 BW |
55 | #include "openvswitch/dynamic-string.h" |
56 | #include "openvswitch/meta-flow.h" | |
57 | #include "openvswitch/list.h" | |
58 | #include "openvswitch/ofp-actions.h" | |
1fc11c59 | 59 | #include "openvswitch/ofp-ed-props.h" |
b598f214 BW |
60 | #include "openvswitch/vlog.h" |
61 | #include "ovs-lldp.h" | |
a36de779 | 62 | #include "ovs-router.h" |
b598f214 BW |
63 | #include "packets.h" |
64 | #include "tnl-neigh-cache.h" | |
a36de779 | 65 | #include "tnl-ports.h" |
9583bc14 | 66 | #include "tunnel.h" |
ee89ea7b | 67 | #include "util.h" |
c3594cc3 | 68 | #include "uuid.h" |
9583bc14 | 69 | |
46c88433 | 70 | COVERAGE_DEFINE(xlate_actions); |
0f032e95 | 71 | COVERAGE_DEFINE(xlate_actions_oversize); |
7d031d7e | 72 | COVERAGE_DEFINE(xlate_actions_too_many_output); |
9583bc14 EJ |
73 | |
74 | VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate); | |
75 | ||
8a553e9a | 76 | /* Maximum depth of flow table recursion (due to resubmit actions) in a |
790c5d26 BP |
77 | * flow translation. |
78 | * | |
79 | * The goal of limiting the depth of resubmits is to ensure that flow | |
80 | * translation eventually terminates. Only resubmits to the same table or an | |
81 | * earlier table count against the maximum depth. This is because resubmits to | |
82 | * strictly monotonically increasing table IDs will eventually terminate, since | |
83 | * any OpenFlow switch has a finite number of tables. OpenFlow tables are most | |
84 | * commonly traversed in numerically increasing order, so this limit has little | |
85 | * effect on conventionally designed OpenFlow pipelines. | |
86 | * | |
87 | * Outputs to patch ports and to groups also count against the depth limit. */ | |
88 | #define MAX_DEPTH 64 | |
8a553e9a | 89 | |
98b07853 BP |
90 | /* Maximum number of resubmit actions in a flow translation, whether they are |
91 | * recursive or not. */ | |
790c5d26 | 92 | #define MAX_RESUBMITS (MAX_DEPTH * MAX_DEPTH) |
98b07853 | 93 | |
83c2757b ZB |
94 | /* The structure holds an array of IP addresses assigned to a bridge and the |
95 | * number of elements in the array. These data are mutable and are evaluated | |
96 | * when ARP or Neighbor Advertisement packets received on a native tunnel | |
97 | * port are xlated. So 'ref_cnt' and RCU are used for synchronization. */ | |
98 | struct xbridge_addr { | |
99 | struct in6_addr *addr; /* Array of IP addresses of xbridge. */ | |
100 | int n_addr; /* Number of IP addresses. */ | |
101 | struct ovs_refcount ref_cnt; | |
102 | }; | |
103 | ||
46c88433 EJ |
104 | struct xbridge { |
105 | struct hmap_node hmap_node; /* Node in global 'xbridges' map. */ | |
106 | struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */ | |
107 | ||
ca6ba700 | 108 | struct ovs_list xbundles; /* Owned xbundles. */ |
46c88433 EJ |
109 | struct hmap xports; /* Indexed by ofp_port. */ |
110 | ||
111 | char *name; /* Name used in log messages. */ | |
89a8a7f0 | 112 | struct dpif *dpif; /* Datapath interface. */ |
46c88433 | 113 | struct mac_learning *ml; /* Mac learning handle. */ |
6d95c4e8 | 114 | struct mcast_snooping *ms; /* Multicast Snooping handle. */ |
46c88433 EJ |
115 | struct mbridge *mbridge; /* Mirroring. */ |
116 | struct dpif_sflow *sflow; /* SFlow handle, or null. */ | |
117 | struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */ | |
ce3955be | 118 | struct netflow *netflow; /* Netflow handle, or null. */ |
9d189a50 | 119 | struct stp *stp; /* STP or null if disabled. */ |
9efd308e | 120 | struct rstp *rstp; /* RSTP or null if disabled. */ |
46c88433 | 121 | |
46c88433 EJ |
122 | bool has_in_band; /* Bridge has in band control? */ |
123 | bool forward_bpdu; /* Bridge forwards STP BPDUs? */ | |
4b97b70d | 124 | |
b440dd8c JS |
125 | /* Datapath feature support. */ |
126 | struct dpif_backer_support support; | |
83c2757b ZB |
127 | |
128 | struct xbridge_addr *addr; | |
46c88433 EJ |
129 | }; |
130 | ||
131 | struct xbundle { | |
132 | struct hmap_node hmap_node; /* In global 'xbundles' map. */ | |
133 | struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */ | |
134 | ||
ca6ba700 | 135 | struct ovs_list list_node; /* In parent 'xbridges' list. */ |
46c88433 EJ |
136 | struct xbridge *xbridge; /* Parent xbridge. */ |
137 | ||
ca6ba700 | 138 | struct ovs_list xports; /* Contains "struct xport"s. */ |
46c88433 EJ |
139 | |
140 | char *name; /* Name used in log messages. */ | |
141 | struct bond *bond; /* Nonnull iff more than one port. */ | |
142 | struct lacp *lacp; /* LACP handle or null. */ | |
143 | ||
144 | enum port_vlan_mode vlan_mode; /* VLAN mode. */ | |
fed8962a EG |
145 | uint16_t qinq_ethtype; /* Ethertype of dot1q-tunnel interface |
146 | * either 0x8100 or 0x88a8. */ | |
46c88433 EJ |
147 | int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */ |
148 | unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1. | |
149 | * NULL if all VLANs are trunked. */ | |
fed8962a EG |
150 | unsigned long *cvlans; /* Bitmap of allowed customer vlans, |
151 | * NULL if all VLANs are allowed */ | |
88f52d7f EB |
152 | enum port_priority_tags_mode use_priority_tags; |
153 | /* Use 802.1p tag for frames in VLAN 0? */ | |
46c88433 | 154 | bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */ |
c005f976 | 155 | bool protected; /* Protected port mode */ |
46c88433 EJ |
156 | }; |
157 | ||
158 | struct xport { | |
159 | struct hmap_node hmap_node; /* Node in global 'xports' map. */ | |
160 | struct ofport_dpif *ofport; /* Key in global 'xports map. */ | |
161 | ||
162 | struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */ | |
163 | ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */ | |
164 | ||
c3594cc3 ZB |
165 | struct hmap_node uuid_node; /* Node in global 'xports_uuid' map. */ |
166 | struct uuid uuid; /* Key in global 'xports_uuid' map. */ | |
167 | ||
46c88433 EJ |
168 | odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */ |
169 | ||
ca6ba700 | 170 | struct ovs_list bundle_node; /* In parent xbundle (if it exists). */ |
46c88433 EJ |
171 | struct xbundle *xbundle; /* Parent xbundle or null. */ |
172 | ||
173 | struct netdev *netdev; /* 'ofport''s netdev. */ | |
174 | ||
175 | struct xbridge *xbridge; /* Parent bridge. */ | |
176 | struct xport *peer; /* Patch port peer or null. */ | |
177 | ||
178 | enum ofputil_port_config config; /* OpenFlow port configuration. */ | |
dd8cd4b4 | 179 | enum ofputil_port_state state; /* OpenFlow port state. */ |
92cf817b | 180 | int stp_port_no; /* STP port number or -1 if not in use. */ |
f025bcb7 | 181 | struct rstp_port *rstp_port; /* RSTP port or null. */ |
46c88433 | 182 | |
55954f6e EJ |
183 | struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */ |
184 | ||
46c88433 EJ |
185 | bool may_enable; /* May be enabled in bonds. */ |
186 | bool is_tunnel; /* Is a tunnel port. */ | |
875ab130 | 187 | enum netdev_pt_mode pt_mode; /* packet_type handling. */ |
46c88433 EJ |
188 | |
189 | struct cfm *cfm; /* CFM handle or null. */ | |
190 | struct bfd *bfd; /* BFD handle or null. */ | |
0477baa9 | 191 | struct lldp *lldp; /* LLDP handle or null. */ |
46c88433 EJ |
192 | }; |
193 | ||
4d0acc70 EJ |
194 | struct xlate_ctx { |
195 | struct xlate_in *xin; | |
196 | struct xlate_out *xout; | |
197 | ||
0506f184 | 198 | struct xlate_cfg *xcfg; |
46c88433 | 199 | const struct xbridge *xbridge; |
4d0acc70 EJ |
200 | |
201 | /* Flow at the last commit. */ | |
202 | struct flow base_flow; | |
203 | ||
204 | /* Tunnel IP destination address as received. This is stored separately | |
205 | * as the base_flow.tunnel is cleared on init to reflect the datapath | |
206 | * behavior. Used to make sure not to send tunneled output to ourselves, | |
207 | * which might lead to an infinite loop. This could happen easily | |
208 | * if a tunnel is marked as 'ip_remote=flow', and the flow does not | |
209 | * actually set the tun_dst field. */ | |
e4d3706c | 210 | struct in6_addr orig_tunnel_ipv6_dst; |
4d0acc70 | 211 | |
84cf3c1f JR |
212 | /* Stack for the push and pop actions. See comment above nx_stack_push() |
213 | * in nx-match.c for info on how the stack is stored. */ | |
4d0acc70 EJ |
214 | struct ofpbuf stack; |
215 | ||
216 | /* The rule that we are currently translating, or NULL. */ | |
217 | struct rule_dpif *rule; | |
218 | ||
49a73e0c BP |
219 | /* Flow translation populates this with wildcards relevant in translation. |
220 | * When 'xin->wc' is nonnull, this is the same pointer. When 'xin->wc' is | |
c0e638aa | 221 | * null, this is a pointer to a temporary buffer. */ |
49a73e0c BP |
222 | struct flow_wildcards *wc; |
223 | ||
1520ef4f BP |
224 | /* Output buffer for datapath actions. When 'xin->odp_actions' is nonnull, |
225 | * this is the same pointer. When 'xin->odp_actions' is null, this points | |
226 | * to a scratch ofpbuf. This allows code to add actions to | |
227 | * 'ctx->odp_actions' without worrying about whether the caller really | |
228 | * wants actions. */ | |
229 | struct ofpbuf *odp_actions; | |
230 | ||
790c5d26 BP |
231 | /* Statistics maintained by xlate_table_action(). |
232 | * | |
2d9b49dd | 233 | * These statistics limit the amount of work that a single flow |
790c5d26 BP |
234 | * translation can perform. The goal of the first of these, 'depth', is |
235 | * primarily to prevent translation from performing an infinite amount of | |
236 | * work. It counts the current depth of nested "resubmit"s (and a few | |
237 | * other activities); when a resubmit returns, it decreases. Resubmits to | |
238 | * tables in strictly monotonically increasing order don't contribute to | |
239 | * 'depth' because they cannot cause a flow translation to take an infinite | |
240 | * amount of time (because the number of tables is finite). Translation | |
241 | * aborts when 'depth' exceeds MAX_DEPTH. | |
242 | * | |
243 | * 'resubmits', on the other hand, prevents flow translation from | |
244 | * performing an extraordinarily large while still finite amount of work. | |
245 | * It counts the total number of resubmits (and a few other activities) | |
246 | * that have been executed. Returning from a resubmit does not affect this | |
247 | * counter. Thus, this limits the amount of work that a particular | |
248 | * translation can perform. Translation aborts when 'resubmits' exceeds | |
249 | * MAX_RESUBMITS (which is much larger than MAX_DEPTH). | |
250 | */ | |
790c5d26 | 251 | int depth; /* Current resubmit nesting depth. */ |
98b07853 | 252 | int resubmits; /* Total number of resubmits. */ |
029ca940 | 253 | bool in_action_set; /* Currently translating action_set, if true. */ |
331c07ac YHW |
254 | bool in_packet_out; /* Currently translating a packet_out msg, if |
255 | * true. */ | |
1fc11c59 JS |
256 | bool pending_encap; /* True when waiting to commit a pending |
257 | * encap action. */ | |
88ec1e0a JS |
258 | bool pending_decap; /* True when waiting to commit a pending |
259 | * decap action. */ | |
1fc11c59 JS |
260 | struct ofpbuf *encap_data; /* May contain a pointer to an ofpbuf with |
261 | * context for the datapath encap action.*/ | |
98b07853 | 262 | |
4d0acc70 | 263 | uint8_t table_id; /* OpenFlow table ID where flow was found. */ |
8b1e5560 JR |
264 | ovs_be64 rule_cookie; /* Cookie of the rule being translated. */ |
265 | uint32_t orig_skb_priority; /* Priority when packet arrived. */ | |
4d0acc70 | 266 | uint32_t sflow_n_outputs; /* Number of output ports. */ |
4e022ec0 | 267 | odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */ |
2031ef97 | 268 | ofp_port_t nf_output_iface; /* Output interface index for NetFlow. */ |
4d0acc70 | 269 | bool exit; /* No further actions should be processed. */ |
3d6151f3 | 270 | mirror_mask_t mirrors; /* Bitmap of associated mirrors. */ |
1356dbd1 | 271 | int mirror_snaplen; /* Max size of a mirror packet in byte. */ |
7fdb60a7 | 272 | |
1d361a81 BP |
273 | /* Freezing Translation |
274 | * ==================== | |
e672ff9b | 275 | * |
1d361a81 BP |
276 | * At some point during translation, the code may recognize the need to halt |
277 | * and checkpoint the translation in a way that it can be restarted again | |
278 | * later. We call the checkpointing process "freezing" and the restarting | |
279 | * process "thawing". | |
e672ff9b | 280 | * |
1d361a81 | 281 | * The use cases for freezing are: |
e672ff9b | 282 | * |
1d361a81 BP |
283 | * - "Recirculation", where the translation process discovers that it |
284 | * doesn't have enough information to complete translation without | |
285 | * actually executing the actions that have already been translated, | |
286 | * which provides the additionally needed information. In these | |
287 | * situations, translation freezes translation and assigns the frozen | |
288 | * data a unique "recirculation ID", which it associates with the data | |
289 | * in a table in userspace (see ofproto-dpif-rid.h). It also adds a | |
290 | * OVS_ACTION_ATTR_RECIRC action specifying that ID to the datapath | |
291 | * actions. When a packet hits that action, the datapath looks its | |
292 | * flow up again using the ID. If there's a miss, it comes back to | |
293 | * userspace, which find the recirculation table entry for the ID, | |
294 | * thaws the associated frozen data, and continues translation from | |
295 | * that point given the additional information that is now known. | |
e672ff9b | 296 | * |
1d361a81 BP |
297 | * The archetypal example is MPLS. As MPLS is implemented in |
298 | * OpenFlow, the protocol that follows the last MPLS label becomes | |
299 | * known only when that label is popped by an OpenFlow action. That | |
300 | * means that Open vSwitch can't extract the headers beyond the MPLS | |
301 | * labels until the pop action is executed. Thus, at that point | |
302 | * translation uses the recirculation process to extract the headers | |
303 | * beyond the MPLS labels. | |
e672ff9b | 304 | * |
1d361a81 BP |
305 | * (OVS also uses OVS_ACTION_ATTR_RECIRC to implement hashing for |
306 | * output to bonds. OVS pre-populates all the datapath flows for bond | |
307 | * output in the datapath, though, which means that the elaborate | |
308 | * process of coming back to userspace for a second round of | |
309 | * translation isn't needed, and so bonds don't follow the above | |
310 | * process.) | |
e672ff9b | 311 | * |
77ab5fd2 BP |
312 | * - "Continuation". A continuation is a way for an OpenFlow controller |
313 | * to interpose on a packet's traversal of the OpenFlow tables. When | |
314 | * the translation process encounters a "controller" action with the | |
315 | * "pause" flag, it freezes translation, serializes the frozen data, | |
316 | * and sends it to an OpenFlow controller. The controller then | |
317 | * examines and possibly modifies the frozen data and eventually sends | |
318 | * it back to the switch, which thaws it and continues translation. | |
e672ff9b | 319 | * |
1d361a81 BP |
320 | * The main problem of freezing translation is preserving state, so that |
321 | * when the translation is thawed later it resumes from where it left off, | |
322 | * without disruption. In particular, actions must be preserved as follows: | |
323 | * | |
324 | * - If we're freezing because an action needed more information, the | |
325 | * action that prompted it. | |
326 | * | |
327 | * - Any actions remaining to be translated within the current flow. | |
328 | * | |
329 | * - If translation was frozen within a NXAST_RESUBMIT, then any actions | |
330 | * following the resubmit action. Resubmit actions can be nested, so | |
331 | * this has to go all the way up the control stack. | |
e672ff9b JR |
332 | * |
333 | * - The OpenFlow 1.1+ action set. | |
334 | * | |
335 | * State that actions and flow table lookups can depend on, such as the | |
336 | * following, must also be preserved: | |
337 | * | |
338 | * - Metadata fields (input port, registers, OF1.1+ metadata, ...). | |
339 | * | |
1d361a81 | 340 | * - The stack used by NXAST_STACK_PUSH and NXAST_STACK_POP actions. |
e672ff9b JR |
341 | * |
342 | * - The table ID and cookie of the flow being translated at each level | |
1d361a81 BP |
343 | * of the control stack, because these can become visible through |
344 | * OFPAT_CONTROLLER actions (and other ways). | |
e672ff9b JR |
345 | * |
346 | * Translation allows for the control of this state preservation via these | |
1d361a81 BP |
347 | * members. When a need to freeze translation is identified, the |
348 | * translation process: | |
e672ff9b | 349 | * |
1d361a81 | 350 | * 1. Sets 'freezing' to true. |
e672ff9b JR |
351 | * |
352 | * 2. Sets 'exit' to true to tell later steps that we're exiting from the | |
353 | * translation process. | |
354 | * | |
1d361a81 BP |
355 | * 3. Adds an OFPACT_UNROLL_XLATE action to 'frozen_actions', and points |
356 | * frozen_actions.header to the action to make it easy to find it later. | |
357 | * This action holds the current table ID and cookie so that they can be | |
358 | * restored during a post-recirculation upcall translation. | |
e672ff9b JR |
359 | * |
360 | * 4. Adds the action that prompted recirculation and any actions following | |
1d361a81 | 361 | * it within the same flow to 'frozen_actions', so that they can be |
8a5fb3b4 | 362 | * executed during a post-recirculation upcall translation. |
e672ff9b JR |
363 | * |
364 | * 5. Returns. | |
365 | * | |
366 | * 6. The action that prompted recirculation might be nested in a stack of | |
367 | * nested "resubmit"s that have actions remaining. Each of these notices | |
1d361a81 BP |
368 | * that we're exiting and freezing and responds by adding more |
369 | * OFPACT_UNROLL_XLATE actions to 'frozen_actions', as necessary, | |
370 | * followed by any actions that were yet unprocessed. | |
e672ff9b | 371 | * |
1d361a81 BP |
372 | * If we're freezing because of recirculation, the caller generates a |
373 | * recirculation ID and associates all the state produced by this process | |
374 | * with it. For post-recirculation upcall translation, the caller passes it | |
375 | * back in for the new translation to execute. The process yielded a set of | |
376 | * ofpacts that can be translated directly, so it is not much of a special | |
377 | * case at that point. | |
e672ff9b | 378 | */ |
1d361a81 | 379 | bool freezing; |
53cc166a JR |
380 | bool recirc_update_dp_hash; /* Generated recirculation will be preceded |
381 | * by datapath HASH action to get an updated | |
382 | * dp_hash after recirculation. */ | |
383 | uint32_t dp_hash_alg; | |
384 | uint32_t dp_hash_basis; | |
1d361a81 | 385 | struct ofpbuf frozen_actions; |
77ab5fd2 | 386 | const struct ofpact_controller *pause; |
e672ff9b | 387 | |
e12ec36b SH |
388 | /* True if a packet was but is no longer MPLS (due to an MPLS pop action). |
389 | * This is a trigger for recirculation in cases where translating an action | |
390 | * or looking up a flow requires access to the fields of the packet after | |
391 | * the MPLS label stack that was originally present. */ | |
392 | bool was_mpls; | |
393 | ||
07659514 JS |
394 | /* True if conntrack has been performed on this packet during processing |
395 | * on the current bridge. This is used to determine whether conntrack | |
1d361a81 | 396 | * state from the datapath should be honored after thawing. */ |
07659514 JS |
397 | bool conntracked; |
398 | ||
9ac0aada JR |
399 | /* Pointer to an embedded NAT action in a conntrack action, or NULL. */ |
400 | struct ofpact_nat *ct_nat_action; | |
401 | ||
7fdb60a7 SH |
402 | /* OpenFlow 1.1+ action set. |
403 | * | |
404 | * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS. | |
405 | * When translation is otherwise complete, ofpacts_execute_action_set() | |
406 | * converts it to a set of "struct ofpact"s that can be translated into | |
ed9c9e3e | 407 | * datapath actions. */ |
c61f3870 | 408 | bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ |
7fdb60a7 | 409 | struct ofpbuf action_set; /* Action set. */ |
fff1b9c0 JR |
410 | |
411 | enum xlate_error error; /* Translation failed. */ | |
4d0acc70 EJ |
412 | }; |
413 | ||
f0fb825a EG |
414 | /* Structure to track VLAN manipulation */ |
415 | struct xvlan_single { | |
416 | uint16_t tpid; | |
417 | uint16_t vid; | |
418 | uint16_t pcp; | |
419 | }; | |
420 | ||
421 | struct xvlan { | |
422 | struct xvlan_single v[FLOW_MAX_VLAN_HEADERS]; | |
423 | }; | |
424 | ||
fff1b9c0 JR |
425 | const char *xlate_strerror(enum xlate_error error) |
426 | { | |
427 | switch (error) { | |
428 | case XLATE_OK: | |
429 | return "OK"; | |
430 | case XLATE_BRIDGE_NOT_FOUND: | |
431 | return "Bridge not found"; | |
432 | case XLATE_RECURSION_TOO_DEEP: | |
433 | return "Recursion too deep"; | |
434 | case XLATE_TOO_MANY_RESUBMITS: | |
435 | return "Too many resubmits"; | |
436 | case XLATE_STACK_TOO_DEEP: | |
437 | return "Stack too deep"; | |
438 | case XLATE_NO_RECIRCULATION_CONTEXT: | |
439 | return "No recirculation context"; | |
440 | case XLATE_RECIRCULATION_CONFLICT: | |
441 | return "Recirculation conflict"; | |
442 | case XLATE_TOO_MANY_MPLS_LABELS: | |
443 | return "Too many MPLS labels"; | |
8d8ab6c2 JG |
444 | case XLATE_INVALID_TUNNEL_METADATA: |
445 | return "Invalid tunnel metadata"; | |
7873e106 ZB |
446 | case XLATE_UNSUPPORTED_PACKET_TYPE: |
447 | return "Unsupported packet type"; | |
a13a0209 AT |
448 | case XLATE_CONGESTION_DROP: |
449 | return "Congestion Drop"; | |
450 | case XLATE_FORWARDING_DISABLED: | |
451 | return "Forwarding is disabled"; | |
452 | case XLATE_MAX: | |
453 | break; | |
fff1b9c0 JR |
454 | } |
455 | return "Unknown error"; | |
456 | } | |
457 | ||
ed9c9e3e | 458 | static void xlate_action_set(struct xlate_ctx *ctx); |
704bb0bf | 459 | static void xlate_commit_actions(struct xlate_ctx *ctx); |
ed9c9e3e | 460 | |
8bdb2bdb | 461 | static void |
48f704f4 AZ |
462 | patch_port_output(struct xlate_ctx *ctx, const struct xport *in_dev, |
463 | struct xport *out_dev); | |
8bdb2bdb | 464 | |
1d741d6d | 465 | static void |
1d361a81 | 466 | ctx_trigger_freeze(struct xlate_ctx *ctx) |
1d741d6d JR |
467 | { |
468 | ctx->exit = true; | |
1d361a81 | 469 | ctx->freezing = true; |
1d741d6d JR |
470 | } |
471 | ||
53cc166a JR |
472 | static void |
473 | ctx_trigger_recirculate_with_hash(struct xlate_ctx *ctx, uint32_t type, | |
474 | uint32_t basis) | |
475 | { | |
476 | ctx->exit = true; | |
477 | ctx->freezing = true; | |
478 | ctx->recirc_update_dp_hash = true; | |
479 | ctx->dp_hash_alg = type; | |
480 | ctx->dp_hash_basis = basis; | |
481 | } | |
482 | ||
1d741d6d | 483 | static bool |
1d361a81 | 484 | ctx_first_frozen_action(const struct xlate_ctx *ctx) |
1d741d6d | 485 | { |
1d361a81 | 486 | return !ctx->frozen_actions.size; |
e672ff9b JR |
487 | } |
488 | ||
3293cb85 | 489 | static void |
1d361a81 | 490 | ctx_cancel_freeze(struct xlate_ctx *ctx) |
3293cb85 | 491 | { |
1d361a81 BP |
492 | if (ctx->freezing) { |
493 | ctx->freezing = false; | |
53cc166a | 494 | ctx->recirc_update_dp_hash = false; |
1d361a81 BP |
495 | ofpbuf_clear(&ctx->frozen_actions); |
496 | ctx->frozen_actions.header = NULL; | |
5b34f8fc | 497 | ctx->pause = NULL; |
3293cb85 BP |
498 | } |
499 | } | |
500 | ||
77ab5fd2 | 501 | static void finish_freezing(struct xlate_ctx *ctx); |
e672ff9b | 502 | |
9583bc14 EJ |
503 | /* A controller may use OFPP_NONE as the ingress port to indicate that |
504 | * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for | |
505 | * when an input bundle is needed for validation (e.g., mirroring or | |
506 | * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have | |
3548d242 BP |
507 | * any 'port' structs, so care must be taken when dealing with it. */ |
508 | static struct xbundle ofpp_none_bundle = { | |
509 | .name = "OFPP_NONE", | |
510 | .vlan_mode = PORT_VLAN_TRUNK | |
511 | }; | |
9583bc14 | 512 | |
55954f6e EJ |
513 | /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from |
514 | * 'priority' (the datapath's term for QoS queue) to the dscp bits which all | |
515 | * traffic egressing the 'ofport' with that priority should be marked with. */ | |
516 | struct skb_priority_to_dscp { | |
517 | struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */ | |
518 | uint32_t skb_priority; /* Priority of this queue (see struct flow). */ | |
519 | ||
520 | uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */ | |
521 | }; | |
522 | ||
84f0f298 RW |
523 | /* Xlate config contains hash maps of all bridges, bundles and ports. |
524 | * Xcfgp contains the pointer to the current xlate configuration. | |
525 | * When the main thread needs to change the configuration, it copies xcfgp to | |
526 | * new_xcfg and edits new_xcfg. This enables the use of RCU locking which | |
527 | * does not block handler and revalidator threads. */ | |
528 | struct xlate_cfg { | |
529 | struct hmap xbridges; | |
530 | struct hmap xbundles; | |
531 | struct hmap xports; | |
c3594cc3 | 532 | struct hmap xports_uuid; |
84f0f298 | 533 | }; |
b1b72f2d | 534 | static OVSRCU_TYPE(struct xlate_cfg *) xcfgp = OVSRCU_INITIALIZER(NULL); |
f439f23b | 535 | static struct xlate_cfg *new_xcfg = NULL; |
46c88433 | 536 | |
96c3a6e5 | 537 | typedef void xlate_actions_handler(const struct ofpact *, size_t ofpacts_len, |
f5634764 | 538 | struct xlate_ctx *, bool, bool); |
46c88433 | 539 | static bool may_receive(const struct xport *, struct xlate_ctx *); |
9583bc14 | 540 | static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
f5634764 | 541 | struct xlate_ctx *, bool, bool); |
96c3a6e5 | 542 | static void clone_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
f5634764 | 543 | struct xlate_ctx *, bool, bool); |
adcf00ba | 544 | static void xlate_normal(struct xlate_ctx *); |
efc93e66 BP |
545 | static void xlate_normal_flood(struct xlate_ctx *ct, |
546 | struct xbundle *in_xbundle, struct xvlan *); | |
6d328fa2 SH |
547 | static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port, |
548 | uint8_t table_id, bool may_packet_in, | |
feee58b9 | 549 | bool honor_table_miss, bool with_ct_orig, |
96c3a6e5 AZ |
550 | bool is_last_action, xlate_actions_handler *); |
551 | ||
2d9b49dd BP |
552 | static bool input_vid_is_valid(const struct xlate_ctx *, |
553 | uint16_t vid, struct xbundle *); | |
f0fb825a EG |
554 | static void xvlan_copy(struct xvlan *dst, const struct xvlan *src); |
555 | static void xvlan_pop(struct xvlan *src); | |
fed8962a | 556 | static void xvlan_push_uninit(struct xvlan *src); |
f0fb825a | 557 | static void xvlan_extract(const struct flow *, struct xvlan *); |
ffbe41db EB |
558 | static void xvlan_put(struct flow *, const struct xvlan *, |
559 | enum port_priority_tags_mode); | |
f0fb825a EG |
560 | static void xvlan_input_translate(const struct xbundle *, |
561 | const struct xvlan *in, | |
562 | struct xvlan *xvlan); | |
563 | static void xvlan_output_translate(const struct xbundle *, | |
564 | const struct xvlan *xvlan, | |
565 | struct xvlan *out); | |
46c88433 | 566 | static void output_normal(struct xlate_ctx *, const struct xbundle *, |
f0fb825a | 567 | const struct xvlan *); |
e93ef1c7 JR |
568 | |
569 | /* Optional bond recirculation parameter to compose_output_action(). */ | |
570 | struct xlate_bond_recirc { | |
571 | uint32_t recirc_id; /* !0 Use recirculation instead of output. */ | |
572 | uint8_t hash_alg; /* !0 Compute hash for recirc before. */ | |
573 | uint32_t hash_basis; /* Compute hash for recirc before. */ | |
574 | }; | |
575 | ||
576 | static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port, | |
feee58b9 | 577 | const struct xlate_bond_recirc *xr, |
11938578 | 578 | bool is_last_action, bool truncate); |
9583bc14 | 579 | |
84f0f298 RW |
580 | static struct xbridge *xbridge_lookup(struct xlate_cfg *, |
581 | const struct ofproto_dpif *); | |
290835f9 BP |
582 | static struct xbridge *xbridge_lookup_by_uuid(struct xlate_cfg *, |
583 | const struct uuid *); | |
84f0f298 RW |
584 | static struct xbundle *xbundle_lookup(struct xlate_cfg *, |
585 | const struct ofbundle *); | |
586 | static struct xport *xport_lookup(struct xlate_cfg *, | |
587 | const struct ofport_dpif *); | |
00135b86 ZB |
588 | static struct xport *xport_lookup_by_uuid(struct xlate_cfg *, |
589 | const struct uuid *); | |
46c88433 | 590 | static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port); |
55954f6e EJ |
591 | static struct skb_priority_to_dscp *get_skb_priority(const struct xport *, |
592 | uint32_t skb_priority); | |
593 | static void clear_skb_priorities(struct xport *); | |
16194afd | 594 | static size_t count_skb_priorities(const struct xport *); |
55954f6e EJ |
595 | static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, |
596 | uint8_t *dscp); | |
46c88433 | 597 | |
84f0f298 RW |
598 | static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *); |
599 | static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *); | |
600 | static void xlate_xport_init(struct xlate_cfg *, struct xport *); | |
9efd308e | 601 | static void xlate_xbridge_set(struct xbridge *, struct dpif *, |
9efd308e DV |
602 | const struct mac_learning *, struct stp *, |
603 | struct rstp *, const struct mcast_snooping *, | |
604 | const struct mbridge *, | |
605 | const struct dpif_sflow *, | |
606 | const struct dpif_ipfix *, | |
2f47cdf4 | 607 | const struct netflow *, |
84f0f298 | 608 | bool forward_bpdu, bool has_in_band, |
83c2757b ZB |
609 | const struct dpif_backer_support *, |
610 | const struct xbridge_addr *); | |
84f0f298 | 611 | static void xlate_xbundle_set(struct xbundle *xbundle, |
f0fb825a | 612 | enum port_vlan_mode vlan_mode, |
fed8962a EG |
613 | uint16_t qinq_ethtype, int vlan, |
614 | unsigned long *trunks, unsigned long *cvlans, | |
88f52d7f | 615 | enum port_priority_tags_mode, |
84f0f298 | 616 | const struct bond *bond, const struct lacp *lacp, |
c005f976 | 617 | bool floodable, bool protected); |
84f0f298 RW |
618 | static void xlate_xport_set(struct xport *xport, odp_port_t odp_port, |
619 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 DF |
620 | const struct bfd *bfd, const struct lldp *lldp, |
621 | int stp_port_no, const struct rstp_port *rstp_port, | |
84f0f298 RW |
622 | enum ofputil_port_config config, |
623 | enum ofputil_port_state state, bool is_tunnel, | |
624 | bool may_enable); | |
625 | static void xlate_xbridge_remove(struct xlate_cfg *, struct xbridge *); | |
626 | static void xlate_xbundle_remove(struct xlate_cfg *, struct xbundle *); | |
627 | static void xlate_xport_remove(struct xlate_cfg *, struct xport *); | |
628 | static void xlate_xbridge_copy(struct xbridge *); | |
629 | static void xlate_xbundle_copy(struct xbridge *, struct xbundle *); | |
630 | static void xlate_xport_copy(struct xbridge *, struct xbundle *, | |
631 | struct xport *); | |
632 | static void xlate_xcfg_free(struct xlate_cfg *); | |
2d9b49dd BP |
633 | \f |
634 | /* Tracing helpers. */ | |
635 | ||
636 | /* If tracing is enabled in 'ctx', creates a new trace node and appends it to | |
637 | * the list of nodes maintained in ctx->xin. The new node has type 'type' and | |
638 | * its text is created from 'format' by treating it as a printf format string. | |
639 | * Returns the list of nodes embedded within the new trace node; ordinarily, | |
640 | * the calleer can ignore this, but it is useful if the caller needs to nest | |
641 | * more trace nodes within the new node. | |
642 | * | |
643 | * If tracing is not enabled, does nothing and returns NULL. */ | |
644 | static struct ovs_list * OVS_PRINTF_FORMAT(3, 4) | |
645 | xlate_report(const struct xlate_ctx *ctx, enum oftrace_node_type type, | |
646 | const char *format, ...) | |
34dd0d78 | 647 | { |
2d9b49dd BP |
648 | struct ovs_list *subtrace = NULL; |
649 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
c1b3756c | 650 | va_list args; |
c1b3756c | 651 | va_start(args, format); |
2d9b49dd BP |
652 | char *text = xvasprintf(format, args); |
653 | subtrace = &oftrace_report(ctx->xin->trace, type, text)->subs; | |
c1b3756c | 654 | va_end(args); |
2d9b49dd | 655 | free(text); |
34dd0d78 | 656 | } |
2d9b49dd | 657 | return subtrace; |
34dd0d78 | 658 | } |
84f0f298 | 659 | |
2d9b49dd BP |
660 | /* This is like xlate_report() for errors that are serious enough that we |
661 | * should log them even if we are not tracing. */ | |
662 | static void OVS_PRINTF_FORMAT(2, 3) | |
663 | xlate_report_error(const struct xlate_ctx *ctx, const char *format, ...) | |
664 | { | |
665 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
666 | if (!OVS_UNLIKELY(ctx->xin->trace) | |
667 | && (!ctx->xin->packet || VLOG_DROP_WARN(&rl))) { | |
668 | return; | |
669 | } | |
670 | ||
671 | struct ds s = DS_EMPTY_INITIALIZER; | |
672 | va_list args; | |
673 | va_start(args, format); | |
674 | ds_put_format_valist(&s, format, args); | |
675 | va_end(args); | |
676 | ||
677 | if (ctx->xin->trace) { | |
678 | oftrace_report(ctx->xin->trace, OFT_ERROR, ds_cstr(&s)); | |
679 | } else { | |
a4c3463d BP |
680 | ds_put_format(&s, " on bridge %s while processing ", |
681 | ctx->xbridge->name); | |
50f96b10 | 682 | flow_format(&s, &ctx->base_flow, NULL); |
2d9b49dd BP |
683 | VLOG_WARN("%s", ds_cstr(&s)); |
684 | } | |
685 | ds_destroy(&s); | |
686 | } | |
687 | ||
f89547e2 ZW |
688 | /* This is like xlate_report() for messages that should be logged |
689 | at the info level (even when not tracing). */ | |
690 | static void OVS_PRINTF_FORMAT(2, 3) | |
691 | xlate_report_info(const struct xlate_ctx *ctx, const char *format, ...) | |
692 | { | |
693 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
694 | if (!OVS_UNLIKELY(ctx->xin->trace) | |
695 | && (!ctx->xin->packet || VLOG_DROP_INFO(&rl))) { | |
696 | return; | |
697 | } | |
698 | ||
699 | struct ds s = DS_EMPTY_INITIALIZER; | |
700 | va_list args; | |
701 | va_start(args, format); | |
702 | ds_put_format_valist(&s, format, args); | |
703 | va_end(args); | |
704 | ||
705 | if (ctx->xin->trace) { | |
706 | oftrace_report(ctx->xin->trace, OFT_WARN, ds_cstr(&s)); | |
707 | } else { | |
708 | ds_put_format(&s, " on bridge %s while processing ", | |
709 | ctx->xbridge->name); | |
710 | flow_format(&s, &ctx->base_flow, NULL); | |
711 | VLOG_INFO("%s", ds_cstr(&s)); | |
712 | } | |
713 | ds_destroy(&s); | |
714 | } | |
715 | ||
2d9b49dd BP |
716 | /* This is like xlate_report() for messages that should be logged at debug |
717 | * level (even if we are not tracing) because they can be valuable for | |
718 | * debugging. */ | |
719 | static void OVS_PRINTF_FORMAT(3, 4) | |
720 | xlate_report_debug(const struct xlate_ctx *ctx, enum oftrace_node_type type, | |
721 | const char *format, ...) | |
722 | { | |
723 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
724 | if (!OVS_UNLIKELY(ctx->xin->trace) | |
725 | && (!ctx->xin->packet || VLOG_DROP_DBG(&rl))) { | |
726 | return; | |
727 | } | |
728 | ||
729 | struct ds s = DS_EMPTY_INITIALIZER; | |
730 | va_list args; | |
731 | va_start(args, format); | |
732 | ds_put_format_valist(&s, format, args); | |
733 | va_end(args); | |
fff1b9c0 | 734 | |
2d9b49dd BP |
735 | if (ctx->xin->trace) { |
736 | oftrace_report(ctx->xin->trace, type, ds_cstr(&s)); | |
737 | } else { | |
738 | VLOG_DBG("bridge %s: %s", ctx->xbridge->name, ds_cstr(&s)); | |
739 | } | |
740 | ds_destroy(&s); | |
741 | } | |
fff1b9c0 | 742 | |
2d9b49dd BP |
743 | /* If tracing is enabled in 'ctx', appends a node of the given 'type' to the |
744 | * trace, whose text is 'title' followed by a formatted version of the | |
745 | * 'ofpacts_len' OpenFlow actions in 'ofpacts'. | |
746 | * | |
747 | * If tracing is not enabled, does nothing. */ | |
748 | static void | |
749 | xlate_report_actions(const struct xlate_ctx *ctx, enum oftrace_node_type type, | |
750 | const char *title, | |
d6bef3cc BP |
751 | const struct ofpact *ofpacts, size_t ofpacts_len) |
752 | { | |
2d9b49dd | 753 | if (OVS_UNLIKELY(ctx->xin->trace)) { |
d6bef3cc | 754 | struct ds s = DS_EMPTY_INITIALIZER; |
2d9b49dd | 755 | ds_put_format(&s, "%s: ", title); |
efefbcae BP |
756 | struct ofpact_format_params fp = { .s = &s }; |
757 | ofpacts_format(ofpacts, ofpacts_len, &fp); | |
2d9b49dd | 758 | oftrace_report(ctx->xin->trace, type, ds_cstr(&s)); |
d6bef3cc BP |
759 | ds_destroy(&s); |
760 | } | |
761 | } | |
762 | ||
2d9b49dd BP |
763 | /* If tracing is enabled in 'ctx', appends a node of type OFT_DETAIL to the |
764 | * trace, whose the message is a formatted version of the OpenFlow action set. | |
765 | * 'verb' should be "was" or "is", depending on whether the action set reported | |
766 | * is the new action set or the old one. | |
767 | * | |
768 | * If tracing is not enabled, does nothing. */ | |
769 | static void | |
770 | xlate_report_action_set(const struct xlate_ctx *ctx, const char *verb) | |
771 | { | |
772 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
773 | struct ofpbuf action_list; | |
774 | ofpbuf_init(&action_list, 0); | |
775 | ofpacts_execute_action_set(&action_list, &ctx->action_set); | |
776 | if (action_list.size) { | |
777 | struct ds s = DS_EMPTY_INITIALIZER; | |
efefbcae BP |
778 | struct ofpact_format_params fp = { .s = &s }; |
779 | ofpacts_format(action_list.data, action_list.size, &fp); | |
2d9b49dd BP |
780 | xlate_report(ctx, OFT_DETAIL, "action set %s: %s", |
781 | verb, ds_cstr(&s)); | |
782 | ds_destroy(&s); | |
783 | } else { | |
784 | xlate_report(ctx, OFT_DETAIL, "action set %s empty", verb); | |
785 | } | |
786 | ofpbuf_uninit(&action_list); | |
787 | } | |
788 | } | |
789 | ||
790 | ||
791 | /* If tracing is enabled in 'ctx', appends a node representing 'rule' (in | |
792 | * OpenFlow table 'table_id') to the trace and makes this node the parent for | |
793 | * future trace nodes. The caller should save ctx->xin->trace before calling | |
794 | * this function, then after tracing all of the activities under the table, | |
795 | * restore its previous value. | |
796 | * | |
797 | * If tracing is not enabled, does nothing. */ | |
798 | static void | |
799 | xlate_report_table(const struct xlate_ctx *ctx, struct rule_dpif *rule, | |
800 | uint8_t table_id) | |
801 | { | |
802 | if (OVS_LIKELY(!ctx->xin->trace)) { | |
803 | return; | |
804 | } | |
805 | ||
806 | struct ds s = DS_EMPTY_INITIALIZER; | |
807 | ds_put_format(&s, "%2d. ", table_id); | |
808 | if (rule == ctx->xin->ofproto->miss_rule) { | |
809 | ds_put_cstr(&s, "No match, and a \"packet-in\" is called for."); | |
810 | } else if (rule == ctx->xin->ofproto->no_packet_in_rule) { | |
811 | ds_put_cstr(&s, "No match."); | |
812 | } else if (rule == ctx->xin->ofproto->drop_frags_rule) { | |
813 | ds_put_cstr(&s, "Packets are IP fragments and " | |
814 | "the fragment handling mode is \"drop\"."); | |
815 | } else { | |
816 | minimatch_format(&rule->up.cr.match, | |
817 | ofproto_get_tun_tab(&ctx->xin->ofproto->up), | |
50f96b10 | 818 | NULL, &s, OFP_DEFAULT_PRIORITY); |
2d9b49dd BP |
819 | if (ds_last(&s) != ' ') { |
820 | ds_put_cstr(&s, ", "); | |
821 | } | |
822 | ds_put_format(&s, "priority %d", rule->up.cr.priority); | |
823 | if (rule->up.flow_cookie) { | |
824 | ds_put_format(&s, ", cookie %#"PRIx64, | |
825 | ntohll(rule->up.flow_cookie)); | |
826 | } | |
827 | } | |
828 | ctx->xin->trace = &oftrace_report(ctx->xin->trace, OFT_TABLE, | |
829 | ds_cstr(&s))->subs; | |
830 | ds_destroy(&s); | |
831 | } | |
832 | ||
833 | /* If tracing is enabled in 'ctx', adds an OFT_DETAIL trace node to 'ctx' | |
834 | * reporting the value of subfield 'sf'. | |
835 | * | |
836 | * If tracing is not enabled, does nothing. */ | |
837 | static void | |
838 | xlate_report_subfield(const struct xlate_ctx *ctx, | |
839 | const struct mf_subfield *sf) | |
840 | { | |
841 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
842 | struct ds s = DS_EMPTY_INITIALIZER; | |
843 | mf_format_subfield(sf, &s); | |
844 | ds_put_cstr(&s, " is now "); | |
845 | ||
846 | if (sf->ofs == 0 && sf->n_bits >= sf->field->n_bits) { | |
847 | union mf_value value; | |
848 | mf_get_value(sf->field, &ctx->xin->flow, &value); | |
50f96b10 | 849 | mf_format(sf->field, &value, NULL, NULL, &s); |
2d9b49dd BP |
850 | } else { |
851 | union mf_subvalue cst; | |
852 | mf_read_subfield(sf, &ctx->xin->flow, &cst); | |
853 | ds_put_hex(&s, &cst, sizeof cst); | |
854 | } | |
855 | ||
856 | xlate_report(ctx, OFT_DETAIL, "%s", ds_cstr(&s)); | |
857 | ||
858 | ds_destroy(&s); | |
859 | } | |
860 | } | |
861 | \f | |
84f0f298 RW |
862 | static void |
863 | xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
864 | { | |
417e7e66 | 865 | ovs_list_init(&xbridge->xbundles); |
84f0f298 RW |
866 | hmap_init(&xbridge->xports); |
867 | hmap_insert(&xcfg->xbridges, &xbridge->hmap_node, | |
868 | hash_pointer(xbridge->ofproto, 0)); | |
869 | } | |
870 | ||
871 | static void | |
872 | xlate_xbundle_init(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
873 | { | |
417e7e66 BW |
874 | ovs_list_init(&xbundle->xports); |
875 | ovs_list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); | |
84f0f298 RW |
876 | hmap_insert(&xcfg->xbundles, &xbundle->hmap_node, |
877 | hash_pointer(xbundle->ofbundle, 0)); | |
878 | } | |
879 | ||
880 | static void | |
881 | xlate_xport_init(struct xlate_cfg *xcfg, struct xport *xport) | |
882 | { | |
883 | hmap_init(&xport->skb_priorities); | |
884 | hmap_insert(&xcfg->xports, &xport->hmap_node, | |
885 | hash_pointer(xport->ofport, 0)); | |
886 | hmap_insert(&xport->xbridge->xports, &xport->ofp_node, | |
887 | hash_ofp_port(xport->ofp_port)); | |
c3594cc3 ZB |
888 | hmap_insert(&xcfg->xports_uuid, &xport->uuid_node, |
889 | uuid_hash(&xport->uuid)); | |
84f0f298 RW |
890 | } |
891 | ||
83c2757b ZB |
892 | static struct xbridge_addr * |
893 | xbridge_addr_create(struct xbridge *xbridge) | |
894 | { | |
895 | struct xbridge_addr *xbridge_addr = xbridge->addr; | |
896 | struct in6_addr *addr = NULL, *mask = NULL; | |
897 | struct netdev *dev; | |
898 | int err, n_addr = 0; | |
899 | ||
900 | err = netdev_open(xbridge->name, NULL, &dev); | |
901 | if (!err) { | |
902 | err = netdev_get_addr_list(dev, &addr, &mask, &n_addr); | |
903 | if (!err) { | |
904 | if (!xbridge->addr || | |
905 | n_addr != xbridge->addr->n_addr || | |
906 | (xbridge->addr->addr && memcmp(addr, xbridge->addr->addr, | |
907 | sizeof(*addr) * n_addr))) { | |
908 | xbridge_addr = xzalloc(sizeof *xbridge_addr); | |
909 | xbridge_addr->addr = addr; | |
910 | xbridge_addr->n_addr = n_addr; | |
911 | ovs_refcount_init(&xbridge_addr->ref_cnt); | |
912 | } else { | |
913 | free(addr); | |
914 | } | |
915 | free(mask); | |
916 | } | |
917 | netdev_close(dev); | |
918 | } | |
919 | ||
920 | return xbridge_addr; | |
921 | } | |
922 | ||
923 | static struct xbridge_addr * | |
924 | xbridge_addr_ref(const struct xbridge_addr *addr_) | |
925 | { | |
926 | struct xbridge_addr *addr = CONST_CAST(struct xbridge_addr *, addr_); | |
927 | if (addr) { | |
928 | ovs_refcount_ref(&addr->ref_cnt); | |
929 | } | |
930 | return addr; | |
931 | } | |
932 | ||
933 | static void | |
934 | xbridge_addr_unref(struct xbridge_addr *addr) | |
935 | { | |
936 | if (addr && ovs_refcount_unref_relaxed(&addr->ref_cnt) == 1) { | |
937 | free(addr->addr); | |
938 | free(addr); | |
939 | } | |
940 | } | |
941 | ||
84f0f298 RW |
942 | static void |
943 | xlate_xbridge_set(struct xbridge *xbridge, | |
944 | struct dpif *dpif, | |
ec89fc6f | 945 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 946 | struct rstp *rstp, const struct mcast_snooping *ms, |
ec89fc6f | 947 | const struct mbridge *mbridge, |
46c88433 | 948 | const struct dpif_sflow *sflow, |
ce3955be | 949 | const struct dpif_ipfix *ipfix, |
2f47cdf4 | 950 | const struct netflow *netflow, |
4b97b70d | 951 | bool forward_bpdu, bool has_in_band, |
83c2757b ZB |
952 | const struct dpif_backer_support *support, |
953 | const struct xbridge_addr *addr) | |
46c88433 | 954 | { |
46c88433 EJ |
955 | if (xbridge->ml != ml) { |
956 | mac_learning_unref(xbridge->ml); | |
957 | xbridge->ml = mac_learning_ref(ml); | |
958 | } | |
959 | ||
6d95c4e8 FL |
960 | if (xbridge->ms != ms) { |
961 | mcast_snooping_unref(xbridge->ms); | |
962 | xbridge->ms = mcast_snooping_ref(ms); | |
963 | } | |
964 | ||
46c88433 EJ |
965 | if (xbridge->mbridge != mbridge) { |
966 | mbridge_unref(xbridge->mbridge); | |
967 | xbridge->mbridge = mbridge_ref(mbridge); | |
968 | } | |
969 | ||
970 | if (xbridge->sflow != sflow) { | |
971 | dpif_sflow_unref(xbridge->sflow); | |
972 | xbridge->sflow = dpif_sflow_ref(sflow); | |
973 | } | |
974 | ||
975 | if (xbridge->ipfix != ipfix) { | |
976 | dpif_ipfix_unref(xbridge->ipfix); | |
977 | xbridge->ipfix = dpif_ipfix_ref(ipfix); | |
978 | } | |
979 | ||
9d189a50 EJ |
980 | if (xbridge->stp != stp) { |
981 | stp_unref(xbridge->stp); | |
982 | xbridge->stp = stp_ref(stp); | |
983 | } | |
984 | ||
9efd308e DV |
985 | if (xbridge->rstp != rstp) { |
986 | rstp_unref(xbridge->rstp); | |
987 | xbridge->rstp = rstp_ref(rstp); | |
988 | } | |
989 | ||
ce3955be EJ |
990 | if (xbridge->netflow != netflow) { |
991 | netflow_unref(xbridge->netflow); | |
992 | xbridge->netflow = netflow_ref(netflow); | |
993 | } | |
994 | ||
83c2757b ZB |
995 | if (xbridge->addr != addr) { |
996 | xbridge_addr_unref(xbridge->addr); | |
997 | xbridge->addr = xbridge_addr_ref(addr); | |
998 | } | |
999 | ||
89a8a7f0 | 1000 | xbridge->dpif = dpif; |
46c88433 EJ |
1001 | xbridge->forward_bpdu = forward_bpdu; |
1002 | xbridge->has_in_band = has_in_band; | |
b440dd8c | 1003 | xbridge->support = *support; |
46c88433 EJ |
1004 | } |
1005 | ||
84f0f298 RW |
1006 | static void |
1007 | xlate_xbundle_set(struct xbundle *xbundle, | |
fed8962a EG |
1008 | enum port_vlan_mode vlan_mode, uint16_t qinq_ethtype, |
1009 | int vlan, unsigned long *trunks, unsigned long *cvlans, | |
88f52d7f | 1010 | enum port_priority_tags_mode use_priority_tags, |
84f0f298 | 1011 | const struct bond *bond, const struct lacp *lacp, |
c005f976 | 1012 | bool floodable, bool protected) |
84f0f298 RW |
1013 | { |
1014 | ovs_assert(xbundle->xbridge); | |
1015 | ||
1016 | xbundle->vlan_mode = vlan_mode; | |
fed8962a | 1017 | xbundle->qinq_ethtype = qinq_ethtype; |
84f0f298 RW |
1018 | xbundle->vlan = vlan; |
1019 | xbundle->trunks = trunks; | |
fed8962a | 1020 | xbundle->cvlans = cvlans; |
84f0f298 RW |
1021 | xbundle->use_priority_tags = use_priority_tags; |
1022 | xbundle->floodable = floodable; | |
c005f976 | 1023 | xbundle->protected = protected; |
84f0f298 RW |
1024 | |
1025 | if (xbundle->bond != bond) { | |
1026 | bond_unref(xbundle->bond); | |
1027 | xbundle->bond = bond_ref(bond); | |
1028 | } | |
1029 | ||
1030 | if (xbundle->lacp != lacp) { | |
1031 | lacp_unref(xbundle->lacp); | |
1032 | xbundle->lacp = lacp_ref(lacp); | |
1033 | } | |
1034 | } | |
1035 | ||
1036 | static void | |
1037 | xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
1038 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 | 1039 | const struct bfd *bfd, const struct lldp *lldp, int stp_port_no, |
f025bcb7 | 1040 | const struct rstp_port* rstp_port, |
84f0f298 RW |
1041 | enum ofputil_port_config config, enum ofputil_port_state state, |
1042 | bool is_tunnel, bool may_enable) | |
1043 | { | |
1044 | xport->config = config; | |
1045 | xport->state = state; | |
1046 | xport->stp_port_no = stp_port_no; | |
1047 | xport->is_tunnel = is_tunnel; | |
875ab130 | 1048 | xport->pt_mode = netdev_get_pt_mode(netdev); |
84f0f298 RW |
1049 | xport->may_enable = may_enable; |
1050 | xport->odp_port = odp_port; | |
1051 | ||
f025bcb7 JR |
1052 | if (xport->rstp_port != rstp_port) { |
1053 | rstp_port_unref(xport->rstp_port); | |
1054 | xport->rstp_port = rstp_port_ref(rstp_port); | |
1055 | } | |
1056 | ||
84f0f298 RW |
1057 | if (xport->cfm != cfm) { |
1058 | cfm_unref(xport->cfm); | |
1059 | xport->cfm = cfm_ref(cfm); | |
1060 | } | |
1061 | ||
1062 | if (xport->bfd != bfd) { | |
1063 | bfd_unref(xport->bfd); | |
1064 | xport->bfd = bfd_ref(bfd); | |
1065 | } | |
1066 | ||
0477baa9 DF |
1067 | if (xport->lldp != lldp) { |
1068 | lldp_unref(xport->lldp); | |
1069 | xport->lldp = lldp_ref(lldp); | |
1070 | } | |
1071 | ||
84f0f298 RW |
1072 | if (xport->netdev != netdev) { |
1073 | netdev_close(xport->netdev); | |
1074 | xport->netdev = netdev_ref(netdev); | |
1075 | } | |
1076 | } | |
1077 | ||
1078 | static void | |
1079 | xlate_xbridge_copy(struct xbridge *xbridge) | |
1080 | { | |
1081 | struct xbundle *xbundle; | |
1082 | struct xport *xport; | |
1083 | struct xbridge *new_xbridge = xzalloc(sizeof *xbridge); | |
1084 | new_xbridge->ofproto = xbridge->ofproto; | |
1085 | new_xbridge->name = xstrdup(xbridge->name); | |
1086 | xlate_xbridge_init(new_xcfg, new_xbridge); | |
1087 | ||
1088 | xlate_xbridge_set(new_xbridge, | |
34dd0d78 | 1089 | xbridge->dpif, xbridge->ml, xbridge->stp, |
9efd308e DV |
1090 | xbridge->rstp, xbridge->ms, xbridge->mbridge, |
1091 | xbridge->sflow, xbridge->ipfix, xbridge->netflow, | |
b440dd8c | 1092 | xbridge->forward_bpdu, xbridge->has_in_band, |
83c2757b | 1093 | &xbridge->support, xbridge->addr); |
84f0f298 RW |
1094 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
1095 | xlate_xbundle_copy(new_xbridge, xbundle); | |
1096 | } | |
1097 | ||
1098 | /* Copy xports which are not part of a xbundle */ | |
1099 | HMAP_FOR_EACH (xport, ofp_node, &xbridge->xports) { | |
1100 | if (!xport->xbundle) { | |
1101 | xlate_xport_copy(new_xbridge, NULL, xport); | |
1102 | } | |
1103 | } | |
1104 | } | |
1105 | ||
1106 | static void | |
1107 | xlate_xbundle_copy(struct xbridge *xbridge, struct xbundle *xbundle) | |
1108 | { | |
1109 | struct xport *xport; | |
1110 | struct xbundle *new_xbundle = xzalloc(sizeof *xbundle); | |
1111 | new_xbundle->ofbundle = xbundle->ofbundle; | |
1112 | new_xbundle->xbridge = xbridge; | |
1113 | new_xbundle->name = xstrdup(xbundle->name); | |
1114 | xlate_xbundle_init(new_xcfg, new_xbundle); | |
1115 | ||
fed8962a EG |
1116 | xlate_xbundle_set(new_xbundle, xbundle->vlan_mode, xbundle->qinq_ethtype, |
1117 | xbundle->vlan, xbundle->trunks, xbundle->cvlans, | |
84f0f298 | 1118 | xbundle->use_priority_tags, xbundle->bond, xbundle->lacp, |
c005f976 | 1119 | xbundle->floodable, xbundle->protected); |
84f0f298 RW |
1120 | LIST_FOR_EACH (xport, bundle_node, &xbundle->xports) { |
1121 | xlate_xport_copy(xbridge, new_xbundle, xport); | |
1122 | } | |
1123 | } | |
1124 | ||
1125 | static void | |
1126 | xlate_xport_copy(struct xbridge *xbridge, struct xbundle *xbundle, | |
1127 | struct xport *xport) | |
1128 | { | |
1129 | struct skb_priority_to_dscp *pdscp, *new_pdscp; | |
1130 | struct xport *new_xport = xzalloc(sizeof *xport); | |
1131 | new_xport->ofport = xport->ofport; | |
1132 | new_xport->ofp_port = xport->ofp_port; | |
1133 | new_xport->xbridge = xbridge; | |
c3594cc3 | 1134 | new_xport->uuid = xport->uuid; |
84f0f298 RW |
1135 | xlate_xport_init(new_xcfg, new_xport); |
1136 | ||
1137 | xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm, | |
0477baa9 DF |
1138 | xport->bfd, xport->lldp, xport->stp_port_no, |
1139 | xport->rstp_port, xport->config, xport->state, | |
1140 | xport->is_tunnel, xport->may_enable); | |
84f0f298 RW |
1141 | |
1142 | if (xport->peer) { | |
1143 | struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport); | |
1144 | if (peer) { | |
1145 | new_xport->peer = peer; | |
1146 | new_xport->peer->peer = new_xport; | |
1147 | } | |
1148 | } | |
1149 | ||
1150 | if (xbundle) { | |
1151 | new_xport->xbundle = xbundle; | |
417e7e66 | 1152 | ovs_list_insert(&new_xport->xbundle->xports, &new_xport->bundle_node); |
84f0f298 RW |
1153 | } |
1154 | ||
1155 | HMAP_FOR_EACH (pdscp, hmap_node, &xport->skb_priorities) { | |
1156 | new_pdscp = xmalloc(sizeof *pdscp); | |
1157 | new_pdscp->skb_priority = pdscp->skb_priority; | |
1158 | new_pdscp->dscp = pdscp->dscp; | |
1159 | hmap_insert(&new_xport->skb_priorities, &new_pdscp->hmap_node, | |
1160 | hash_int(new_pdscp->skb_priority, 0)); | |
1161 | } | |
1162 | } | |
1163 | ||
1164 | /* Sets the current xlate configuration to new_xcfg and frees the old xlate | |
1165 | * configuration in xcfgp. | |
1166 | * | |
1167 | * This needs to be called after editing the xlate configuration. | |
1168 | * | |
1169 | * Functions that edit the new xlate configuration are | |
6cd20a22 | 1170 | * xlate_<ofproto/bundle/ofport>_set and xlate_<ofproto/bundle/ofport>_remove. |
84f0f298 RW |
1171 | * |
1172 | * A sample workflow: | |
1173 | * | |
586cd310 BP |
1174 | * xlate_txn_start(); |
1175 | * ... | |
1176 | * edit_xlate_configuration(); | |
1177 | * ... | |
1178 | * xlate_txn_commit(); | |
1179 | * | |
1180 | * The ovsrcu_synchronize() call here also ensures that the upcall threads | |
1181 | * retain no references to anything in the previous configuration. | |
1182 | */ | |
46c88433 | 1183 | void |
84f0f298 RW |
1184 | xlate_txn_commit(void) |
1185 | { | |
1186 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
1187 | ||
1188 | ovsrcu_set(&xcfgp, new_xcfg); | |
40a9c4c2 AW |
1189 | ovsrcu_synchronize(); |
1190 | xlate_xcfg_free(xcfg); | |
84f0f298 RW |
1191 | new_xcfg = NULL; |
1192 | } | |
1193 | ||
1194 | /* Copies the current xlate configuration in xcfgp to new_xcfg. | |
1195 | * | |
1196 | * This needs to be called prior to editing the xlate configuration. */ | |
1197 | void | |
1198 | xlate_txn_start(void) | |
1199 | { | |
1200 | struct xbridge *xbridge; | |
1201 | struct xlate_cfg *xcfg; | |
1202 | ||
1203 | ovs_assert(!new_xcfg); | |
1204 | ||
1205 | new_xcfg = xmalloc(sizeof *new_xcfg); | |
1206 | hmap_init(&new_xcfg->xbridges); | |
1207 | hmap_init(&new_xcfg->xbundles); | |
1208 | hmap_init(&new_xcfg->xports); | |
c3594cc3 | 1209 | hmap_init(&new_xcfg->xports_uuid); |
84f0f298 RW |
1210 | |
1211 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
1212 | if (!xcfg) { | |
1213 | return; | |
1214 | } | |
1215 | ||
1216 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
1217 | xlate_xbridge_copy(xbridge); | |
1218 | } | |
1219 | } | |
1220 | ||
1221 | ||
1222 | static void | |
1223 | xlate_xcfg_free(struct xlate_cfg *xcfg) | |
1224 | { | |
1225 | struct xbridge *xbridge, *next_xbridge; | |
1226 | ||
1227 | if (!xcfg) { | |
1228 | return; | |
1229 | } | |
1230 | ||
1231 | HMAP_FOR_EACH_SAFE (xbridge, next_xbridge, hmap_node, &xcfg->xbridges) { | |
1232 | xlate_xbridge_remove(xcfg, xbridge); | |
1233 | } | |
1234 | ||
1235 | hmap_destroy(&xcfg->xbridges); | |
1236 | hmap_destroy(&xcfg->xbundles); | |
1237 | hmap_destroy(&xcfg->xports); | |
c3594cc3 | 1238 | hmap_destroy(&xcfg->xports_uuid); |
84f0f298 RW |
1239 | free(xcfg); |
1240 | } | |
1241 | ||
1242 | void | |
1243 | xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, | |
34dd0d78 | 1244 | struct dpif *dpif, |
84f0f298 | 1245 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 1246 | struct rstp *rstp, const struct mcast_snooping *ms, |
84f0f298 RW |
1247 | const struct mbridge *mbridge, |
1248 | const struct dpif_sflow *sflow, | |
1249 | const struct dpif_ipfix *ipfix, | |
2f47cdf4 | 1250 | const struct netflow *netflow, |
b440dd8c JS |
1251 | bool forward_bpdu, bool has_in_band, |
1252 | const struct dpif_backer_support *support) | |
84f0f298 RW |
1253 | { |
1254 | struct xbridge *xbridge; | |
83c2757b | 1255 | struct xbridge_addr *xbridge_addr, *old_addr; |
84f0f298 RW |
1256 | |
1257 | ovs_assert(new_xcfg); | |
1258 | ||
1259 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
1260 | if (!xbridge) { | |
1261 | xbridge = xzalloc(sizeof *xbridge); | |
1262 | xbridge->ofproto = ofproto; | |
1263 | ||
1264 | xlate_xbridge_init(new_xcfg, xbridge); | |
1265 | } | |
1266 | ||
1267 | free(xbridge->name); | |
1268 | xbridge->name = xstrdup(name); | |
1269 | ||
83c2757b ZB |
1270 | xbridge_addr = xbridge_addr_create(xbridge); |
1271 | old_addr = xbridge->addr; | |
1272 | ||
34dd0d78 | 1273 | xlate_xbridge_set(xbridge, dpif, ml, stp, rstp, ms, mbridge, sflow, ipfix, |
83c2757b ZB |
1274 | netflow, forward_bpdu, has_in_band, support, |
1275 | xbridge_addr); | |
1276 | ||
1277 | if (xbridge_addr != old_addr) { | |
1278 | xbridge_addr_unref(xbridge_addr); | |
1279 | } | |
84f0f298 RW |
1280 | } |
1281 | ||
1282 | static void | |
1283 | xlate_xbridge_remove(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
46c88433 | 1284 | { |
46c88433 EJ |
1285 | struct xbundle *xbundle, *next_xbundle; |
1286 | struct xport *xport, *next_xport; | |
1287 | ||
1288 | if (!xbridge) { | |
1289 | return; | |
1290 | } | |
1291 | ||
1292 | HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) { | |
84f0f298 | 1293 | xlate_xport_remove(xcfg, xport); |
46c88433 EJ |
1294 | } |
1295 | ||
1296 | LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) { | |
84f0f298 | 1297 | xlate_xbundle_remove(xcfg, xbundle); |
46c88433 EJ |
1298 | } |
1299 | ||
84f0f298 | 1300 | hmap_remove(&xcfg->xbridges, &xbridge->hmap_node); |
795cc5c1 | 1301 | mac_learning_unref(xbridge->ml); |
6d95c4e8 | 1302 | mcast_snooping_unref(xbridge->ms); |
795cc5c1 EJ |
1303 | mbridge_unref(xbridge->mbridge); |
1304 | dpif_sflow_unref(xbridge->sflow); | |
1305 | dpif_ipfix_unref(xbridge->ipfix); | |
3570f7e4 | 1306 | netflow_unref(xbridge->netflow); |
795cc5c1 | 1307 | stp_unref(xbridge->stp); |
9efd308e | 1308 | rstp_unref(xbridge->rstp); |
83c2757b | 1309 | xbridge_addr_unref(xbridge->addr); |
795cc5c1 | 1310 | hmap_destroy(&xbridge->xports); |
46c88433 EJ |
1311 | free(xbridge->name); |
1312 | free(xbridge); | |
1313 | } | |
1314 | ||
84f0f298 RW |
1315 | void |
1316 | xlate_remove_ofproto(struct ofproto_dpif *ofproto) | |
1317 | { | |
1318 | struct xbridge *xbridge; | |
1319 | ||
1320 | ovs_assert(new_xcfg); | |
1321 | ||
1322 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
1323 | xlate_xbridge_remove(new_xcfg, xbridge); | |
1324 | } | |
1325 | ||
46c88433 EJ |
1326 | void |
1327 | xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
f0fb825a | 1328 | const char *name, enum port_vlan_mode vlan_mode, |
fed8962a EG |
1329 | uint16_t qinq_ethtype, int vlan, |
1330 | unsigned long *trunks, unsigned long *cvlans, | |
88f52d7f | 1331 | enum port_priority_tags_mode use_priority_tags, |
46c88433 | 1332 | const struct bond *bond, const struct lacp *lacp, |
c005f976 | 1333 | bool floodable, bool protected) |
46c88433 | 1334 | { |
84f0f298 | 1335 | struct xbundle *xbundle; |
46c88433 | 1336 | |
84f0f298 RW |
1337 | ovs_assert(new_xcfg); |
1338 | ||
1339 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
46c88433 EJ |
1340 | if (!xbundle) { |
1341 | xbundle = xzalloc(sizeof *xbundle); | |
1342 | xbundle->ofbundle = ofbundle; | |
84f0f298 | 1343 | xbundle->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 1344 | |
84f0f298 | 1345 | xlate_xbundle_init(new_xcfg, xbundle); |
46c88433 EJ |
1346 | } |
1347 | ||
46c88433 EJ |
1348 | free(xbundle->name); |
1349 | xbundle->name = xstrdup(name); | |
1350 | ||
fed8962a | 1351 | xlate_xbundle_set(xbundle, vlan_mode, qinq_ethtype, vlan, trunks, cvlans, |
c005f976 | 1352 | use_priority_tags, bond, lacp, floodable, protected); |
46c88433 EJ |
1353 | } |
1354 | ||
84f0f298 RW |
1355 | static void |
1356 | xlate_xbundle_remove(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
46c88433 | 1357 | { |
5f03c983 | 1358 | struct xport *xport; |
46c88433 EJ |
1359 | |
1360 | if (!xbundle) { | |
1361 | return; | |
1362 | } | |
1363 | ||
5f03c983 | 1364 | LIST_FOR_EACH_POP (xport, bundle_node, &xbundle->xports) { |
46c88433 EJ |
1365 | xport->xbundle = NULL; |
1366 | } | |
1367 | ||
84f0f298 | 1368 | hmap_remove(&xcfg->xbundles, &xbundle->hmap_node); |
417e7e66 | 1369 | ovs_list_remove(&xbundle->list_node); |
46c88433 EJ |
1370 | bond_unref(xbundle->bond); |
1371 | lacp_unref(xbundle->lacp); | |
1372 | free(xbundle->name); | |
1373 | free(xbundle); | |
1374 | } | |
1375 | ||
84f0f298 RW |
1376 | void |
1377 | xlate_bundle_remove(struct ofbundle *ofbundle) | |
1378 | { | |
1379 | struct xbundle *xbundle; | |
1380 | ||
1381 | ovs_assert(new_xcfg); | |
1382 | ||
1383 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
1384 | xlate_xbundle_remove(new_xcfg, xbundle); | |
1385 | } | |
1386 | ||
46c88433 EJ |
1387 | void |
1388 | xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
1389 | struct ofport_dpif *ofport, ofp_port_t ofp_port, | |
1390 | odp_port_t odp_port, const struct netdev *netdev, | |
1391 | const struct cfm *cfm, const struct bfd *bfd, | |
0477baa9 DF |
1392 | const struct lldp *lldp, struct ofport_dpif *peer, |
1393 | int stp_port_no, const struct rstp_port *rstp_port, | |
55954f6e | 1394 | const struct ofproto_port_queue *qdscp_list, size_t n_qdscp, |
dd8cd4b4 SH |
1395 | enum ofputil_port_config config, |
1396 | enum ofputil_port_state state, bool is_tunnel, | |
9d189a50 | 1397 | bool may_enable) |
46c88433 | 1398 | { |
55954f6e | 1399 | size_t i; |
84f0f298 RW |
1400 | struct xport *xport; |
1401 | ||
1402 | ovs_assert(new_xcfg); | |
46c88433 | 1403 | |
84f0f298 | 1404 | xport = xport_lookup(new_xcfg, ofport); |
46c88433 EJ |
1405 | if (!xport) { |
1406 | xport = xzalloc(sizeof *xport); | |
1407 | xport->ofport = ofport; | |
84f0f298 | 1408 | xport->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 1409 | xport->ofp_port = ofp_port; |
c3594cc3 | 1410 | uuid_generate(&xport->uuid); |
46c88433 | 1411 | |
84f0f298 | 1412 | xlate_xport_init(new_xcfg, xport); |
46c88433 EJ |
1413 | } |
1414 | ||
1415 | ovs_assert(xport->ofp_port == ofp_port); | |
1416 | ||
0477baa9 DF |
1417 | xlate_xport_set(xport, odp_port, netdev, cfm, bfd, lldp, |
1418 | stp_port_no, rstp_port, config, state, is_tunnel, | |
1419 | may_enable); | |
46c88433 EJ |
1420 | |
1421 | if (xport->peer) { | |
1422 | xport->peer->peer = NULL; | |
1423 | } | |
84f0f298 | 1424 | xport->peer = xport_lookup(new_xcfg, peer); |
46c88433 EJ |
1425 | if (xport->peer) { |
1426 | xport->peer->peer = xport; | |
1427 | } | |
1428 | ||
1429 | if (xport->xbundle) { | |
417e7e66 | 1430 | ovs_list_remove(&xport->bundle_node); |
46c88433 | 1431 | } |
84f0f298 | 1432 | xport->xbundle = xbundle_lookup(new_xcfg, ofbundle); |
46c88433 | 1433 | if (xport->xbundle) { |
417e7e66 | 1434 | ovs_list_insert(&xport->xbundle->xports, &xport->bundle_node); |
46c88433 | 1435 | } |
55954f6e EJ |
1436 | |
1437 | clear_skb_priorities(xport); | |
1438 | for (i = 0; i < n_qdscp; i++) { | |
1439 | struct skb_priority_to_dscp *pdscp; | |
1440 | uint32_t skb_priority; | |
1441 | ||
89a8a7f0 EJ |
1442 | if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue, |
1443 | &skb_priority)) { | |
55954f6e EJ |
1444 | continue; |
1445 | } | |
1446 | ||
1447 | pdscp = xmalloc(sizeof *pdscp); | |
1448 | pdscp->skb_priority = skb_priority; | |
1449 | pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK; | |
1450 | hmap_insert(&xport->skb_priorities, &pdscp->hmap_node, | |
1451 | hash_int(pdscp->skb_priority, 0)); | |
1452 | } | |
46c88433 EJ |
1453 | } |
1454 | ||
84f0f298 RW |
1455 | static void |
1456 | xlate_xport_remove(struct xlate_cfg *xcfg, struct xport *xport) | |
46c88433 | 1457 | { |
46c88433 EJ |
1458 | if (!xport) { |
1459 | return; | |
1460 | } | |
1461 | ||
1462 | if (xport->peer) { | |
1463 | xport->peer->peer = NULL; | |
1464 | xport->peer = NULL; | |
1465 | } | |
1466 | ||
e621a12d | 1467 | if (xport->xbundle) { |
417e7e66 | 1468 | ovs_list_remove(&xport->bundle_node); |
e621a12d EJ |
1469 | } |
1470 | ||
55954f6e EJ |
1471 | clear_skb_priorities(xport); |
1472 | hmap_destroy(&xport->skb_priorities); | |
1473 | ||
84f0f298 | 1474 | hmap_remove(&xcfg->xports, &xport->hmap_node); |
c3594cc3 | 1475 | hmap_remove(&xcfg->xports_uuid, &xport->uuid_node); |
46c88433 EJ |
1476 | hmap_remove(&xport->xbridge->xports, &xport->ofp_node); |
1477 | ||
1478 | netdev_close(xport->netdev); | |
f025bcb7 | 1479 | rstp_port_unref(xport->rstp_port); |
46c88433 EJ |
1480 | cfm_unref(xport->cfm); |
1481 | bfd_unref(xport->bfd); | |
0477baa9 | 1482 | lldp_unref(xport->lldp); |
46c88433 EJ |
1483 | free(xport); |
1484 | } | |
1485 | ||
84f0f298 RW |
1486 | void |
1487 | xlate_ofport_remove(struct ofport_dpif *ofport) | |
1488 | { | |
1489 | struct xport *xport; | |
1490 | ||
1491 | ovs_assert(new_xcfg); | |
1492 | ||
1493 | xport = xport_lookup(new_xcfg, ofport); | |
8416e50f VD |
1494 | if (xport) { |
1495 | tnl_neigh_flush(netdev_get_name(xport->netdev)); | |
1496 | } | |
84f0f298 RW |
1497 | xlate_xport_remove(new_xcfg, xport); |
1498 | } | |
1499 | ||
ef377a58 | 1500 | static struct ofproto_dpif * |
d40533fc BP |
1501 | xlate_lookup_ofproto_(const struct dpif_backer *backer, |
1502 | const struct flow *flow, | |
1503 | ofp_port_t *ofp_in_port, const struct xport **xportp, | |
1504 | char **errorp) | |
ef377a58 | 1505 | { |
e672ff9b | 1506 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
ef377a58 | 1507 | const struct xport *xport; |
f9038ef6 | 1508 | |
00135b86 ZB |
1509 | /* If packet is recirculated, xport can be retrieved from frozen state. */ |
1510 | if (flow->recirc_id) { | |
1511 | const struct recirc_id_node *recirc_id_node; | |
1512 | ||
1513 | recirc_id_node = recirc_id_node_find(flow->recirc_id); | |
1514 | ||
1515 | if (OVS_UNLIKELY(!recirc_id_node)) { | |
d40533fc BP |
1516 | if (errorp) { |
1517 | *errorp = xasprintf("no recirculation data for recirc_id " | |
1518 | "%"PRIu32, flow->recirc_id); | |
1519 | } | |
00135b86 ZB |
1520 | return NULL; |
1521 | } | |
1522 | ||
323ae1e8 BP |
1523 | ofp_port_t in_port = recirc_id_node->state.metadata.in_port; |
1524 | if (in_port != OFPP_NONE && in_port != OFPP_CONTROLLER) { | |
00135b86 ZB |
1525 | struct uuid xport_uuid = recirc_id_node->state.xport_uuid; |
1526 | xport = xport_lookup_by_uuid(xcfg, &xport_uuid); | |
1527 | if (xport && xport->xbridge && xport->xbridge->ofproto) { | |
1528 | goto out; | |
1529 | } | |
323ae1e8 BP |
1530 | } else { |
1531 | /* OFPP_NONE and OFPP_CONTROLLER are not real ports. They indicate | |
1532 | * that the packet originated from the controller via an OpenFlow | |
1533 | * "packet-out". The right thing to do is to find just the | |
1534 | * ofproto. There is no xport, which is OK. | |
1535 | * | |
1536 | * OFPP_NONE can also indicate that a bond caused recirculation. */ | |
1537 | struct uuid uuid = recirc_id_node->state.ofproto_uuid; | |
1538 | const struct xbridge *bridge = xbridge_lookup_by_uuid(xcfg, &uuid); | |
1539 | if (bridge && bridge->ofproto) { | |
1540 | if (errorp) { | |
1541 | *errorp = NULL; | |
1542 | } | |
1543 | *xportp = NULL; | |
1544 | if (ofp_in_port) { | |
1545 | *ofp_in_port = in_port; | |
1546 | } | |
1547 | return bridge->ofproto; | |
1548 | } | |
00135b86 ZB |
1549 | } |
1550 | } | |
1551 | ||
e672ff9b JR |
1552 | xport = xport_lookup(xcfg, tnl_port_should_receive(flow) |
1553 | ? tnl_port_receive(flow) | |
1554 | : odp_port_to_ofport(backer, flow->in_port.odp_port)); | |
1555 | if (OVS_UNLIKELY(!xport)) { | |
d40533fc BP |
1556 | if (errorp) { |
1557 | *errorp = (tnl_port_should_receive(flow) | |
1558 | ? xstrdup("no OpenFlow tunnel port for this packet") | |
1559 | : xasprintf("no OpenFlow tunnel port for datapath " | |
1560 | "port %"PRIu32, flow->in_port.odp_port)); | |
1561 | } | |
e672ff9b | 1562 | return NULL; |
ef377a58 | 1563 | } |
00135b86 ZB |
1564 | |
1565 | out: | |
d40533fc BP |
1566 | if (errorp) { |
1567 | *errorp = NULL; | |
1568 | } | |
e672ff9b | 1569 | *xportp = xport; |
f9038ef6 | 1570 | if (ofp_in_port) { |
e672ff9b | 1571 | *ofp_in_port = xport->ofp_port; |
f9038ef6 | 1572 | } |
e672ff9b | 1573 | return xport->xbridge->ofproto; |
ef377a58 JR |
1574 | } |
1575 | ||
1576 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively) | |
1577 | * returns the corresponding struct ofproto_dpif and OpenFlow port number. */ | |
1578 | struct ofproto_dpif * | |
1579 | xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow, | |
d40533fc | 1580 | ofp_port_t *ofp_in_port, char **errorp) |
ef377a58 JR |
1581 | { |
1582 | const struct xport *xport; | |
1583 | ||
d40533fc | 1584 | return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport, errorp); |
ef377a58 JR |
1585 | } |
1586 | ||
cc377352 | 1587 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively), |
bfc691bb | 1588 | * optionally populates 'ofprotop' with the ofproto_dpif, 'ofp_in_port' with the |
cc377352 | 1589 | * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate |
dcc2c6cd JR |
1590 | * handles for those protocols if they're enabled. Caller may use the returned |
1591 | * pointers until quiescing, for longer term use additional references must | |
1592 | * be taken. | |
8449c4d6 | 1593 | * |
f9038ef6 | 1594 | * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto. |
ef377a58 | 1595 | */ |
8449c4d6 | 1596 | int |
5c476ea3 JR |
1597 | xlate_lookup(const struct dpif_backer *backer, const struct flow *flow, |
1598 | struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix, | |
1599 | struct dpif_sflow **sflow, struct netflow **netflow, | |
1600 | ofp_port_t *ofp_in_port) | |
8449c4d6 | 1601 | { |
ef377a58 | 1602 | struct ofproto_dpif *ofproto; |
84f0f298 | 1603 | const struct xport *xport; |
8449c4d6 | 1604 | |
d40533fc | 1605 | ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport, NULL); |
8449c4d6 | 1606 | |
f9038ef6 | 1607 | if (!ofproto) { |
cc377352 | 1608 | return ENODEV; |
8449c4d6 | 1609 | } |
8449c4d6 | 1610 | |
ef377a58 JR |
1611 | if (ofprotop) { |
1612 | *ofprotop = ofproto; | |
8449c4d6 EJ |
1613 | } |
1614 | ||
1dfdb9b3 | 1615 | if (ipfix) { |
f9038ef6 | 1616 | *ipfix = xport ? xport->xbridge->ipfix : NULL; |
1dfdb9b3 EJ |
1617 | } |
1618 | ||
1619 | if (sflow) { | |
f9038ef6 | 1620 | *sflow = xport ? xport->xbridge->sflow : NULL; |
1dfdb9b3 EJ |
1621 | } |
1622 | ||
1623 | if (netflow) { | |
f9038ef6 | 1624 | *netflow = xport ? xport->xbridge->netflow : NULL; |
1dfdb9b3 | 1625 | } |
f9038ef6 | 1626 | |
cc377352 | 1627 | return 0; |
8449c4d6 EJ |
1628 | } |
1629 | ||
46c88433 | 1630 | static struct xbridge * |
84f0f298 | 1631 | xbridge_lookup(struct xlate_cfg *xcfg, const struct ofproto_dpif *ofproto) |
46c88433 | 1632 | { |
84f0f298 | 1633 | struct hmap *xbridges; |
46c88433 EJ |
1634 | struct xbridge *xbridge; |
1635 | ||
84f0f298 | 1636 | if (!ofproto || !xcfg) { |
5e6af486 EJ |
1637 | return NULL; |
1638 | } | |
1639 | ||
84f0f298 RW |
1640 | xbridges = &xcfg->xbridges; |
1641 | ||
46c88433 | 1642 | HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0), |
84f0f298 | 1643 | xbridges) { |
46c88433 EJ |
1644 | if (xbridge->ofproto == ofproto) { |
1645 | return xbridge; | |
1646 | } | |
1647 | } | |
1648 | return NULL; | |
1649 | } | |
1650 | ||
290835f9 BP |
1651 | static struct xbridge * |
1652 | xbridge_lookup_by_uuid(struct xlate_cfg *xcfg, const struct uuid *uuid) | |
1653 | { | |
1654 | struct xbridge *xbridge; | |
1655 | ||
1656 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
07a3cd5c | 1657 | if (uuid_equals(&xbridge->ofproto->uuid, uuid)) { |
290835f9 BP |
1658 | return xbridge; |
1659 | } | |
1660 | } | |
1661 | return NULL; | |
1662 | } | |
1663 | ||
46c88433 | 1664 | static struct xbundle * |
84f0f298 | 1665 | xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle) |
46c88433 | 1666 | { |
84f0f298 | 1667 | struct hmap *xbundles; |
46c88433 EJ |
1668 | struct xbundle *xbundle; |
1669 | ||
84f0f298 | 1670 | if (!ofbundle || !xcfg) { |
5e6af486 EJ |
1671 | return NULL; |
1672 | } | |
1673 | ||
84f0f298 RW |
1674 | xbundles = &xcfg->xbundles; |
1675 | ||
46c88433 | 1676 | HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), |
84f0f298 | 1677 | xbundles) { |
46c88433 EJ |
1678 | if (xbundle->ofbundle == ofbundle) { |
1679 | return xbundle; | |
1680 | } | |
1681 | } | |
1682 | return NULL; | |
1683 | } | |
1684 | ||
1685 | static struct xport * | |
84f0f298 | 1686 | xport_lookup(struct xlate_cfg *xcfg, const struct ofport_dpif *ofport) |
46c88433 | 1687 | { |
84f0f298 | 1688 | struct hmap *xports; |
46c88433 EJ |
1689 | struct xport *xport; |
1690 | ||
84f0f298 | 1691 | if (!ofport || !xcfg) { |
5e6af486 EJ |
1692 | return NULL; |
1693 | } | |
1694 | ||
84f0f298 RW |
1695 | xports = &xcfg->xports; |
1696 | ||
46c88433 | 1697 | HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0), |
84f0f298 | 1698 | xports) { |
46c88433 EJ |
1699 | if (xport->ofport == ofport) { |
1700 | return xport; | |
1701 | } | |
1702 | } | |
1703 | return NULL; | |
1704 | } | |
1705 | ||
00135b86 ZB |
1706 | static struct xport * |
1707 | xport_lookup_by_uuid(struct xlate_cfg *xcfg, const struct uuid *uuid) | |
1708 | { | |
1709 | struct hmap *xports; | |
1710 | struct xport *xport; | |
1711 | ||
1712 | if (uuid_is_zero(uuid) || !xcfg) { | |
1713 | return NULL; | |
1714 | } | |
1715 | ||
1716 | xports = &xcfg->xports_uuid; | |
1717 | ||
1718 | HMAP_FOR_EACH_IN_BUCKET (xport, uuid_node, uuid_hash(uuid), xports) { | |
1719 | if (uuid_equals(&xport->uuid, uuid)) { | |
1720 | return xport; | |
1721 | } | |
1722 | } | |
1723 | return NULL; | |
1724 | } | |
1725 | ||
40085e56 EJ |
1726 | static struct stp_port * |
1727 | xport_get_stp_port(const struct xport *xport) | |
1728 | { | |
92cf817b | 1729 | return xport->xbridge->stp && xport->stp_port_no != -1 |
40085e56 EJ |
1730 | ? stp_get_port(xport->xbridge->stp, xport->stp_port_no) |
1731 | : NULL; | |
1732 | } | |
9d189a50 | 1733 | |
0d1cee12 | 1734 | static bool |
9d189a50 EJ |
1735 | xport_stp_learn_state(const struct xport *xport) |
1736 | { | |
40085e56 | 1737 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1738 | return sp |
1739 | ? stp_learn_in_state(stp_port_get_state(sp)) | |
1740 | : true; | |
9d189a50 EJ |
1741 | } |
1742 | ||
1743 | static bool | |
1744 | xport_stp_forward_state(const struct xport *xport) | |
1745 | { | |
40085e56 | 1746 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1747 | return sp |
1748 | ? stp_forward_in_state(stp_port_get_state(sp)) | |
1749 | : true; | |
9d189a50 EJ |
1750 | } |
1751 | ||
0d1cee12 | 1752 | static bool |
bacdb85a | 1753 | xport_stp_should_forward_bpdu(const struct xport *xport) |
0d1cee12 K |
1754 | { |
1755 | struct stp_port *sp = xport_get_stp_port(xport); | |
bacdb85a | 1756 | return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED); |
0d1cee12 K |
1757 | } |
1758 | ||
9d189a50 EJ |
1759 | /* Returns true if STP should process 'flow'. Sets fields in 'wc' that |
1760 | * were used to make the determination.*/ | |
1761 | static bool | |
1762 | stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc) | |
1763 | { | |
bbbca389 | 1764 | /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */ |
9d189a50 | 1765 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
bbbca389 | 1766 | return is_stp(flow); |
9d189a50 EJ |
1767 | } |
1768 | ||
1769 | static void | |
cf62fa4c | 1770 | stp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9d189a50 | 1771 | { |
40085e56 | 1772 | struct stp_port *sp = xport_get_stp_port(xport); |
cf62fa4c PS |
1773 | struct dp_packet payload = *packet; |
1774 | struct eth_header *eth = dp_packet_data(&payload); | |
9d189a50 EJ |
1775 | |
1776 | /* Sink packets on ports that have STP disabled when the bridge has | |
1777 | * STP enabled. */ | |
1778 | if (!sp || stp_port_get_state(sp) == STP_DISABLED) { | |
1779 | return; | |
1780 | } | |
1781 | ||
1782 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1783 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1784 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9d189a50 EJ |
1785 | } |
1786 | ||
cf62fa4c PS |
1787 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1788 | stp_received_bpdu(sp, dp_packet_data(&payload), dp_packet_size(&payload)); | |
9d189a50 EJ |
1789 | } |
1790 | } | |
1791 | ||
f025bcb7 JR |
1792 | static enum rstp_state |
1793 | xport_get_rstp_port_state(const struct xport *xport) | |
9efd308e | 1794 | { |
f025bcb7 JR |
1795 | return xport->rstp_port |
1796 | ? rstp_port_get_state(xport->rstp_port) | |
1797 | : RSTP_DISABLED; | |
9efd308e DV |
1798 | } |
1799 | ||
1800 | static bool | |
1801 | xport_rstp_learn_state(const struct xport *xport) | |
1802 | { | |
4b5f1996 DV |
1803 | return xport->xbridge->rstp && xport->rstp_port |
1804 | ? rstp_learn_in_state(xport_get_rstp_port_state(xport)) | |
1805 | : true; | |
9efd308e DV |
1806 | } |
1807 | ||
1808 | static bool | |
1809 | xport_rstp_forward_state(const struct xport *xport) | |
1810 | { | |
4b5f1996 DV |
1811 | return xport->xbridge->rstp && xport->rstp_port |
1812 | ? rstp_forward_in_state(xport_get_rstp_port_state(xport)) | |
1813 | : true; | |
9efd308e DV |
1814 | } |
1815 | ||
1816 | static bool | |
1817 | xport_rstp_should_manage_bpdu(const struct xport *xport) | |
1818 | { | |
f025bcb7 | 1819 | return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport)); |
9efd308e DV |
1820 | } |
1821 | ||
1822 | static void | |
cf62fa4c | 1823 | rstp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9efd308e | 1824 | { |
cf62fa4c PS |
1825 | struct dp_packet payload = *packet; |
1826 | struct eth_header *eth = dp_packet_data(&payload); | |
9efd308e | 1827 | |
f025bcb7 JR |
1828 | /* Sink packets on ports that have no RSTP. */ |
1829 | if (!xport->rstp_port) { | |
9efd308e DV |
1830 | return; |
1831 | } | |
1832 | ||
1833 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1834 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1835 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9efd308e DV |
1836 | } |
1837 | ||
f2f78d5c MM |
1838 | int len = ETH_HEADER_LEN + LLC_HEADER_LEN; |
1839 | if (eth->eth_type == htons(ETH_TYPE_VLAN)) { | |
1840 | len += VLAN_HEADER_LEN; | |
1841 | } | |
1842 | if (dp_packet_try_pull(&payload, len)) { | |
cf62fa4c PS |
1843 | rstp_port_received_bpdu(xport->rstp_port, dp_packet_data(&payload), |
1844 | dp_packet_size(&payload)); | |
9efd308e DV |
1845 | } |
1846 | } | |
1847 | ||
46c88433 EJ |
1848 | static struct xport * |
1849 | get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1850 | { | |
1851 | struct xport *xport; | |
1852 | ||
1853 | HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port), | |
1854 | &xbridge->xports) { | |
1855 | if (xport->ofp_port == ofp_port) { | |
1856 | return xport; | |
1857 | } | |
1858 | } | |
1859 | return NULL; | |
1860 | } | |
1861 | ||
1862 | static odp_port_t | |
1863 | ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1864 | { | |
1865 | const struct xport *xport = get_ofp_port(xbridge, ofp_port); | |
1866 | return xport ? xport->odp_port : ODPP_NONE; | |
1867 | } | |
1868 | ||
dd8cd4b4 SH |
1869 | static bool |
1870 | odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port) | |
1871 | { | |
086fa873 BP |
1872 | struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1873 | return xport && xport->may_enable; | |
dd8cd4b4 SH |
1874 | } |
1875 | ||
1e684d7d | 1876 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1877 | group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *, |
1878 | int depth); | |
1879 | ||
1880 | static bool | |
1881 | group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) | |
1882 | { | |
1883 | struct group_dpif *group; | |
dd8cd4b4 | 1884 | |
5d08a275 | 1885 | group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, |
1f4a8933 | 1886 | ctx->xin->tables_version, false); |
db88b35c | 1887 | if (group) { |
76973237 | 1888 | return group_first_live_bucket(ctx, group, depth) != NULL; |
dc25893e | 1889 | } |
dd8cd4b4 | 1890 | |
dc25893e | 1891 | return false; |
dd8cd4b4 SH |
1892 | } |
1893 | ||
1894 | #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ | |
1895 | ||
1896 | static bool | |
1897 | bucket_is_alive(const struct xlate_ctx *ctx, | |
1e684d7d | 1898 | struct ofputil_bucket *bucket, int depth) |
dd8cd4b4 SH |
1899 | { |
1900 | if (depth >= MAX_LIVENESS_RECURSION) { | |
2d9b49dd BP |
1901 | xlate_report_error(ctx, "bucket chaining exceeded %d links", |
1902 | MAX_LIVENESS_RECURSION); | |
dd8cd4b4 SH |
1903 | return false; |
1904 | } | |
1905 | ||
fdb1999b AZ |
1906 | return (!ofputil_bucket_has_liveness(bucket) |
1907 | || (bucket->watch_port != OFPP_ANY | |
44810e6d | 1908 | && bucket->watch_port != OFPP_CONTROLLER |
fdb1999b AZ |
1909 | && odp_port_is_alive(ctx, bucket->watch_port)) |
1910 | || (bucket->watch_group != OFPG_ANY | |
44810e6d VDA |
1911 | && group_is_alive(ctx, bucket->watch_group, depth + 1)) |
1912 | || (bucket->watch_port == OFPP_CONTROLLER | |
1913 | && ofproto_is_alive(&ctx->xbridge->ofproto->up))); | |
dd8cd4b4 SH |
1914 | } |
1915 | ||
fac4786a BP |
1916 | static void |
1917 | xlate_report_bucket_not_live(const struct xlate_ctx *ctx, | |
1918 | const struct ofputil_bucket *bucket) | |
1919 | { | |
1920 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
1921 | struct ds s = DS_EMPTY_INITIALIZER; | |
1922 | if (bucket->watch_port != OFPP_ANY) { | |
1923 | ds_put_cstr(&s, "port "); | |
1924 | ofputil_format_port(bucket->watch_port, NULL, &s); | |
1925 | } | |
1926 | if (bucket->watch_group != OFPG_ANY) { | |
1927 | if (s.length) { | |
1928 | ds_put_cstr(&s, " and "); | |
1929 | } | |
1930 | ds_put_format(&s, "port %"PRIu32, bucket->watch_group); | |
1931 | } | |
1932 | ||
1933 | xlate_report(ctx, OFT_DETAIL, "bucket %"PRIu32": not live due to %s", | |
1934 | bucket->bucket_id, ds_cstr(&s)); | |
1935 | ||
1936 | ds_destroy(&s); | |
1937 | } | |
1938 | } | |
1939 | ||
1e684d7d | 1940 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1941 | group_first_live_bucket(const struct xlate_ctx *ctx, |
1942 | const struct group_dpif *group, int depth) | |
1943 | { | |
1944 | struct ofputil_bucket *bucket; | |
07a3cd5c | 1945 | LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { |
dd8cd4b4 SH |
1946 | if (bucket_is_alive(ctx, bucket, depth)) { |
1947 | return bucket; | |
1948 | } | |
fac4786a | 1949 | xlate_report_bucket_not_live(ctx, bucket); |
dd8cd4b4 SH |
1950 | } |
1951 | ||
1952 | return NULL; | |
1953 | } | |
1954 | ||
1e684d7d | 1955 | static struct ofputil_bucket * |
fe7e5749 SH |
1956 | group_best_live_bucket(const struct xlate_ctx *ctx, |
1957 | const struct group_dpif *group, | |
1958 | uint32_t basis) | |
1959 | { | |
1e684d7d | 1960 | struct ofputil_bucket *best_bucket = NULL; |
fe7e5749 | 1961 | uint32_t best_score = 0; |
fe7e5749 | 1962 | |
1e684d7d | 1963 | struct ofputil_bucket *bucket; |
07a3cd5c | 1964 | LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { |
fe7e5749 | 1965 | if (bucket_is_alive(ctx, bucket, 0)) { |
c09cb861 LS |
1966 | uint32_t score = |
1967 | (hash_int(bucket->bucket_id, basis) & 0xffff) * bucket->weight; | |
fe7e5749 SH |
1968 | if (score >= best_score) { |
1969 | best_bucket = bucket; | |
1970 | best_score = score; | |
1971 | } | |
fac4786a BP |
1972 | xlate_report(ctx, OFT_DETAIL, "bucket %"PRIu32": score %"PRIu32, |
1973 | bucket->bucket_id, score); | |
1974 | } else { | |
1975 | xlate_report_bucket_not_live(ctx, bucket); | |
fe7e5749 | 1976 | } |
fe7e5749 SH |
1977 | } |
1978 | ||
1979 | return best_bucket; | |
1980 | } | |
1981 | ||
9583bc14 | 1982 | static bool |
46c88433 | 1983 | xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan) |
9583bc14 EJ |
1984 | { |
1985 | return (bundle->vlan_mode != PORT_VLAN_ACCESS | |
1986 | && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan))); | |
1987 | } | |
1988 | ||
fed8962a EG |
1989 | static bool |
1990 | xbundle_allows_cvlan(const struct xbundle *bundle, uint16_t vlan) | |
1991 | { | |
1992 | return (!bundle->cvlans || bitmap_is_set(bundle->cvlans, vlan)); | |
1993 | } | |
1994 | ||
9583bc14 | 1995 | static bool |
f0fb825a | 1996 | xbundle_includes_vlan(const struct xbundle *xbundle, const struct xvlan *xvlan) |
46c88433 | 1997 | { |
f0fb825a EG |
1998 | switch (xbundle->vlan_mode) { |
1999 | case PORT_VLAN_ACCESS: | |
2000 | return xvlan->v[0].vid == xbundle->vlan && xvlan->v[1].vid == 0; | |
2001 | ||
2002 | case PORT_VLAN_TRUNK: | |
2003 | case PORT_VLAN_NATIVE_UNTAGGED: | |
2004 | case PORT_VLAN_NATIVE_TAGGED: | |
2005 | return xbundle_trunks_vlan(xbundle, xvlan->v[0].vid); | |
2006 | ||
fed8962a EG |
2007 | case PORT_VLAN_DOT1Q_TUNNEL: |
2008 | return xvlan->v[0].vid == xbundle->vlan && | |
2009 | xbundle_allows_cvlan(xbundle, xvlan->v[1].vid); | |
2010 | ||
f0fb825a EG |
2011 | default: |
2012 | OVS_NOT_REACHED(); | |
2013 | } | |
46c88433 EJ |
2014 | } |
2015 | ||
2016 | static mirror_mask_t | |
2017 | xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle) | |
2018 | { | |
2019 | return xbundle != &ofpp_none_bundle | |
2020 | ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle) | |
2021 | : 0; | |
2022 | } | |
2023 | ||
2024 | static mirror_mask_t | |
2025 | xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 2026 | { |
46c88433 EJ |
2027 | return xbundle != &ofpp_none_bundle |
2028 | ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle) | |
2029 | : 0; | |
9583bc14 EJ |
2030 | } |
2031 | ||
46c88433 EJ |
2032 | static mirror_mask_t |
2033 | xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 2034 | { |
46c88433 EJ |
2035 | return xbundle != &ofpp_none_bundle |
2036 | ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle) | |
2037 | : 0; | |
2038 | } | |
2039 | ||
2040 | static struct xbundle * | |
2d9b49dd BP |
2041 | lookup_input_bundle__(const struct xbridge *xbridge, |
2042 | ofp_port_t in_port, struct xport **in_xportp) | |
46c88433 EJ |
2043 | { |
2044 | struct xport *xport; | |
9583bc14 EJ |
2045 | |
2046 | /* Find the port and bundle for the received packet. */ | |
46c88433 EJ |
2047 | xport = get_ofp_port(xbridge, in_port); |
2048 | if (in_xportp) { | |
2049 | *in_xportp = xport; | |
9583bc14 | 2050 | } |
46c88433 EJ |
2051 | if (xport && xport->xbundle) { |
2052 | return xport->xbundle; | |
9583bc14 EJ |
2053 | } |
2054 | ||
6362203b YT |
2055 | /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+), |
2056 | * which a controller may use as the ingress port for traffic that | |
2057 | * it is sourcing. */ | |
2058 | if (in_port == OFPP_CONTROLLER || in_port == OFPP_NONE) { | |
9583bc14 EJ |
2059 | return &ofpp_none_bundle; |
2060 | } | |
2d9b49dd BP |
2061 | return NULL; |
2062 | } | |
9583bc14 | 2063 | |
2d9b49dd BP |
2064 | static struct xbundle * |
2065 | lookup_input_bundle(const struct xlate_ctx *ctx, | |
2066 | ofp_port_t in_port, struct xport **in_xportp) | |
2067 | { | |
2068 | struct xbundle *xbundle = lookup_input_bundle__(ctx->xbridge, | |
2069 | in_port, in_xportp); | |
2070 | if (!xbundle) { | |
2071 | /* Odd. A few possible reasons here: | |
2072 | * | |
2073 | * - We deleted a port but there are still a few packets queued up | |
2074 | * from it. | |
2075 | * | |
2076 | * - Someone externally added a port (e.g. "ovs-dpctl add-if") that | |
2077 | * we don't know about. | |
2078 | * | |
2079 | * - The ofproto client didn't configure the port as part of a bundle. | |
2080 | * This is particularly likely to happen if a packet was received on | |
2081 | * the port after it was created, but before the client had a chance | |
2082 | * to configure its bundle. | |
2083 | */ | |
94783c7c | 2084 | xlate_report_error(ctx, "received packet on unknown port %"PRIu32, |
2d9b49dd | 2085 | in_port); |
9583bc14 | 2086 | } |
2d9b49dd | 2087 | return xbundle; |
9583bc14 EJ |
2088 | } |
2089 | ||
faa624b4 BP |
2090 | /* Mirrors the packet represented by 'ctx' to appropriate mirror destinations, |
2091 | * given the packet is ingressing or egressing on 'xbundle', which has ingress | |
2092 | * or egress (as appropriate) mirrors 'mirrors'. */ | |
9583bc14 | 2093 | static void |
7efbc3b7 BP |
2094 | mirror_packet(struct xlate_ctx *ctx, struct xbundle *xbundle, |
2095 | mirror_mask_t mirrors) | |
9583bc14 | 2096 | { |
f0fb825a EG |
2097 | struct xvlan in_xvlan; |
2098 | struct xvlan xvlan; | |
2099 | ||
faa624b4 BP |
2100 | /* Figure out what VLAN the packet is in (because mirrors can select |
2101 | * packets on basis of VLAN). */ | |
f0fb825a EG |
2102 | xvlan_extract(&ctx->xin->flow, &in_xvlan); |
2103 | if (!input_vid_is_valid(ctx, in_xvlan.v[0].vid, xbundle)) { | |
9583bc14 EJ |
2104 | return; |
2105 | } | |
f0fb825a | 2106 | xvlan_input_translate(xbundle, &in_xvlan, &xvlan); |
9583bc14 | 2107 | |
7efbc3b7 | 2108 | const struct xbridge *xbridge = ctx->xbridge; |
9583bc14 | 2109 | |
7efbc3b7 BP |
2110 | /* Don't mirror to destinations that we've already mirrored to. */ |
2111 | mirrors &= ~ctx->mirrors; | |
9583bc14 EJ |
2112 | if (!mirrors) { |
2113 | return; | |
2114 | } | |
2115 | ||
a3954fd8 BP |
2116 | /* 'mirrors' is a bit-mask of candidates for mirroring. Iterate through |
2117 | * the candidates, adding the ones that really should be mirrored to | |
2118 | * 'used_mirrors', as long as some candidates remain. */ | |
2119 | mirror_mask_t used_mirrors = 0; | |
9583bc14 | 2120 | while (mirrors) { |
7efbc3b7 | 2121 | const unsigned long *vlans; |
ec7ceaed EJ |
2122 | mirror_mask_t dup_mirrors; |
2123 | struct ofbundle *out; | |
ec7ceaed | 2124 | int out_vlan; |
1356dbd1 | 2125 | int snaplen; |
ec7ceaed | 2126 | |
faa624b4 | 2127 | /* Get the details of the mirror represented by the rightmost 1-bit. */ |
500db308 BP |
2128 | ovs_assert(mirror_get(xbridge->mbridge, raw_ctz(mirrors), |
2129 | &vlans, &dup_mirrors, | |
2130 | &out, &snaplen, &out_vlan)); | |
ec7ceaed | 2131 | |
1356dbd1 | 2132 | |
faa624b4 BP |
2133 | /* If this mirror selects on the basis of VLAN, and it does not select |
2134 | * 'vlan', then discard this mirror and go on to the next one. */ | |
ec7ceaed | 2135 | if (vlans) { |
f0fb825a | 2136 | ctx->wc->masks.vlans[0].tci |= htons(VLAN_CFI | VLAN_VID_MASK); |
9583bc14 | 2137 | } |
f0fb825a | 2138 | if (vlans && !bitmap_is_set(vlans, xvlan.v[0].vid)) { |
9583bc14 EJ |
2139 | mirrors = zero_rightmost_1bit(mirrors); |
2140 | continue; | |
2141 | } | |
2142 | ||
a3954fd8 BP |
2143 | /* We sent a packet to this mirror. */ |
2144 | used_mirrors |= rightmost_1bit(mirrors); | |
2145 | ||
faa624b4 BP |
2146 | /* Record the mirror, and the mirrors that output to the same |
2147 | * destination, so that we don't mirror to them again. This must be | |
2148 | * done now to ensure that output_normal(), below, doesn't recursively | |
2149 | * output to the same mirrors. */ | |
3d6151f3 | 2150 | ctx->mirrors |= dup_mirrors; |
1356dbd1 | 2151 | ctx->mirror_snaplen = snaplen; |
faa624b4 BP |
2152 | |
2153 | /* Send the packet to the mirror. */ | |
ec7ceaed | 2154 | if (out) { |
0506f184 | 2155 | struct xbundle *out_xbundle = xbundle_lookup(ctx->xcfg, out); |
46c88433 | 2156 | if (out_xbundle) { |
f0fb825a | 2157 | output_normal(ctx, out_xbundle, &xvlan); |
46c88433 | 2158 | } |
f0fb825a | 2159 | } else if (xvlan.v[0].vid != out_vlan |
7efbc3b7 | 2160 | && !eth_addr_is_reserved(ctx->xin->flow.dl_dst)) { |
71f21279 | 2161 | struct xbundle *xb; |
f0fb825a | 2162 | uint16_t old_vid = xvlan.v[0].vid; |
9583bc14 | 2163 | |
f0fb825a | 2164 | xvlan.v[0].vid = out_vlan; |
71f21279 BP |
2165 | LIST_FOR_EACH (xb, list_node, &xbridge->xbundles) { |
2166 | if (xbundle_includes_vlan(xb, &xvlan) | |
2167 | && !xbundle_mirror_out(xbridge, xb)) { | |
2168 | output_normal(ctx, xb, &xvlan); | |
9583bc14 EJ |
2169 | } |
2170 | } | |
f0fb825a | 2171 | xvlan.v[0].vid = old_vid; |
9583bc14 | 2172 | } |
faa624b4 BP |
2173 | |
2174 | /* output_normal() could have recursively output (to different | |
2175 | * mirrors), so make sure that we don't send duplicates. */ | |
2176 | mirrors &= ~ctx->mirrors; | |
1356dbd1 | 2177 | ctx->mirror_snaplen = 0; |
9583bc14 | 2178 | } |
a3954fd8 BP |
2179 | |
2180 | if (used_mirrors) { | |
2181 | if (ctx->xin->resubmit_stats) { | |
2182 | mirror_update_stats(xbridge->mbridge, used_mirrors, | |
2183 | ctx->xin->resubmit_stats->n_packets, | |
2184 | ctx->xin->resubmit_stats->n_bytes); | |
2185 | } | |
2186 | if (ctx->xin->xcache) { | |
2187 | struct xc_entry *entry; | |
2188 | ||
2189 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_MIRROR); | |
2190 | entry->mirror.mbridge = mbridge_ref(xbridge->mbridge); | |
2191 | entry->mirror.mirrors = used_mirrors; | |
2192 | } | |
2193 | } | |
9583bc14 EJ |
2194 | } |
2195 | ||
7efbc3b7 BP |
2196 | static void |
2197 | mirror_ingress_packet(struct xlate_ctx *ctx) | |
2198 | { | |
2199 | if (mbridge_has_mirrors(ctx->xbridge->mbridge)) { | |
7efbc3b7 | 2200 | struct xbundle *xbundle = lookup_input_bundle( |
2d9b49dd | 2201 | ctx, ctx->xin->flow.in_port.ofp_port, NULL); |
7efbc3b7 BP |
2202 | if (xbundle) { |
2203 | mirror_packet(ctx, xbundle, | |
2204 | xbundle_mirror_src(ctx->xbridge, xbundle)); | |
2205 | } | |
2206 | } | |
2207 | } | |
2208 | ||
46c88433 | 2209 | /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'. |
2d9b49dd | 2210 | * If so, returns true. Otherwise, returns false. |
9583bc14 EJ |
2211 | * |
2212 | * 'vid' should be the VID obtained from the 802.1Q header that was received as | |
2213 | * part of a packet (specify 0 if there was no 802.1Q header), in the range | |
2214 | * 0...4095. */ | |
2215 | static bool | |
2d9b49dd BP |
2216 | input_vid_is_valid(const struct xlate_ctx *ctx, |
2217 | uint16_t vid, struct xbundle *in_xbundle) | |
9583bc14 EJ |
2218 | { |
2219 | /* Allow any VID on the OFPP_NONE port. */ | |
46c88433 | 2220 | if (in_xbundle == &ofpp_none_bundle) { |
9583bc14 EJ |
2221 | return true; |
2222 | } | |
2223 | ||
46c88433 | 2224 | switch (in_xbundle->vlan_mode) { |
9583bc14 EJ |
2225 | case PORT_VLAN_ACCESS: |
2226 | if (vid) { | |
2d9b49dd BP |
2227 | xlate_report_error(ctx, "dropping VLAN %"PRIu16" tagged " |
2228 | "packet received on port %s configured as VLAN " | |
fd13c6b5 | 2229 | "%d access port", vid, in_xbundle->name, |
2d9b49dd | 2230 | in_xbundle->vlan); |
9583bc14 EJ |
2231 | return false; |
2232 | } | |
2233 | return true; | |
2234 | ||
2235 | case PORT_VLAN_NATIVE_UNTAGGED: | |
2236 | case PORT_VLAN_NATIVE_TAGGED: | |
2237 | if (!vid) { | |
2238 | /* Port must always carry its native VLAN. */ | |
2239 | return true; | |
2240 | } | |
2241 | /* Fall through. */ | |
2242 | case PORT_VLAN_TRUNK: | |
f0fb825a | 2243 | if (!xbundle_trunks_vlan(in_xbundle, vid)) { |
2d9b49dd BP |
2244 | xlate_report_error(ctx, "dropping VLAN %"PRIu16" packet " |
2245 | "received on port %s not configured for " | |
2246 | "trunking VLAN %"PRIu16, | |
2247 | vid, in_xbundle->name, vid); | |
9583bc14 EJ |
2248 | return false; |
2249 | } | |
2250 | return true; | |
2251 | ||
fed8962a EG |
2252 | case PORT_VLAN_DOT1Q_TUNNEL: |
2253 | if (!xbundle_allows_cvlan(in_xbundle, vid)) { | |
2254 | xlate_report_error(ctx, "dropping VLAN %"PRIu16" packet received " | |
2255 | "on dot1q-tunnel port %s that excludes this " | |
2256 | "VLAN", vid, in_xbundle->name); | |
2257 | return false; | |
2258 | } | |
2259 | return true; | |
2260 | ||
9583bc14 | 2261 | default: |
428b2edd | 2262 | OVS_NOT_REACHED(); |
9583bc14 EJ |
2263 | } |
2264 | ||
2265 | } | |
2266 | ||
f0fb825a EG |
2267 | static void |
2268 | xvlan_copy(struct xvlan *dst, const struct xvlan *src) | |
2269 | { | |
2270 | *dst = *src; | |
2271 | } | |
2272 | ||
2273 | static void | |
2274 | xvlan_pop(struct xvlan *src) | |
2275 | { | |
2276 | memmove(&src->v[0], &src->v[1], sizeof(src->v) - sizeof(src->v[0])); | |
2277 | memset(&src->v[FLOW_MAX_VLAN_HEADERS - 1], 0, | |
2278 | sizeof(src->v[FLOW_MAX_VLAN_HEADERS - 1])); | |
2279 | } | |
2280 | ||
fed8962a EG |
2281 | static void |
2282 | xvlan_push_uninit(struct xvlan *src) | |
2283 | { | |
2284 | memmove(&src->v[1], &src->v[0], sizeof(src->v) - sizeof(src->v[0])); | |
2285 | memset(&src->v[0], 0, sizeof(src->v[0])); | |
2286 | } | |
2287 | ||
f0fb825a EG |
2288 | /* Extract VLAN information (headers) from flow */ |
2289 | static void | |
2290 | xvlan_extract(const struct flow *flow, struct xvlan *xvlan) | |
2291 | { | |
2292 | int i; | |
2293 | memset(xvlan, 0, sizeof(*xvlan)); | |
2294 | for (i = 0; i < FLOW_MAX_VLAN_HEADERS; i++) { | |
2295 | if (!eth_type_vlan(flow->vlans[i].tpid) || | |
2296 | !(flow->vlans[i].tci & htons(VLAN_CFI))) { | |
2297 | break; | |
2298 | } | |
2299 | xvlan->v[i].tpid = ntohs(flow->vlans[i].tpid); | |
2300 | xvlan->v[i].vid = vlan_tci_to_vid(flow->vlans[i].tci); | |
2301 | xvlan->v[i].pcp = ntohs(flow->vlans[i].tci) & VLAN_PCP_MASK; | |
2302 | } | |
2303 | } | |
2304 | ||
2305 | /* Put VLAN information (headers) to flow */ | |
2306 | static void | |
ffbe41db EB |
2307 | xvlan_put(struct flow *flow, const struct xvlan *xvlan, |
2308 | enum port_priority_tags_mode use_priority_tags) | |
f0fb825a EG |
2309 | { |
2310 | ovs_be16 tci; | |
2311 | int i; | |
2312 | for (i = 0; i < FLOW_MAX_VLAN_HEADERS; i++) { | |
2313 | tci = htons(xvlan->v[i].vid | (xvlan->v[i].pcp & VLAN_PCP_MASK)); | |
ffbe41db EB |
2314 | if (tci || ((use_priority_tags == PORT_PRIORITY_TAGS_ALWAYS) && |
2315 | xvlan->v[i].tpid)) { | |
f0fb825a EG |
2316 | tci |= htons(VLAN_CFI); |
2317 | flow->vlans[i].tpid = xvlan->v[i].tpid ? | |
2318 | htons(xvlan->v[i].tpid) : | |
2319 | htons(ETH_TYPE_VLAN_8021Q); | |
2320 | } | |
2321 | flow->vlans[i].tci = tci; | |
2322 | } | |
2323 | } | |
2324 | ||
2325 | /* Given 'in_xvlan', extracted from the input 802.1Q headers received as part | |
2326 | * of a packet, and 'in_xbundle', the bundle on which the packet was received, | |
2327 | * returns the VLANs of the packet during bridge internal processing. */ | |
2328 | static void | |
2329 | xvlan_input_translate(const struct xbundle *in_xbundle, | |
2330 | const struct xvlan *in_xvlan, struct xvlan *xvlan) | |
2331 | { | |
2332 | ||
2333 | switch (in_xbundle->vlan_mode) { | |
2334 | case PORT_VLAN_ACCESS: | |
2335 | memset(xvlan, 0, sizeof(*xvlan)); | |
2336 | xvlan->v[0].tpid = in_xvlan->v[0].tpid ? in_xvlan->v[0].tpid : | |
2337 | ETH_TYPE_VLAN_8021Q; | |
2338 | xvlan->v[0].vid = in_xbundle->vlan; | |
2339 | xvlan->v[0].pcp = in_xvlan->v[0].pcp; | |
2340 | break; | |
2341 | ||
2342 | case PORT_VLAN_TRUNK: | |
2343 | xvlan_copy(xvlan, in_xvlan); | |
2344 | break; | |
2345 | ||
2346 | case PORT_VLAN_NATIVE_UNTAGGED: | |
2347 | case PORT_VLAN_NATIVE_TAGGED: | |
2348 | xvlan_copy(xvlan, in_xvlan); | |
2349 | if (!in_xvlan->v[0].vid) { | |
2350 | xvlan->v[0].tpid = in_xvlan->v[0].tpid ? in_xvlan->v[0].tpid : | |
2351 | ETH_TYPE_VLAN_8021Q; | |
2352 | xvlan->v[0].vid = in_xbundle->vlan; | |
2353 | xvlan->v[0].pcp = in_xvlan->v[0].pcp; | |
2354 | } | |
2355 | break; | |
2356 | ||
fed8962a EG |
2357 | case PORT_VLAN_DOT1Q_TUNNEL: |
2358 | xvlan_copy(xvlan, in_xvlan); | |
2359 | xvlan_push_uninit(xvlan); | |
2360 | xvlan->v[0].tpid = in_xbundle->qinq_ethtype; | |
2361 | xvlan->v[0].vid = in_xbundle->vlan; | |
2362 | xvlan->v[0].pcp = 0; | |
2363 | break; | |
2364 | ||
f0fb825a EG |
2365 | default: |
2366 | OVS_NOT_REACHED(); | |
2367 | } | |
2368 | } | |
2369 | ||
2370 | /* Given 'xvlan', the VLANs of a packet during internal processing, and | |
2371 | * 'out_xbundle', a bundle on which the packet is to be output, returns the | |
2372 | * VLANs that should be included in output packet. */ | |
2373 | static void | |
2374 | xvlan_output_translate(const struct xbundle *out_xbundle, | |
2375 | const struct xvlan *xvlan, struct xvlan *out_xvlan) | |
9583bc14 | 2376 | { |
46c88433 | 2377 | switch (out_xbundle->vlan_mode) { |
9583bc14 | 2378 | case PORT_VLAN_ACCESS: |
f0fb825a EG |
2379 | memset(out_xvlan, 0, sizeof(*out_xvlan)); |
2380 | break; | |
9583bc14 EJ |
2381 | |
2382 | case PORT_VLAN_TRUNK: | |
2383 | case PORT_VLAN_NATIVE_TAGGED: | |
f0fb825a EG |
2384 | xvlan_copy(out_xvlan, xvlan); |
2385 | break; | |
9583bc14 EJ |
2386 | |
2387 | case PORT_VLAN_NATIVE_UNTAGGED: | |
f0fb825a EG |
2388 | xvlan_copy(out_xvlan, xvlan); |
2389 | if (xvlan->v[0].vid == out_xbundle->vlan) { | |
2390 | xvlan_pop(out_xvlan); | |
2391 | } | |
2392 | break; | |
9583bc14 | 2393 | |
fed8962a EG |
2394 | case PORT_VLAN_DOT1Q_TUNNEL: |
2395 | xvlan_copy(out_xvlan, xvlan); | |
2396 | xvlan_pop(out_xvlan); | |
2397 | break; | |
2398 | ||
9583bc14 | 2399 | default: |
428b2edd | 2400 | OVS_NOT_REACHED(); |
9583bc14 EJ |
2401 | } |
2402 | } | |
2403 | ||
fed8962a EG |
2404 | /* If output xbundle is dot1q-tunnel, set mask bits of cvlan */ |
2405 | static void | |
2406 | check_and_set_cvlan_mask(struct flow_wildcards *wc, | |
2407 | const struct xbundle *xbundle) | |
2408 | { | |
2409 | if (xbundle->vlan_mode == PORT_VLAN_DOT1Q_TUNNEL && xbundle->cvlans) { | |
2410 | wc->masks.vlans[1].tci = htons(0xffff); | |
2411 | } | |
2412 | } | |
2413 | ||
9583bc14 | 2414 | static void |
46c88433 | 2415 | output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, |
f0fb825a | 2416 | const struct xvlan *xvlan) |
9583bc14 | 2417 | { |
9583bc14 | 2418 | uint16_t vid; |
f0fb825a | 2419 | union flow_vlan_hdr old_vlans[FLOW_MAX_VLAN_HEADERS]; |
46c88433 | 2420 | struct xport *xport; |
e93ef1c7 JR |
2421 | struct xlate_bond_recirc xr; |
2422 | bool use_recirc = false; | |
f0fb825a | 2423 | struct xvlan out_xvlan; |
9583bc14 | 2424 | |
fed8962a EG |
2425 | check_and_set_cvlan_mask(ctx->wc, out_xbundle); |
2426 | ||
f0fb825a EG |
2427 | xvlan_output_translate(out_xbundle, xvlan, &out_xvlan); |
2428 | if (out_xbundle->use_priority_tags) { | |
2429 | out_xvlan.v[0].pcp = ntohs(ctx->xin->flow.vlans[0].tci) & | |
2430 | VLAN_PCP_MASK; | |
2431 | } | |
2432 | vid = out_xvlan.v[0].vid; | |
417e7e66 | 2433 | if (ovs_list_is_empty(&out_xbundle->xports)) { |
46c88433 EJ |
2434 | /* Partially configured bundle with no slaves. Drop the packet. */ |
2435 | return; | |
2436 | } else if (!out_xbundle->bond) { | |
417e7e66 | 2437 | xport = CONTAINER_OF(ovs_list_front(&out_xbundle->xports), struct xport, |
46c88433 | 2438 | bundle_node); |
9583bc14 | 2439 | } else { |
49a73e0c | 2440 | struct flow_wildcards *wc = ctx->wc; |
84f0f298 | 2441 | struct ofport_dpif *ofport; |
adcf00ba | 2442 | |
a80aba3a AZ |
2443 | if (ctx->xbridge->support.odp.recirc) { |
2444 | /* In case recirculation is not actually in use, 'xr.recirc_id' | |
2445 | * will be set to '0', since a valid 'recirc_id' can | |
82f9f1f5 AZ |
2446 | * not be zero. */ |
2447 | bond_update_post_recirc_rules(out_xbundle->bond, | |
2448 | &xr.recirc_id, | |
2449 | &xr.hash_basis); | |
2450 | if (xr.recirc_id) { | |
2451 | /* Use recirculation instead of output. */ | |
2452 | use_recirc = true; | |
e93ef1c7 | 2453 | xr.hash_alg = OVS_HASH_ALG_L4; |
54ecb5a2 AZ |
2454 | /* Recirculation does not require unmasking hash fields. */ |
2455 | wc = NULL; | |
adcf00ba AZ |
2456 | } |
2457 | } | |
46c88433 | 2458 | |
54ecb5a2 AZ |
2459 | ofport = bond_choose_output_slave(out_xbundle->bond, |
2460 | &ctx->xin->flow, wc, vid); | |
0506f184 | 2461 | xport = xport_lookup(ctx->xcfg, ofport); |
46c88433 EJ |
2462 | |
2463 | if (!xport) { | |
9583bc14 EJ |
2464 | /* No slaves enabled, so drop packet. */ |
2465 | return; | |
2466 | } | |
d6fc5f57 | 2467 | |
e93ef1c7 | 2468 | /* If use_recirc is set, the main thread will handle stats |
b256dc52 | 2469 | * accounting for this bond. */ |
e93ef1c7 | 2470 | if (!use_recirc) { |
b256dc52 JS |
2471 | if (ctx->xin->resubmit_stats) { |
2472 | bond_account(out_xbundle->bond, &ctx->xin->flow, vid, | |
2473 | ctx->xin->resubmit_stats->n_bytes); | |
2474 | } | |
2475 | if (ctx->xin->xcache) { | |
2476 | struct xc_entry *entry; | |
2477 | struct flow *flow; | |
2478 | ||
2479 | flow = &ctx->xin->flow; | |
2480 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND); | |
901a517e JR |
2481 | entry->bond.bond = bond_ref(out_xbundle->bond); |
2482 | entry->bond.flow = xmemdup(flow, sizeof *flow); | |
2483 | entry->bond.vid = vid; | |
b256dc52 | 2484 | } |
d6fc5f57 | 2485 | } |
9583bc14 EJ |
2486 | } |
2487 | ||
f0fb825a | 2488 | memcpy(&old_vlans, &ctx->xin->flow.vlans, sizeof(old_vlans)); |
ffbe41db | 2489 | xvlan_put(&ctx->xin->flow, &out_xvlan, out_xbundle->use_priority_tags); |
9583bc14 | 2490 | |
feee58b9 | 2491 | compose_output_action(ctx, xport->ofp_port, use_recirc ? &xr : NULL, |
11938578 | 2492 | false, false); |
f0fb825a | 2493 | memcpy(&ctx->xin->flow.vlans, &old_vlans, sizeof(old_vlans)); |
9583bc14 EJ |
2494 | } |
2495 | ||
2496 | /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after | |
2497 | * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to | |
2498 | * indicate this; newer upstream kernels use gratuitous ARP requests. */ | |
2499 | static bool | |
2500 | is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) | |
2501 | { | |
2502 | if (flow->dl_type != htons(ETH_TYPE_ARP)) { | |
2503 | return false; | |
2504 | } | |
2505 | ||
2506 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
2507 | if (!eth_addr_is_broadcast(flow->dl_dst)) { | |
2508 | return false; | |
2509 | } | |
2510 | ||
2511 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
2512 | if (flow->nw_proto == ARP_OP_REPLY) { | |
2513 | return true; | |
2514 | } else if (flow->nw_proto == ARP_OP_REQUEST) { | |
2515 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
2516 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
2517 | ||
2518 | return flow->nw_src == flow->nw_dst; | |
2519 | } else { | |
2520 | return false; | |
2521 | } | |
2522 | } | |
2523 | ||
ff69c24a FL |
2524 | /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or |
2525 | * dropped. Returns true if they may be forwarded, false if they should be | |
2526 | * dropped. | |
2527 | * | |
2528 | * 'in_port' must be the xport that corresponds to flow->in_port. | |
2529 | * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). | |
2530 | * | |
2531 | * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as | |
2532 | * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as | |
2533 | * checked by input_vid_is_valid(). | |
2534 | * | |
2535 | * May also add tags to '*tags', although the current implementation only does | |
2536 | * so in one special case. | |
2537 | */ | |
2538 | static bool | |
2539 | is_admissible(struct xlate_ctx *ctx, struct xport *in_port, | |
2540 | uint16_t vlan) | |
2541 | { | |
2542 | struct xbundle *in_xbundle = in_port->xbundle; | |
2543 | const struct xbridge *xbridge = ctx->xbridge; | |
2544 | struct flow *flow = &ctx->xin->flow; | |
2545 | ||
2546 | /* Drop frames for reserved multicast addresses | |
2547 | * only if forward_bpdu option is absent. */ | |
2548 | if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { | |
2d9b49dd BP |
2549 | xlate_report(ctx, OFT_DETAIL, |
2550 | "packet has reserved destination MAC, dropping"); | |
ff69c24a FL |
2551 | return false; |
2552 | } | |
2553 | ||
2554 | if (in_xbundle->bond) { | |
2555 | struct mac_entry *mac; | |
2556 | ||
2557 | switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, | |
2558 | flow->dl_dst)) { | |
2559 | case BV_ACCEPT: | |
2560 | break; | |
2561 | ||
2562 | case BV_DROP: | |
2d9b49dd BP |
2563 | xlate_report(ctx, OFT_DETAIL, |
2564 | "bonding refused admissibility, dropping"); | |
ff69c24a FL |
2565 | return false; |
2566 | ||
2567 | case BV_DROP_IF_MOVED: | |
2568 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
2569 | mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan); | |
9d078ec2 BP |
2570 | if (mac |
2571 | && mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle | |
49a73e0c | 2572 | && (!is_gratuitous_arp(flow, ctx->wc) |
9d078ec2 | 2573 | || mac_entry_is_grat_arp_locked(mac))) { |
ff69c24a | 2574 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
2d9b49dd BP |
2575 | xlate_report(ctx, OFT_DETAIL, |
2576 | "SLB bond thinks this packet looped back, " | |
ff69c24a FL |
2577 | "dropping"); |
2578 | return false; | |
2579 | } | |
2580 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
2581 | break; | |
2582 | } | |
2583 | } | |
2584 | ||
2585 | return true; | |
2586 | } | |
2587 | ||
2d9b49dd BP |
2588 | static bool |
2589 | update_learning_table__(const struct xbridge *xbridge, | |
2590 | struct xbundle *in_xbundle, struct eth_addr dl_src, | |
2591 | int vlan, bool is_grat_arp) | |
2592 | { | |
2593 | return (in_xbundle == &ofpp_none_bundle | |
2594 | || !mac_learning_update(xbridge->ml, dl_src, vlan, | |
2595 | is_grat_arp, | |
2596 | in_xbundle->bond != NULL, | |
2597 | in_xbundle->ofbundle)); | |
2598 | } | |
2599 | ||
ee047520 | 2600 | static void |
2d9b49dd | 2601 | update_learning_table(const struct xlate_ctx *ctx, |
064799a1 JR |
2602 | struct xbundle *in_xbundle, struct eth_addr dl_src, |
2603 | int vlan, bool is_grat_arp) | |
ee047520 | 2604 | { |
2d9b49dd BP |
2605 | if (!update_learning_table__(ctx->xbridge, in_xbundle, dl_src, vlan, |
2606 | is_grat_arp)) { | |
2607 | xlate_report_debug(ctx, OFT_DETAIL, "learned that "ETH_ADDR_FMT" is " | |
2608 | "on port %s in VLAN %d", | |
2609 | ETH_ADDR_ARGS(dl_src), in_xbundle->name, vlan); | |
ee047520 | 2610 | } |
9583bc14 EJ |
2611 | } |
2612 | ||
86e2dcdd FL |
2613 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2614 | * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */ | |
2615 | static void | |
2d9b49dd | 2616 | update_mcast_snooping_table4__(const struct xlate_ctx *ctx, |
06994f87 TLSC |
2617 | const struct flow *flow, |
2618 | struct mcast_snooping *ms, int vlan, | |
2619 | struct xbundle *in_xbundle, | |
2620 | const struct dp_packet *packet) | |
86e2dcdd FL |
2621 | OVS_REQ_WRLOCK(ms->rwlock) |
2622 | { | |
46445c63 | 2623 | const struct igmp_header *igmp; |
e3102e42 | 2624 | int count; |
46445c63 | 2625 | size_t offset; |
06994f87 | 2626 | ovs_be32 ip4 = flow->igmp_group_ip4; |
86e2dcdd | 2627 | |
46445c63 EC |
2628 | offset = (char *) dp_packet_l4(packet) - (char *) dp_packet_data(packet); |
2629 | igmp = dp_packet_at(packet, offset, IGMP_HEADER_LEN); | |
2630 | if (!igmp || csum(igmp, dp_packet_l4_size(packet)) != 0) { | |
2d9b49dd BP |
2631 | xlate_report_debug(ctx, OFT_DETAIL, |
2632 | "multicast snooping received bad IGMP " | |
2633 | "checksum on port %s in VLAN %d", | |
2634 | in_xbundle->name, vlan); | |
46445c63 EC |
2635 | return; |
2636 | } | |
2637 | ||
86e2dcdd FL |
2638 | switch (ntohs(flow->tp_src)) { |
2639 | case IGMP_HOST_MEMBERSHIP_REPORT: | |
2640 | case IGMPV2_HOST_MEMBERSHIP_REPORT: | |
964a4d5f | 2641 | if (mcast_snooping_add_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
2d9b49dd BP |
2642 | xlate_report_debug(ctx, OFT_DETAIL, |
2643 | "multicast snooping learned that " | |
2644 | IP_FMT" is on port %s in VLAN %d", | |
2645 | IP_ARGS(ip4), in_xbundle->name, vlan); | |
86e2dcdd FL |
2646 | } |
2647 | break; | |
2648 | case IGMP_HOST_LEAVE_MESSAGE: | |
964a4d5f | 2649 | if (mcast_snooping_leave_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
2d9b49dd BP |
2650 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping leaving " |
2651 | IP_FMT" is on port %s in VLAN %d", | |
2652 | IP_ARGS(ip4), in_xbundle->name, vlan); | |
86e2dcdd FL |
2653 | } |
2654 | break; | |
2655 | case IGMP_HOST_MEMBERSHIP_QUERY: | |
2656 | if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan, | |
2d9b49dd BP |
2657 | in_xbundle->ofbundle)) { |
2658 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping query " | |
2659 | "from "IP_FMT" is on port %s in VLAN %d", | |
2660 | IP_ARGS(flow->nw_src), in_xbundle->name, vlan); | |
86e2dcdd FL |
2661 | } |
2662 | break; | |
e3102e42 | 2663 | case IGMPV3_HOST_MEMBERSHIP_REPORT: |
2d9b49dd BP |
2664 | count = mcast_snooping_add_report(ms, packet, vlan, |
2665 | in_xbundle->ofbundle); | |
2666 | if (count) { | |
2667 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping processed " | |
2668 | "%d addresses on port %s in VLAN %d", | |
2669 | count, in_xbundle->name, vlan); | |
e3102e42 TLSC |
2670 | } |
2671 | break; | |
86e2dcdd FL |
2672 | } |
2673 | } | |
2674 | ||
06994f87 | 2675 | static void |
2d9b49dd | 2676 | update_mcast_snooping_table6__(const struct xlate_ctx *ctx, |
06994f87 TLSC |
2677 | const struct flow *flow, |
2678 | struct mcast_snooping *ms, int vlan, | |
2679 | struct xbundle *in_xbundle, | |
2680 | const struct dp_packet *packet) | |
2681 | OVS_REQ_WRLOCK(ms->rwlock) | |
2682 | { | |
46445c63 | 2683 | const struct mld_header *mld; |
06994f87 | 2684 | int count; |
46445c63 EC |
2685 | size_t offset; |
2686 | ||
2687 | offset = (char *) dp_packet_l4(packet) - (char *) dp_packet_data(packet); | |
2688 | mld = dp_packet_at(packet, offset, MLD_HEADER_LEN); | |
2689 | ||
2690 | if (!mld || | |
2691 | packet_csum_upperlayer6(dp_packet_l3(packet), | |
2692 | mld, IPPROTO_ICMPV6, | |
2693 | dp_packet_l4_size(packet)) != 0) { | |
2d9b49dd BP |
2694 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping received " |
2695 | "bad MLD checksum on port %s in VLAN %d", | |
2696 | in_xbundle->name, vlan); | |
46445c63 EC |
2697 | return; |
2698 | } | |
06994f87 TLSC |
2699 | |
2700 | switch (ntohs(flow->tp_src)) { | |
2701 | case MLD_QUERY: | |
2702 | if (!ipv6_addr_equals(&flow->ipv6_src, &in6addr_any) | |
2703 | && mcast_snooping_add_mrouter(ms, vlan, in_xbundle->ofbundle)) { | |
2d9b49dd BP |
2704 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping query on " |
2705 | "port %s in VLAN %d", in_xbundle->name, vlan); | |
06994f87 TLSC |
2706 | } |
2707 | break; | |
2708 | case MLD_REPORT: | |
2709 | case MLD_DONE: | |
2710 | case MLD2_REPORT: | |
2711 | count = mcast_snooping_add_mld(ms, packet, vlan, in_xbundle->ofbundle); | |
2712 | if (count) { | |
2d9b49dd BP |
2713 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping processed " |
2714 | "%d addresses on port %s in VLAN %d", | |
2715 | count, in_xbundle->name, vlan); | |
06994f87 TLSC |
2716 | } |
2717 | break; | |
2718 | } | |
2719 | } | |
2720 | ||
86e2dcdd FL |
2721 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2722 | * was received on 'in_xbundle' in 'vlan'. */ | |
2723 | static void | |
2d9b49dd | 2724 | update_mcast_snooping_table(const struct xlate_ctx *ctx, |
86e2dcdd | 2725 | const struct flow *flow, int vlan, |
e3102e42 TLSC |
2726 | struct xbundle *in_xbundle, |
2727 | const struct dp_packet *packet) | |
86e2dcdd | 2728 | { |
2d9b49dd | 2729 | struct mcast_snooping *ms = ctx->xbridge->ms; |
86e2dcdd | 2730 | struct xbundle *mcast_xbundle; |
f4ae6e23 | 2731 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2732 | |
2733 | /* Don't learn the OFPP_NONE port. */ | |
2734 | if (in_xbundle == &ofpp_none_bundle) { | |
2735 | return; | |
2736 | } | |
2737 | ||
2738 | /* Don't learn from flood ports */ | |
2739 | mcast_xbundle = NULL; | |
2740 | ovs_rwlock_wrlock(&ms->rwlock); | |
f4ae6e23 | 2741 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
0506f184 | 2742 | mcast_xbundle = xbundle_lookup(ctx->xcfg, fport->port); |
86e2dcdd FL |
2743 | if (mcast_xbundle == in_xbundle) { |
2744 | break; | |
2745 | } | |
2746 | } | |
2747 | ||
2748 | if (!mcast_xbundle || mcast_xbundle != in_xbundle) { | |
06994f87 | 2749 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2d9b49dd | 2750 | update_mcast_snooping_table4__(ctx, flow, ms, vlan, |
06994f87 TLSC |
2751 | in_xbundle, packet); |
2752 | } else { | |
2d9b49dd | 2753 | update_mcast_snooping_table6__(ctx, flow, ms, vlan, |
06994f87 TLSC |
2754 | in_xbundle, packet); |
2755 | } | |
86e2dcdd FL |
2756 | } |
2757 | ovs_rwlock_unlock(&ms->rwlock); | |
2758 | } | |
efc93e66 BP |
2759 | \f |
2760 | /* A list of multicast output ports. | |
2761 | * | |
2762 | * We accumulate output ports and then do all the outputs afterward. It would | |
2763 | * be more natural to do the outputs one at a time as we discover the need for | |
2764 | * each one, but this can cause a deadlock because we need to take the | |
2765 | * mcast_snooping's rwlock for reading to iterate through the port lists and | |
2766 | * doing an output, if it goes to a patch port, can eventually come back to the | |
2767 | * same mcast_snooping and attempt to take the write lock (see | |
2768 | * https://github.com/openvswitch/ovs-issues/issues/153). */ | |
2769 | struct mcast_output { | |
2770 | /* Discrete ports. */ | |
2771 | struct xbundle **xbundles; | |
2772 | size_t n, allocated; | |
2773 | ||
2774 | /* If set, flood to all ports. */ | |
2775 | bool flood; | |
2776 | }; | |
2777 | #define MCAST_OUTPUT_INIT { NULL, 0, 0, false } | |
2778 | ||
2779 | /* Add 'mcast_bundle' to 'out'. */ | |
2780 | static void | |
2781 | mcast_output_add(struct mcast_output *out, struct xbundle *mcast_xbundle) | |
2782 | { | |
2783 | if (out->n >= out->allocated) { | |
2784 | out->xbundles = x2nrealloc(out->xbundles, &out->allocated, | |
2785 | sizeof *out->xbundles); | |
2786 | } | |
2787 | out->xbundles[out->n++] = mcast_xbundle; | |
2788 | } | |
2789 | ||
2790 | /* Outputs the packet in 'ctx' to all of the output ports in 'out', given input | |
2791 | * bundle 'in_xbundle' and the current 'xvlan'. */ | |
2792 | static void | |
2793 | mcast_output_finish(struct xlate_ctx *ctx, struct mcast_output *out, | |
2794 | struct xbundle *in_xbundle, struct xvlan *xvlan) | |
2795 | { | |
2796 | if (out->flood) { | |
2797 | xlate_normal_flood(ctx, in_xbundle, xvlan); | |
2798 | } else { | |
2799 | for (size_t i = 0; i < out->n; i++) { | |
2800 | output_normal(ctx, out->xbundles[i], xvlan); | |
2801 | } | |
2802 | } | |
2803 | ||
2804 | free(out->xbundles); | |
2805 | } | |
86e2dcdd FL |
2806 | |
2807 | /* send the packet to ports having the multicast group learned */ | |
2808 | static void | |
2809 | xlate_normal_mcast_send_group(struct xlate_ctx *ctx, | |
2810 | struct mcast_snooping *ms OVS_UNUSED, | |
2811 | struct mcast_group *grp, | |
f0fb825a | 2812 | struct xbundle *in_xbundle, |
efc93e66 | 2813 | struct mcast_output *out) |
86e2dcdd FL |
2814 | OVS_REQ_RDLOCK(ms->rwlock) |
2815 | { | |
86e2dcdd FL |
2816 | struct mcast_group_bundle *b; |
2817 | struct xbundle *mcast_xbundle; | |
2818 | ||
86e2dcdd | 2819 | LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) { |
0506f184 | 2820 | mcast_xbundle = xbundle_lookup(ctx->xcfg, b->port); |
86e2dcdd | 2821 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { |
2d9b49dd | 2822 | xlate_report(ctx, OFT_DETAIL, "forwarding to mcast group port"); |
efc93e66 | 2823 | mcast_output_add(out, mcast_xbundle); |
86e2dcdd | 2824 | } else if (!mcast_xbundle) { |
2d9b49dd BP |
2825 | xlate_report(ctx, OFT_WARN, |
2826 | "mcast group port is unknown, dropping"); | |
86e2dcdd | 2827 | } else { |
2d9b49dd BP |
2828 | xlate_report(ctx, OFT_DETAIL, |
2829 | "mcast group port is input port, dropping"); | |
86e2dcdd FL |
2830 | } |
2831 | } | |
2832 | } | |
2833 | ||
2834 | /* send the packet to ports connected to multicast routers */ | |
2835 | static void | |
2836 | xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx, | |
2837 | struct mcast_snooping *ms, | |
f0fb825a | 2838 | struct xbundle *in_xbundle, |
efc93e66 BP |
2839 | const struct xvlan *xvlan, |
2840 | struct mcast_output *out) | |
86e2dcdd FL |
2841 | OVS_REQ_RDLOCK(ms->rwlock) |
2842 | { | |
86e2dcdd FL |
2843 | struct mcast_mrouter_bundle *mrouter; |
2844 | struct xbundle *mcast_xbundle; | |
2845 | ||
86e2dcdd | 2846 | LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) { |
0506f184 | 2847 | mcast_xbundle = xbundle_lookup(ctx->xcfg, mrouter->port); |
94a881c1 | 2848 | if (mcast_xbundle && mcast_xbundle != in_xbundle |
f0fb825a | 2849 | && mrouter->vlan == xvlan->v[0].vid) { |
2d9b49dd | 2850 | xlate_report(ctx, OFT_DETAIL, "forwarding to mcast router port"); |
efc93e66 | 2851 | mcast_output_add(out, mcast_xbundle); |
86e2dcdd | 2852 | } else if (!mcast_xbundle) { |
2d9b49dd BP |
2853 | xlate_report(ctx, OFT_WARN, |
2854 | "mcast router port is unknown, dropping"); | |
f0fb825a | 2855 | } else if (mrouter->vlan != xvlan->v[0].vid) { |
2d9b49dd BP |
2856 | xlate_report(ctx, OFT_DETAIL, |
2857 | "mcast router is on another vlan, dropping"); | |
86e2dcdd | 2858 | } else { |
2d9b49dd BP |
2859 | xlate_report(ctx, OFT_DETAIL, |
2860 | "mcast router port is input port, dropping"); | |
86e2dcdd FL |
2861 | } |
2862 | } | |
2863 | } | |
2864 | ||
2865 | /* send the packet to ports flagged to be flooded */ | |
2866 | static void | |
2867 | xlate_normal_mcast_send_fports(struct xlate_ctx *ctx, | |
2868 | struct mcast_snooping *ms, | |
f0fb825a | 2869 | struct xbundle *in_xbundle, |
efc93e66 | 2870 | struct mcast_output *out) |
86e2dcdd FL |
2871 | OVS_REQ_RDLOCK(ms->rwlock) |
2872 | { | |
f4ae6e23 | 2873 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2874 | struct xbundle *mcast_xbundle; |
2875 | ||
f4ae6e23 | 2876 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
0506f184 | 2877 | mcast_xbundle = xbundle_lookup(ctx->xcfg, fport->port); |
86e2dcdd | 2878 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { |
2d9b49dd | 2879 | xlate_report(ctx, OFT_DETAIL, "forwarding to mcast flood port"); |
efc93e66 | 2880 | mcast_output_add(out, mcast_xbundle); |
86e2dcdd | 2881 | } else if (!mcast_xbundle) { |
2d9b49dd BP |
2882 | xlate_report(ctx, OFT_WARN, |
2883 | "mcast flood port is unknown, dropping"); | |
86e2dcdd | 2884 | } else { |
2d9b49dd BP |
2885 | xlate_report(ctx, OFT_DETAIL, |
2886 | "mcast flood port is input port, dropping"); | |
86e2dcdd FL |
2887 | } |
2888 | } | |
2889 | } | |
2890 | ||
8e04a33f FL |
2891 | /* forward the Reports to configured ports */ |
2892 | static void | |
2893 | xlate_normal_mcast_send_rports(struct xlate_ctx *ctx, | |
2894 | struct mcast_snooping *ms, | |
f0fb825a | 2895 | struct xbundle *in_xbundle, |
efc93e66 | 2896 | struct mcast_output *out) |
8e04a33f FL |
2897 | OVS_REQ_RDLOCK(ms->rwlock) |
2898 | { | |
8e04a33f FL |
2899 | struct mcast_port_bundle *rport; |
2900 | struct xbundle *mcast_xbundle; | |
2901 | ||
8e04a33f | 2902 | LIST_FOR_EACH(rport, node, &ms->rport_list) { |
0506f184 | 2903 | mcast_xbundle = xbundle_lookup(ctx->xcfg, rport->port); |
c781bd52 LJ |
2904 | if (mcast_xbundle |
2905 | && mcast_xbundle != in_xbundle | |
2906 | && mcast_xbundle->ofbundle != in_xbundle->ofbundle) { | |
2d9b49dd BP |
2907 | xlate_report(ctx, OFT_DETAIL, |
2908 | "forwarding report to mcast flagged port"); | |
efc93e66 | 2909 | mcast_output_add(out, mcast_xbundle); |
8e04a33f | 2910 | } else if (!mcast_xbundle) { |
2d9b49dd BP |
2911 | xlate_report(ctx, OFT_WARN, |
2912 | "mcast port is unknown, dropping the report"); | |
8e04a33f | 2913 | } else { |
2d9b49dd BP |
2914 | xlate_report(ctx, OFT_DETAIL, |
2915 | "mcast port is input port, dropping the Report"); | |
8e04a33f FL |
2916 | } |
2917 | } | |
2918 | } | |
2919 | ||
682800a4 FL |
2920 | static void |
2921 | xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, | |
f0fb825a | 2922 | struct xvlan *xvlan) |
682800a4 FL |
2923 | { |
2924 | struct xbundle *xbundle; | |
2925 | ||
2926 | LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { | |
2927 | if (xbundle != in_xbundle | |
c781bd52 | 2928 | && xbundle->ofbundle != in_xbundle->ofbundle |
f0fb825a | 2929 | && xbundle_includes_vlan(xbundle, xvlan) |
682800a4 FL |
2930 | && xbundle->floodable |
2931 | && !xbundle_mirror_out(ctx->xbridge, xbundle)) { | |
f0fb825a | 2932 | output_normal(ctx, xbundle, xvlan); |
682800a4 FL |
2933 | } |
2934 | } | |
2031ef97 | 2935 | ctx->nf_output_iface = NF_OUT_FLOOD; |
682800a4 FL |
2936 | } |
2937 | ||
a75636c8 BP |
2938 | static bool |
2939 | is_ip_local_multicast(const struct flow *flow, struct flow_wildcards *wc) | |
2940 | { | |
2941 | if (flow->dl_type == htons(ETH_TYPE_IP)) { | |
2942 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
2943 | return ip_is_local_multicast(flow->nw_dst); | |
2944 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
2945 | memset(&wc->masks.ipv6_dst, 0xff, sizeof wc->masks.ipv6_dst); | |
2946 | return ipv6_is_all_hosts(&flow->ipv6_dst); | |
2947 | } else { | |
2948 | return false; | |
2949 | } | |
2950 | } | |
2951 | ||
9583bc14 EJ |
2952 | static void |
2953 | xlate_normal(struct xlate_ctx *ctx) | |
2954 | { | |
49a73e0c | 2955 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 2956 | struct flow *flow = &ctx->xin->flow; |
46c88433 EJ |
2957 | struct xbundle *in_xbundle; |
2958 | struct xport *in_port; | |
9583bc14 | 2959 | struct mac_entry *mac; |
d6d5bbc9 | 2960 | void *mac_port; |
f0fb825a EG |
2961 | struct xvlan in_xvlan; |
2962 | struct xvlan xvlan; | |
9583bc14 | 2963 | uint16_t vlan; |
9583bc14 | 2964 | |
33bf9176 BP |
2965 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
2966 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
f0fb825a | 2967 | wc->masks.vlans[0].tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
9583bc14 | 2968 | |
2d9b49dd | 2969 | in_xbundle = lookup_input_bundle(ctx, flow->in_port.ofp_port, &in_port); |
46c88433 | 2970 | if (!in_xbundle) { |
2d9b49dd | 2971 | xlate_report(ctx, OFT_WARN, "no input bundle, dropping"); |
9583bc14 EJ |
2972 | return; |
2973 | } | |
2974 | ||
2975 | /* Drop malformed frames. */ | |
f0fb825a EG |
2976 | if (eth_type_vlan(flow->dl_type) && |
2977 | !(flow->vlans[0].tci & htons(VLAN_CFI))) { | |
9583bc14 | 2978 | if (ctx->xin->packet != NULL) { |
2d9b49dd BP |
2979 | xlate_report_error(ctx, "dropping packet with partial " |
2980 | "VLAN tag received on port %s", | |
2981 | in_xbundle->name); | |
9583bc14 | 2982 | } |
2d9b49dd | 2983 | xlate_report(ctx, OFT_WARN, "partial VLAN tag, dropping"); |
9583bc14 EJ |
2984 | return; |
2985 | } | |
2986 | ||
2987 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 2988 | if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) { |
9583bc14 | 2989 | if (ctx->xin->packet != NULL) { |
2d9b49dd BP |
2990 | xlate_report_error(ctx, "dropping packet received on port %s, " |
2991 | "which is reserved exclusively for mirroring", | |
2992 | in_xbundle->name); | |
9583bc14 | 2993 | } |
2d9b49dd BP |
2994 | xlate_report(ctx, OFT_WARN, |
2995 | "input port is mirror output port, dropping"); | |
9583bc14 EJ |
2996 | return; |
2997 | } | |
2998 | ||
2999 | /* Check VLAN. */ | |
f0fb825a EG |
3000 | xvlan_extract(flow, &in_xvlan); |
3001 | if (!input_vid_is_valid(ctx, in_xvlan.v[0].vid, in_xbundle)) { | |
2d9b49dd BP |
3002 | xlate_report(ctx, OFT_WARN, |
3003 | "disallowed VLAN VID for this input port, dropping"); | |
9583bc14 EJ |
3004 | return; |
3005 | } | |
f0fb825a EG |
3006 | xvlan_input_translate(in_xbundle, &in_xvlan, &xvlan); |
3007 | vlan = xvlan.v[0].vid; | |
9583bc14 EJ |
3008 | |
3009 | /* Check other admissibility requirements. */ | |
3010 | if (in_port && !is_admissible(ctx, in_port, vlan)) { | |
3011 | return; | |
3012 | } | |
3013 | ||
3014 | /* Learn source MAC. */ | |
064799a1 | 3015 | bool is_grat_arp = is_gratuitous_arp(flow, wc); |
875ab130 BP |
3016 | if (ctx->xin->allow_side_effects |
3017 | && flow->packet_type == htonl(PT_ETH) | |
3018 | && in_port->pt_mode != NETDEV_PT_LEGACY_L3 | |
3019 | ) { | |
2d9b49dd | 3020 | update_learning_table(ctx, in_xbundle, flow->dl_src, vlan, |
064799a1 | 3021 | is_grat_arp); |
9583bc14 | 3022 | } |
064799a1 | 3023 | if (ctx->xin->xcache && in_xbundle != &ofpp_none_bundle) { |
b256dc52 JS |
3024 | struct xc_entry *entry; |
3025 | ||
064799a1 | 3026 | /* Save just enough info to update mac learning table later. */ |
b256dc52 | 3027 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL); |
901a517e JR |
3028 | entry->normal.ofproto = ctx->xbridge->ofproto; |
3029 | entry->normal.in_port = flow->in_port.ofp_port; | |
3030 | entry->normal.dl_src = flow->dl_src; | |
3031 | entry->normal.vlan = vlan; | |
3032 | entry->normal.is_gratuitous_arp = is_grat_arp; | |
b256dc52 | 3033 | } |
9583bc14 EJ |
3034 | |
3035 | /* Determine output bundle. */ | |
86e2dcdd FL |
3036 | if (mcast_snooping_enabled(ctx->xbridge->ms) |
3037 | && !eth_addr_is_broadcast(flow->dl_dst) | |
3038 | && eth_addr_is_multicast(flow->dl_dst) | |
06994f87 | 3039 | && is_ip_any(flow)) { |
86e2dcdd | 3040 | struct mcast_snooping *ms = ctx->xbridge->ms; |
06994f87 | 3041 | struct mcast_group *grp = NULL; |
86e2dcdd | 3042 | |
a75636c8 | 3043 | if (is_igmp(flow, wc)) { |
1bc24169 BP |
3044 | /* |
3045 | * IGMP packets need to take the slow path, in order to be | |
3046 | * processed for mdb updates. That will prevent expires | |
3047 | * firing off even after hosts have sent reports. | |
3048 | */ | |
3049 | ctx->xout->slow |= SLOW_ACTION; | |
3050 | ||
a75636c8 | 3051 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); |
d29f137b TLSC |
3052 | if (mcast_snooping_is_membership(flow->tp_src) || |
3053 | mcast_snooping_is_query(flow->tp_src)) { | |
df70a773 | 3054 | if (ctx->xin->allow_side_effects && ctx->xin->packet) { |
2d9b49dd | 3055 | update_mcast_snooping_table(ctx, flow, vlan, |
e3102e42 | 3056 | in_xbundle, ctx->xin->packet); |
d29f137b | 3057 | } |
86e2dcdd | 3058 | } |
d6d5bbc9 | 3059 | |
86e2dcdd | 3060 | if (mcast_snooping_is_membership(flow->tp_src)) { |
efc93e66 BP |
3061 | struct mcast_output out = MCAST_OUTPUT_INIT; |
3062 | ||
86e2dcdd | 3063 | ovs_rwlock_rdlock(&ms->rwlock); |
efc93e66 BP |
3064 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, &xvlan, |
3065 | &out); | |
8e04a33f FL |
3066 | /* RFC4541: section 2.1.1, item 1: A snooping switch should |
3067 | * forward IGMP Membership Reports only to those ports where | |
3068 | * multicast routers are attached. Alternatively stated: a | |
3069 | * snooping switch should not forward IGMP Membership Reports | |
3070 | * to ports on which only hosts are attached. | |
3071 | * An administrative control may be provided to override this | |
3072 | * restriction, allowing the report messages to be flooded to | |
3073 | * other ports. */ | |
efc93e66 | 3074 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, &out); |
86e2dcdd | 3075 | ovs_rwlock_unlock(&ms->rwlock); |
efc93e66 BP |
3076 | |
3077 | mcast_output_finish(ctx, &out, in_xbundle, &xvlan); | |
86e2dcdd | 3078 | } else { |
2d9b49dd | 3079 | xlate_report(ctx, OFT_DETAIL, "multicast traffic, flooding"); |
f0fb825a | 3080 | xlate_normal_flood(ctx, in_xbundle, &xvlan); |
86e2dcdd FL |
3081 | } |
3082 | return; | |
a75636c8 | 3083 | } else if (is_mld(flow, wc)) { |
06994f87 | 3084 | ctx->xout->slow |= SLOW_ACTION; |
df70a773 | 3085 | if (ctx->xin->allow_side_effects && ctx->xin->packet) { |
2d9b49dd | 3086 | update_mcast_snooping_table(ctx, flow, vlan, |
06994f87 TLSC |
3087 | in_xbundle, ctx->xin->packet); |
3088 | } | |
a75636c8 | 3089 | if (is_mld_report(flow, wc)) { |
efc93e66 BP |
3090 | struct mcast_output out = MCAST_OUTPUT_INIT; |
3091 | ||
06994f87 | 3092 | ovs_rwlock_rdlock(&ms->rwlock); |
efc93e66 BP |
3093 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, &xvlan, |
3094 | &out); | |
3095 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, &out); | |
06994f87 | 3096 | ovs_rwlock_unlock(&ms->rwlock); |
efc93e66 BP |
3097 | |
3098 | mcast_output_finish(ctx, &out, in_xbundle, &xvlan); | |
06994f87 | 3099 | } else { |
2d9b49dd | 3100 | xlate_report(ctx, OFT_DETAIL, "MLD query, flooding"); |
f0fb825a | 3101 | xlate_normal_flood(ctx, in_xbundle, &xvlan); |
06994f87 | 3102 | } |
86e2dcdd | 3103 | } else { |
a75636c8 | 3104 | if (is_ip_local_multicast(flow, wc)) { |
86e2dcdd FL |
3105 | /* RFC4541: section 2.1.2, item 2: Packets with a dst IP |
3106 | * address in the 224.0.0.x range which are not IGMP must | |
3107 | * be forwarded on all ports */ | |
2d9b49dd BP |
3108 | xlate_report(ctx, OFT_DETAIL, |
3109 | "RFC4541: section 2.1.2, item 2, flooding"); | |
f0fb825a | 3110 | xlate_normal_flood(ctx, in_xbundle, &xvlan); |
86e2dcdd FL |
3111 | return; |
3112 | } | |
3113 | } | |
3114 | ||
3115 | /* forwarding to group base ports */ | |
efc93e66 BP |
3116 | struct mcast_output out = MCAST_OUTPUT_INIT; |
3117 | ||
86e2dcdd | 3118 | ovs_rwlock_rdlock(&ms->rwlock); |
06994f87 TLSC |
3119 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
3120 | grp = mcast_snooping_lookup4(ms, flow->nw_dst, vlan); | |
3121 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
3122 | grp = mcast_snooping_lookup(ms, &flow->ipv6_dst, vlan); | |
3123 | } | |
86e2dcdd | 3124 | if (grp) { |
efc93e66 BP |
3125 | xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, &out); |
3126 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, &out); | |
3127 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, &xvlan, | |
3128 | &out); | |
9583bc14 | 3129 | } else { |
86e2dcdd | 3130 | if (mcast_snooping_flood_unreg(ms)) { |
2d9b49dd BP |
3131 | xlate_report(ctx, OFT_DETAIL, |
3132 | "unregistered multicast, flooding"); | |
efc93e66 | 3133 | out.flood = true; |
86e2dcdd | 3134 | } else { |
efc93e66 BP |
3135 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, &xvlan, |
3136 | &out); | |
3137 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, &out); | |
86e2dcdd | 3138 | } |
9583bc14 | 3139 | } |
86e2dcdd | 3140 | ovs_rwlock_unlock(&ms->rwlock); |
efc93e66 BP |
3141 | |
3142 | mcast_output_finish(ctx, &out, in_xbundle, &xvlan); | |
9583bc14 | 3143 | } else { |
86e2dcdd FL |
3144 | ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); |
3145 | mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); | |
9d078ec2 | 3146 | mac_port = mac ? mac_entry_get_port(ctx->xbridge->ml, mac) : NULL; |
86e2dcdd FL |
3147 | ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); |
3148 | ||
3149 | if (mac_port) { | |
0506f184 | 3150 | struct xbundle *mac_xbundle = xbundle_lookup(ctx->xcfg, mac_port); |
84dd881b DL |
3151 | |
3152 | if (mac_xbundle && xbundle_mirror_out(ctx->xbridge, mac_xbundle)) { | |
3153 | xlate_report(ctx, OFT_WARN, | |
3154 | "learned port is a mirror port, dropping"); | |
3155 | return; | |
3156 | } | |
3157 | ||
c781bd52 LJ |
3158 | if (mac_xbundle |
3159 | && mac_xbundle != in_xbundle | |
3160 | && mac_xbundle->ofbundle != in_xbundle->ofbundle) { | |
2d9b49dd | 3161 | xlate_report(ctx, OFT_DETAIL, "forwarding to learned port"); |
f0fb825a | 3162 | output_normal(ctx, mac_xbundle, &xvlan); |
86e2dcdd | 3163 | } else if (!mac_xbundle) { |
2d9b49dd BP |
3164 | xlate_report(ctx, OFT_WARN, |
3165 | "learned port is unknown, dropping"); | |
86e2dcdd | 3166 | } else { |
2d9b49dd BP |
3167 | xlate_report(ctx, OFT_DETAIL, |
3168 | "learned port is input port, dropping"); | |
86e2dcdd FL |
3169 | } |
3170 | } else { | |
2d9b49dd BP |
3171 | xlate_report(ctx, OFT_DETAIL, |
3172 | "no learned MAC for destination, flooding"); | |
f0fb825a | 3173 | xlate_normal_flood(ctx, in_xbundle, &xvlan); |
86e2dcdd | 3174 | } |
9583bc14 EJ |
3175 | } |
3176 | } | |
3177 | ||
a6092018 BP |
3178 | /* Appends a "sample" action for sFlow or IPFIX to 'ctx->odp_actions'. The |
3179 | * 'probability' is the number of packets out of UINT32_MAX to sample. The | |
8de6ff3e JP |
3180 | * 'cookie' is passed back in the callback for each sampled packet. |
3181 | * 'tunnel_out_port', if not ODPP_NONE, is added as the | |
3182 | * OVS_USERSPACE_ATTR_EGRESS_TUN_PORT attribute. If 'include_actions', | |
3183 | * an OVS_USERSPACE_ATTR_ACTIONS attribute is added. If | |
3184 | * 'emit_set_tunnel', sample(sampling_port=1) would translate into | |
3185 | * datapath sample action set(tunnel(...)), sample(...) and it is used | |
3186 | * for sampling egress tunnel information. | |
9583bc14 EJ |
3187 | */ |
3188 | static size_t | |
a6092018 | 3189 | compose_sample_action(struct xlate_ctx *ctx, |
9583bc14 | 3190 | const uint32_t probability, |
8de6ff3e | 3191 | const struct user_action_cookie *cookie, |
7321bda3 NM |
3192 | const odp_port_t tunnel_out_port, |
3193 | bool include_actions) | |
9583bc14 | 3194 | { |
b97f2c3a BY |
3195 | if (probability == 0) { |
3196 | /* No need to generate sampling or the inner action. */ | |
3197 | return 0; | |
3198 | } | |
3199 | ||
31b29c2e AZ |
3200 | /* If the slow path meter is configured by the controller, |
3201 | * insert a meter action before the user space action. */ | |
3202 | struct ofproto *ofproto = &ctx->xin->ofproto->up; | |
3203 | uint32_t meter_id = ofproto->slowpath_meter_id; | |
3204 | ||
3205 | /* When meter action is not required, avoid generate sample action | |
3206 | * for 100% sampling rate. */ | |
3207 | bool is_sample = probability < UINT32_MAX || meter_id != UINT32_MAX; | |
dc2c9ce3 | 3208 | size_t sample_offset = 0, actions_offset = 0; |
72471622 BY |
3209 | if (is_sample) { |
3210 | sample_offset = nl_msg_start_nested(ctx->odp_actions, | |
3211 | OVS_ACTION_ATTR_SAMPLE); | |
3212 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, | |
3213 | probability); | |
3214 | actions_offset = nl_msg_start_nested(ctx->odp_actions, | |
3215 | OVS_SAMPLE_ATTR_ACTIONS); | |
3216 | } | |
9583bc14 | 3217 | |
31b29c2e AZ |
3218 | if (meter_id != UINT32_MAX) { |
3219 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_METER, meter_id); | |
3220 | } | |
3221 | ||
a6092018 BP |
3222 | odp_port_t odp_port = ofp_port_to_odp_port( |
3223 | ctx->xbridge, ctx->xin->flow.in_port.ofp_port); | |
769b5034 | 3224 | uint32_t pid = dpif_port_get_pid(ctx->xbridge->dpif, odp_port); |
3276f2f2 LH |
3225 | size_t cookie_offset = odp_put_userspace_action(pid, cookie, |
3226 | sizeof *cookie, | |
3227 | tunnel_out_port, | |
3228 | include_actions, | |
3229 | ctx->odp_actions); | |
89a8a7f0 | 3230 | |
72471622 BY |
3231 | if (is_sample) { |
3232 | nl_msg_end_nested(ctx->odp_actions, actions_offset); | |
3233 | nl_msg_end_nested(ctx->odp_actions, sample_offset); | |
3234 | } | |
9583bc14 | 3235 | |
9583bc14 EJ |
3236 | return cookie_offset; |
3237 | } | |
3238 | ||
a6092018 BP |
3239 | /* If sFLow is not enabled, returns 0 without doing anything. |
3240 | * | |
3241 | * If sFlow is enabled, appends a template "sample" action to the ODP actions | |
3242 | * in 'ctx'. This action is a template because some of the information needed | |
3243 | * to fill it out is not available until flow translation is complete. In this | |
3244 | * case, this functions returns an offset, which is always nonzero, to pass | |
3245 | * later to fix_sflow_action() to fill in the rest of the template. */ | |
9583bc14 | 3246 | static size_t |
a6092018 | 3247 | compose_sflow_action(struct xlate_ctx *ctx) |
9583bc14 | 3248 | { |
a6092018 BP |
3249 | struct dpif_sflow *sflow = ctx->xbridge->sflow; |
3250 | if (!sflow || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { | |
9583bc14 EJ |
3251 | return 0; |
3252 | } | |
3253 | ||
24a4bbe1 IM |
3254 | struct user_action_cookie cookie; |
3255 | ||
3256 | memset(&cookie, 0, sizeof cookie); | |
3257 | cookie.type = USER_ACTION_COOKIE_SFLOW; | |
3258 | cookie.ofp_in_port = ctx->xin->flow.in_port.ofp_port; | |
3259 | cookie.ofproto_uuid = ctx->xbridge->ofproto->uuid; | |
3260 | ||
a6092018 | 3261 | return compose_sample_action(ctx, dpif_sflow_get_probability(sflow), |
8de6ff3e | 3262 | &cookie, ODPP_NONE, true); |
9583bc14 EJ |
3263 | } |
3264 | ||
f69f713b BY |
3265 | /* If flow IPFIX is enabled, make sure IPFIX flow sample action |
3266 | * at egress point of tunnel port is just in front of corresponding | |
3267 | * output action. If bridge IPFIX is enabled, this appends an IPFIX | |
3268 | * sample action to 'ctx->odp_actions'. */ | |
9583bc14 | 3269 | static void |
a6092018 | 3270 | compose_ipfix_action(struct xlate_ctx *ctx, odp_port_t output_odp_port) |
9583bc14 | 3271 | { |
a6092018 | 3272 | struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; |
8b7ea2d4 | 3273 | odp_port_t tunnel_out_port = ODPP_NONE; |
9583bc14 | 3274 | |
a6092018 | 3275 | if (!ipfix || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
3276 | return; |
3277 | } | |
3278 | ||
8b7ea2d4 WZ |
3279 | /* For input case, output_odp_port is ODPP_NONE, which is an invalid port |
3280 | * number. */ | |
3281 | if (output_odp_port == ODPP_NONE && | |
a6092018 | 3282 | !dpif_ipfix_get_bridge_exporter_input_sampling(ipfix)) { |
8b7ea2d4 WZ |
3283 | return; |
3284 | } | |
3285 | ||
f69f713b | 3286 | /* For output case, output_odp_port is valid. */ |
8b7ea2d4 | 3287 | if (output_odp_port != ODPP_NONE) { |
a6092018 | 3288 | if (!dpif_ipfix_get_bridge_exporter_output_sampling(ipfix)) { |
8b7ea2d4 WZ |
3289 | return; |
3290 | } | |
3291 | /* If tunnel sampling is enabled, put an additional option attribute: | |
3292 | * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT | |
3293 | */ | |
a6092018 | 3294 | if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(ipfix) && |
cd32509e | 3295 | dpif_ipfix_is_tunnel_port(ipfix, output_odp_port) ) { |
8b7ea2d4 WZ |
3296 | tunnel_out_port = output_odp_port; |
3297 | } | |
3298 | } | |
3299 | ||
24a4bbe1 IM |
3300 | struct user_action_cookie cookie; |
3301 | ||
3302 | memset(&cookie, 0, sizeof cookie); | |
3303 | cookie.type = USER_ACTION_COOKIE_IPFIX; | |
3304 | cookie.ofp_in_port = ctx->xin->flow.in_port.ofp_port; | |
3305 | cookie.ofproto_uuid = ctx->xbridge->ofproto->uuid; | |
3306 | cookie.ipfix.output_odp_port = output_odp_port; | |
3307 | ||
a6092018 BP |
3308 | compose_sample_action(ctx, |
3309 | dpif_ipfix_get_bridge_exporter_probability(ipfix), | |
8de6ff3e | 3310 | &cookie, tunnel_out_port, false); |
9583bc14 EJ |
3311 | } |
3312 | ||
a6092018 BP |
3313 | /* Fix "sample" action according to data collected while composing ODP actions, |
3314 | * as described in compose_sflow_action(). | |
3315 | * | |
8de6ff3e JP |
3316 | * 'user_cookie_offset' must be the offset returned by |
3317 | * compose_sflow_action(). */ | |
9583bc14 | 3318 | static void |
a6092018 | 3319 | fix_sflow_action(struct xlate_ctx *ctx, unsigned int user_cookie_offset) |
9583bc14 EJ |
3320 | { |
3321 | const struct flow *base = &ctx->base_flow; | |
8de6ff3e | 3322 | struct user_action_cookie *cookie; |
9583bc14 | 3323 | |
8de6ff3e | 3324 | cookie = ofpbuf_at(ctx->odp_actions, user_cookie_offset, sizeof *cookie); |
9583bc14 EJ |
3325 | ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); |
3326 | ||
f0fb825a | 3327 | cookie->sflow.vlan_tci = base->vlans[0].tci; |
a6092018 BP |
3328 | |
3329 | /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output | |
3330 | * port information") for the interpretation of cookie->output. */ | |
3331 | switch (ctx->sflow_n_outputs) { | |
3332 | case 0: | |
3333 | /* 0x40000000 | 256 means "packet dropped for unknown reason". */ | |
3334 | cookie->sflow.output = 0x40000000 | 256; | |
3335 | break; | |
3336 | ||
3337 | case 1: | |
3338 | cookie->sflow.output = dpif_sflow_odp_port_to_ifindex( | |
3339 | ctx->xbridge->sflow, ctx->sflow_odp_port); | |
3340 | if (cookie->sflow.output) { | |
3341 | break; | |
3342 | } | |
3343 | /* Fall through. */ | |
3344 | default: | |
3345 | /* 0x80000000 means "multiple output ports. */ | |
3346 | cookie->sflow.output = 0x80000000 | ctx->sflow_n_outputs; | |
3347 | break; | |
3348 | } | |
9583bc14 EJ |
3349 | } |
3350 | ||
515793d5 BP |
3351 | static bool |
3352 | process_special(struct xlate_ctx *ctx, const struct xport *xport) | |
db7d4e46 | 3353 | { |
515793d5 | 3354 | const struct flow *flow = &ctx->xin->flow; |
49a73e0c | 3355 | struct flow_wildcards *wc = ctx->wc; |
46c88433 | 3356 | const struct xbridge *xbridge = ctx->xbridge; |
515793d5 BP |
3357 | const struct dp_packet *packet = ctx->xin->packet; |
3358 | enum slow_path_reason slow; | |
a8448cb1 | 3359 | bool lacp_may_enable; |
642dc74d | 3360 | |
46c88433 | 3361 | if (!xport) { |
515793d5 | 3362 | slow = 0; |
46c88433 | 3363 | } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) { |
db7d4e46 | 3364 | if (packet) { |
46c88433 | 3365 | cfm_process_heartbeat(xport->cfm, packet); |
db7d4e46 | 3366 | } |
515793d5 | 3367 | slow = SLOW_CFM; |
fab52e16 | 3368 | } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) { |
db7d4e46 | 3369 | if (packet) { |
46c88433 | 3370 | bfd_process_packet(xport->bfd, flow, packet); |
60d02c72 AW |
3371 | /* If POLL received, immediately sends FINAL back. */ |
3372 | if (bfd_should_send_packet(xport->bfd)) { | |
6d308b28 | 3373 | ofproto_dpif_monitor_port_send_soon(xport->ofport); |
60d02c72 | 3374 | } |
db7d4e46 | 3375 | } |
515793d5 | 3376 | slow = SLOW_BFD; |
46c88433 | 3377 | } else if (xport->xbundle && xport->xbundle->lacp |
db7d4e46 JP |
3378 | && flow->dl_type == htons(ETH_TYPE_LACP)) { |
3379 | if (packet) { | |
a8448cb1 NK |
3380 | lacp_may_enable = lacp_process_packet(xport->xbundle->lacp, |
3381 | xport->ofport, packet); | |
3382 | /* Update LACP status in bond-slave to avoid packet-drops until | |
3383 | * LACP state machine is run by the main thread. */ | |
3384 | if (xport->xbundle->bond && lacp_may_enable) { | |
3385 | bond_slave_set_may_enable(xport->xbundle->bond, xport->ofport, | |
3386 | lacp_may_enable); | |
3387 | } | |
db7d4e46 | 3388 | } |
515793d5 | 3389 | slow = SLOW_LACP; |
9efd308e DV |
3390 | } else if ((xbridge->stp || xbridge->rstp) && |
3391 | stp_should_process_flow(flow, wc)) { | |
db7d4e46 | 3392 | if (packet) { |
f025bcb7 JR |
3393 | xbridge->stp |
3394 | ? stp_process_packet(xport, packet) | |
3395 | : rstp_process_packet(xport, packet); | |
db7d4e46 | 3396 | } |
515793d5 | 3397 | slow = SLOW_STP; |
19aef6ef | 3398 | } else if (xport->lldp && lldp_should_process_flow(xport->lldp, flow)) { |
0477baa9 DF |
3399 | if (packet) { |
3400 | lldp_process_packet(xport->lldp, packet); | |
3401 | } | |
515793d5 | 3402 | slow = SLOW_LLDP; |
db7d4e46 | 3403 | } else { |
515793d5 BP |
3404 | slow = 0; |
3405 | } | |
3406 | ||
3407 | if (slow) { | |
3408 | ctx->xout->slow |= slow; | |
3409 | return true; | |
3410 | } else { | |
3411 | return false; | |
db7d4e46 JP |
3412 | } |
3413 | } | |
3414 | ||
a36de779 | 3415 | static int |
0506f184 HH |
3416 | tnl_route_lookup_flow(const struct xlate_ctx *ctx, |
3417 | const struct flow *oflow, | |
a8704b50 PS |
3418 | struct in6_addr *ip, struct in6_addr *src, |
3419 | struct xport **out_port) | |
a36de779 PS |
3420 | { |
3421 | char out_dev[IFNAMSIZ]; | |
3422 | struct xbridge *xbridge; | |
c2b878e0 TLSC |
3423 | struct in6_addr gw; |
3424 | struct in6_addr dst; | |
a36de779 | 3425 | |
c2b878e0 | 3426 | dst = flow_tnl_dst(&oflow->tunnel); |
ed52ca57 | 3427 | if (!ovs_router_lookup(oflow->pkt_mark, &dst, out_dev, src, &gw)) { |
a36de779 PS |
3428 | return -ENOENT; |
3429 | } | |
3430 | ||
c2b878e0 TLSC |
3431 | if (ipv6_addr_is_set(&gw) && |
3432 | (!IN6_IS_ADDR_V4MAPPED(&gw) || in6_addr_get_mapped_ipv4(&gw))) { | |
a36de779 PS |
3433 | *ip = gw; |
3434 | } else { | |
c2b878e0 | 3435 | *ip = dst; |
a36de779 PS |
3436 | } |
3437 | ||
0506f184 | 3438 | HMAP_FOR_EACH (xbridge, hmap_node, &ctx->xcfg->xbridges) { |
a36de779 PS |
3439 | if (!strncmp(xbridge->name, out_dev, IFNAMSIZ)) { |
3440 | struct xport *port; | |
3441 | ||
3442 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
3443 | if (!strncmp(netdev_get_name(port->netdev), out_dev, IFNAMSIZ)) { | |
3444 | *out_port = port; | |
3445 | return 0; | |
3446 | } | |
3447 | } | |
3448 | } | |
3449 | } | |
dc0bd12f YS |
3450 | |
3451 | /* If tunnel IP isn't configured on bridges, then we search all ports. */ | |
3452 | HMAP_FOR_EACH (xbridge, hmap_node, &ctx->xcfg->xbridges) { | |
3453 | struct xport *port; | |
3454 | ||
3455 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
3456 | if (!strncmp(netdev_get_name(port->netdev), | |
3457 | out_dev, IFNAMSIZ)) { | |
3458 | *out_port = port; | |
3459 | return 0; | |
3460 | } | |
3461 | } | |
3462 | } | |
a36de779 PS |
3463 | return -ENOENT; |
3464 | } | |
3465 | ||
3466 | static int | |
cdd42eda JG |
3467 | compose_table_xlate(struct xlate_ctx *ctx, const struct xport *out_dev, |
3468 | struct dp_packet *packet) | |
a36de779 | 3469 | { |
cdd42eda | 3470 | struct xbridge *xbridge = out_dev->xbridge; |
ea41c034 | 3471 | ovs_version_t version = ofproto_dpif_get_tables_version(xbridge->ofproto); |
a36de779 PS |
3472 | struct ofpact_output output; |
3473 | struct flow flow; | |
3474 | ||
3475 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); | |
cf62fa4c | 3476 | flow_extract(packet, &flow); |
cdd42eda JG |
3477 | flow.in_port.ofp_port = out_dev->ofp_port; |
3478 | output.port = OFPP_TABLE; | |
a36de779 PS |
3479 | output.max_len = 0; |
3480 | ||
ea41c034 | 3481 | return ofproto_dpif_execute_actions__(xbridge->ofproto, version, &flow, |
1f4a8933 | 3482 | NULL, &output.ofpact, sizeof output, |
2d9b49dd | 3483 | ctx->depth, ctx->resubmits, packet); |
a36de779 PS |
3484 | } |
3485 | ||
c2b878e0 TLSC |
3486 | static void |
3487 | tnl_send_nd_request(struct xlate_ctx *ctx, const struct xport *out_dev, | |
3488 | const struct eth_addr eth_src, | |
3489 | struct in6_addr * ipv6_src, struct in6_addr * ipv6_dst) | |
3490 | { | |
3491 | struct dp_packet packet; | |
3492 | ||
3493 | dp_packet_init(&packet, 0); | |
16187903 | 3494 | compose_nd_ns(&packet, eth_src, ipv6_src, ipv6_dst); |
c2b878e0 TLSC |
3495 | compose_table_xlate(ctx, out_dev, &packet); |
3496 | dp_packet_uninit(&packet); | |
3497 | } | |
3498 | ||
a36de779 | 3499 | static void |
cdd42eda | 3500 | tnl_send_arp_request(struct xlate_ctx *ctx, const struct xport *out_dev, |
74ff3298 | 3501 | const struct eth_addr eth_src, |
a36de779 PS |
3502 | ovs_be32 ip_src, ovs_be32 ip_dst) |
3503 | { | |
cf62fa4c | 3504 | struct dp_packet packet; |
a36de779 | 3505 | |
cf62fa4c | 3506 | dp_packet_init(&packet, 0); |
eb0b295e BP |
3507 | compose_arp(&packet, ARP_OP_REQUEST, |
3508 | eth_src, eth_addr_zero, true, ip_src, ip_dst); | |
a36de779 | 3509 | |
cdd42eda | 3510 | compose_table_xlate(ctx, out_dev, &packet); |
cf62fa4c | 3511 | dp_packet_uninit(&packet); |
a36de779 PS |
3512 | } |
3513 | ||
7c12dfc5 SC |
3514 | static void |
3515 | propagate_tunnel_data_to_flow__(struct flow *dst_flow, | |
3516 | const struct flow *src_flow, | |
3517 | struct eth_addr dmac, struct eth_addr smac, | |
3518 | struct in6_addr s_ip6, ovs_be32 s_ip, | |
3519 | bool is_tnl_ipv6, uint8_t nw_proto) | |
3520 | { | |
3521 | dst_flow->dl_dst = dmac; | |
3522 | dst_flow->dl_src = smac; | |
3523 | ||
3524 | dst_flow->packet_type = htonl(PT_ETH); | |
3525 | dst_flow->nw_dst = src_flow->tunnel.ip_dst; | |
3526 | dst_flow->nw_src = src_flow->tunnel.ip_src; | |
3527 | dst_flow->ipv6_dst = src_flow->tunnel.ipv6_dst; | |
3528 | dst_flow->ipv6_src = src_flow->tunnel.ipv6_src; | |
3529 | ||
9cce7e71 | 3530 | dst_flow->nw_frag = 0; /* Tunnel packets are unfragmented. */ |
7c12dfc5 SC |
3531 | dst_flow->nw_tos = src_flow->tunnel.ip_tos; |
3532 | dst_flow->nw_ttl = src_flow->tunnel.ip_ttl; | |
3533 | dst_flow->tp_dst = src_flow->tunnel.tp_dst; | |
3534 | dst_flow->tp_src = src_flow->tunnel.tp_src; | |
3535 | ||
3536 | if (is_tnl_ipv6) { | |
3537 | dst_flow->dl_type = htons(ETH_TYPE_IPV6); | |
3538 | if (ipv6_mask_is_any(&dst_flow->ipv6_src) | |
3539 | && !ipv6_mask_is_any(&s_ip6)) { | |
3540 | dst_flow->ipv6_src = s_ip6; | |
3541 | } | |
3542 | } else { | |
3543 | dst_flow->dl_type = htons(ETH_TYPE_IP); | |
3544 | if (dst_flow->nw_src == 0 && s_ip) { | |
3545 | dst_flow->nw_src = s_ip; | |
3546 | } | |
3547 | } | |
3548 | dst_flow->nw_proto = nw_proto; | |
3549 | } | |
3550 | ||
3551 | /* | |
3552 | * Populate the 'flow' and 'base_flow' L3 fields to do the post tunnel push | |
3553 | * translations. | |
3554 | */ | |
3555 | static void | |
3556 | propagate_tunnel_data_to_flow(struct xlate_ctx *ctx, struct eth_addr dmac, | |
3557 | struct eth_addr smac, struct in6_addr s_ip6, | |
3558 | ovs_be32 s_ip, bool is_tnl_ipv6, | |
3559 | enum ovs_vport_type tnl_type) | |
3560 | { | |
3561 | struct flow *base_flow, *flow; | |
3562 | flow = &ctx->xin->flow; | |
3563 | base_flow = &ctx->base_flow; | |
3564 | uint8_t nw_proto = 0; | |
3565 | ||
3566 | switch (tnl_type) { | |
3567 | case OVS_VPORT_TYPE_GRE: | |
c387d817 GR |
3568 | case OVS_VPORT_TYPE_ERSPAN: |
3569 | case OVS_VPORT_TYPE_IP6ERSPAN: | |
3570 | case OVS_VPORT_TYPE_IP6GRE: | |
7c12dfc5 SC |
3571 | nw_proto = IPPROTO_GRE; |
3572 | break; | |
3573 | case OVS_VPORT_TYPE_VXLAN: | |
3574 | case OVS_VPORT_TYPE_GENEVE: | |
3c6d05a0 | 3575 | case OVS_VPORT_TYPE_GTPU: |
7c12dfc5 SC |
3576 | nw_proto = IPPROTO_UDP; |
3577 | break; | |
3578 | case OVS_VPORT_TYPE_LISP: | |
3579 | case OVS_VPORT_TYPE_STT: | |
3580 | case OVS_VPORT_TYPE_UNSPEC: | |
3581 | case OVS_VPORT_TYPE_NETDEV: | |
3582 | case OVS_VPORT_TYPE_INTERNAL: | |
3583 | case __OVS_VPORT_TYPE_MAX: | |
3584 | default: | |
3585 | OVS_NOT_REACHED(); | |
7c12dfc5 SC |
3586 | } |
3587 | /* | |
3588 | * Update base_flow first followed by flow as the dst_flow gets modified | |
3589 | * in the function. | |
3590 | */ | |
3591 | propagate_tunnel_data_to_flow__(base_flow, flow, dmac, smac, s_ip6, s_ip, | |
3592 | is_tnl_ipv6, nw_proto); | |
3593 | propagate_tunnel_data_to_flow__(flow, flow, dmac, smac, s_ip6, s_ip, | |
3594 | is_tnl_ipv6, nw_proto); | |
3595 | } | |
3596 | ||
a36de779 | 3597 | static int |
11938578 AZ |
3598 | native_tunnel_output(struct xlate_ctx *ctx, const struct xport *xport, |
3599 | const struct flow *flow, odp_port_t tunnel_odp_port, | |
3600 | bool truncate) | |
a36de779 | 3601 | { |
4975aa3e | 3602 | struct netdev_tnl_build_header_params tnl_params; |
a36de779 PS |
3603 | struct ovs_action_push_tnl tnl_push_data; |
3604 | struct xport *out_dev = NULL; | |
c2b878e0 TLSC |
3605 | ovs_be32 s_ip = 0, d_ip = 0; |
3606 | struct in6_addr s_ip6 = in6addr_any; | |
3607 | struct in6_addr d_ip6 = in6addr_any; | |
74ff3298 JR |
3608 | struct eth_addr smac; |
3609 | struct eth_addr dmac; | |
a36de779 | 3610 | int err; |
c2b878e0 TLSC |
3611 | char buf_sip6[INET6_ADDRSTRLEN]; |
3612 | char buf_dip6[INET6_ADDRSTRLEN]; | |
a36de779 | 3613 | |
283d8662 ZB |
3614 | /* Store sFlow data. */ |
3615 | uint32_t sflow_n_outputs = ctx->sflow_n_outputs; | |
3616 | ||
7c12dfc5 SC |
3617 | /* Structures to backup Ethernet and IP of base_flow. */ |
3618 | struct flow old_base_flow; | |
3619 | struct flow old_flow; | |
3620 | ||
3621 | /* Backup flow & base_flow data. */ | |
3622 | memcpy(&old_base_flow, &ctx->base_flow, sizeof old_base_flow); | |
3623 | memcpy(&old_flow, &ctx->xin->flow, sizeof old_flow); | |
5a0e4aec | 3624 | |
8e4e4588 | 3625 | if (flow->tunnel.ip_src) { |
3626 | in6_addr_set_mapped_ipv4(&s_ip6, flow->tunnel.ip_src); | |
3627 | } | |
7c12dfc5 | 3628 | |
0506f184 | 3629 | err = tnl_route_lookup_flow(ctx, flow, &d_ip6, &s_ip6, &out_dev); |
a36de779 | 3630 | if (err) { |
2d9b49dd | 3631 | xlate_report(ctx, OFT_WARN, "native tunnel routing failed"); |
a36de779 PS |
3632 | return err; |
3633 | } | |
c2b878e0 | 3634 | |
2d9b49dd | 3635 | xlate_report(ctx, OFT_DETAIL, "tunneling to %s via %s", |
c2b878e0 TLSC |
3636 | ipv6_string_mapped(buf_dip6, &d_ip6), |
3637 | netdev_get_name(out_dev->netdev)); | |
a36de779 PS |
3638 | |
3639 | /* Use mac addr of bridge port of the peer. */ | |
74ff3298 | 3640 | err = netdev_get_etheraddr(out_dev->netdev, &smac); |
a36de779 | 3641 | if (err) { |
2d9b49dd BP |
3642 | xlate_report(ctx, OFT_WARN, |
3643 | "tunnel output device lacks Ethernet address"); | |
a36de779 PS |
3644 | return err; |
3645 | } | |
3646 | ||
c2b878e0 TLSC |
3647 | d_ip = in6_addr_get_mapped_ipv4(&d_ip6); |
3648 | if (d_ip) { | |
a8704b50 | 3649 | s_ip = in6_addr_get_mapped_ipv4(&s_ip6); |
a36de779 PS |
3650 | } |
3651 | ||
c2b878e0 | 3652 | err = tnl_neigh_lookup(out_dev->xbridge->name, &d_ip6, &dmac); |
a36de779 | 3653 | if (err) { |
2d9b49dd BP |
3654 | xlate_report(ctx, OFT_DETAIL, |
3655 | "neighbor cache miss for %s on bridge %s, " | |
c2b878e0 TLSC |
3656 | "sending %s request", |
3657 | buf_dip6, out_dev->xbridge->name, d_ip ? "ARP" : "ND"); | |
3658 | if (d_ip) { | |
3659 | tnl_send_arp_request(ctx, out_dev, smac, s_ip, d_ip); | |
3660 | } else { | |
3661 | tnl_send_nd_request(ctx, out_dev, smac, &s_ip6, &d_ip6); | |
3662 | } | |
a36de779 PS |
3663 | return err; |
3664 | } | |
c2b878e0 | 3665 | |
a36de779 PS |
3666 | if (ctx->xin->xcache) { |
3667 | struct xc_entry *entry; | |
3668 | ||
53902038 | 3669 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_NEIGH); |
901a517e JR |
3670 | ovs_strlcpy(entry->tnl_neigh_cache.br_name, out_dev->xbridge->name, |
3671 | sizeof entry->tnl_neigh_cache.br_name); | |
3672 | entry->tnl_neigh_cache.d_ipv6 = d_ip6; | |
a36de779 | 3673 | } |
81de18ec | 3674 | |
2d9b49dd | 3675 | xlate_report(ctx, OFT_DETAIL, "tunneling from "ETH_ADDR_FMT" %s" |
c2b878e0 TLSC |
3676 | " to "ETH_ADDR_FMT" %s", |
3677 | ETH_ADDR_ARGS(smac), ipv6_string_mapped(buf_sip6, &s_ip6), | |
3678 | ETH_ADDR_ARGS(dmac), buf_dip6); | |
3679 | ||
4975aa3e PS |
3680 | netdev_init_tnl_build_header_params(&tnl_params, flow, &s_ip6, dmac, smac); |
3681 | err = tnl_port_build_header(xport->ofport, &tnl_push_data, &tnl_params); | |
a36de779 PS |
3682 | if (err) { |
3683 | return err; | |
3684 | } | |
81765c00 BP |
3685 | tnl_push_data.tnl_port = tunnel_odp_port; |
3686 | tnl_push_data.out_port = out_dev->odp_port; | |
beb75a40 | 3687 | |
7c12dfc5 SC |
3688 | /* After tunnel header has been added, MAC and IP data of flow and |
3689 | * base_flow need to be set properly, since there is not recirculation | |
3690 | * any more when sending packet to tunnel. */ | |
beb75a40 | 3691 | |
11938578 AZ |
3692 | propagate_tunnel_data_to_flow(ctx, dmac, smac, s_ip6, |
3693 | s_ip, tnl_params.is_ipv6, | |
3694 | tnl_push_data.tnl_type); | |
7c12dfc5 | 3695 | |
11938578 AZ |
3696 | size_t clone_ofs = 0; |
3697 | size_t push_action_size; | |
7c12dfc5 | 3698 | |
11938578 AZ |
3699 | clone_ofs = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CLONE); |
3700 | odp_put_tnl_push_action(ctx->odp_actions, &tnl_push_data); | |
3701 | push_action_size = ctx->odp_actions->size; | |
3702 | ||
3703 | if (!truncate) { | |
3704 | const struct dpif_flow_stats *backup_resubmit_stats; | |
3705 | struct xlate_cache *backup_xcache; | |
3706 | struct flow_wildcards *backup_wc, wc; | |
3707 | bool backup_side_effects; | |
3708 | const struct dp_packet *backup_packet; | |
3709 | ||
3710 | memset(&wc, 0 , sizeof wc); | |
3711 | backup_wc = ctx->wc; | |
3712 | ctx->wc = &wc; | |
3713 | ctx->xin->wc = NULL; | |
3714 | backup_resubmit_stats = ctx->xin->resubmit_stats; | |
3715 | backup_xcache = ctx->xin->xcache; | |
3716 | backup_side_effects = ctx->xin->allow_side_effects; | |
3717 | backup_packet = ctx->xin->packet; | |
3718 | ||
3719 | ctx->xin->resubmit_stats = NULL; | |
3720 | ctx->xin->xcache = xlate_cache_new(); /* Use new temporary cache. */ | |
3721 | ctx->xin->allow_side_effects = false; | |
3722 | ctx->xin->packet = NULL; | |
3723 | ||
3724 | /* Push the cache entry for the tunnel first. */ | |
3725 | struct xc_entry *entry; | |
3726 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TUNNEL_HEADER); | |
3727 | entry->tunnel_hdr.hdr_size = tnl_push_data.header_len; | |
3728 | entry->tunnel_hdr.operation = ADD; | |
3729 | ||
3730 | patch_port_output(ctx, xport, out_dev); | |
3731 | ||
3732 | /* Similar to the stats update in revalidation, the x_cache entries | |
3733 | * are populated by the previous translation are used to update the | |
3734 | * stats correctly. | |
7c12dfc5 | 3735 | */ |
11938578 AZ |
3736 | if (backup_resubmit_stats) { |
3737 | struct dpif_flow_stats stats = *backup_resubmit_stats; | |
16441315 | 3738 | xlate_push_stats(ctx->xin->xcache, &stats, false); |
11938578 AZ |
3739 | } |
3740 | xlate_cache_steal_entries(backup_xcache, ctx->xin->xcache); | |
3741 | ||
3742 | if (ctx->odp_actions->size > push_action_size) { | |
3743 | nl_msg_end_non_empty_nested(ctx->odp_actions, clone_ofs); | |
3744 | } else { | |
3745 | nl_msg_cancel_nested(ctx->odp_actions, clone_ofs); | |
11938578 AZ |
3746 | } |
3747 | ||
3748 | /* Restore context status. */ | |
3749 | ctx->xin->resubmit_stats = backup_resubmit_stats; | |
3750 | xlate_cache_delete(ctx->xin->xcache); | |
3751 | ctx->xin->xcache = backup_xcache; | |
3752 | ctx->xin->allow_side_effects = backup_side_effects; | |
3753 | ctx->xin->packet = backup_packet; | |
3754 | ctx->wc = backup_wc; | |
3755 | } else { | |
3756 | /* In order to maintain accurate stats, use recirc for | |
3757 | * natvie tunneling. */ | |
7c12dfc5 SC |
3758 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, 0); |
3759 | nl_msg_end_nested(ctx->odp_actions, clone_ofs); | |
3760 | } | |
11938578 | 3761 | |
7c12dfc5 SC |
3762 | /* Restore the flows after the translation. */ |
3763 | memcpy(&ctx->xin->flow, &old_flow, sizeof ctx->xin->flow); | |
3764 | memcpy(&ctx->base_flow, &old_base_flow, sizeof ctx->base_flow); | |
283d8662 ZB |
3765 | |
3766 | /* Restore sFlow data. */ | |
3767 | ctx->sflow_n_outputs = sflow_n_outputs; | |
3768 | ||
a36de779 PS |
3769 | return 0; |
3770 | } | |
3771 | ||
704bb0bf JS |
3772 | static void |
3773 | xlate_commit_actions(struct xlate_ctx *ctx) | |
3774 | { | |
3775 | bool use_masked = ctx->xbridge->support.masked_set_action; | |
3776 | ||
3777 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, | |
3778 | ctx->odp_actions, ctx->wc, | |
1fc11c59 | 3779 | use_masked, ctx->pending_encap, |
88ec1e0a | 3780 | ctx->pending_decap, ctx->encap_data); |
f839892a | 3781 | ctx->pending_encap = false; |
88ec1e0a | 3782 | ctx->pending_decap = false; |
1fc11c59 JS |
3783 | ofpbuf_delete(ctx->encap_data); |
3784 | ctx->encap_data = NULL; | |
704bb0bf JS |
3785 | } |
3786 | ||
07659514 | 3787 | static void |
72fe7578 | 3788 | clear_conntrack(struct xlate_ctx *ctx) |
07659514 | 3789 | { |
72fe7578 | 3790 | ctx->conntracked = false; |
6846e91e | 3791 | flow_clear_conntrack(&ctx->xin->flow); |
07659514 JS |
3792 | } |
3793 | ||
58d636ee BK |
3794 | static bool |
3795 | xlate_flow_is_protected(const struct xlate_ctx *ctx, const struct flow *flow, const struct xport *xport_out) | |
3796 | { | |
3797 | const struct xport *xport_in; | |
3798 | ||
3799 | if (!xport_out) { | |
3800 | return false; | |
3801 | } | |
3802 | ||
3803 | xport_in = get_ofp_port(ctx->xbridge, flow->in_port.ofp_port); | |
3804 | ||
3805 | return (xport_in && xport_in->xbundle && xport_out->xbundle && | |
3806 | xport_in->xbundle->protected && xport_out->xbundle->protected); | |
3807 | } | |
3808 | ||
48f704f4 | 3809 | /* Function handles when a packet is sent from one bridge to another bridge. |
8bdb2bdb | 3810 | * |
48f704f4 AZ |
3811 | * The bridges are internally connected, either with patch ports or with |
3812 | * tunnel ports. | |
3813 | * | |
3814 | * The output action to another bridge causes translation to continue within | |
3815 | * the next bridge. This process can be recursive; the next bridge can | |
3816 | * output yet to another bridge. | |
3817 | * | |
3818 | * The translated actions from the second bridge onwards are enclosed within | |
3819 | * the clone action, so that any modification to the packet will not be visible | |
3820 | * to the remaining actions of the originating bridge. | |
8bdb2bdb SC |
3821 | */ |
3822 | static void | |
48f704f4 AZ |
3823 | patch_port_output(struct xlate_ctx *ctx, const struct xport *in_dev, |
3824 | struct xport *out_dev) | |
8bdb2bdb SC |
3825 | { |
3826 | struct flow *flow = &ctx->xin->flow; | |
3827 | struct flow old_flow = ctx->xin->flow; | |
3828 | struct flow_tnl old_flow_tnl_wc = ctx->wc->masks.tunnel; | |
3829 | bool old_conntrack = ctx->conntracked; | |
3830 | bool old_was_mpls = ctx->was_mpls; | |
3831 | ovs_version_t old_version = ctx->xin->tables_version; | |
3832 | struct ofpbuf old_stack = ctx->stack; | |
3833 | uint8_t new_stack[1024]; | |
3834 | struct ofpbuf old_action_set = ctx->action_set; | |
3835 | struct ovs_list *old_trace = ctx->xin->trace; | |
3836 | uint64_t actset_stub[1024 / 8]; | |
3837 | ||
3838 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); | |
3839 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); | |
3840 | flow->in_port.ofp_port = out_dev->ofp_port; | |
3841 | flow->metadata = htonll(0); | |
3842 | memset(&flow->tunnel, 0, sizeof flow->tunnel); | |
86bfb29a | 3843 | memset(&ctx->wc->masks.tunnel, 0, sizeof ctx->wc->masks.tunnel); |
8bdb2bdb SC |
3844 | flow->tunnel.metadata.tab = |
3845 | ofproto_get_tun_tab(&out_dev->xbridge->ofproto->up); | |
3846 | ctx->wc->masks.tunnel.metadata.tab = flow->tunnel.metadata.tab; | |
3847 | memset(flow->regs, 0, sizeof flow->regs); | |
3848 | flow->actset_output = OFPP_UNSET; | |
3849 | clear_conntrack(ctx); | |
3850 | ctx->xin->trace = xlate_report(ctx, OFT_BRIDGE, "bridge(\"%s\")", | |
3851 | out_dev->xbridge->name); | |
3852 | mirror_mask_t old_mirrors = ctx->mirrors; | |
3853 | bool independent_mirrors = out_dev->xbridge != ctx->xbridge; | |
3854 | if (independent_mirrors) { | |
3855 | ctx->mirrors = 0; | |
3856 | } | |
3857 | ctx->xbridge = out_dev->xbridge; | |
3858 | ||
3859 | /* The bridge is now known so obtain its table version. */ | |
3860 | ctx->xin->tables_version | |
3861 | = ofproto_dpif_get_tables_version(ctx->xbridge->ofproto); | |
3862 | ||
3863 | if (!process_special(ctx, out_dev) && may_receive(out_dev, ctx)) { | |
3864 | if (xport_stp_forward_state(out_dev) && | |
3865 | xport_rstp_forward_state(out_dev)) { | |
3866 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true, | |
96c3a6e5 | 3867 | false, true, clone_xlate_actions); |
8bdb2bdb SC |
3868 | if (!ctx->freezing) { |
3869 | xlate_action_set(ctx); | |
3870 | } | |
3871 | if (ctx->freezing) { | |
3872 | finish_freezing(ctx); | |
3873 | } | |
3874 | } else { | |
3875 | /* Forwarding is disabled by STP and RSTP. Let OFPP_NORMAL and | |
3876 | * the learning action look at the packet, then drop it. */ | |
3877 | struct flow old_base_flow = ctx->base_flow; | |
3878 | size_t old_size = ctx->odp_actions->size; | |
3879 | mirror_mask_t old_mirrors2 = ctx->mirrors; | |
3880 | ||
3881 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true, | |
96c3a6e5 | 3882 | false, true, clone_xlate_actions); |
8bdb2bdb SC |
3883 | ctx->mirrors = old_mirrors2; |
3884 | ctx->base_flow = old_base_flow; | |
3885 | ctx->odp_actions->size = old_size; | |
3886 | ||
3887 | /* Undo changes that may have been done for freezing. */ | |
3888 | ctx_cancel_freeze(ctx); | |
3889 | } | |
3890 | } | |
3891 | ||
3892 | ctx->xin->trace = old_trace; | |
3893 | if (independent_mirrors) { | |
3894 | ctx->mirrors = old_mirrors; | |
3895 | } | |
3896 | ctx->xin->flow = old_flow; | |
3897 | ctx->xbridge = in_dev->xbridge; | |
3898 | ofpbuf_uninit(&ctx->action_set); | |
3899 | ctx->action_set = old_action_set; | |
3900 | ofpbuf_uninit(&ctx->stack); | |
3901 | ctx->stack = old_stack; | |
3902 | ||
3903 | /* Restore calling bridge's lookup version. */ | |
3904 | ctx->xin->tables_version = old_version; | |
3905 | ||
3906 | /* Restore to calling bridge tunneling information */ | |
3907 | ctx->wc->masks.tunnel = old_flow_tnl_wc; | |
3908 | ||
3909 | /* The out bridge popping MPLS should have no effect on the original | |
3910 | * bridge. */ | |
3911 | ctx->was_mpls = old_was_mpls; | |
3912 | ||
3913 | /* The out bridge's conntrack execution should have no effect on the | |
3914 | * original bridge. */ | |
3915 | ctx->conntracked = old_conntrack; | |
3916 | ||
3917 | /* The fact that the out bridge exits (for any reason) does not mean | |
3918 | * that the original bridge should exit. Specifically, if the out | |
3919 | * bridge freezes translation, the original bridge must continue | |
3920 | * processing with the original, not the frozen packet! */ | |
3921 | ctx->exit = false; | |
3922 | ||
3923 | /* Out bridge errors do not propagate back. */ | |
3924 | ctx->error = XLATE_OK; | |
3925 | ||
3926 | if (ctx->xin->resubmit_stats) { | |
3927 | netdev_vport_inc_tx(in_dev->netdev, ctx->xin->resubmit_stats); | |
3928 | netdev_vport_inc_rx(out_dev->netdev, ctx->xin->resubmit_stats); | |
3929 | if (out_dev->bfd) { | |
3930 | bfd_account_rx(out_dev->bfd, ctx->xin->resubmit_stats); | |
3931 | } | |
3932 | } | |
3933 | if (ctx->xin->xcache) { | |
3934 | struct xc_entry *entry; | |
3935 | ||
3936 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
3937 | entry->dev.tx = netdev_ref(in_dev->netdev); | |
3938 | entry->dev.rx = netdev_ref(out_dev->netdev); | |
3939 | entry->dev.bfd = bfd_ref(out_dev->bfd); | |
3940 | } | |
3941 | } | |
3942 | ||
081617f0 JS |
3943 | static bool |
3944 | check_output_prerequisites(struct xlate_ctx *ctx, | |
3945 | const struct xport *xport, | |
3946 | struct flow *flow, | |
3947 | bool check_stp) | |
9583bc14 | 3948 | { |
49a73e0c | 3949 | struct flow_wildcards *wc = ctx->wc; |
9583bc14 | 3950 | |
46c88433 | 3951 | if (!xport) { |
2d9b49dd | 3952 | xlate_report(ctx, OFT_WARN, "Nonexistent output port"); |
081617f0 | 3953 | return false; |
46c88433 | 3954 | } else if (xport->config & OFPUTIL_PC_NO_FWD) { |
2d9b49dd | 3955 | xlate_report(ctx, OFT_DETAIL, "OFPPC_NO_FWD set, skipping output"); |
081617f0 | 3956 | return false; |
1356dbd1 | 3957 | } else if (ctx->mirror_snaplen != 0 && xport->odp_port == ODPP_NONE) { |
2d9b49dd BP |
3958 | xlate_report(ctx, OFT_WARN, |
3959 | "Mirror truncate to ODPP_NONE, skipping output"); | |
081617f0 | 3960 | return false; |
58d636ee | 3961 | } else if (xlate_flow_is_protected(ctx, flow, xport)) { |
2d9b49dd BP |
3962 | xlate_report(ctx, OFT_WARN, |
3963 | "Flow is between protected ports, skipping output."); | |
081617f0 | 3964 | return false; |
0d1cee12 | 3965 | } else if (check_stp) { |
bbbca389 | 3966 | if (is_stp(&ctx->base_flow)) { |
9efd308e DV |
3967 | if (!xport_stp_should_forward_bpdu(xport) && |
3968 | !xport_rstp_should_manage_bpdu(xport)) { | |
3969 | if (ctx->xbridge->stp != NULL) { | |
2d9b49dd BP |
3970 | xlate_report(ctx, OFT_WARN, |
3971 | "STP not in listening state, " | |
3972 | "skipping bpdu output"); | |
9efd308e | 3973 | } else if (ctx->xbridge->rstp != NULL) { |
2d9b49dd BP |
3974 | xlate_report(ctx, OFT_WARN, |
3975 | "RSTP not managing BPDU in this state, " | |
3976 | "skipping bpdu output"); | |
9efd308e | 3977 | } |
081617f0 | 3978 | return false; |
0d1cee12 | 3979 | } |
67818616 MV |
3980 | } else if ((xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) |
3981 | || (xport->bfd && bfd_should_process_flow(xport->bfd, flow, | |
3982 | wc))) { | |
3983 | /* Pass; STP should not block link health detection. */ | |
9efd308e DV |
3984 | } else if (!xport_stp_forward_state(xport) || |
3985 | !xport_rstp_forward_state(xport)) { | |
3986 | if (ctx->xbridge->stp != NULL) { | |
2d9b49dd BP |
3987 | xlate_report(ctx, OFT_WARN, |
3988 | "STP not in forwarding state, skipping output"); | |
9efd308e | 3989 | } else if (ctx->xbridge->rstp != NULL) { |
2d9b49dd BP |
3990 | xlate_report(ctx, OFT_WARN, |
3991 | "RSTP not in forwarding state, skipping output"); | |
9efd308e | 3992 | } |
081617f0 | 3993 | return false; |
0d1cee12 | 3994 | } |
9583bc14 | 3995 | } |
5dbfe239 ZB |
3996 | |
3997 | if (xport->pt_mode == NETDEV_PT_LEGACY_L2 && | |
3998 | flow->packet_type != htonl(PT_ETH)) { | |
3999 | xlate_report(ctx, OFT_WARN, "Trying to send non-Ethernet packet " | |
4000 | "through legacy L2 port. Dropping packet."); | |
4001 | return false; | |
4002 | } | |
4003 | ||
081617f0 JS |
4004 | return true; |
4005 | } | |
4006 | ||
83c2757b ZB |
4007 | /* Function verifies if destination address of received Neighbor Advertisement |
4008 | * message stored in 'flow' is correct. It should be either FF02::1:FFXX:XXXX | |
4009 | * where XX:XXXX stands for the last 24 bits of 'ipv6_addr' or it should match | |
4010 | * 'ipv6_addr'. */ | |
4011 | static bool | |
4012 | is_nd_dst_correct(const struct flow *flow, const struct in6_addr *ipv6_addr) | |
4013 | { | |
4014 | const uint8_t *flow_ipv6_addr = (uint8_t *) &flow->ipv6_dst; | |
4015 | const uint8_t *addr = (uint8_t *) ipv6_addr; | |
4016 | ||
68fd9251 | 4017 | return (IN6_IS_ADDR_MC_LINKLOCAL(&flow->ipv6_dst) && |
83c2757b ZB |
4018 | flow_ipv6_addr[11] == 0x01 && |
4019 | flow_ipv6_addr[12] == 0xff && | |
4020 | flow_ipv6_addr[13] == addr[13] && | |
4021 | flow_ipv6_addr[14] == addr[14] && | |
4022 | flow_ipv6_addr[15] == addr[15]) || | |
4023 | IN6_ARE_ADDR_EQUAL(&flow->ipv6_dst, ipv6_addr); | |
4024 | } | |
4025 | ||
dc0bd12f YS |
4026 | static bool |
4027 | is_neighbor_reply_matched(const struct flow *flow, struct in6_addr *ip_addr) | |
4028 | { | |
4029 | return ((IN6_IS_ADDR_V4MAPPED(ip_addr) && | |
4030 | flow->dl_type == htons(ETH_TYPE_ARP) && | |
4031 | in6_addr_get_mapped_ipv4(ip_addr) == flow->nw_dst) || | |
4032 | (!IN6_IS_ADDR_V4MAPPED(ip_addr) && | |
4033 | is_nd_dst_correct(flow, ip_addr))); | |
4034 | } | |
4035 | ||
83c2757b ZB |
4036 | /* Function verifies if the ARP reply or Neighbor Advertisement represented by |
4037 | * 'flow' addresses the 'xbridge' of 'ctx'. Returns true if the ARP TA or | |
4038 | * neighbor discovery destination is in the list of configured IP addresses of | |
4039 | * the bridge. Otherwise, it returns false. */ | |
4040 | static bool | |
4041 | is_neighbor_reply_correct(const struct xlate_ctx *ctx, const struct flow *flow) | |
4042 | { | |
4043 | bool ret = false; | |
4044 | int i; | |
4045 | struct xbridge_addr *xbridge_addr = xbridge_addr_ref(ctx->xbridge->addr); | |
4046 | ||
4047 | /* Verify if 'nw_dst' of ARP or 'ipv6_dst' of ICMPV6 is in the list. */ | |
4048 | for (i = 0; xbridge_addr && i < xbridge_addr->n_addr; i++) { | |
4049 | struct in6_addr *ip_addr = &xbridge_addr->addr[i]; | |
dc0bd12f | 4050 | if (is_neighbor_reply_matched(flow, ip_addr)) { |
83c2757b ZB |
4051 | /* Found a match. */ |
4052 | ret = true; | |
4053 | break; | |
4054 | } | |
4055 | } | |
4056 | ||
4057 | xbridge_addr_unref(xbridge_addr); | |
dc0bd12f YS |
4058 | |
4059 | /* If not found in bridge's IPs, search in its ports. */ | |
4060 | if (!ret) { | |
4061 | struct in6_addr *ip_addr, *mask; | |
4062 | struct xport *port; | |
4063 | int error, n_in6; | |
4064 | ||
4065 | HMAP_FOR_EACH (port, ofp_node, &ctx->xbridge->xports) { | |
4066 | error = netdev_get_addr_list(port->netdev, &ip_addr, | |
4067 | &mask, &n_in6); | |
26228f44 DS |
4068 | if (!error) { |
4069 | ret = is_neighbor_reply_matched(flow, ip_addr); | |
4070 | free(ip_addr); | |
4071 | free(mask); | |
4072 | if (ret) { | |
4073 | /* Found a match. */ | |
4074 | break; | |
4075 | } | |
dc0bd12f YS |
4076 | } |
4077 | } | |
4078 | } | |
83c2757b ZB |
4079 | return ret; |
4080 | } | |
4081 | ||
081617f0 | 4082 | static bool |
dc0bd12f YS |
4083 | terminate_native_tunnel(struct xlate_ctx *ctx, struct flow *flow, |
4084 | struct flow_wildcards *wc, odp_port_t *tnl_port) | |
081617f0 JS |
4085 | { |
4086 | *tnl_port = ODPP_NONE; | |
4087 | ||
4088 | /* XXX: Write better Filter for tunnel port. We can use in_port | |
4089 | * in tunnel-port flow to avoid these checks completely. */ | |
dc0bd12f | 4090 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
081617f0 | 4091 | *tnl_port = tnl_port_map_lookup(flow, wc); |
83c2757b ZB |
4092 | |
4093 | /* If no tunnel port was found and it's about an ARP or ICMPv6 packet, | |
4094 | * do tunnel neighbor snooping. */ | |
4095 | if (*tnl_port == ODPP_NONE && | |
4096 | (flow->dl_type == htons(ETH_TYPE_ARP) || | |
4097 | flow->nw_proto == IPPROTO_ICMPV6) && | |
4098 | is_neighbor_reply_correct(ctx, flow)) { | |
4099 | tnl_neigh_snoop(flow, wc, ctx->xbridge->name); | |
4100 | } | |
081617f0 JS |
4101 | } |
4102 | ||
4103 | return *tnl_port != ODPP_NONE; | |
4104 | } | |
4105 | ||
4106 | static void | |
4107 | compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, | |
feee58b9 | 4108 | const struct xlate_bond_recirc *xr, bool check_stp, |
11938578 | 4109 | bool is_last_action OVS_UNUSED, bool truncate) |
081617f0 JS |
4110 | { |
4111 | const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); | |
4112 | struct flow_wildcards *wc = ctx->wc; | |
4113 | struct flow *flow = &ctx->xin->flow; | |
4114 | struct flow_tnl flow_tnl; | |
4115 | union flow_vlan_hdr flow_vlans[FLOW_MAX_VLAN_HEADERS]; | |
4116 | uint8_t flow_nw_tos; | |
4117 | odp_port_t out_port, odp_port, odp_tnl_port; | |
4118 | bool is_native_tunnel = false; | |
4119 | uint8_t dscp; | |
5dbfe239 ZB |
4120 | struct eth_addr flow_dl_dst = flow->dl_dst; |
4121 | struct eth_addr flow_dl_src = flow->dl_src; | |
4122 | ovs_be32 flow_packet_type = flow->packet_type; | |
4123 | ovs_be16 flow_dl_type = flow->dl_type; | |
081617f0 JS |
4124 | |
4125 | /* If 'struct flow' gets additional metadata, we'll need to zero it out | |
4126 | * before traversing a patch port. */ | |
3c6d05a0 | 4127 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 42); |
081617f0 JS |
4128 | memset(&flow_tnl, 0, sizeof flow_tnl); |
4129 | ||
4130 | if (!check_output_prerequisites(ctx, xport, flow, check_stp)) { | |
4131 | return; | |
4132 | } | |
9583bc14 | 4133 | |
875ab130 BP |
4134 | if (flow->packet_type == htonl(PT_ETH)) { |
4135 | /* Strip Ethernet header for legacy L3 port. */ | |
4136 | if (xport->pt_mode == NETDEV_PT_LEGACY_L3) { | |
4137 | flow->packet_type = PACKET_TYPE_BE(OFPHTN_ETHERTYPE, | |
4138 | ntohs(flow->dl_type)); | |
4139 | } | |
beb75a40 JS |
4140 | } |
4141 | ||
46c88433 | 4142 | if (xport->peer) { |
5cb92182 BP |
4143 | if (truncate) { |
4144 | xlate_report_error(ctx, "Cannot truncate output to patch port"); | |
4145 | } | |
48f704f4 | 4146 | patch_port_output(ctx, xport, xport->peer); |
8bdb2bdb | 4147 | return; |
9583bc14 EJ |
4148 | } |
4149 | ||
f0fb825a | 4150 | memcpy(flow_vlans, flow->vlans, sizeof flow_vlans); |
33bf9176 | 4151 | flow_nw_tos = flow->nw_tos; |
9583bc14 | 4152 | |
16194afd DDP |
4153 | if (count_skb_priorities(xport)) { |
4154 | memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority); | |
4155 | if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { | |
4156 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
4157 | flow->nw_tos &= ~IP_DSCP_MASK; | |
4158 | flow->nw_tos |= dscp; | |
4159 | } | |
9583bc14 EJ |
4160 | } |
4161 | ||
46c88433 | 4162 | if (xport->is_tunnel) { |
c2b878e0 | 4163 | struct in6_addr dst; |
9583bc14 EJ |
4164 | /* Save tunnel metadata so that changes made due to |
4165 | * the Logical (tunnel) Port are not visible for any further | |
4166 | * matches, while explicit set actions on tunnel metadata are. | |
4167 | */ | |
a36de779 | 4168 | flow_tnl = flow->tunnel; |
49a73e0c | 4169 | odp_port = tnl_port_send(xport->ofport, flow, ctx->wc); |
4e022ec0 | 4170 | if (odp_port == ODPP_NONE) { |
2d9b49dd | 4171 | xlate_report(ctx, OFT_WARN, "Tunneling decided against output"); |
9583bc14 EJ |
4172 | goto out; /* restore flow_nw_tos */ |
4173 | } | |
c2b878e0 TLSC |
4174 | dst = flow_tnl_dst(&flow->tunnel); |
4175 | if (ipv6_addr_equals(&dst, &ctx->orig_tunnel_ipv6_dst)) { | |
2d9b49dd | 4176 | xlate_report(ctx, OFT_WARN, "Not tunneling to our own address"); |
9583bc14 EJ |
4177 | goto out; /* restore flow_nw_tos */ |
4178 | } | |
4179 | if (ctx->xin->resubmit_stats) { | |
46c88433 | 4180 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
9583bc14 | 4181 | } |
b256dc52 JS |
4182 | if (ctx->xin->xcache) { |
4183 | struct xc_entry *entry; | |
4184 | ||
4185 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
901a517e | 4186 | entry->dev.tx = netdev_ref(xport->netdev); |
b256dc52 | 4187 | } |
9583bc14 | 4188 | out_port = odp_port; |
a36de779 | 4189 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
2d9b49dd | 4190 | xlate_report(ctx, OFT_DETAIL, "output to native tunnel"); |
081617f0 | 4191 | is_native_tunnel = true; |
a36de779 | 4192 | } else { |
c6d87201 WT |
4193 | const char *tnl_type; |
4194 | ||
2d9b49dd | 4195 | xlate_report(ctx, OFT_DETAIL, "output to kernel tunnel"); |
c6d87201 WT |
4196 | tnl_type = tnl_port_get_type(xport->ofport); |
4197 | commit_odp_tunnel_action(flow, &ctx->base_flow, | |
4198 | ctx->odp_actions, tnl_type); | |
a36de779 PS |
4199 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ |
4200 | } | |
9583bc14 | 4201 | } else { |
46c88433 | 4202 | odp_port = xport->odp_port; |
7614e5d0 | 4203 | out_port = odp_port; |
9583bc14 | 4204 | } |
9583bc14 | 4205 | |
4e022ec0 | 4206 | if (out_port != ODPP_NONE) { |
081617f0 | 4207 | /* Commit accumulated flow updates before output. */ |
704bb0bf | 4208 | xlate_commit_actions(ctx); |
adcf00ba | 4209 | |
9df65060 VDA |
4210 | if (xr && bond_use_lb_output_action(xport->xbundle->bond)) { |
4211 | /* | |
4212 | * If bond mode is balance-tcp and optimize balance tcp is enabled | |
4213 | * then use the hash directly for slave selection and avoid | |
4214 | * recirculation. | |
4215 | * | |
4216 | * Currently support for netdev datapath only. | |
4217 | */ | |
4218 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_LB_OUTPUT, | |
4219 | xr->recirc_id); | |
4220 | } else if (xr) { | |
081617f0 | 4221 | /* Recirculate the packet. */ |
347bf289 | 4222 | struct ovs_action_hash *act_hash; |
adcf00ba | 4223 | |
347bf289 | 4224 | /* Hash action. */ |
6a0b0d3b JS |
4225 | enum ovs_hash_alg hash_alg = xr->hash_alg; |
4226 | if (hash_alg > ctx->xbridge->support.max_hash_alg) { | |
4227 | /* Algorithm supported by all datapaths. */ | |
4228 | hash_alg = OVS_HASH_ALG_L4; | |
4229 | } | |
1520ef4f | 4230 | act_hash = nl_msg_put_unspec_uninit(ctx->odp_actions, |
347bf289 AZ |
4231 | OVS_ACTION_ATTR_HASH, |
4232 | sizeof *act_hash); | |
6a0b0d3b | 4233 | act_hash->hash_alg = hash_alg; |
62ac1f20 | 4234 | act_hash->hash_basis = xr->hash_basis; |
347bf289 AZ |
4235 | |
4236 | /* Recirc action. */ | |
1520ef4f | 4237 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, |
347bf289 | 4238 | xr->recirc_id); |
081617f0 JS |
4239 | } else if (is_native_tunnel) { |
4240 | /* Output to native tunnel port. */ | |
11938578 | 4241 | native_tunnel_output(ctx, xport, flow, odp_port, truncate); |
081617f0 | 4242 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ |
a36de779 | 4243 | |
dc0bd12f | 4244 | } else if (terminate_native_tunnel(ctx, flow, wc, |
081617f0 JS |
4245 | &odp_tnl_port)) { |
4246 | /* Intercept packet to be received on native tunnel port. */ | |
4247 | nl_msg_put_odp_port(ctx->odp_actions, OVS_ACTION_ATTR_TUNNEL_POP, | |
4248 | odp_tnl_port); | |
a36de779 | 4249 | |
081617f0 JS |
4250 | } else { |
4251 | /* Tunnel push-pop action is not compatible with | |
4252 | * IPFIX action. */ | |
4253 | compose_ipfix_action(ctx, out_port); | |
4254 | ||
4255 | /* Handle truncation of the mirrored packet. */ | |
4256 | if (ctx->mirror_snaplen > 0 && | |
4257 | ctx->mirror_snaplen < UINT16_MAX) { | |
4258 | struct ovs_action_trunc *trunc; | |
4259 | ||
4260 | trunc = nl_msg_put_unspec_uninit(ctx->odp_actions, | |
4261 | OVS_ACTION_ATTR_TRUNC, | |
4262 | sizeof *trunc); | |
4263 | trunc->max_len = ctx->mirror_snaplen; | |
4264 | if (!ctx->xbridge->support.trunc) { | |
4265 | ctx->xout->slow |= SLOW_ACTION; | |
1356dbd1 WT |
4266 | } |
4267 | } | |
081617f0 JS |
4268 | |
4269 | nl_msg_put_odp_port(ctx->odp_actions, | |
4270 | OVS_ACTION_ATTR_OUTPUT, | |
4271 | out_port); | |
adcf00ba | 4272 | } |
9583bc14 | 4273 | |
6cbbf4fa EJ |
4274 | ctx->sflow_odp_port = odp_port; |
4275 | ctx->sflow_n_outputs++; | |
2031ef97 | 4276 | ctx->nf_output_iface = ofp_port; |
6cbbf4fa EJ |
4277 | } |
4278 | ||
7efbc3b7 BP |
4279 | if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { |
4280 | mirror_packet(ctx, xport->xbundle, | |
4281 | xbundle_mirror_dst(xport->xbundle->xbridge, | |
4282 | xport->xbundle)); | |
4283 | } | |
4284 | ||
6cbbf4fa | 4285 | out: |
9583bc14 | 4286 | /* Restore flow */ |
f0fb825a | 4287 | memcpy(flow->vlans, flow_vlans, sizeof flow->vlans); |
33bf9176 | 4288 | flow->nw_tos = flow_nw_tos; |
5dbfe239 ZB |
4289 | flow->dl_dst = flow_dl_dst; |
4290 | flow->dl_src = flow_dl_src; | |
4291 | flow->packet_type = flow_packet_type; | |
4292 | flow->dl_type = flow_dl_type; | |
9583bc14 EJ |
4293 | } |
4294 | ||
4295 | static void | |
e93ef1c7 | 4296 | compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
feee58b9 | 4297 | const struct xlate_bond_recirc *xr, |
11938578 | 4298 | bool is_last_action, bool truncate) |
9583bc14 | 4299 | { |
11938578 AZ |
4300 | compose_output_action__(ctx, ofp_port, xr, true, |
4301 | is_last_action, truncate); | |
9583bc14 EJ |
4302 | } |
4303 | ||
bb61b33d | 4304 | static void |
feee58b9 | 4305 | xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule, |
96c3a6e5 AZ |
4306 | bool deepens, bool is_last_action, |
4307 | xlate_actions_handler *actions_xlator) | |
bb61b33d BP |
4308 | { |
4309 | struct rule_dpif *old_rule = ctx->rule; | |
8b1e5560 | 4310 | ovs_be64 old_cookie = ctx->rule_cookie; |
dc723c44 | 4311 | const struct rule_actions *actions; |
bb61b33d BP |
4312 | |
4313 | if (ctx->xin->resubmit_stats) { | |
16441315 | 4314 | rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats, false); |
bb61b33d BP |
4315 | } |
4316 | ||
98b07853 | 4317 | ctx->resubmits++; |
790c5d26 | 4318 | |
790c5d26 | 4319 | ctx->depth += deepens; |
bb61b33d | 4320 | ctx->rule = rule; |
07a3cd5c BP |
4321 | ctx->rule_cookie = rule->up.flow_cookie; |
4322 | actions = rule_get_actions(&rule->up); | |
96c3a6e5 | 4323 | actions_xlator(actions->ofpacts, actions->ofpacts_len, ctx, |
f5634764 | 4324 | is_last_action, false); |
8b1e5560 | 4325 | ctx->rule_cookie = old_cookie; |
bb61b33d | 4326 | ctx->rule = old_rule; |
790c5d26 | 4327 | ctx->depth -= deepens; |
bb61b33d BP |
4328 | } |
4329 | ||
bd3240ba SH |
4330 | static bool |
4331 | xlate_resubmit_resource_check(struct xlate_ctx *ctx) | |
9583bc14 | 4332 | { |
790c5d26 | 4333 | if (ctx->depth >= MAX_DEPTH) { |
2d9b49dd | 4334 | xlate_report_error(ctx, "over max translation depth %d", MAX_DEPTH); |
fff1b9c0 | 4335 | ctx->error = XLATE_RECURSION_TOO_DEEP; |
790c5d26 | 4336 | } else if (ctx->resubmits >= MAX_RESUBMITS) { |
2d9b49dd | 4337 | xlate_report_error(ctx, "over %d resubmit actions", MAX_RESUBMITS); |
fff1b9c0 | 4338 | ctx->error = XLATE_TOO_MANY_RESUBMITS; |
1520ef4f | 4339 | } else if (ctx->odp_actions->size > UINT16_MAX) { |
2d9b49dd | 4340 | xlate_report_error(ctx, "resubmits yielded over 64 kB of actions"); |
fff1b9c0 JR |
4341 | /* NOT an error, as we'll be slow-pathing the flow in this case? */ |
4342 | ctx->exit = true; /* XXX: translation still terminated! */ | |
6fd6ed71 | 4343 | } else if (ctx->stack.size >= 65536) { |
2d9b49dd | 4344 | xlate_report_error(ctx, "resubmits yielded over 64 kB of stack"); |
fff1b9c0 | 4345 | ctx->error = XLATE_STACK_TOO_DEEP; |
98b07853 | 4346 | } else { |
bd3240ba SH |
4347 | return true; |
4348 | } | |
4349 | ||
4350 | return false; | |
4351 | } | |
4352 | ||
2cd20955 JR |
4353 | static void |
4354 | tuple_swap_flow(struct flow *flow, bool ipv4) | |
4355 | { | |
4356 | uint8_t nw_proto = flow->nw_proto; | |
4357 | flow->nw_proto = flow->ct_nw_proto; | |
4358 | flow->ct_nw_proto = nw_proto; | |
4359 | ||
4360 | if (ipv4) { | |
4361 | ovs_be32 nw_src = flow->nw_src; | |
4362 | flow->nw_src = flow->ct_nw_src; | |
4363 | flow->ct_nw_src = nw_src; | |
4364 | ||
4365 | ovs_be32 nw_dst = flow->nw_dst; | |
4366 | flow->nw_dst = flow->ct_nw_dst; | |
4367 | flow->ct_nw_dst = nw_dst; | |
4368 | } else { | |
4369 | struct in6_addr ipv6_src = flow->ipv6_src; | |
4370 | flow->ipv6_src = flow->ct_ipv6_src; | |
4371 | flow->ct_ipv6_src = ipv6_src; | |
4372 | ||
4373 | struct in6_addr ipv6_dst = flow->ipv6_dst; | |
4374 | flow->ipv6_dst = flow->ct_ipv6_dst; | |
4375 | flow->ct_ipv6_dst = ipv6_dst; | |
4376 | } | |
4377 | ||
4378 | ovs_be16 tp_src = flow->tp_src; | |
4379 | flow->tp_src = flow->ct_tp_src; | |
4380 | flow->ct_tp_src = tp_src; | |
4381 | ||
4382 | ovs_be16 tp_dst = flow->tp_dst; | |
4383 | flow->tp_dst = flow->ct_tp_dst; | |
4384 | flow->ct_tp_dst = tp_dst; | |
4385 | } | |
4386 | ||
4387 | static void | |
4388 | tuple_swap(struct flow *flow, struct flow_wildcards *wc) | |
4389 | { | |
4390 | bool ipv4 = (flow->dl_type == htons(ETH_TYPE_IP)); | |
4391 | ||
4392 | tuple_swap_flow(flow, ipv4); | |
4393 | tuple_swap_flow(&wc->masks, ipv4); | |
4394 | } | |
4395 | ||
bd3240ba | 4396 | static void |
6d328fa2 | 4397 | xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, |
2cd20955 | 4398 | bool may_packet_in, bool honor_table_miss, |
96c3a6e5 AZ |
4399 | bool with_ct_orig, bool is_last_action, |
4400 | xlate_actions_handler *xlator) | |
bd3240ba | 4401 | { |
e12ec36b SH |
4402 | /* Check if we need to recirculate before matching in a table. */ |
4403 | if (ctx->was_mpls) { | |
4404 | ctx_trigger_freeze(ctx); | |
4405 | return; | |
4406 | } | |
bd3240ba | 4407 | if (xlate_resubmit_resource_check(ctx)) { |
9583bc14 | 4408 | uint8_t old_table_id = ctx->table_id; |
3f207910 | 4409 | struct rule_dpif *rule; |
9583bc14 EJ |
4410 | |
4411 | ctx->table_id = table_id; | |
4412 | ||
2cd20955 JR |
4413 | /* Swap packet fields with CT 5-tuple if requested. */ |
4414 | if (with_ct_orig) { | |
4415 | /* Do not swap if there is no CT tuple, or if key is not IP. */ | |
4416 | if (ctx->xin->flow.ct_nw_proto == 0 || | |
4417 | !is_ip_any(&ctx->xin->flow)) { | |
4418 | xlate_report_error(ctx, | |
4419 | "resubmit(ct) with non-tracked or non-IP packet!"); | |
7b950521 | 4420 | ctx->table_id = old_table_id; |
2cd20955 JR |
4421 | return; |
4422 | } | |
4423 | tuple_swap(&ctx->xin->flow, ctx->wc); | |
4424 | } | |
34dd0d78 | 4425 | rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, |
1f4a8933 | 4426 | ctx->xin->tables_version, |
c0e638aa | 4427 | &ctx->xin->flow, ctx->wc, |
34dd0d78 JR |
4428 | ctx->xin->resubmit_stats, |
4429 | &ctx->table_id, in_port, | |
a027899e JR |
4430 | may_packet_in, honor_table_miss, |
4431 | ctx->xin->xcache); | |
2cd20955 JR |
4432 | /* Swap back. */ |
4433 | if (with_ct_orig) { | |
4434 | tuple_swap(&ctx->xin->flow, ctx->wc); | |
4435 | } | |
ad3efdcb | 4436 | |
a2143702 | 4437 | if (rule) { |
83709dfa JR |
4438 | /* Fill in the cache entry here instead of xlate_recursively |
4439 | * to make the reference counting more explicit. We take a | |
4440 | * reference in the lookups above if we are going to cache the | |
4441 | * rule. */ | |
4442 | if (ctx->xin->xcache) { | |
4443 | struct xc_entry *entry; | |
4444 | ||
4445 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE); | |
901a517e | 4446 | entry->rule = rule; |
07a3cd5c | 4447 | ofproto_rule_ref(&rule->up); |
83709dfa | 4448 | } |
2d9b49dd BP |
4449 | |
4450 | struct ovs_list *old_trace = ctx->xin->trace; | |
4451 | xlate_report_table(ctx, rule, table_id); | |
feee58b9 | 4452 | xlate_recursively(ctx, rule, table_id <= old_table_id, |
96c3a6e5 | 4453 | is_last_action, xlator); |
2d9b49dd | 4454 | ctx->xin->trace = old_trace; |
ad3efdcb EJ |
4455 | } |
4456 | ||
9583bc14 | 4457 | ctx->table_id = old_table_id; |
98b07853 | 4458 | return; |
9583bc14 EJ |
4459 | } |
4460 | } | |
4461 | ||
76973237 | 4462 | /* Consumes the group reference, which is only taken if xcache exists. */ |
f4fb341b | 4463 | static void |
1e684d7d RW |
4464 | xlate_group_stats(struct xlate_ctx *ctx, struct group_dpif *group, |
4465 | struct ofputil_bucket *bucket) | |
4466 | { | |
4467 | if (ctx->xin->resubmit_stats) { | |
4468 | group_dpif_credit_stats(group, bucket, ctx->xin->resubmit_stats); | |
4469 | } | |
4470 | if (ctx->xin->xcache) { | |
4471 | struct xc_entry *entry; | |
4472 | ||
4473 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_GROUP); | |
901a517e JR |
4474 | entry->group.group = group; |
4475 | entry->group.bucket = bucket; | |
1e684d7d RW |
4476 | } |
4477 | } | |
4478 | ||
4479 | static void | |
feee58b9 AZ |
4480 | xlate_group_bucket(struct xlate_ctx *ctx, struct ofputil_bucket *bucket, |
4481 | bool is_last_action) | |
f4fb341b | 4482 | { |
fac4786a BP |
4483 | struct ovs_list *old_trace = ctx->xin->trace; |
4484 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
4485 | char *s = xasprintf("bucket %"PRIu32, bucket->bucket_id); | |
4486 | ctx->xin->trace = &oftrace_report(ctx->xin->trace, OFT_BUCKET, | |
4487 | s)->subs; | |
4488 | free(s); | |
4489 | } | |
4490 | ||
f4fb341b | 4491 | uint64_t action_list_stub[1024 / 8]; |
0a2869d5 BP |
4492 | struct ofpbuf action_list = OFPBUF_STUB_INITIALIZER(action_list_stub); |
4493 | struct ofpbuf action_set = ofpbuf_const_initializer(bucket->ofpacts, | |
4494 | bucket->ofpacts_len); | |
5b09e569 | 4495 | struct flow old_flow = ctx->xin->flow; |
e12ec36b | 4496 | bool old_was_mpls = ctx->was_mpls; |
f4fb341b | 4497 | |
f4fb341b | 4498 | ofpacts_execute_action_set(&action_list, &action_set); |
790c5d26 | 4499 | ctx->depth++; |
f5634764 KG |
4500 | do_xlate_actions(action_list.data, action_list.size, ctx, is_last_action, |
4501 | true); | |
790c5d26 | 4502 | ctx->depth--; |
f4fb341b | 4503 | |
f4fb341b | 4504 | ofpbuf_uninit(&action_list); |
5b09e569 | 4505 | |
77ab5fd2 | 4506 | /* Check if need to freeze. */ |
1d361a81 | 4507 | if (ctx->freezing) { |
77ab5fd2 | 4508 | finish_freezing(ctx); |
e672ff9b JR |
4509 | } |
4510 | ||
5b09e569 JR |
4511 | /* Roll back flow to previous state. |
4512 | * This is equivalent to cloning the packet for each bucket. | |
4513 | * | |
4514 | * As a side effect any subsequently applied actions will | |
4515 | * also effectively be applied to a clone of the packet taken | |
4516 | * just before applying the all or indirect group. | |
4517 | * | |
4518 | * Note that group buckets are action sets, hence they cannot modify the | |
4519 | * main action set. Also any stack actions are ignored when executing an | |
afc0c379 | 4520 | * action set, so group buckets cannot directly change the stack either. |
5b09e569 | 4521 | * However, we do allow resubmit actions in group buckets, which could |
afc0c379 BP |
4522 | * recursively execute actions that do modify the action set or change the |
4523 | * stack. The controller must be careful about what it does to the | |
4524 | * action_set and stack in the tables resubmitted to from group buckets. */ | |
5b09e569 JR |
4525 | ctx->xin->flow = old_flow; |
4526 | ||
e12ec36b SH |
4527 | /* The group bucket popping MPLS should have no effect after bucket |
4528 | * execution. */ | |
4529 | ctx->was_mpls = old_was_mpls; | |
4530 | ||
5b09e569 JR |
4531 | /* The fact that the group bucket exits (for any reason) does not mean that |
4532 | * the translation after the group action should exit. Specifically, if | |
1d361a81 BP |
4533 | * the group bucket freezes translation, the actions after the group action |
4534 | * must continue processing with the original, not the frozen packet! */ | |
5b09e569 | 4535 | ctx->exit = false; |
b99654b0 VDA |
4536 | |
4537 | /* Context error in a bucket should not impact processing of other buckets | |
4538 | * or actions. This is similar to cloning a packet for group buckets. | |
4539 | * There is no need to restore the error back to old value due to the fact | |
4540 | * that we actually processed group action which can happen only when there | |
4541 | * is no previous context error. | |
4542 | * | |
4543 | * Exception to above is errors which are system limits to protect | |
4544 | * translation from running too long or occupy too much space. These errors | |
4545 | * should not be masked. XLATE_RECURSION_TOO_DEEP, XLATE_TOO_MANY_RESUBMITS | |
4546 | * and XLATE_STACK_TOO_DEEP fall in this category. */ | |
4547 | if (ctx->error == XLATE_TOO_MANY_MPLS_LABELS || | |
4548 | ctx->error == XLATE_UNSUPPORTED_PACKET_TYPE) { | |
4549 | /* reset the error and continue processing other buckets */ | |
4550 | ctx->error = XLATE_OK; | |
4551 | } | |
f4fb341b | 4552 | |
fac4786a | 4553 | ctx->xin->trace = old_trace; |
f4fb341b SH |
4554 | } |
4555 | ||
a04e5888 BP |
4556 | static struct ofputil_bucket * |
4557 | pick_ff_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
dd8cd4b4 | 4558 | { |
a04e5888 | 4559 | return group_first_live_bucket(ctx, group, 0); |
dd8cd4b4 SH |
4560 | } |
4561 | ||
a04e5888 BP |
4562 | static struct ofputil_bucket * |
4563 | pick_default_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
fe7e5749 | 4564 | { |
a04e5888 BP |
4565 | flow_mask_hash_fields(&ctx->xin->flow, ctx->wc, |
4566 | NX_HASH_FIELDS_SYMMETRIC_L4); | |
4567 | return group_best_live_bucket(ctx, group, | |
4568 | flow_hash_symmetric_l4(&ctx->xin->flow, 0)); | |
fe7e5749 SH |
4569 | } |
4570 | ||
a04e5888 BP |
4571 | static struct ofputil_bucket * |
4572 | pick_hash_fields_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
0c4b9393 | 4573 | { |
07a3cd5c BP |
4574 | const struct field_array *fields = &group->up.props.fields; |
4575 | const uint8_t *mask_values = fields->values; | |
4576 | uint32_t basis = hash_uint64(group->up.props.selection_method_param); | |
0c4b9393 | 4577 | |
07a3cd5c | 4578 | size_t i; |
e8dba719 JR |
4579 | BITMAP_FOR_EACH_1 (i, MFF_N_IDS, fields->used.bm) { |
4580 | const struct mf_field *mf = mf_from_id(i); | |
0c4b9393 | 4581 | |
5bcd4754 | 4582 | /* Skip fields for which prerequisites are not met. */ |
e8dba719 JR |
4583 | if (!mf_are_prereqs_ok(mf, &ctx->xin->flow, ctx->wc)) { |
4584 | /* Skip the mask bytes for this field. */ | |
4585 | mask_values += mf->n_bytes; | |
4586 | continue; | |
4587 | } | |
0c4b9393 | 4588 | |
e8dba719 JR |
4589 | union mf_value value; |
4590 | union mf_value mask; | |
0c4b9393 | 4591 | |
e8dba719 JR |
4592 | mf_get_value(mf, &ctx->xin->flow, &value); |
4593 | /* Mask the value. */ | |
4594 | for (int j = 0; j < mf->n_bytes; j++) { | |
4595 | mask.b[j] = *mask_values++; | |
4596 | value.b[j] &= mask.b[j]; | |
4597 | } | |
4598 | basis = hash_bytes(&value, mf->n_bytes, basis); | |
1cb20095 | 4599 | |
e8dba719 JR |
4600 | /* For tunnels, hash in whether the field is present. */ |
4601 | if (mf_is_tun_metadata(mf)) { | |
4602 | basis = hash_boolean(mf_is_set(mf, &ctx->xin->flow), basis); | |
0c4b9393 | 4603 | } |
e8dba719 JR |
4604 | |
4605 | mf_mask_field_masked(mf, &mask, ctx->wc); | |
0c4b9393 SH |
4606 | } |
4607 | ||
a04e5888 | 4608 | return group_best_live_bucket(ctx, group, basis); |
0c4b9393 SH |
4609 | } |
4610 | ||
a04e5888 BP |
4611 | static struct ofputil_bucket * |
4612 | pick_dp_hash_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
53cc166a | 4613 | { |
2e3fd24c JS |
4614 | uint32_t dp_hash = ctx->xin->flow.dp_hash; |
4615 | ||
53cc166a JR |
4616 | /* dp_hash value 0 is special since it means that the dp_hash has not been |
4617 | * computed, as all computed dp_hash values are non-zero. Therefore | |
4618 | * compare to zero can be used to decide if the dp_hash value is valid | |
4619 | * without masking the dp_hash field. */ | |
2e3fd24c JS |
4620 | if (!dp_hash) { |
4621 | enum ovs_hash_alg hash_alg = group->hash_alg; | |
4622 | if (hash_alg > ctx->xbridge->support.max_hash_alg) { | |
4623 | /* Algorithm supported by all datapaths. */ | |
4624 | hash_alg = OVS_HASH_ALG_L4; | |
4625 | } | |
4626 | ctx_trigger_recirculate_with_hash(ctx, hash_alg, group->hash_basis); | |
a04e5888 | 4627 | return NULL; |
53cc166a | 4628 | } else { |
2e3fd24c JS |
4629 | uint32_t hash_mask = group->hash_mask; |
4630 | ctx->wc->masks.dp_hash |= hash_mask; | |
4631 | ||
4632 | /* Starting from the original masked dp_hash value iterate over the | |
4633 | * hash mapping table to find the first live bucket. As the buckets | |
4634 | * are quasi-randomly spread over the hash values, this maintains | |
4635 | * a distribution according to bucket weights even when some buckets | |
4636 | * are non-live. */ | |
4637 | for (int i = 0; i <= hash_mask; i++) { | |
4638 | struct ofputil_bucket *b = | |
4639 | group->hash_map[(dp_hash + i) & hash_mask]; | |
4640 | if (bucket_is_alive(ctx, b, 0)) { | |
4641 | return b; | |
4642 | } | |
53cc166a | 4643 | } |
2e3fd24c | 4644 | |
a04e5888 | 4645 | return NULL; |
53cc166a JR |
4646 | } |
4647 | } | |
4648 | ||
a04e5888 BP |
4649 | static struct ofputil_bucket * |
4650 | pick_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
7565c3e4 | 4651 | { |
e12ec36b SH |
4652 | /* Select groups may access flow keys beyond L2 in order to |
4653 | * select a bucket. Recirculate as appropriate to make this possible. | |
4654 | */ | |
4655 | if (ctx->was_mpls) { | |
4656 | ctx_trigger_freeze(ctx); | |
991f1f63 | 4657 | return NULL; |
e12ec36b SH |
4658 | } |
4659 | ||
2e3fd24c JS |
4660 | switch (group->selection_method) { |
4661 | case SEL_METHOD_DEFAULT: | |
a04e5888 | 4662 | return pick_default_select_group(ctx, group); |
2e3fd24c JS |
4663 | break; |
4664 | case SEL_METHOD_HASH: | |
a04e5888 | 4665 | return pick_hash_fields_select_group(ctx, group); |
2e3fd24c JS |
4666 | break; |
4667 | case SEL_METHOD_DP_HASH: | |
a04e5888 | 4668 | return pick_dp_hash_select_group(ctx, group); |
2e3fd24c JS |
4669 | break; |
4670 | default: | |
4671 | /* Parsing of groups ensures this never happens */ | |
7565c3e4 SH |
4672 | OVS_NOT_REACHED(); |
4673 | } | |
2e3fd24c JS |
4674 | |
4675 | return NULL; | |
7565c3e4 SH |
4676 | } |
4677 | ||
f4fb341b | 4678 | static void |
feee58b9 AZ |
4679 | xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group, |
4680 | bool is_last_action) | |
f4fb341b | 4681 | { |
a04e5888 | 4682 | if (group->up.type == OFPGT11_ALL || group->up.type == OFPGT11_INDIRECT) { |
03121ac4 | 4683 | struct ovs_list *last_bucket = group->up.buckets.prev; |
a04e5888 BP |
4684 | struct ofputil_bucket *bucket; |
4685 | LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { | |
4686 | bool is_last_bucket = &bucket->list_node == last_bucket; | |
4687 | xlate_group_bucket(ctx, bucket, is_last_action && is_last_bucket); | |
4688 | } | |
4689 | xlate_group_stats(ctx, group, NULL); | |
4690 | } else { | |
4691 | struct ofputil_bucket *bucket; | |
4692 | if (group->up.type == OFPGT11_SELECT) { | |
4693 | bucket = pick_select_group(ctx, group); | |
4694 | } else if (group->up.type == OFPGT11_FF) { | |
4695 | bucket = pick_ff_group(ctx, group); | |
4696 | } else { | |
4697 | OVS_NOT_REACHED(); | |
4698 | } | |
fac4786a | 4699 | |
a04e5888 | 4700 | if (bucket) { |
fac4786a BP |
4701 | xlate_report(ctx, OFT_DETAIL, "using bucket %"PRIu32, |
4702 | bucket->bucket_id); | |
a04e5888 BP |
4703 | xlate_group_bucket(ctx, bucket, is_last_action); |
4704 | xlate_group_stats(ctx, group, bucket); | |
fac4786a BP |
4705 | } else { |
4706 | xlate_report(ctx, OFT_DETAIL, "no live bucket"); | |
4707 | if (ctx->xin->xcache) { | |
4708 | ofproto_group_unref(&group->up); | |
4709 | } | |
a04e5888 | 4710 | } |
f4fb341b | 4711 | } |
f4fb341b SH |
4712 | } |
4713 | ||
4714 | static bool | |
feee58b9 AZ |
4715 | xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id, |
4716 | bool is_last_action) | |
f4fb341b | 4717 | { |
0eb48fe1 | 4718 | if (xlate_resubmit_resource_check(ctx)) { |
f4fb341b | 4719 | struct group_dpif *group; |
f4fb341b | 4720 | |
76973237 JR |
4721 | /* Take ref only if xcache exists. */ |
4722 | group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, | |
1f4a8933 | 4723 | ctx->xin->tables_version, ctx->xin->xcache); |
db88b35c JR |
4724 | if (!group) { |
4725 | /* XXX: Should set ctx->error ? */ | |
2d9b49dd BP |
4726 | xlate_report(ctx, OFT_WARN, "output to nonexistent group %"PRIu32, |
4727 | group_id); | |
f4fb341b SH |
4728 | return true; |
4729 | } | |
feee58b9 | 4730 | xlate_group_action__(ctx, group, is_last_action); |
f4fb341b SH |
4731 | } |
4732 | ||
4733 | return false; | |
4734 | } | |
4735 | ||
9583bc14 EJ |
4736 | static void |
4737 | xlate_ofpact_resubmit(struct xlate_ctx *ctx, | |
feee58b9 AZ |
4738 | const struct ofpact_resubmit *resubmit, |
4739 | bool is_last_action) | |
9583bc14 | 4740 | { |
4e022ec0 | 4741 | ofp_port_t in_port; |
9583bc14 | 4742 | uint8_t table_id; |
adcf00ba AZ |
4743 | bool may_packet_in = false; |
4744 | bool honor_table_miss = false; | |
4745 | ||
4746 | if (ctx->rule && rule_dpif_is_internal(ctx->rule)) { | |
4747 | /* Still allow missed packets to be sent to the controller | |
4748 | * if resubmitting from an internal table. */ | |
4749 | may_packet_in = true; | |
4750 | honor_table_miss = true; | |
4751 | } | |
9583bc14 EJ |
4752 | |
4753 | in_port = resubmit->in_port; | |
4754 | if (in_port == OFPP_IN_PORT) { | |
4e022ec0 | 4755 | in_port = ctx->xin->flow.in_port.ofp_port; |
9583bc14 EJ |
4756 | } |
4757 | ||
4758 | table_id = resubmit->table_id; | |
4759 | if (table_id == 255) { | |
4760 | table_id = ctx->table_id; | |
4761 | } | |
4762 | ||
adcf00ba | 4763 | xlate_table_action(ctx, in_port, table_id, may_packet_in, |
feee58b9 | 4764 | honor_table_miss, resubmit->with_ct_orig, |
96c3a6e5 | 4765 | is_last_action, do_xlate_actions); |
9583bc14 EJ |
4766 | } |
4767 | ||
4768 | static void | |
feee58b9 AZ |
4769 | flood_packet_to_port(struct xlate_ctx *ctx, const struct xport *xport, |
4770 | bool all, bool is_last_action) | |
9583bc14 | 4771 | { |
feee58b9 AZ |
4772 | if (!xport) { |
4773 | return; | |
4774 | } | |
4775 | ||
4776 | if (all) { | |
4777 | compose_output_action__(ctx, xport->ofp_port, NULL, false, | |
11938578 | 4778 | is_last_action, false); |
feee58b9 | 4779 | } else { |
11938578 AZ |
4780 | compose_output_action(ctx, xport->ofp_port, NULL, is_last_action, |
4781 | false); | |
feee58b9 AZ |
4782 | } |
4783 | } | |
4784 | ||
4785 | static void | |
4786 | flood_packets(struct xlate_ctx *ctx, bool all, bool is_last_action) | |
4787 | { | |
4788 | const struct xport *xport, *last = NULL; | |
9583bc14 | 4789 | |
feee58b9 | 4790 | /* Use 'last' the keep track of the last output port. */ |
46c88433 EJ |
4791 | HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) { |
4792 | if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
4793 | continue; |
4794 | } | |
4795 | ||
feee58b9 AZ |
4796 | if (all || !(xport->config & OFPUTIL_PC_NO_FLOOD)) { |
4797 | /* 'last' is not the last port, send a packet out, and | |
4798 | * update 'last'. */ | |
4799 | flood_packet_to_port(ctx, last, all, false); | |
4800 | last = xport; | |
9583bc14 EJ |
4801 | } |
4802 | } | |
4803 | ||
feee58b9 AZ |
4804 | /* Send the packet to the 'last' port. */ |
4805 | flood_packet_to_port(ctx, last, all, is_last_action); | |
2031ef97 | 4806 | ctx->nf_output_iface = NF_OUT_FLOOD; |
9583bc14 EJ |
4807 | } |
4808 | ||
74c4530d JP |
4809 | static void |
4810 | put_controller_user_action(struct xlate_ctx *ctx, | |
4811 | bool dont_send, bool continuation, | |
4812 | uint32_t recirc_id, int len, | |
4813 | enum ofp_packet_in_reason reason, | |
4814 | uint16_t controller_id) | |
4815 | { | |
4816 | struct user_action_cookie cookie; | |
4817 | ||
4818 | memset(&cookie, 0, sizeof cookie); | |
4819 | cookie.type = USER_ACTION_COOKIE_CONTROLLER; | |
4820 | cookie.ofp_in_port = OFPP_NONE, | |
4821 | cookie.ofproto_uuid = ctx->xbridge->ofproto->uuid; | |
4822 | cookie.controller.dont_send = dont_send; | |
4823 | cookie.controller.continuation = continuation; | |
4824 | cookie.controller.reason = reason; | |
4825 | cookie.controller.recirc_id = recirc_id; | |
4826 | put_32aligned_be64(&cookie.controller.rule_cookie, ctx->rule_cookie); | |
4827 | cookie.controller.controller_id = controller_id; | |
4828 | cookie.controller.max_len = len; | |
4829 | ||
4830 | odp_port_t odp_port = ofp_port_to_odp_port(ctx->xbridge, | |
4831 | ctx->xin->flow.in_port.ofp_port); | |
769b5034 | 4832 | uint32_t pid = dpif_port_get_pid(ctx->xbridge->dpif, odp_port); |
74c4530d JP |
4833 | odp_put_userspace_action(pid, &cookie, sizeof cookie, ODPP_NONE, |
4834 | false, ctx->odp_actions); | |
4835 | } | |
4836 | ||
9583bc14 | 4837 | static void |
d39ec23d JP |
4838 | xlate_controller_action(struct xlate_ctx *ctx, int len, |
4839 | enum ofp_packet_in_reason reason, | |
4840 | uint16_t controller_id, | |
206ddb9a | 4841 | uint32_t provider_meter_id, |
d39ec23d | 4842 | const uint8_t *userdata, size_t userdata_len) |
9583bc14 | 4843 | { |
b476e2f2 | 4844 | xlate_commit_actions(ctx); |
beb75a40 | 4845 | |
9bfe9334 BP |
4846 | /* A packet sent by an action in a table-miss rule is considered an |
4847 | * explicit table miss. OpenFlow before 1.3 doesn't have that concept so | |
4848 | * it will get translated back to OFPR_ACTION for those versions. */ | |
4849 | if (reason == OFPR_ACTION | |
07a3cd5c | 4850 | && ctx->rule && rule_is_table_miss(&ctx->rule->up)) { |
9bfe9334 BP |
4851 | reason = OFPR_EXPLICIT_MISS; |
4852 | } | |
4853 | ||
d39ec23d JP |
4854 | struct frozen_state state = { |
4855 | .table_id = ctx->table_id, | |
4856 | .ofproto_uuid = ctx->xbridge->ofproto->uuid, | |
4857 | .stack = ctx->stack.data, | |
4858 | .stack_size = ctx->stack.size, | |
4859 | .mirrors = ctx->mirrors, | |
4860 | .conntracked = ctx->conntracked, | |
aeb6566d | 4861 | .was_mpls = ctx->was_mpls, |
d39ec23d JP |
4862 | .ofpacts = NULL, |
4863 | .ofpacts_len = 0, | |
4864 | .action_set = NULL, | |
4865 | .action_set_len = 0, | |
4866 | .userdata = CONST_CAST(uint8_t *, userdata), | |
4867 | .userdata_len = userdata_len, | |
9bfe9334 | 4868 | }; |
d39ec23d | 4869 | frozen_metadata_from_flow(&state.metadata, &ctx->xin->flow); |
9583bc14 | 4870 | |
d39ec23d JP |
4871 | uint32_t recirc_id = recirc_alloc_id_ctx(&state); |
4872 | if (!recirc_id) { | |
4873 | xlate_report_error(ctx, "Failed to allocate recirculation id"); | |
4874 | ctx->error = XLATE_NO_RECIRCULATION_CONTEXT; | |
4875 | return; | |
4876 | } | |
4877 | recirc_refs_add(&ctx->xout->recircs, recirc_id); | |
df70a773 | 4878 | |
206ddb9a JP |
4879 | /* If the controller action didn't request a meter (indicated by a |
4880 | * 'meter_id' argument other than NX_CTLR_NO_METER), see if one was | |
4881 | * configured through the "controller" virtual meter. | |
4882 | * | |
4883 | * Internally, ovs-vswitchd uses UINT32_MAX to indicate no meter is | |
4884 | * configured. */ | |
4885 | uint32_t meter_id; | |
4886 | if (provider_meter_id == UINT32_MAX) { | |
4887 | meter_id = ctx->xbridge->ofproto->up.controller_meter_id; | |
4888 | } else { | |
4889 | meter_id = provider_meter_id; | |
4890 | } | |
4891 | ||
d39ec23d JP |
4892 | size_t offset; |
4893 | size_t ac_offset; | |
d39ec23d JP |
4894 | if (meter_id != UINT32_MAX) { |
4895 | /* If controller meter is configured, generate clone(meter, userspace) | |
4896 | * action. */ | |
4897 | offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_SAMPLE); | |
4898 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, | |
4899 | UINT32_MAX); | |
4900 | ac_offset = nl_msg_start_nested(ctx->odp_actions, | |
4901 | OVS_SAMPLE_ATTR_ACTIONS); | |
4902 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_METER, meter_id); | |
df70a773 | 4903 | } |
3b4fff43 | 4904 | |
d39ec23d JP |
4905 | /* Generate the datapath flows even if we don't send the packet-in |
4906 | * so that debugging more closely represents normal state. */ | |
74c4530d | 4907 | bool dont_send = false; |
d39ec23d | 4908 | if (!ctx->xin->allow_side_effects && !ctx->xin->xcache) { |
74c4530d | 4909 | dont_send = true; |
d39ec23d | 4910 | } |
74c4530d JP |
4911 | put_controller_user_action(ctx, dont_send, false, recirc_id, len, |
4912 | reason, controller_id); | |
d39ec23d JP |
4913 | |
4914 | if (meter_id != UINT32_MAX) { | |
4915 | nl_msg_end_nested(ctx->odp_actions, ac_offset); | |
4916 | nl_msg_end_nested(ctx->odp_actions, offset); | |
4917 | } | |
9583bc14 EJ |
4918 | } |
4919 | ||
e6bc8e74 YHW |
4920 | /* Creates a frozen state, and allocates a unique recirc id for the given |
4921 | * state. Returns a non-zero recirc id if it is allocated successfully. | |
4922 | * Returns 0 otherwise. | |
4923 | **/ | |
4924 | static uint32_t | |
77ab5fd2 BP |
4925 | finish_freezing__(struct xlate_ctx *ctx, uint8_t table) |
4926 | { | |
1d361a81 | 4927 | ovs_assert(ctx->freezing); |
7bbdd84f | 4928 | |
1d361a81 | 4929 | struct frozen_state state = { |
07659514 | 4930 | .table_id = table, |
07a3cd5c | 4931 | .ofproto_uuid = ctx->xbridge->ofproto->uuid, |
5c1b2314 | 4932 | .stack = ctx->stack.data, |
84cf3c1f | 4933 | .stack_size = ctx->stack.size, |
29bae541 | 4934 | .mirrors = ctx->mirrors, |
07659514 | 4935 | .conntracked = ctx->conntracked, |
aeb6566d | 4936 | .was_mpls = ctx->was_mpls, |
00135b86 | 4937 | .xport_uuid = ctx->xin->xport_uuid, |
1d361a81 BP |
4938 | .ofpacts = ctx->frozen_actions.data, |
4939 | .ofpacts_len = ctx->frozen_actions.size, | |
417509fa | 4940 | .action_set = ctx->action_set.data, |
8a5fb3b4 | 4941 | .action_set_len = ctx->action_set.size, |
74c4530d JP |
4942 | .userdata = ctx->pause ? CONST_CAST(uint8_t *,ctx->pause->userdata) |
4943 | : NULL, | |
4944 | .userdata_len = ctx->pause ? ctx->pause->userdata_len : 0, | |
2082425c | 4945 | }; |
77ab5fd2 | 4946 | frozen_metadata_from_flow(&state.metadata, &ctx->xin->flow); |
2082425c | 4947 | |
74c4530d JP |
4948 | /* Allocate a unique recirc id for the given metadata state in the |
4949 | * flow. An existing id, with a new reference to the corresponding | |
4950 | * recirculation context, will be returned if possible. | |
4951 | * The life-cycle of this recirc id is managed by associating it | |
4952 | * with the udpif key ('ukey') created for each new datapath flow. */ | |
4953 | uint32_t recirc_id = recirc_alloc_id_ctx(&state); | |
4954 | if (!recirc_id) { | |
4955 | xlate_report_error(ctx, "Failed to allocate recirculation id"); | |
4956 | ctx->error = XLATE_NO_RECIRCULATION_CONTEXT; | |
4957 | return 0; | |
4958 | } | |
4959 | recirc_refs_add(&ctx->xout->recircs, recirc_id); | |
4960 | ||
77ab5fd2 | 4961 | if (ctx->pause) { |
74c4530d | 4962 | if (!ctx->xin->allow_side_effects && !ctx->xin->xcache) { |
e6bc8e74 | 4963 | return 0; |
77ab5fd2 | 4964 | } |
7bbdd84f | 4965 | |
74c4530d JP |
4966 | put_controller_user_action(ctx, false, true, recirc_id, |
4967 | ctx->pause->max_len, | |
4968 | ctx->pause->reason, | |
4969 | ctx->pause->controller_id); | |
4970 | } else { | |
53cc166a JR |
4971 | if (ctx->recirc_update_dp_hash) { |
4972 | struct ovs_action_hash *act_hash; | |
4973 | ||
4974 | /* Hash action. */ | |
4975 | act_hash = nl_msg_put_unspec_uninit(ctx->odp_actions, | |
4976 | OVS_ACTION_ATTR_HASH, | |
4977 | sizeof *act_hash); | |
2e3fd24c JS |
4978 | act_hash->hash_alg = ctx->dp_hash_alg; |
4979 | act_hash->hash_basis = ctx->dp_hash_basis; | |
53cc166a | 4980 | } |
74c4530d | 4981 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, recirc_id); |
77ab5fd2 | 4982 | } |
e672ff9b | 4983 | |
1d361a81 BP |
4984 | /* Undo changes done by freezing. */ |
4985 | ctx_cancel_freeze(ctx); | |
74c4530d | 4986 | return recirc_id; |
7bbdd84f SH |
4987 | } |
4988 | ||
1d361a81 | 4989 | /* Called only when we're freezing. */ |
07659514 | 4990 | static void |
77ab5fd2 | 4991 | finish_freezing(struct xlate_ctx *ctx) |
07659514 JS |
4992 | { |
4993 | xlate_commit_actions(ctx); | |
77ab5fd2 | 4994 | finish_freezing__(ctx, 0); |
07659514 JS |
4995 | } |
4996 | ||
e37b8437 JS |
4997 | /* Fork the pipeline here. The current packet will continue processing the |
4998 | * current action list. A clone of the current packet will recirculate, skip | |
4999 | * the remainder of the current action list and asynchronously resume pipeline | |
5000 | * processing in 'table' with the current metadata and action set. */ | |
5001 | static void | |
5fdd80cc YHW |
5002 | compose_recirculate_and_fork(struct xlate_ctx *ctx, uint8_t table, |
5003 | const uint16_t zone) | |
e37b8437 | 5004 | { |
e6bc8e74 | 5005 | uint32_t recirc_id; |
1d361a81 | 5006 | ctx->freezing = true; |
e6bc8e74 YHW |
5007 | recirc_id = finish_freezing__(ctx, table); |
5008 | ||
5009 | if (OVS_UNLIKELY(ctx->xin->trace) && recirc_id) { | |
5010 | if (oftrace_add_recirc_node(ctx->xin->recirc_queue, | |
5011 | OFT_RECIRC_CONNTRACK, &ctx->xin->flow, | |
d072d2de DC |
5012 | ctx->ct_nat_action, ctx->xin->packet, |
5013 | recirc_id, zone)) { | |
e6bc8e74 YHW |
5014 | xlate_report(ctx, OFT_DETAIL, "A clone of the packet is forked to " |
5015 | "recirculate. The forked pipeline will be resumed at " | |
5016 | "table %u.", table); | |
5017 | } else { | |
5018 | xlate_report(ctx, OFT_DETAIL, "Failed to trace the conntrack " | |
5019 | "forked pipeline with recirc_id = %d.", recirc_id); | |
5020 | } | |
5021 | } | |
e37b8437 JS |
5022 | } |
5023 | ||
8bfd0fda BP |
5024 | static void |
5025 | compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) | |
9583bc14 | 5026 | { |
33bf9176 | 5027 | struct flow *flow = &ctx->xin->flow; |
8bfd0fda | 5028 | int n; |
33bf9176 | 5029 | |
8bfd0fda | 5030 | ovs_assert(eth_type_mpls(mpls->ethertype)); |
b0a17866 | 5031 | |
49a73e0c | 5032 | n = flow_count_mpls_labels(flow, ctx->wc); |
8bfd0fda | 5033 | if (!n) { |
704bb0bf | 5034 | xlate_commit_actions(ctx); |
8bfd0fda BP |
5035 | } else if (n >= FLOW_MAX_MPLS_LABELS) { |
5036 | if (ctx->xin->packet != NULL) { | |
2d9b49dd BP |
5037 | xlate_report_error(ctx, "dropping packet on which an MPLS push " |
5038 | "action can't be performed as it would have " | |
5039 | "more MPLS LSEs than the %d supported.", | |
5040 | FLOW_MAX_MPLS_LABELS); | |
9583bc14 | 5041 | } |
fff1b9c0 | 5042 | ctx->error = XLATE_TOO_MANY_MPLS_LABELS; |
8bfd0fda | 5043 | return; |
9583bc14 | 5044 | } |
b0a17866 | 5045 | |
742c0ac3 JR |
5046 | /* Update flow's MPLS stack, and clear L3/4 fields to mark them invalid. */ |
5047 | flow_push_mpls(flow, n, mpls->ethertype, ctx->wc, true); | |
9583bc14 EJ |
5048 | } |
5049 | ||
8bfd0fda | 5050 | static void |
9cfef3d0 | 5051 | compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) |
9583bc14 | 5052 | { |
8bfd0fda | 5053 | struct flow *flow = &ctx->xin->flow; |
49a73e0c | 5054 | int n = flow_count_mpls_labels(flow, ctx->wc); |
33bf9176 | 5055 | |
49a73e0c | 5056 | if (flow_pop_mpls(flow, n, eth_type, ctx->wc)) { |
8bf009bf | 5057 | if (!eth_type_mpls(eth_type) && ctx->xbridge->support.odp.recirc) { |
e12ec36b | 5058 | ctx->was_mpls = true; |
7bbdd84f SH |
5059 | } |
5060 | } else if (n >= FLOW_MAX_MPLS_LABELS) { | |
8bfd0fda | 5061 | if (ctx->xin->packet != NULL) { |
2d9b49dd BP |
5062 | xlate_report_error(ctx, "dropping packet on which an " |
5063 | "MPLS pop action can't be performed as it has " | |
5064 | "more MPLS LSEs than the %d supported.", | |
5065 | FLOW_MAX_MPLS_LABELS); | |
8bfd0fda | 5066 | } |
fff1b9c0 | 5067 | ctx->error = XLATE_TOO_MANY_MPLS_LABELS; |
1520ef4f | 5068 | ofpbuf_clear(ctx->odp_actions); |
9583bc14 EJ |
5069 | } |
5070 | } | |
5071 | ||
5072 | static bool | |
5073 | compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids) | |
5074 | { | |
33bf9176 BP |
5075 | struct flow *flow = &ctx->xin->flow; |
5076 | ||
5077 | if (!is_ip_any(flow)) { | |
9583bc14 EJ |
5078 | return false; |
5079 | } | |
5080 | ||
49a73e0c | 5081 | ctx->wc->masks.nw_ttl = 0xff; |
33bf9176 BP |
5082 | if (flow->nw_ttl > 1) { |
5083 | flow->nw_ttl--; | |
9583bc14 EJ |
5084 | return false; |
5085 | } else { | |
5086 | size_t i; | |
5087 | ||
5088 | for (i = 0; i < ids->n_controllers; i++) { | |
d39ec23d | 5089 | xlate_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, |
206ddb9a | 5090 | ids->cnt_ids[i], UINT32_MAX, NULL, 0); |
9583bc14 EJ |
5091 | } |
5092 | ||
5093 | /* Stop processing for current table. */ | |
2d9b49dd BP |
5094 | xlate_report(ctx, OFT_WARN, "IPv%d decrement TTL exception", |
5095 | flow->dl_type == htons(ETH_TYPE_IP) ? 4 : 6); | |
9583bc14 EJ |
5096 | return true; |
5097 | } | |
5098 | } | |
5099 | ||
8bfd0fda | 5100 | static void |
097d4939 JR |
5101 | compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label) |
5102 | { | |
8bfd0fda | 5103 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 5104 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK); |
8bfd0fda | 5105 | set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label); |
097d4939 | 5106 | } |
097d4939 JR |
5107 | } |
5108 | ||
8bfd0fda | 5109 | static void |
097d4939 JR |
5110 | compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc) |
5111 | { | |
8bfd0fda | 5112 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 5113 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TC_MASK); |
8bfd0fda | 5114 | set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc); |
097d4939 | 5115 | } |
097d4939 JR |
5116 | } |
5117 | ||
491e05c2 YY |
5118 | static bool |
5119 | compose_dec_nsh_ttl_action(struct xlate_ctx *ctx) | |
5120 | { | |
5121 | struct flow *flow = &ctx->xin->flow; | |
5122 | ||
5123 | if ((flow->packet_type == htonl(PT_NSH)) || | |
5124 | (flow->dl_type == htons(ETH_TYPE_NSH))) { | |
5125 | ctx->wc->masks.nsh.ttl = 0xff; | |
5126 | if (flow->nsh.ttl > 1) { | |
5127 | flow->nsh.ttl--; | |
5128 | return false; | |
5129 | } else { | |
5130 | xlate_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, | |
206ddb9a | 5131 | 0, UINT32_MAX, NULL, 0); |
491e05c2 YY |
5132 | } |
5133 | } | |
5134 | ||
5135 | /* Stop processing for current table. */ | |
5136 | xlate_report(ctx, OFT_WARN, "NSH decrement TTL exception"); | |
5137 | return true; | |
5138 | } | |
5139 | ||
8bfd0fda | 5140 | static void |
9cfef3d0 | 5141 | compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl) |
9583bc14 | 5142 | { |
8bfd0fda | 5143 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 5144 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda | 5145 | set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl); |
b0a17866 | 5146 | } |
9583bc14 EJ |
5147 | } |
5148 | ||
5149 | static bool | |
9cfef3d0 | 5150 | compose_dec_mpls_ttl_action(struct xlate_ctx *ctx) |
9583bc14 | 5151 | { |
33bf9176 | 5152 | struct flow *flow = &ctx->xin->flow; |
1dd35f8a | 5153 | |
8bfd0fda | 5154 | if (eth_type_mpls(flow->dl_type)) { |
22d38fca JR |
5155 | uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]); |
5156 | ||
49a73e0c | 5157 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda BP |
5158 | if (ttl > 1) { |
5159 | ttl--; | |
5160 | set_mpls_lse_ttl(&flow->mpls_lse[0], ttl); | |
5161 | return false; | |
5162 | } else { | |
d39ec23d | 5163 | xlate_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0, |
206ddb9a | 5164 | UINT32_MAX, NULL, 0); |
8bfd0fda | 5165 | } |
9583bc14 | 5166 | } |
22d38fca JR |
5167 | |
5168 | /* Stop processing for current table. */ | |
2d9b49dd | 5169 | xlate_report(ctx, OFT_WARN, "MPLS decrement TTL exception"); |
22d38fca | 5170 | return true; |
9583bc14 EJ |
5171 | } |
5172 | ||
81f71381 YHW |
5173 | static void |
5174 | xlate_delete_field(struct xlate_ctx *ctx, | |
5175 | struct flow *flow, | |
5176 | const struct ofpact_delete_field *odf) | |
5177 | { | |
5178 | struct ds s = DS_EMPTY_INITIALIZER; | |
5179 | ||
5180 | /* Currently, only tun_metadata is allowed for delete_field action. */ | |
5181 | tun_metadata_delete(&flow->tunnel, odf->field); | |
5182 | ||
5183 | ds_put_format(&s, "delete %s", odf->field->name); | |
5184 | xlate_report(ctx, OFT_DETAIL, "%s", ds_cstr(&s)); | |
5185 | ds_destroy(&s); | |
5186 | } | |
5187 | ||
8b496c72 BP |
5188 | /* Emits an action that outputs to 'port', within 'ctx'. |
5189 | * | |
5190 | * 'controller_len' affects only packets sent to an OpenFlow controller. It | |
5191 | * is the maximum number of bytes of the packet to send. UINT16_MAX means to | |
5192 | * send the whole packet (and 0 means to omit the packet entirely). | |
5193 | * | |
5194 | * 'may_packet_in' determines whether the packet may be sent to an OpenFlow | |
5195 | * controller. If it is false, then the packet is never sent to the OpenFlow | |
5196 | * controller. | |
5197 | * | |
5198 | * 'is_last_action' should be true if this output is the last OpenFlow action | |
5199 | * to be processed, which enables certain optimizations. | |
5200 | * | |
5201 | * 'truncate' should be true if the packet to be output is being truncated, | |
5202 | * which suppresses certain optimizations. */ | |
9583bc14 | 5203 | static void |
8b496c72 BP |
5204 | xlate_output_action(struct xlate_ctx *ctx, ofp_port_t port, |
5205 | uint16_t controller_len, bool may_packet_in, | |
f5634764 KG |
5206 | bool is_last_action, bool truncate, |
5207 | bool group_bucket_action) | |
9583bc14 | 5208 | { |
2031ef97 | 5209 | ofp_port_t prev_nf_output_iface = ctx->nf_output_iface; |
9583bc14 | 5210 | |
2031ef97 | 5211 | ctx->nf_output_iface = NF_OUT_DROP; |
9583bc14 EJ |
5212 | |
5213 | switch (port) { | |
5214 | case OFPP_IN_PORT: | |
feee58b9 | 5215 | compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port, NULL, |
8bbbda3c | 5216 | is_last_action, truncate); |
9583bc14 EJ |
5217 | break; |
5218 | case OFPP_TABLE: | |
4e022ec0 | 5219 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
11938578 | 5220 | 0, may_packet_in, true, false, false, |
96c3a6e5 | 5221 | do_xlate_actions); |
9583bc14 EJ |
5222 | break; |
5223 | case OFPP_NORMAL: | |
5224 | xlate_normal(ctx); | |
5225 | break; | |
5226 | case OFPP_FLOOD: | |
feee58b9 | 5227 | flood_packets(ctx, false, is_last_action); |
9583bc14 EJ |
5228 | break; |
5229 | case OFPP_ALL: | |
feee58b9 | 5230 | flood_packets(ctx, true, is_last_action); |
9583bc14 EJ |
5231 | break; |
5232 | case OFPP_CONTROLLER: | |
d39ec23d JP |
5233 | xlate_controller_action(ctx, controller_len, |
5234 | (ctx->in_packet_out ? OFPR_PACKET_OUT | |
f5634764 | 5235 | : group_bucket_action ? OFPR_GROUP |
d39ec23d JP |
5236 | : ctx->in_action_set ? OFPR_ACTION_SET |
5237 | : OFPR_ACTION), | |
206ddb9a | 5238 | 0, UINT32_MAX, NULL, 0); |
9583bc14 EJ |
5239 | break; |
5240 | case OFPP_NONE: | |
5241 | break; | |
5242 | case OFPP_LOCAL: | |
5243 | default: | |
4e022ec0 | 5244 | if (port != ctx->xin->flow.in_port.ofp_port) { |
8bbbda3c | 5245 | compose_output_action(ctx, port, NULL, is_last_action, truncate); |
9583bc14 | 5246 | } else { |
f89547e2 | 5247 | xlate_report_info(ctx, "skipping output to input port"); |
9583bc14 EJ |
5248 | } |
5249 | break; | |
5250 | } | |
5251 | ||
5252 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
2031ef97 BP |
5253 | ctx->nf_output_iface = NF_OUT_FLOOD; |
5254 | } else if (ctx->nf_output_iface == NF_OUT_DROP) { | |
5255 | ctx->nf_output_iface = prev_nf_output_iface; | |
9583bc14 | 5256 | } else if (prev_nf_output_iface != NF_OUT_DROP && |
2031ef97 BP |
5257 | ctx->nf_output_iface != NF_OUT_FLOOD) { |
5258 | ctx->nf_output_iface = NF_OUT_MULTI; | |
9583bc14 EJ |
5259 | } |
5260 | } | |
5261 | ||
5262 | static void | |
5263 | xlate_output_reg_action(struct xlate_ctx *ctx, | |
feee58b9 | 5264 | const struct ofpact_output_reg *or, |
f5634764 KG |
5265 | bool is_last_action, |
5266 | bool group_bucket_action) | |
9583bc14 EJ |
5267 | { |
5268 | uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow); | |
5269 | if (port <= UINT16_MAX) { | |
2d9b49dd BP |
5270 | xlate_report(ctx, OFT_DETAIL, "output port is %"PRIu64, port); |
5271 | ||
9583bc14 EJ |
5272 | union mf_subvalue value; |
5273 | ||
5274 | memset(&value, 0xff, sizeof value); | |
49a73e0c | 5275 | mf_write_subfield_flow(&or->src, &value, &ctx->wc->masks); |
8b496c72 | 5276 | xlate_output_action(ctx, u16_to_ofp(port), or->max_len, |
f5634764 KG |
5277 | false, is_last_action, false, |
5278 | group_bucket_action); | |
2d9b49dd BP |
5279 | } else { |
5280 | xlate_report(ctx, OFT_WARN, "output port %"PRIu64" is out of range", | |
5281 | port); | |
9583bc14 EJ |
5282 | } |
5283 | } | |
5284 | ||
aaca4fe0 WT |
5285 | static void |
5286 | xlate_output_trunc_action(struct xlate_ctx *ctx, | |
feee58b9 | 5287 | ofp_port_t port, uint32_t max_len, |
f5634764 KG |
5288 | bool is_last_action, |
5289 | bool group_bucket_action) | |
aaca4fe0 WT |
5290 | { |
5291 | bool support_trunc = ctx->xbridge->support.trunc; | |
5292 | struct ovs_action_trunc *trunc; | |
29718ad4 | 5293 | char name[OFP_MAX_PORT_NAME_LEN]; |
aaca4fe0 WT |
5294 | |
5295 | switch (port) { | |
5296 | case OFPP_TABLE: | |
5297 | case OFPP_NORMAL: | |
5298 | case OFPP_FLOOD: | |
5299 | case OFPP_ALL: | |
5300 | case OFPP_CONTROLLER: | |
5301 | case OFPP_NONE: | |
50f96b10 | 5302 | ofputil_port_to_string(port, NULL, name, sizeof name); |
2d9b49dd BP |
5303 | xlate_report(ctx, OFT_WARN, |
5304 | "output_trunc does not support port: %s", name); | |
aaca4fe0 WT |
5305 | break; |
5306 | case OFPP_LOCAL: | |
5307 | case OFPP_IN_PORT: | |
5308 | default: | |
5309 | if (port != ctx->xin->flow.in_port.ofp_port) { | |
5310 | const struct xport *xport = get_ofp_port(ctx->xbridge, port); | |
5311 | ||
5312 | if (xport == NULL || xport->odp_port == ODPP_NONE) { | |
5313 | /* Since truncate happens at its following output action, if | |
5314 | * the output port is a patch port, the behavior is somehow | |
49f17344 | 5315 | * unpredictable. For simplicity, disallow this case. */ |
50f96b10 | 5316 | ofputil_port_to_string(port, NULL, name, sizeof name); |
2d9b49dd BP |
5317 | xlate_report_error(ctx, "output_trunc does not support " |
5318 | "patch port %s", name); | |
aaca4fe0 WT |
5319 | break; |
5320 | } | |
5321 | ||
5322 | trunc = nl_msg_put_unspec_uninit(ctx->odp_actions, | |
5323 | OVS_ACTION_ATTR_TRUNC, | |
5324 | sizeof *trunc); | |
5325 | trunc->max_len = max_len; | |
f5634764 KG |
5326 | xlate_output_action(ctx, port, 0, false, is_last_action, true, |
5327 | group_bucket_action); | |
aaca4fe0 WT |
5328 | if (!support_trunc) { |
5329 | ctx->xout->slow |= SLOW_ACTION; | |
5330 | } | |
5331 | } else { | |
f89547e2 | 5332 | xlate_report_info(ctx, "skipping output to input port"); |
aaca4fe0 WT |
5333 | } |
5334 | break; | |
5335 | } | |
5336 | } | |
5337 | ||
9583bc14 EJ |
5338 | static void |
5339 | xlate_enqueue_action(struct xlate_ctx *ctx, | |
feee58b9 | 5340 | const struct ofpact_enqueue *enqueue, |
f5634764 KG |
5341 | bool is_last_action, |
5342 | bool group_bucket_action) | |
9583bc14 | 5343 | { |
4e022ec0 | 5344 | ofp_port_t ofp_port = enqueue->port; |
9583bc14 EJ |
5345 | uint32_t queue_id = enqueue->queue; |
5346 | uint32_t flow_priority, priority; | |
5347 | int error; | |
5348 | ||
5349 | /* Translate queue to priority. */ | |
89a8a7f0 | 5350 | error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority); |
9583bc14 EJ |
5351 | if (error) { |
5352 | /* Fall back to ordinary output action. */ | |
8b496c72 | 5353 | xlate_output_action(ctx, enqueue->port, 0, false, |
f5634764 KG |
5354 | is_last_action, false, |
5355 | group_bucket_action); | |
9583bc14 EJ |
5356 | return; |
5357 | } | |
5358 | ||
5359 | /* Check output port. */ | |
5360 | if (ofp_port == OFPP_IN_PORT) { | |
4e022ec0 AW |
5361 | ofp_port = ctx->xin->flow.in_port.ofp_port; |
5362 | } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
5363 | return; |
5364 | } | |
5365 | ||
5366 | /* Add datapath actions. */ | |
5367 | flow_priority = ctx->xin->flow.skb_priority; | |
5368 | ctx->xin->flow.skb_priority = priority; | |
11938578 | 5369 | compose_output_action(ctx, ofp_port, NULL, is_last_action, false); |
9583bc14 EJ |
5370 | ctx->xin->flow.skb_priority = flow_priority; |
5371 | ||
5372 | /* Update NetFlow output port. */ | |
2031ef97 BP |
5373 | if (ctx->nf_output_iface == NF_OUT_DROP) { |
5374 | ctx->nf_output_iface = ofp_port; | |
5375 | } else if (ctx->nf_output_iface != NF_OUT_FLOOD) { | |
5376 | ctx->nf_output_iface = NF_OUT_MULTI; | |
9583bc14 EJ |
5377 | } |
5378 | } | |
5379 | ||
5380 | static void | |
5381 | xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id) | |
5382 | { | |
5383 | uint32_t skb_priority; | |
5384 | ||
89a8a7f0 | 5385 | if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) { |
9583bc14 EJ |
5386 | ctx->xin->flow.skb_priority = skb_priority; |
5387 | } else { | |
5388 | /* Couldn't translate queue to a priority. Nothing to do. A warning | |
5389 | * has already been logged. */ | |
5390 | } | |
5391 | } | |
5392 | ||
5393 | static bool | |
46c88433 | 5394 | slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_) |
9583bc14 | 5395 | { |
46c88433 EJ |
5396 | const struct xbridge *xbridge = xbridge_; |
5397 | struct xport *port; | |
9583bc14 EJ |
5398 | |
5399 | switch (ofp_port) { | |
5400 | case OFPP_IN_PORT: | |
5401 | case OFPP_TABLE: | |
5402 | case OFPP_NORMAL: | |
5403 | case OFPP_FLOOD: | |
5404 | case OFPP_ALL: | |
5405 | case OFPP_NONE: | |
5406 | return true; | |
5407 | case OFPP_CONTROLLER: /* Not supported by the bundle action. */ | |
5408 | return false; | |
5409 | default: | |
46c88433 | 5410 | port = get_ofp_port(xbridge, ofp_port); |
9583bc14 EJ |
5411 | return port ? port->may_enable : false; |
5412 | } | |
5413 | } | |
5414 | ||
5415 | static void | |
5416 | xlate_bundle_action(struct xlate_ctx *ctx, | |
feee58b9 | 5417 | const struct ofpact_bundle *bundle, |
f5634764 KG |
5418 | bool is_last_action, |
5419 | bool group_bucket_action) | |
9583bc14 | 5420 | { |
4e022ec0 | 5421 | ofp_port_t port; |
9583bc14 | 5422 | |
49a73e0c | 5423 | port = bundle_execute(bundle, &ctx->xin->flow, ctx->wc, slave_enabled_cb, |
46c88433 | 5424 | CONST_CAST(struct xbridge *, ctx->xbridge)); |
9583bc14 | 5425 | if (bundle->dst.field) { |
49a73e0c | 5426 | nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow, ctx->wc); |
2d9b49dd | 5427 | xlate_report_subfield(ctx, &bundle->dst); |
9583bc14 | 5428 | } else { |
f5634764 KG |
5429 | xlate_output_action(ctx, port, 0, false, is_last_action, false, |
5430 | group_bucket_action); | |
9583bc14 EJ |
5431 | } |
5432 | } | |
5433 | ||
4165b5e0 JS |
5434 | static void |
5435 | xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) | |
5436 | { | |
49a73e0c | 5437 | learn_mask(learn, ctx->wc); |
9583bc14 | 5438 | |
df70a773 | 5439 | if (ctx->xin->xcache || ctx->xin->allow_side_effects) { |
4165b5e0 JS |
5440 | uint64_t ofpacts_stub[1024 / 8]; |
5441 | struct ofputil_flow_mod fm; | |
2c7ee524 | 5442 | struct ofproto_flow_mod ofm__, *ofm; |
4165b5e0 | 5443 | struct ofpbuf ofpacts; |
2c7ee524 JR |
5444 | enum ofperr error; |
5445 | ||
5446 | if (ctx->xin->xcache) { | |
3f3b97b0 | 5447 | ofm = xmalloc(sizeof *ofm); |
2c7ee524 JR |
5448 | } else { |
5449 | ofm = &ofm__; | |
5450 | } | |
4165b5e0 JS |
5451 | |
5452 | ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); | |
2c7ee524 | 5453 | learn_execute(learn, &ctx->xin->flow, &fm, &ofpacts); |
2d9b49dd BP |
5454 | if (OVS_UNLIKELY(ctx->xin->trace)) { |
5455 | struct ds s = DS_EMPTY_INITIALIZER; | |
5456 | ds_put_format(&s, "table=%"PRIu8" ", fm.table_id); | |
6a6b7060 BP |
5457 | minimatch_format(&fm.match, |
5458 | ofproto_get_tun_tab(&ctx->xin->ofproto->up), | |
5459 | NULL, &s, OFP_DEFAULT_PRIORITY); | |
2d9b49dd BP |
5460 | ds_chomp(&s, ' '); |
5461 | ds_put_format(&s, " priority=%d", fm.priority); | |
5462 | if (fm.new_cookie) { | |
5463 | ds_put_format(&s, " cookie=%#"PRIx64, ntohll(fm.new_cookie)); | |
5464 | } | |
5465 | if (fm.idle_timeout != OFP_FLOW_PERMANENT) { | |
5466 | ds_put_format(&s, " idle=%"PRIu16, fm.idle_timeout); | |
5467 | } | |
5468 | if (fm.hard_timeout != OFP_FLOW_PERMANENT) { | |
5469 | ds_put_format(&s, " hard=%"PRIu16, fm.hard_timeout); | |
5470 | } | |
5471 | if (fm.flags & NX_LEARN_F_SEND_FLOW_REM) { | |
5472 | ds_put_cstr(&s, " send_flow_rem"); | |
5473 | } | |
5474 | ds_put_cstr(&s, " actions="); | |
efefbcae BP |
5475 | struct ofpact_format_params fp = { .s = &s }; |
5476 | ofpacts_format(fm.ofpacts, fm.ofpacts_len, &fp); | |
2d9b49dd BP |
5477 | xlate_report(ctx, OFT_DETAIL, "%s", ds_cstr(&s)); |
5478 | ds_destroy(&s); | |
5479 | } | |
2c7ee524 JR |
5480 | error = ofproto_dpif_flow_mod_init_for_learn(ctx->xbridge->ofproto, |
5481 | &fm, ofm); | |
4165b5e0 | 5482 | ofpbuf_uninit(&ofpacts); |
2c7ee524 | 5483 | |
3f3b97b0 | 5484 | if (!error) { |
4c71600d | 5485 | bool success = true; |
3f3b97b0 | 5486 | if (ctx->xin->allow_side_effects) { |
4c71600d DDP |
5487 | error = ofproto_flow_mod_learn(ofm, ctx->xin->xcache != NULL, |
5488 | learn->limit, &success); | |
5489 | } else if (learn->limit) { | |
5490 | if (!ofm->temp_rule | |
5491 | || ofm->temp_rule->state != RULE_INSERTED) { | |
5492 | /* The learned rule expired and there are no packets, so | |
5493 | * we cannot learn again. Since the translated actions | |
5494 | * depend on the result of learning, we tell the caller | |
5495 | * that there's no point in caching this result. */ | |
5496 | ctx->xout->avoid_caching = true; | |
5497 | } | |
3f3b97b0 DDP |
5498 | } |
5499 | ||
4c71600d DDP |
5500 | if (learn->flags & NX_LEARN_F_WRITE_RESULT) { |
5501 | nxm_reg_load(&learn->result_dst, success ? 1 : 0, | |
5502 | &ctx->xin->flow, ctx->wc); | |
5503 | xlate_report_subfield(ctx, &learn->result_dst); | |
5504 | } | |
5505 | ||
5506 | if (success && ctx->xin->xcache) { | |
3f3b97b0 DDP |
5507 | struct xc_entry *entry; |
5508 | ||
5509 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN); | |
5510 | entry->learn.ofm = ofm; | |
4c71600d | 5511 | entry->learn.limit = learn->limit; |
3f3b97b0 | 5512 | ofm = NULL; |
1bddcb5d YS |
5513 | } else { |
5514 | ofproto_flow_mod_uninit(ofm); | |
3f3b97b0 | 5515 | } |
4c71600d DDP |
5516 | |
5517 | if (OVS_UNLIKELY(ctx->xin->trace && !success)) { | |
5518 | xlate_report(ctx, OFT_DETAIL, "Limit exceeded, learn failed"); | |
5519 | } | |
3f3b97b0 DDP |
5520 | } |
5521 | ||
3f76c123 | 5522 | if (ofm != &ofm__) { |
3f3b97b0 | 5523 | free(ofm); |
2c7ee524 JR |
5524 | } |
5525 | ||
5526 | if (error) { | |
2d9b49dd BP |
5527 | xlate_report_error(ctx, "LEARN action execution failed (%s).", |
5528 | ofperr_to_string(error)); | |
2c7ee524 | 5529 | } |
6a6b7060 BP |
5530 | |
5531 | minimatch_destroy(&fm.match); | |
2d9b49dd BP |
5532 | } else { |
5533 | xlate_report(ctx, OFT_WARN, | |
5534 | "suppressing side effects, so learn action ignored"); | |
b256dc52 JS |
5535 | } |
5536 | } | |
5537 | ||
5538 | static void | |
5539 | xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags, | |
5540 | uint16_t idle_timeout, uint16_t hard_timeout) | |
5541 | { | |
5542 | if (tcp_flags & (TCP_FIN | TCP_RST)) { | |
07a3cd5c | 5543 | ofproto_rule_reduce_timeouts(&rule->up, idle_timeout, hard_timeout); |
b256dc52 | 5544 | } |
9583bc14 EJ |
5545 | } |
5546 | ||
9583bc14 EJ |
5547 | static void |
5548 | xlate_fin_timeout(struct xlate_ctx *ctx, | |
5549 | const struct ofpact_fin_timeout *oft) | |
5550 | { | |
b256dc52 | 5551 | if (ctx->rule) { |
df70a773 JR |
5552 | if (ctx->xin->allow_side_effects) { |
5553 | xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags, | |
5554 | oft->fin_idle_timeout, oft->fin_hard_timeout); | |
5555 | } | |
b256dc52 JS |
5556 | if (ctx->xin->xcache) { |
5557 | struct xc_entry *entry; | |
5558 | ||
5559 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT); | |
83709dfa JR |
5560 | /* XC_RULE already holds a reference on the rule, none is taken |
5561 | * here. */ | |
901a517e JR |
5562 | entry->fin.rule = ctx->rule; |
5563 | entry->fin.idle = oft->fin_idle_timeout; | |
5564 | entry->fin.hard = oft->fin_hard_timeout; | |
b256dc52 | 5565 | } |
9583bc14 EJ |
5566 | } |
5567 | } | |
5568 | ||
5569 | static void | |
5570 | xlate_sample_action(struct xlate_ctx *ctx, | |
5571 | const struct ofpact_sample *os) | |
5572 | { | |
f69f713b BY |
5573 | odp_port_t output_odp_port = ODPP_NONE; |
5574 | odp_port_t tunnel_out_port = ODPP_NONE; | |
5575 | struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; | |
5576 | bool emit_set_tunnel = false; | |
5577 | ||
5578 | if (!ipfix || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { | |
5579 | return; | |
5580 | } | |
5581 | ||
e824d78d JR |
5582 | /* Scale the probability from 16-bit to 32-bit while representing |
5583 | * the same percentage. */ | |
5584 | uint32_t probability = (os->probability << 16) | os->probability; | |
5585 | ||
f69f713b BY |
5586 | /* If ofp_port in flow sample action is equel to ofp_port, |
5587 | * this sample action is a input port action. */ | |
5588 | if (os->sampling_port != OFPP_NONE && | |
5589 | os->sampling_port != ctx->xin->flow.in_port.ofp_port) { | |
5590 | output_odp_port = ofp_port_to_odp_port(ctx->xbridge, | |
5591 | os->sampling_port); | |
5592 | if (output_odp_port == ODPP_NONE) { | |
2d9b49dd BP |
5593 | xlate_report_error(ctx, "can't use unknown port %d in flow sample " |
5594 | "action", os->sampling_port); | |
f69f713b BY |
5595 | return; |
5596 | } | |
5597 | ||
5598 | if (dpif_ipfix_get_flow_exporter_tunnel_sampling(ipfix, | |
5599 | os->collector_set_id) | |
cd32509e | 5600 | && dpif_ipfix_is_tunnel_port(ipfix, output_odp_port)) { |
f69f713b BY |
5601 | tunnel_out_port = output_odp_port; |
5602 | emit_set_tunnel = true; | |
5603 | } | |
5604 | } | |
5605 | ||
5606 | xlate_commit_actions(ctx); | |
5607 | /* If 'emit_set_tunnel', sample(sampling_port=1) would translate | |
5608 | * into datapath sample action set(tunnel(...)), sample(...) and | |
5609 | * it is used for sampling egress tunnel information. */ | |
5610 | if (emit_set_tunnel) { | |
5611 | const struct xport *xport = get_ofp_port(ctx->xbridge, | |
5612 | os->sampling_port); | |
5613 | ||
5614 | if (xport && xport->is_tunnel) { | |
5615 | struct flow *flow = &ctx->xin->flow; | |
5616 | tnl_port_send(xport->ofport, flow, ctx->wc); | |
5617 | if (!ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { | |
5618 | struct flow_tnl flow_tnl = flow->tunnel; | |
c6d87201 | 5619 | const char *tnl_type; |
f69f713b | 5620 | |
c6d87201 | 5621 | tnl_type = tnl_port_get_type(xport->ofport); |
f69f713b | 5622 | commit_odp_tunnel_action(flow, &ctx->base_flow, |
c6d87201 | 5623 | ctx->odp_actions, tnl_type); |
f69f713b BY |
5624 | flow->tunnel = flow_tnl; |
5625 | } | |
5626 | } else { | |
2d9b49dd BP |
5627 | xlate_report_error(ctx, |
5628 | "sampling_port:%d should be a tunnel port.", | |
5629 | os->sampling_port); | |
f69f713b BY |
5630 | } |
5631 | } | |
e824d78d | 5632 | |
24a4bbe1 IM |
5633 | struct user_action_cookie cookie; |
5634 | ||
5635 | memset(&cookie, 0, sizeof cookie); | |
5636 | cookie.type = USER_ACTION_COOKIE_FLOW_SAMPLE; | |
5637 | cookie.ofp_in_port = ctx->xin->flow.in_port.ofp_port; | |
5638 | cookie.ofproto_uuid = ctx->xbridge->ofproto->uuid; | |
5639 | cookie.flow_sample.probability = os->probability; | |
5640 | cookie.flow_sample.collector_set_id = os->collector_set_id; | |
5641 | cookie.flow_sample.obs_domain_id = os->obs_domain_id; | |
5642 | cookie.flow_sample.obs_point_id = os->obs_point_id; | |
5643 | cookie.flow_sample.output_odp_port = output_odp_port; | |
5644 | cookie.flow_sample.direction = os->direction; | |
5645 | ||
8de6ff3e | 5646 | compose_sample_action(ctx, probability, &cookie, tunnel_out_port, false); |
9583bc14 EJ |
5647 | } |
5648 | ||
eee69393 AZ |
5649 | /* Determine if an datapath action translated from the openflow action |
5650 | * can be reversed by another datapath action. | |
5651 | * | |
5652 | * Openflow actions that do not emit datapath actions are trivially | |
5653 | * reversible. Reversiblity of other actions depends on nature of | |
5654 | * action and their translation. */ | |
5655 | static bool | |
5656 | reversible_actions(const struct ofpact *ofpacts, size_t ofpacts_len) | |
bef503e8 | 5657 | { |
eee69393 | 5658 | const struct ofpact *a; |
bef503e8 | 5659 | |
eee69393 AZ |
5660 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
5661 | switch (a->type) { | |
5662 | case OFPACT_BUNDLE: | |
5663 | case OFPACT_CLEAR_ACTIONS: | |
5664 | case OFPACT_CLONE: | |
5665 | case OFPACT_CONJUNCTION: | |
5666 | case OFPACT_CONTROLLER: | |
5667 | case OFPACT_CT_CLEAR: | |
5668 | case OFPACT_DEBUG_RECIRC: | |
a934a3dd | 5669 | case OFPACT_DEBUG_SLOW: |
eee69393 AZ |
5670 | case OFPACT_DEC_MPLS_TTL: |
5671 | case OFPACT_DEC_TTL: | |
5672 | case OFPACT_ENQUEUE: | |
5673 | case OFPACT_EXIT: | |
5674 | case OFPACT_FIN_TIMEOUT: | |
5675 | case OFPACT_GOTO_TABLE: | |
5676 | case OFPACT_GROUP: | |
5677 | case OFPACT_LEARN: | |
5678 | case OFPACT_MULTIPATH: | |
5679 | case OFPACT_NOTE: | |
5680 | case OFPACT_OUTPUT: | |
5681 | case OFPACT_OUTPUT_REG: | |
5682 | case OFPACT_POP_MPLS: | |
5683 | case OFPACT_POP_QUEUE: | |
5684 | case OFPACT_PUSH_MPLS: | |
5685 | case OFPACT_PUSH_VLAN: | |
5686 | case OFPACT_REG_MOVE: | |
5687 | case OFPACT_RESUBMIT: | |
5688 | case OFPACT_SAMPLE: | |
5689 | case OFPACT_SET_ETH_DST: | |
5690 | case OFPACT_SET_ETH_SRC: | |
5691 | case OFPACT_SET_FIELD: | |
5692 | case OFPACT_SET_IP_DSCP: | |
5693 | case OFPACT_SET_IP_ECN: | |
5694 | case OFPACT_SET_IP_TTL: | |
5695 | case OFPACT_SET_IPV4_DST: | |
5696 | case OFPACT_SET_IPV4_SRC: | |
5697 | case OFPACT_SET_L4_DST_PORT: | |
5698 | case OFPACT_SET_L4_SRC_PORT: | |
5699 | case OFPACT_SET_MPLS_LABEL: | |
5700 | case OFPACT_SET_MPLS_TC: | |
5701 | case OFPACT_SET_MPLS_TTL: | |
5702 | case OFPACT_SET_QUEUE: | |
5703 | case OFPACT_SET_TUNNEL: | |
5704 | case OFPACT_SET_VLAN_PCP: | |
5705 | case OFPACT_SET_VLAN_VID: | |
5706 | case OFPACT_STACK_POP: | |
5707 | case OFPACT_STACK_PUSH: | |
5708 | case OFPACT_STRIP_VLAN: | |
5709 | case OFPACT_UNROLL_XLATE: | |
5710 | case OFPACT_WRITE_ACTIONS: | |
5711 | case OFPACT_WRITE_METADATA: | |
5b34f8fc | 5712 | case OFPACT_CHECK_PKT_LARGER: |
81f71381 | 5713 | case OFPACT_DELETE_FIELD: |
eee69393 AZ |
5714 | break; |
5715 | ||
5716 | case OFPACT_CT: | |
5717 | case OFPACT_METER: | |
5718 | case OFPACT_NAT: | |
5719 | case OFPACT_OUTPUT_TRUNC: | |
2142be1f BP |
5720 | case OFPACT_ENCAP: |
5721 | case OFPACT_DECAP: | |
491e05c2 | 5722 | case OFPACT_DEC_NSH_TTL: |
eee69393 | 5723 | return false; |
9c2a44dc | 5724 | } |
456024cb | 5725 | } |
eee69393 | 5726 | return true; |
bef503e8 AZ |
5727 | } |
5728 | ||
5729 | static void | |
c9f0a445 | 5730 | clone_xlate_actions(const struct ofpact *actions, size_t actions_len, |
f5634764 KG |
5731 | struct xlate_ctx *ctx, bool is_last_action, |
5732 | bool group_bucket_action OVS_UNUSED) | |
7ae62a67 | 5733 | { |
b827b231 BP |
5734 | struct ofpbuf old_stack = ctx->stack; |
5735 | union mf_subvalue new_stack[1024 / sizeof(union mf_subvalue)]; | |
5736 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); | |
5737 | ofpbuf_put(&ctx->stack, old_stack.data, old_stack.size); | |
5738 | ||
5739 | struct ofpbuf old_action_set = ctx->action_set; | |
5740 | uint64_t actset_stub[1024 / 8]; | |
5741 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); | |
5742 | ofpbuf_put(&ctx->action_set, old_action_set.data, old_action_set.size); | |
5743 | ||
eee69393 | 5744 | size_t offset, ac_offset; |
eee69393 AZ |
5745 | struct flow old_flow = ctx->xin->flow; |
5746 | ||
feee58b9 | 5747 | if (reversible_actions(actions, actions_len) || is_last_action) { |
eee69393 | 5748 | old_flow = ctx->xin->flow; |
f5634764 | 5749 | do_xlate_actions(actions, actions_len, ctx, is_last_action, false); |
118b21d9 EG |
5750 | if (!ctx->freezing) { |
5751 | xlate_action_set(ctx); | |
5752 | } | |
60eebf12 AZ |
5753 | if (ctx->freezing) { |
5754 | finish_freezing(ctx); | |
5755 | } | |
eee69393 AZ |
5756 | goto xlate_done; |
5757 | } | |
5758 | ||
5759 | /* Commit datapath actions before emitting the clone action to | |
5760 | * avoid emitting those actions twice. Once inside | |
5761 | * the clone, another time for the action after clone. */ | |
5762 | xlate_commit_actions(ctx); | |
9c2a44dc | 5763 | struct flow old_base = ctx->base_flow; |
eee69393 AZ |
5764 | bool old_was_mpls = ctx->was_mpls; |
5765 | bool old_conntracked = ctx->conntracked; | |
ba653d2a | 5766 | |
eee69393 AZ |
5767 | /* The actions are not reversible, a datapath clone action is |
5768 | * required to encode the translation. Select the clone action | |
5769 | * based on datapath capabilities. */ | |
5770 | if (ctx->xbridge->support.clone) { /* Use clone action */ | |
5771 | /* Use clone action as datapath clone. */ | |
5772 | offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CLONE); | |
f5634764 | 5773 | do_xlate_actions(actions, actions_len, ctx, true, false); |
118b21d9 EG |
5774 | if (!ctx->freezing) { |
5775 | xlate_action_set(ctx); | |
5776 | } | |
60eebf12 AZ |
5777 | if (ctx->freezing) { |
5778 | finish_freezing(ctx); | |
5779 | } | |
eee69393 AZ |
5780 | nl_msg_end_non_empty_nested(ctx->odp_actions, offset); |
5781 | goto dp_clone_done; | |
5782 | } | |
b827b231 | 5783 | |
eee69393 AZ |
5784 | if (ctx->xbridge->support.sample_nesting > 3) { |
5785 | /* Use sample action as datapath clone. */ | |
5786 | offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_SAMPLE); | |
5787 | ac_offset = nl_msg_start_nested(ctx->odp_actions, | |
5788 | OVS_SAMPLE_ATTR_ACTIONS); | |
f5634764 | 5789 | do_xlate_actions(actions, actions_len, ctx, true, false); |
118b21d9 EG |
5790 | if (!ctx->freezing) { |
5791 | xlate_action_set(ctx); | |
5792 | } | |
60eebf12 AZ |
5793 | if (ctx->freezing) { |
5794 | finish_freezing(ctx); | |
5795 | } | |
eee69393 AZ |
5796 | if (nl_msg_end_non_empty_nested(ctx->odp_actions, ac_offset)) { |
5797 | nl_msg_cancel_nested(ctx->odp_actions, offset); | |
5798 | } else { | |
5799 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, | |
5800 | UINT32_MAX); /* 100% probability. */ | |
5801 | nl_msg_end_nested(ctx->odp_actions, offset); | |
5802 | } | |
5803 | goto dp_clone_done; | |
5804 | } | |
5805 | ||
5806 | /* Datapath does not support clone, skip xlate 'oc' and | |
5807 | * report an error */ | |
5808 | xlate_report_error(ctx, "Failed to compose clone action"); | |
ba653d2a | 5809 | |
eee69393 | 5810 | dp_clone_done: |
ba653d2a BP |
5811 | /* The clone's conntrack execution should have no effect on the original |
5812 | * packet. */ | |
5813 | ctx->conntracked = old_conntracked; | |
bd3c2df3 BP |
5814 | |
5815 | /* Popping MPLS from the clone should have no effect on the original | |
5816 | * packet. */ | |
5817 | ctx->was_mpls = old_was_mpls; | |
eee69393 AZ |
5818 | |
5819 | /* Restore the 'base_flow' for the next action. */ | |
5820 | ctx->base_flow = old_base; | |
5821 | ||
5822 | xlate_done: | |
5823 | ofpbuf_uninit(&ctx->action_set); | |
5824 | ctx->action_set = old_action_set; | |
5825 | ofpbuf_uninit(&ctx->stack); | |
5826 | ctx->stack = old_stack; | |
5827 | ctx->xin->flow = old_flow; | |
7ae62a67 WT |
5828 | } |
5829 | ||
c9f0a445 | 5830 | static void |
feee58b9 AZ |
5831 | compose_clone(struct xlate_ctx *ctx, const struct ofpact_nest *oc, |
5832 | bool is_last_action) | |
c9f0a445 AZ |
5833 | { |
5834 | size_t oc_actions_len = ofpact_nest_get_action_len(oc); | |
5835 | ||
f5634764 KG |
5836 | clone_xlate_actions(oc->actions, oc_actions_len, ctx, is_last_action, |
5837 | false); | |
c9f0a445 AZ |
5838 | } |
5839 | ||
076caa2f JR |
5840 | static void |
5841 | xlate_meter_action(struct xlate_ctx *ctx, const struct ofpact_meter *meter) | |
5842 | { | |
5843 | if (meter->provider_meter_id != UINT32_MAX) { | |
5844 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_METER, | |
5845 | meter->provider_meter_id); | |
5846 | } | |
5847 | } | |
5848 | ||
9583bc14 | 5849 | static bool |
46c88433 | 5850 | may_receive(const struct xport *xport, struct xlate_ctx *ctx) |
9583bc14 | 5851 | { |
bbbca389 | 5852 | if (xport->config & (is_stp(&ctx->xin->flow) |
46c88433 EJ |
5853 | ? OFPUTIL_PC_NO_RECV_STP |
5854 | : OFPUTIL_PC_NO_RECV)) { | |
9583bc14 EJ |
5855 | return false; |
5856 | } | |
5857 | ||
5858 | /* Only drop packets here if both forwarding and learning are | |
5859 | * disabled. If just learning is enabled, we need to have | |
5860 | * OFPP_NORMAL and the learning action have a look at the packet | |
5861 | * before we can drop it. */ | |
9efd308e DV |
5862 | if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) || |
5863 | (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) { | |
9583bc14 EJ |
5864 | return false; |
5865 | } | |
5866 | ||
5867 | return true; | |
5868 | } | |
5869 | ||
7fdb60a7 | 5870 | static void |
7e7e8dbb BP |
5871 | xlate_write_actions__(struct xlate_ctx *ctx, |
5872 | const struct ofpact *ofpacts, size_t ofpacts_len) | |
7fdb60a7 | 5873 | { |
c61f3870 BP |
5874 | /* Maintain actset_output depending on the contents of the action set: |
5875 | * | |
5876 | * - OFPP_UNSET, if there is no "output" action. | |
5877 | * | |
5878 | * - The output port, if there is an "output" action and no "group" | |
5879 | * action. | |
5880 | * | |
5881 | * - OFPP_UNSET, if there is a "group" action. | |
5882 | */ | |
5883 | if (!ctx->action_set_has_group) { | |
7e7e8dbb BP |
5884 | const struct ofpact *a; |
5885 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { | |
5886 | if (a->type == OFPACT_OUTPUT) { | |
5887 | ctx->xin->flow.actset_output = ofpact_get_OUTPUT(a)->port; | |
5888 | } else if (a->type == OFPACT_GROUP) { | |
c61f3870 BP |
5889 | ctx->xin->flow.actset_output = OFPP_UNSET; |
5890 | ctx->action_set_has_group = true; | |
9055ca9a | 5891 | break; |
c61f3870 BP |
5892 | } |
5893 | } | |
5894 | } | |
5895 | ||
7e7e8dbb BP |
5896 | ofpbuf_put(&ctx->action_set, ofpacts, ofpacts_len); |
5897 | } | |
5898 | ||
5899 | static void | |
5900 | xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact_nest *a) | |
5901 | { | |
5902 | xlate_write_actions__(ctx, a->actions, ofpact_nest_get_action_len(a)); | |
7fdb60a7 SH |
5903 | } |
5904 | ||
5905 | static void | |
5906 | xlate_action_set(struct xlate_ctx *ctx) | |
5907 | { | |
2d9b49dd BP |
5908 | uint64_t action_list_stub[1024 / 8]; |
5909 | struct ofpbuf action_list = OFPBUF_STUB_INITIALIZER(action_list_stub); | |
7fdb60a7 | 5910 | ofpacts_execute_action_set(&action_list, &ctx->action_set); |
ed9c9e3e JR |
5911 | /* Clear the action set, as it is not needed any more. */ |
5912 | ofpbuf_clear(&ctx->action_set); | |
2d9b49dd BP |
5913 | if (action_list.size) { |
5914 | ctx->in_action_set = true; | |
5915 | ||
5916 | struct ovs_list *old_trace = ctx->xin->trace; | |
5917 | ctx->xin->trace = xlate_report(ctx, OFT_TABLE, | |
5918 | "--. Executing action set:"); | |
f5634764 | 5919 | do_xlate_actions(action_list.data, action_list.size, ctx, true, false); |
2d9b49dd BP |
5920 | ctx->xin->trace = old_trace; |
5921 | ||
5922 | ctx->in_action_set = false; | |
5923 | } | |
7fdb60a7 SH |
5924 | ofpbuf_uninit(&action_list); |
5925 | } | |
5926 | ||
e672ff9b | 5927 | static void |
1d361a81 | 5928 | freeze_put_unroll_xlate(struct xlate_ctx *ctx) |
e672ff9b | 5929 | { |
1d361a81 | 5930 | struct ofpact_unroll_xlate *unroll = ctx->frozen_actions.header; |
e672ff9b JR |
5931 | |
5932 | /* Restore the table_id and rule cookie for a potential PACKET | |
5933 | * IN if needed. */ | |
5934 | if (!unroll || | |
5935 | (ctx->table_id != unroll->rule_table_id | |
5936 | || ctx->rule_cookie != unroll->rule_cookie)) { | |
1d361a81 | 5937 | unroll = ofpact_put_UNROLL_XLATE(&ctx->frozen_actions); |
e672ff9b JR |
5938 | unroll->rule_table_id = ctx->table_id; |
5939 | unroll->rule_cookie = ctx->rule_cookie; | |
1d361a81 | 5940 | ctx->frozen_actions.header = unroll; |
e672ff9b JR |
5941 | } |
5942 | } | |
5943 | ||
5944 | ||
1d361a81 BP |
5945 | /* Copy actions 'a' through 'end' to ctx->frozen_actions, which will be |
5946 | * executed after thawing. Inserts an UNROLL_XLATE action, if none is already | |
5947 | * present, before any action that may depend on the current table ID or flow | |
5948 | * cookie. */ | |
e672ff9b | 5949 | static void |
1d361a81 | 5950 | freeze_unroll_actions(const struct ofpact *a, const struct ofpact *end, |
e672ff9b JR |
5951 | struct xlate_ctx *ctx) |
5952 | { | |
c2b283b7 | 5953 | for (; a < end; a = ofpact_next(a)) { |
e672ff9b | 5954 | switch (a->type) { |
e672ff9b | 5955 | case OFPACT_OUTPUT_REG: |
aaca4fe0 | 5956 | case OFPACT_OUTPUT_TRUNC: |
e672ff9b JR |
5957 | case OFPACT_GROUP: |
5958 | case OFPACT_OUTPUT: | |
5959 | case OFPACT_CONTROLLER: | |
5960 | case OFPACT_DEC_MPLS_TTL: | |
491e05c2 | 5961 | case OFPACT_DEC_NSH_TTL: |
e672ff9b | 5962 | case OFPACT_DEC_TTL: |
83a31283 BP |
5963 | /* These actions may generate asynchronous messages, which include |
5964 | * table ID and flow cookie information. */ | |
1d361a81 | 5965 | freeze_put_unroll_xlate(ctx); |
e672ff9b JR |
5966 | break; |
5967 | ||
83a31283 BP |
5968 | case OFPACT_RESUBMIT: |
5969 | if (ofpact_get_RESUBMIT(a)->table_id == 0xff) { | |
5970 | /* This resubmit action is relative to the current table, so we | |
5971 | * need to track what table that is.*/ | |
1d361a81 | 5972 | freeze_put_unroll_xlate(ctx); |
83a31283 BP |
5973 | } |
5974 | break; | |
5975 | ||
e672ff9b JR |
5976 | case OFPACT_SET_TUNNEL: |
5977 | case OFPACT_REG_MOVE: | |
5978 | case OFPACT_SET_FIELD: | |
5979 | case OFPACT_STACK_PUSH: | |
5980 | case OFPACT_STACK_POP: | |
5981 | case OFPACT_LEARN: | |
5982 | case OFPACT_WRITE_METADATA: | |
83a31283 | 5983 | case OFPACT_GOTO_TABLE: |
e672ff9b JR |
5984 | case OFPACT_ENQUEUE: |
5985 | case OFPACT_SET_VLAN_VID: | |
5986 | case OFPACT_SET_VLAN_PCP: | |
5987 | case OFPACT_STRIP_VLAN: | |
5988 | case OFPACT_PUSH_VLAN: | |
5989 | case OFPACT_SET_ETH_SRC: | |
5990 | case OFPACT_SET_ETH_DST: | |
5991 | case OFPACT_SET_IPV4_SRC: | |
5992 | case OFPACT_SET_IPV4_DST: | |
5993 | case OFPACT_SET_IP_DSCP: | |
5994 | case OFPACT_SET_IP_ECN: | |
5995 | case OFPACT_SET_IP_TTL: | |
5996 | case OFPACT_SET_L4_SRC_PORT: | |
5997 | case OFPACT_SET_L4_DST_PORT: | |
5998 | case OFPACT_SET_QUEUE: | |
5999 | case OFPACT_POP_QUEUE: | |
6000 | case OFPACT_PUSH_MPLS: | |
6001 | case OFPACT_POP_MPLS: | |
6002 | case OFPACT_SET_MPLS_LABEL: | |
6003 | case OFPACT_SET_MPLS_TC: | |
6004 | case OFPACT_SET_MPLS_TTL: | |
6005 | case OFPACT_MULTIPATH: | |
6006 | case OFPACT_BUNDLE: | |
6007 | case OFPACT_EXIT: | |
6008 | case OFPACT_UNROLL_XLATE: | |
6009 | case OFPACT_FIN_TIMEOUT: | |
6010 | case OFPACT_CLEAR_ACTIONS: | |
6011 | case OFPACT_WRITE_ACTIONS: | |
6012 | case OFPACT_METER: | |
6013 | case OFPACT_SAMPLE: | |
7ae62a67 | 6014 | case OFPACT_CLONE: |
f839892a JS |
6015 | case OFPACT_ENCAP: |
6016 | case OFPACT_DECAP: | |
d4abaff5 | 6017 | case OFPACT_DEBUG_RECIRC: |
a934a3dd | 6018 | case OFPACT_DEBUG_SLOW: |
07659514 | 6019 | case OFPACT_CT: |
72fe7578 | 6020 | case OFPACT_CT_CLEAR: |
9ac0aada | 6021 | case OFPACT_NAT: |
5b34f8fc | 6022 | case OFPACT_CHECK_PKT_LARGER: |
81f71381 | 6023 | case OFPACT_DELETE_FIELD: |
83a31283 | 6024 | /* These may not generate PACKET INs. */ |
e672ff9b JR |
6025 | break; |
6026 | ||
e672ff9b JR |
6027 | case OFPACT_NOTE: |
6028 | case OFPACT_CONJUNCTION: | |
83a31283 | 6029 | /* These need not be copied for restoration. */ |
e672ff9b JR |
6030 | continue; |
6031 | } | |
6032 | /* Copy the action over. */ | |
1d361a81 | 6033 | ofpbuf_put(&ctx->frozen_actions, a, OFPACT_ALIGN(a->len)); |
e672ff9b JR |
6034 | } |
6035 | } | |
6036 | ||
8e53fe8c | 6037 | static void |
f2d105b5 JS |
6038 | put_ct_mark(const struct flow *flow, struct ofpbuf *odp_actions, |
6039 | struct flow_wildcards *wc) | |
8e53fe8c | 6040 | { |
2a754f4a JS |
6041 | if (wc->masks.ct_mark) { |
6042 | struct { | |
6043 | uint32_t key; | |
6044 | uint32_t mask; | |
6045 | } *odp_ct_mark; | |
6046 | ||
6047 | odp_ct_mark = nl_msg_put_unspec_uninit(odp_actions, OVS_CT_ATTR_MARK, | |
6048 | sizeof(*odp_ct_mark)); | |
6049 | odp_ct_mark->key = flow->ct_mark & wc->masks.ct_mark; | |
6050 | odp_ct_mark->mask = wc->masks.ct_mark; | |
8e53fe8c JS |
6051 | } |
6052 | } | |
6053 | ||
9daf2348 | 6054 | static void |
f2d105b5 JS |
6055 | put_ct_label(const struct flow *flow, struct ofpbuf *odp_actions, |
6056 | struct flow_wildcards *wc) | |
9daf2348 | 6057 | { |
2ff8484b | 6058 | if (!ovs_u128_is_zero(wc->masks.ct_label)) { |
9daf2348 JS |
6059 | struct { |
6060 | ovs_u128 key; | |
6061 | ovs_u128 mask; | |
89cf41ec | 6062 | } odp_ct_label; |
9daf2348 | 6063 | |
89cf41ec BP |
6064 | odp_ct_label.key = ovs_u128_and(flow->ct_label, wc->masks.ct_label); |
6065 | odp_ct_label.mask = wc->masks.ct_label; | |
6066 | nl_msg_put_unspec(odp_actions, OVS_CT_ATTR_LABELS, | |
6067 | &odp_ct_label, sizeof odp_ct_label); | |
9daf2348 JS |
6068 | } |
6069 | } | |
6070 | ||
a13a0209 AT |
6071 | static void |
6072 | put_drop_action(struct ofpbuf *odp_actions, enum xlate_error error) | |
6073 | { | |
6074 | nl_msg_put_u32(odp_actions, OVS_ACTION_ATTR_DROP, error); | |
6075 | } | |
6076 | ||
d787ad39 | 6077 | static void |
2d9b49dd BP |
6078 | put_ct_helper(struct xlate_ctx *ctx, |
6079 | struct ofpbuf *odp_actions, struct ofpact_conntrack *ofc) | |
d787ad39 JS |
6080 | { |
6081 | if (ofc->alg) { | |
40c7b2fc JS |
6082 | switch(ofc->alg) { |
6083 | case IPPORT_FTP: | |
d787ad39 | 6084 | nl_msg_put_string(odp_actions, OVS_CT_ATTR_HELPER, "ftp"); |
40c7b2fc JS |
6085 | break; |
6086 | case IPPORT_TFTP: | |
6087 | nl_msg_put_string(odp_actions, OVS_CT_ATTR_HELPER, "tftp"); | |
6088 | break; | |
6089 | default: | |
2d9b49dd | 6090 | xlate_report_error(ctx, "cannot serialize ct_helper %d", ofc->alg); |
40c7b2fc | 6091 | break; |
d787ad39 JS |
6092 | } |
6093 | } | |
6094 | } | |
6095 | ||
187bb41f YHW |
6096 | static void |
6097 | put_ct_timeout(struct ofpbuf *odp_actions, const struct dpif_backer *backer, | |
6098 | const struct flow *flow, struct flow_wildcards *wc, | |
6099 | uint16_t zone_id) | |
6100 | { | |
6101 | bool unwildcard; | |
6102 | char *tp_name = NULL; | |
6103 | ||
6104 | if (ofproto_dpif_ct_zone_timeout_policy_get_name(backer, zone_id, | |
6105 | ntohs(flow->dl_type), flow->nw_proto, &tp_name, &unwildcard)) { | |
6106 | nl_msg_put_string(odp_actions, OVS_CT_ATTR_TIMEOUT, tp_name); | |
6107 | ||
6108 | if (unwildcard) { | |
6109 | /* The underlying datapath requires separate timeout | |
6110 | * policies for different Ethertypes and IP protocols. We | |
6111 | * don't need to unwildcard 'wc->masks.dl_type' since that | |
6112 | * field is always unwildcarded in megaflows. */ | |
6113 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
6114 | } | |
6115 | } | |
6116 | free(tp_name); | |
6117 | } | |
6118 | ||
9ac0aada JR |
6119 | static void |
6120 | put_ct_nat(struct xlate_ctx *ctx) | |
6121 | { | |
6122 | struct ofpact_nat *ofn = ctx->ct_nat_action; | |
6123 | size_t nat_offset; | |
6124 | ||
6125 | if (!ofn) { | |
6126 | return; | |
6127 | } | |
6128 | ||
6129 | nat_offset = nl_msg_start_nested(ctx->odp_actions, OVS_CT_ATTR_NAT); | |
6130 | if (ofn->flags & NX_NAT_F_SRC || ofn->flags & NX_NAT_F_DST) { | |
6131 | nl_msg_put_flag(ctx->odp_actions, ofn->flags & NX_NAT_F_SRC | |
6132 | ? OVS_NAT_ATTR_SRC : OVS_NAT_ATTR_DST); | |
6133 | if (ofn->flags & NX_NAT_F_PERSISTENT) { | |
6134 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PERSISTENT); | |
6135 | } | |
6136 | if (ofn->flags & NX_NAT_F_PROTO_HASH) { | |
6137 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_HASH); | |
6138 | } else if (ofn->flags & NX_NAT_F_PROTO_RANDOM) { | |
6139 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_RANDOM); | |
6140 | } | |
6141 | if (ofn->range_af == AF_INET) { | |
73e8bc23 | 6142 | nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, |
9ac0aada JR |
6143 | ofn->range.addr.ipv4.min); |
6144 | if (ofn->range.addr.ipv4.max && | |
73e8bc23 BP |
6145 | (ntohl(ofn->range.addr.ipv4.max) |
6146 | > ntohl(ofn->range.addr.ipv4.min))) { | |
6147 | nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, | |
6148 | ofn->range.addr.ipv4.max); | |
9ac0aada JR |
6149 | } |
6150 | } else if (ofn->range_af == AF_INET6) { | |
6151 | nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, | |
6152 | &ofn->range.addr.ipv6.min, | |
6153 | sizeof ofn->range.addr.ipv6.min); | |
6154 | if (!ipv6_mask_is_any(&ofn->range.addr.ipv6.max) && | |
6155 | memcmp(&ofn->range.addr.ipv6.max, &ofn->range.addr.ipv6.min, | |
6156 | sizeof ofn->range.addr.ipv6.max) > 0) { | |
6157 | nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, | |
6158 | &ofn->range.addr.ipv6.max, | |
6159 | sizeof ofn->range.addr.ipv6.max); | |
6160 | } | |
6161 | } | |
6162 | if (ofn->range_af != AF_UNSPEC && ofn->range.proto.min) { | |
6163 | nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MIN, | |
6164 | ofn->range.proto.min); | |
6165 | if (ofn->range.proto.max && | |
6166 | ofn->range.proto.max > ofn->range.proto.min) { | |
6167 | nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MAX, | |
6168 | ofn->range.proto.max); | |
6169 | } | |
6170 | } | |
6171 | } | |
6172 | nl_msg_end_nested(ctx->odp_actions, nat_offset); | |
6173 | } | |
6174 | ||
07659514 | 6175 | static void |
feee58b9 AZ |
6176 | compose_conntrack_action(struct xlate_ctx *ctx, struct ofpact_conntrack *ofc, |
6177 | bool is_last_action) | |
07659514 | 6178 | { |
f2d105b5 | 6179 | ovs_u128 old_ct_label_mask = ctx->wc->masks.ct_label; |
f2d105b5 | 6180 | uint32_t old_ct_mark_mask = ctx->wc->masks.ct_mark; |
07659514 JS |
6181 | size_t ct_offset; |
6182 | uint16_t zone; | |
6183 | ||
6184 | /* Ensure that any prior actions are applied before composing the new | |
6185 | * conntrack action. */ | |
6186 | xlate_commit_actions(ctx); | |
6187 | ||
8e53fe8c | 6188 | /* Process nested actions first, to populate the key. */ |
9ac0aada | 6189 | ctx->ct_nat_action = NULL; |
f2d105b5 | 6190 | ctx->wc->masks.ct_mark = 0; |
f6fabcc6 | 6191 | ctx->wc->masks.ct_label = OVS_U128_ZERO; |
feee58b9 | 6192 | do_xlate_actions(ofc->actions, ofpact_ct_get_action_len(ofc), ctx, |
f5634764 | 6193 | is_last_action, false); |
8e53fe8c | 6194 | |
07659514 JS |
6195 | if (ofc->zone_src.field) { |
6196 | zone = mf_get_subfield(&ofc->zone_src, &ctx->xin->flow); | |
6197 | } else { | |
6198 | zone = ofc->zone_imm; | |
6199 | } | |
6200 | ||
6201 | ct_offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CT); | |
6202 | if (ofc->flags & NX_CT_F_COMMIT) { | |
a76a37ef JR |
6203 | nl_msg_put_flag(ctx->odp_actions, ofc->flags & NX_CT_F_FORCE ? |
6204 | OVS_CT_ATTR_FORCE_COMMIT : OVS_CT_ATTR_COMMIT); | |
adfe7a0b JR |
6205 | if (ctx->xbridge->support.ct_eventmask) { |
6206 | nl_msg_put_u32(ctx->odp_actions, OVS_CT_ATTR_EVENTMASK, | |
975954af | 6207 | OVS_CT_EVENTMASK_DEFAULT); |
adfe7a0b | 6208 | } |
187bb41f YHW |
6209 | if (ctx->xbridge->support.ct_timeout) { |
6210 | put_ct_timeout(ctx->odp_actions, ctx->xbridge->ofproto->backer, | |
6211 | &ctx->xin->flow, ctx->wc, zone); | |
6212 | } | |
07659514 JS |
6213 | } |
6214 | nl_msg_put_u16(ctx->odp_actions, OVS_CT_ATTR_ZONE, zone); | |
f2d105b5 JS |
6215 | put_ct_mark(&ctx->xin->flow, ctx->odp_actions, ctx->wc); |
6216 | put_ct_label(&ctx->xin->flow, ctx->odp_actions, ctx->wc); | |
2d9b49dd | 6217 | put_ct_helper(ctx, ctx->odp_actions, ofc); |
9ac0aada | 6218 | put_ct_nat(ctx); |
07659514 JS |
6219 | nl_msg_end_nested(ctx->odp_actions, ct_offset); |
6220 | ||
f2d105b5 | 6221 | ctx->wc->masks.ct_mark = old_ct_mark_mask; |
f2d105b5 | 6222 | ctx->wc->masks.ct_label = old_ct_label_mask; |
8e53fe8c | 6223 | |
f6fabcc6 | 6224 | if (ofc->recirc_table != NX_CT_RECIRC_NONE) { |
07659514 | 6225 | ctx->conntracked = true; |
5fdd80cc | 6226 | compose_recirculate_and_fork(ctx, ofc->recirc_table, zone); |
07659514 | 6227 | } |
f6fabcc6 | 6228 | |
d072d2de DC |
6229 | ctx->ct_nat_action = NULL; |
6230 | ||
f6fabcc6 JP |
6231 | /* The ct_* fields are only available in the scope of the 'recirc_table' |
6232 | * call chain. */ | |
6233 | flow_clear_conntrack(&ctx->xin->flow); | |
5a046d3b YHW |
6234 | xlate_report(ctx, OFT_DETAIL, "Sets the packet to an untracked state, " |
6235 | "and clears all the conntrack fields."); | |
f6fabcc6 | 6236 | ctx->conntracked = false; |
07659514 JS |
6237 | } |
6238 | ||
1fe178d2 EG |
6239 | static void |
6240 | compose_ct_clear_action(struct xlate_ctx *ctx) | |
6241 | { | |
6242 | clear_conntrack(ctx); | |
6243 | /* This action originally existed without dpif support. So to preserve | |
6244 | * compatibility, only append it if the dpif supports it. */ | |
6245 | if (ctx->xbridge->support.ct_clear) { | |
6246 | nl_msg_put_flag(ctx->odp_actions, OVS_ACTION_ATTR_CT_CLEAR); | |
6247 | } | |
6248 | } | |
6249 | ||
5b34f8fc NS |
6250 | /* check_pkt_larger action checks the packet length and stores the |
6251 | * result in the register bit. We translate this action to the | |
6252 | * datapath action - 'check_pkt_len' whose format | |
6253 | * is: 'check_pkt_len(pkt_len, ge(actions), le(actions))'. | |
6254 | * | |
6255 | * We first set the destination register bit to 1 and call | |
6256 | * 'do_xlate_actions' for the case - packet len greater than | |
6257 | * the specified packet length. | |
6258 | * | |
6259 | * We then set the destination register bit to 0 and call | |
6260 | * 'do_xlate_actions' for the case - packet length is lesser or | |
6261 | * equal to the specified packet length. | |
6262 | * | |
6263 | * It is possible for freezing to happen for both the cases. | |
6264 | */ | |
6265 | static void | |
6266 | xlate_check_pkt_larger(struct xlate_ctx *ctx, | |
6267 | struct ofpact_check_pkt_larger *check_pkt_larger, | |
6268 | const struct ofpact *remaining_acts, | |
6269 | size_t remaining_acts_len) | |
6270 | { | |
6271 | union mf_subvalue value; | |
6272 | memset(&value, 0, sizeof value); | |
6273 | if (!ctx->xbridge->support.check_pkt_len) { | |
6274 | uint8_t is_pkt_larger = 0; | |
6275 | if (ctx->xin->packet) { | |
6276 | is_pkt_larger = | |
6277 | dp_packet_size(ctx->xin->packet) > check_pkt_larger->pkt_len; | |
6278 | } | |
6279 | value.u8_val = is_pkt_larger; | |
6280 | mf_write_subfield_flow(&check_pkt_larger->dst, &value, | |
6281 | &ctx->xin->flow); | |
6282 | /* If datapath doesn't support check_pkt_len action, then set the | |
6283 | * SLOW_ACTION flag. If we don't set SLOW_ACTION, we | |
6284 | * will push a flow to the datapath based on the packet length | |
6285 | * in ctx->xin->packet. For subsequent patches which match the | |
6286 | * same flow, datapath will apply the actions without considering | |
6287 | * the packet length. This results in wrong actions being applied. | |
6288 | */ | |
6289 | ctx->xout->slow |= SLOW_ACTION; | |
6290 | return; | |
6291 | } | |
6292 | ||
6293 | struct ofpbuf old_stack = ctx->stack; | |
6294 | union mf_subvalue new_stack[1024 / sizeof(union mf_subvalue)]; | |
6295 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); | |
6296 | ofpbuf_put(&ctx->stack, old_stack.data, old_stack.size); | |
6297 | ||
6298 | struct ofpbuf old_action_set = ctx->action_set; | |
6299 | uint64_t actset_stub[1024 / 8]; | |
6300 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); | |
6301 | ofpbuf_put(&ctx->action_set, old_action_set.data, old_action_set.size); | |
6302 | ||
6303 | struct flow old_flow = ctx->xin->flow; | |
6304 | xlate_commit_actions(ctx); | |
6305 | struct flow old_base = ctx->base_flow; | |
6306 | bool old_was_mpls = ctx->was_mpls; | |
6307 | bool old_conntracked = ctx->conntracked; | |
6308 | ||
6309 | size_t offset = nl_msg_start_nested(ctx->odp_actions, | |
6310 | OVS_ACTION_ATTR_CHECK_PKT_LEN); | |
6311 | nl_msg_put_u16(ctx->odp_actions, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, | |
6312 | check_pkt_larger->pkt_len); | |
6313 | size_t offset_attr = nl_msg_start_nested( | |
6314 | ctx->odp_actions, OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER); | |
6315 | value.u8_val = 1; | |
6316 | mf_write_subfield_flow(&check_pkt_larger->dst, &value, &ctx->xin->flow); | |
6317 | do_xlate_actions(remaining_acts, remaining_acts_len, ctx, true, false); | |
6318 | if (!ctx->freezing) { | |
6319 | xlate_action_set(ctx); | |
6320 | } | |
6321 | if (ctx->freezing) { | |
6322 | finish_freezing(ctx); | |
6323 | } | |
6324 | nl_msg_end_nested(ctx->odp_actions, offset_attr); | |
6325 | ||
6326 | ctx->base_flow = old_base; | |
6327 | ctx->was_mpls = old_was_mpls; | |
6328 | ctx->conntracked = old_conntracked; | |
6329 | ctx->xin->flow = old_flow; | |
6330 | ||
6331 | /* If the flow translation for the IF_GREATER case requires freezing, | |
6332 | * then ctx->exit would be true. Reset to false so that we can | |
6333 | * do flow translation for 'IF_LESS_EQUAL' case. finish_freezing() | |
6334 | * would have taken care of Undoing the changes done for freeze. */ | |
6335 | ctx->exit = false; | |
6336 | ||
6337 | offset_attr = nl_msg_start_nested( | |
6338 | ctx->odp_actions, OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL); | |
6339 | value.u8_val = 0; | |
6340 | mf_write_subfield_flow(&check_pkt_larger->dst, &value, &ctx->xin->flow); | |
6341 | do_xlate_actions(remaining_acts, remaining_acts_len, ctx, true, false); | |
6342 | if (!ctx->freezing) { | |
6343 | xlate_action_set(ctx); | |
6344 | } | |
6345 | if (ctx->freezing) { | |
6346 | finish_freezing(ctx); | |
6347 | } | |
6348 | nl_msg_end_nested(ctx->odp_actions, offset_attr); | |
6349 | nl_msg_end_nested(ctx->odp_actions, offset); | |
6350 | ||
6351 | ofpbuf_uninit(&ctx->action_set); | |
6352 | ctx->action_set = old_action_set; | |
6353 | ofpbuf_uninit(&ctx->stack); | |
6354 | ctx->stack = old_stack; | |
6355 | ctx->base_flow = old_base; | |
6356 | ctx->was_mpls = old_was_mpls; | |
6357 | ctx->conntracked = old_conntracked; | |
6358 | ctx->xin->flow = old_flow; | |
6359 | ctx->exit = true; | |
6360 | } | |
6361 | ||
f839892a JS |
6362 | static void |
6363 | rewrite_flow_encap_ethernet(struct xlate_ctx *ctx, | |
6364 | struct flow *flow, | |
6365 | struct flow_wildcards *wc) | |
6366 | { | |
6367 | wc->masks.packet_type = OVS_BE32_MAX; | |
6368 | if (pt_ns(flow->packet_type) == OFPHTN_ETHERTYPE) { | |
6369 | /* Only adjust the packet_type and zero the dummy Ethernet addresses. */ | |
6370 | ovs_be16 ethertype = pt_ns_type_be(flow->packet_type); | |
6371 | flow->packet_type = htonl(PT_ETH); | |
6372 | flow->dl_src = eth_addr_zero; | |
6373 | flow->dl_dst = eth_addr_zero; | |
6374 | flow->dl_type = ethertype; | |
6375 | } else { | |
1fc11c59 | 6376 | /* Error handling: drop packet. */ |
f839892a | 6377 | xlate_report_debug(ctx, OFT_ACTION, |
1fc11c59 JS |
6378 | "Dropping packet as encap(ethernet) is not " |
6379 | "supported for packet type ethernet."); | |
7873e106 | 6380 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
f839892a JS |
6381 | } |
6382 | } | |
6383 | ||
1fc11c59 JS |
6384 | /* For an MD2 NSH header returns a pointer to an ofpbuf with the encoded |
6385 | * MD2 TLVs provided as encap properties to the encap operation. This | |
f59cb331 | 6386 | * will be stored as encap_data in the ctx and copied into the push_nsh |
1fc11c59 JS |
6387 | * action at the next commit. */ |
6388 | static struct ofpbuf * | |
f59cb331 YY |
6389 | rewrite_flow_push_nsh(struct xlate_ctx *ctx, |
6390 | const struct ofpact_encap *encap, | |
6391 | struct flow *flow, | |
6392 | struct flow_wildcards *wc) | |
1fc11c59 JS |
6393 | { |
6394 | ovs_be32 packet_type = flow->packet_type; | |
6395 | const char *ptr = (char *) encap->props; | |
f59cb331 | 6396 | struct ofpbuf *buf = ofpbuf_new(NSH_CTX_HDRS_MAX_LEN); |
1fc11c59 JS |
6397 | uint8_t md_type = NSH_M_TYPE1; |
6398 | uint8_t np = 0; | |
6399 | int i; | |
6400 | ||
6401 | /* Scan the optional NSH encap TLV properties, if any. */ | |
6402 | for (i = 0; i < encap->n_props; i++) { | |
6403 | struct ofpact_ed_prop *prop_ptr = | |
6404 | ALIGNED_CAST(struct ofpact_ed_prop *, ptr); | |
6405 | if (prop_ptr->prop_class == OFPPPC_NSH) { | |
6406 | switch (prop_ptr->type) { | |
6407 | case OFPPPT_PROP_NSH_MDTYPE: { | |
6408 | struct ofpact_ed_prop_nsh_md_type *prop_md_type = | |
6409 | ALIGNED_CAST(struct ofpact_ed_prop_nsh_md_type *, | |
6410 | prop_ptr); | |
6411 | md_type = prop_md_type->md_type; | |
6412 | break; | |
6413 | } | |
6414 | case OFPPPT_PROP_NSH_TLV: { | |
6415 | struct ofpact_ed_prop_nsh_tlv *tlv_prop = | |
6416 | ALIGNED_CAST(struct ofpact_ed_prop_nsh_tlv *, | |
6417 | prop_ptr); | |
6418 | struct nsh_md2_tlv *md2_ctx = | |
6419 | ofpbuf_put_uninit(buf, sizeof(*md2_ctx)); | |
6420 | md2_ctx->md_class = tlv_prop->tlv_class; | |
6421 | md2_ctx->type = tlv_prop->tlv_type; | |
6422 | md2_ctx->length = tlv_prop->tlv_len; | |
6423 | size_t len = ROUND_UP(md2_ctx->length, 4); | |
6424 | size_t padding = len - md2_ctx->length; | |
6425 | ofpbuf_put(buf, tlv_prop->data, md2_ctx->length); | |
6426 | ofpbuf_put_zeros(buf, padding); | |
6427 | break; | |
6428 | } | |
6429 | default: | |
6430 | /* No other NSH encap properties defined yet. */ | |
6431 | break; | |
6432 | } | |
6433 | } | |
6434 | ptr += ROUND_UP(prop_ptr->len, 8); | |
6435 | } | |
f59cb331 | 6436 | if (buf->size == 0 || buf->size > NSH_CTX_HDRS_MAX_LEN) { |
1fc11c59 JS |
6437 | ofpbuf_delete(buf); |
6438 | buf = NULL; | |
6439 | } | |
6440 | ||
6441 | /* Determine the Next Protocol field for NSH header. */ | |
6442 | switch (ntohl(packet_type)) { | |
6443 | case PT_ETH: | |
6444 | np = NSH_P_ETHERNET; | |
6445 | break; | |
6446 | case PT_IPV4: | |
6447 | np = NSH_P_IPV4; | |
6448 | break; | |
6449 | case PT_IPV6: | |
6450 | np = NSH_P_IPV6; | |
6451 | break; | |
6452 | case PT_NSH: | |
6453 | np = NSH_P_NSH; | |
6454 | break; | |
6455 | default: | |
6456 | /* Error handling: drop packet. */ | |
6457 | xlate_report_debug(ctx, OFT_ACTION, | |
6458 | "Dropping packet as encap(nsh) is not " | |
6459 | "supported for packet type (%d,0x%x)", | |
6460 | pt_ns(packet_type), pt_ns_type(packet_type)); | |
7873e106 | 6461 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
1fc11c59 JS |
6462 | return buf; |
6463 | } | |
6464 | /* Note that we have matched on packet_type! */ | |
6465 | wc->masks.packet_type = OVS_BE32_MAX; | |
6466 | ||
6467 | /* Reset all current flow packet headers. */ | |
6468 | memset(&flow->dl_dst, 0, | |
6469 | sizeof(struct flow) - offsetof(struct flow, dl_dst)); | |
6470 | ||
6471 | /* Populate the flow with the new NSH header. */ | |
6472 | flow->packet_type = htonl(PT_NSH); | |
6473 | flow->dl_type = htons(ETH_TYPE_NSH); | |
17553f27 YY |
6474 | flow->nsh.flags = 0; |
6475 | flow->nsh.ttl = 63; | |
1fc11c59 | 6476 | flow->nsh.np = np; |
17553f27 | 6477 | flow->nsh.path_hdr = htonl(255); |
1fc11c59 JS |
6478 | |
6479 | if (md_type == NSH_M_TYPE1) { | |
6480 | flow->nsh.mdtype = NSH_M_TYPE1; | |
f59cb331 | 6481 | memset(flow->nsh.context, 0, sizeof flow->nsh.context); |
1fc11c59 JS |
6482 | if (buf) { |
6483 | /* Drop any MD2 context TLVs. */ | |
6484 | ofpbuf_delete(buf); | |
6485 | buf = NULL; | |
6486 | } | |
6487 | } else if (md_type == NSH_M_TYPE2) { | |
6488 | flow->nsh.mdtype = NSH_M_TYPE2; | |
6489 | } | |
17553f27 | 6490 | flow->nsh.mdtype &= NSH_MDTYPE_MASK; |
1fc11c59 JS |
6491 | |
6492 | return buf; | |
6493 | } | |
6494 | ||
f839892a JS |
6495 | static void |
6496 | xlate_generic_encap_action(struct xlate_ctx *ctx, | |
6497 | const struct ofpact_encap *encap) | |
6498 | { | |
6499 | struct flow *flow = &ctx->xin->flow; | |
6500 | struct flow_wildcards *wc = ctx->wc; | |
1fc11c59 | 6501 | struct ofpbuf *encap_data = NULL; |
f839892a JS |
6502 | |
6503 | /* Ensure that any pending actions on the inner packet are applied before | |
6504 | * rewriting the flow */ | |
6505 | xlate_commit_actions(ctx); | |
6506 | ||
6507 | /* Rewrite the flow to reflect the effect of pushing the new encap header. */ | |
6508 | switch (ntohl(encap->new_pkt_type)) { | |
6509 | case PT_ETH: | |
6510 | rewrite_flow_encap_ethernet(ctx, flow, wc); | |
6511 | break; | |
1fc11c59 | 6512 | case PT_NSH: |
f59cb331 | 6513 | encap_data = rewrite_flow_push_nsh(ctx, encap, flow, wc); |
1fc11c59 | 6514 | break; |
f839892a | 6515 | default: |
1fc11c59 JS |
6516 | /* New packet type was checked during decoding. */ |
6517 | OVS_NOT_REACHED(); | |
f839892a JS |
6518 | } |
6519 | ||
6520 | if (!ctx->error) { | |
6521 | /* The actual encap datapath action will be generated at next commit. */ | |
6522 | ctx->pending_encap = true; | |
1fc11c59 | 6523 | ctx->encap_data = encap_data; |
f839892a JS |
6524 | } |
6525 | } | |
6526 | ||
6527 | /* Returns true if packet must be recirculated after decapsulation. */ | |
6528 | static bool | |
6529 | xlate_generic_decap_action(struct xlate_ctx *ctx, | |
6530 | const struct ofpact_decap *decap OVS_UNUSED) | |
6531 | { | |
6532 | struct flow *flow = &ctx->xin->flow; | |
6533 | ||
6534 | /* Ensure that any pending actions on the current packet are applied | |
6535 | * before generating the decap action. */ | |
6536 | xlate_commit_actions(ctx); | |
6537 | ||
6538 | /* We assume for now that the new_pkt_type is PT_USE_NEXT_PROTO. */ | |
6539 | switch (ntohl(flow->packet_type)) { | |
6540 | case PT_ETH: | |
6541 | if (flow->vlans[0].tci & htons(VLAN_CFI)) { | |
6542 | /* Error handling: drop packet. */ | |
6543 | xlate_report_debug(ctx, OFT_ACTION, "Dropping packet, cannot " | |
6544 | "decap Ethernet if VLAN is present."); | |
7873e106 | 6545 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
f839892a JS |
6546 | } else { |
6547 | /* Just change the packet_type. | |
6548 | * Delay generating pop_eth to the next commit. */ | |
6549 | flow->packet_type = htonl(PACKET_TYPE(OFPHTN_ETHERTYPE, | |
6550 | ntohs(flow->dl_type))); | |
6551 | ctx->wc->masks.dl_type = OVS_BE16_MAX; | |
6552 | } | |
6553 | return false; | |
1fc11c59 | 6554 | case PT_NSH: |
f59cb331 | 6555 | /* The pop_nsh action is generated at the commit executed as |
1fc11c59 JS |
6556 | * part of freezing the ctx for recirculation. Here we just set |
6557 | * the new packet type based on the NSH next protocol field. */ | |
6558 | switch (flow->nsh.np) { | |
6559 | case NSH_P_ETHERNET: | |
6560 | flow->packet_type = htonl(PT_ETH); | |
6561 | break; | |
6562 | case NSH_P_IPV4: | |
6563 | flow->packet_type = htonl(PT_IPV4); | |
6564 | break; | |
6565 | case NSH_P_IPV6: | |
6566 | flow->packet_type = htonl(PT_IPV6); | |
6567 | break; | |
6568 | case NSH_P_NSH: | |
6569 | flow->packet_type = htonl(PT_NSH); | |
6570 | break; | |
6571 | default: | |
6572 | /* Error handling: drop packet. */ | |
6573 | xlate_report_debug(ctx, OFT_ACTION, | |
6574 | "Dropping packet as NSH next protocol %d " | |
6575 | "is not supported", flow->nsh.np); | |
7873e106 | 6576 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
1fc11c59 JS |
6577 | return false; |
6578 | break; | |
6579 | } | |
6580 | ctx->wc->masks.nsh.np = UINT8_MAX; | |
88ec1e0a | 6581 | ctx->pending_decap = true; |
1fc11c59 JS |
6582 | /* Trigger recirculation. */ |
6583 | return true; | |
f839892a | 6584 | default: |
1fc11c59 JS |
6585 | /* Error handling: drop packet. */ |
6586 | xlate_report_debug( | |
6587 | ctx, OFT_ACTION, | |
6588 | "Dropping packet as the decap() does not support " | |
6589 | "packet type (%d,0x%x)", | |
6590 | pt_ns(flow->packet_type), pt_ns_type(flow->packet_type)); | |
7873e106 | 6591 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
f839892a JS |
6592 | return false; |
6593 | } | |
6594 | } | |
6595 | ||
e12ec36b SH |
6596 | static void |
6597 | recirc_for_mpls(const struct ofpact *a, struct xlate_ctx *ctx) | |
6598 | { | |
6599 | /* No need to recirculate if already exiting. */ | |
6600 | if (ctx->exit) { | |
6601 | return; | |
6602 | } | |
6603 | ||
6604 | /* Do not consider recirculating unless the packet was previously MPLS. */ | |
6605 | if (!ctx->was_mpls) { | |
6606 | return; | |
6607 | } | |
6608 | ||
6609 | /* Special case these actions, only recirculating if necessary. | |
6610 | * This avoids the overhead of recirculation in common use-cases. | |
6611 | */ | |
6612 | switch (a->type) { | |
6613 | ||
6614 | /* Output actions do not require recirculation. */ | |
6615 | case OFPACT_OUTPUT: | |
aaca4fe0 | 6616 | case OFPACT_OUTPUT_TRUNC: |
e12ec36b SH |
6617 | case OFPACT_ENQUEUE: |
6618 | case OFPACT_OUTPUT_REG: | |
6619 | /* Set actions that don't touch L3+ fields do not require recirculation. */ | |
6620 | case OFPACT_SET_VLAN_VID: | |
6621 | case OFPACT_SET_VLAN_PCP: | |
6622 | case OFPACT_SET_ETH_SRC: | |
6623 | case OFPACT_SET_ETH_DST: | |
6624 | case OFPACT_SET_TUNNEL: | |
6625 | case OFPACT_SET_QUEUE: | |
6626 | /* If actions of a group require recirculation that can be detected | |
6627 | * when translating them. */ | |
6628 | case OFPACT_GROUP: | |
6629 | return; | |
6630 | ||
6631 | /* Set field that don't touch L3+ fields don't require recirculation. */ | |
6632 | case OFPACT_SET_FIELD: | |
6633 | if (mf_is_l3_or_higher(ofpact_get_SET_FIELD(a)->field)) { | |
6634 | break; | |
6635 | } | |
6636 | return; | |
6637 | ||
6638 | /* For simplicity, recirculate in all other cases. */ | |
6639 | case OFPACT_CONTROLLER: | |
6640 | case OFPACT_BUNDLE: | |
6641 | case OFPACT_STRIP_VLAN: | |
6642 | case OFPACT_PUSH_VLAN: | |
6643 | case OFPACT_SET_IPV4_SRC: | |
6644 | case OFPACT_SET_IPV4_DST: | |
6645 | case OFPACT_SET_IP_DSCP: | |
6646 | case OFPACT_SET_IP_ECN: | |
6647 | case OFPACT_SET_IP_TTL: | |
6648 | case OFPACT_SET_L4_SRC_PORT: | |
6649 | case OFPACT_SET_L4_DST_PORT: | |
6650 | case OFPACT_REG_MOVE: | |
6651 | case OFPACT_STACK_PUSH: | |
6652 | case OFPACT_STACK_POP: | |
6653 | case OFPACT_DEC_TTL: | |
6654 | case OFPACT_SET_MPLS_LABEL: | |
6655 | case OFPACT_SET_MPLS_TC: | |
6656 | case OFPACT_SET_MPLS_TTL: | |
6657 | case OFPACT_DEC_MPLS_TTL: | |
6658 | case OFPACT_PUSH_MPLS: | |
6659 | case OFPACT_POP_MPLS: | |
6660 | case OFPACT_POP_QUEUE: | |
6661 | case OFPACT_FIN_TIMEOUT: | |
6662 | case OFPACT_RESUBMIT: | |
6663 | case OFPACT_LEARN: | |
6664 | case OFPACT_CONJUNCTION: | |
6665 | case OFPACT_MULTIPATH: | |
6666 | case OFPACT_NOTE: | |
6667 | case OFPACT_EXIT: | |
6668 | case OFPACT_SAMPLE: | |
7ae62a67 | 6669 | case OFPACT_CLONE: |
f839892a JS |
6670 | case OFPACT_ENCAP: |
6671 | case OFPACT_DECAP: | |
491e05c2 | 6672 | case OFPACT_DEC_NSH_TTL: |
e12ec36b SH |
6673 | case OFPACT_UNROLL_XLATE: |
6674 | case OFPACT_CT: | |
72fe7578 | 6675 | case OFPACT_CT_CLEAR: |
e12ec36b SH |
6676 | case OFPACT_NAT: |
6677 | case OFPACT_DEBUG_RECIRC: | |
a934a3dd | 6678 | case OFPACT_DEBUG_SLOW: |
e12ec36b SH |
6679 | case OFPACT_METER: |
6680 | case OFPACT_CLEAR_ACTIONS: | |
6681 | case OFPACT_WRITE_ACTIONS: | |
6682 | case OFPACT_WRITE_METADATA: | |
6683 | case OFPACT_GOTO_TABLE: | |
5b34f8fc | 6684 | case OFPACT_CHECK_PKT_LARGER: |
81f71381 | 6685 | case OFPACT_DELETE_FIELD: |
e12ec36b SH |
6686 | default: |
6687 | break; | |
6688 | } | |
6689 | ||
6690 | /* Recirculate */ | |
6691 | ctx_trigger_freeze(ctx); | |
6692 | } | |
6693 | ||
2d9b49dd BP |
6694 | static void |
6695 | xlate_ofpact_reg_move(struct xlate_ctx *ctx, const struct ofpact_reg_move *a) | |
6696 | { | |
6697 | mf_subfield_copy(&a->src, &a->dst, &ctx->xin->flow, ctx->wc); | |
6698 | xlate_report_subfield(ctx, &a->dst); | |
6699 | } | |
6700 | ||
6701 | static void | |
6702 | xlate_ofpact_stack_pop(struct xlate_ctx *ctx, const struct ofpact_stack *a) | |
6703 | { | |
6704 | if (nxm_execute_stack_pop(a, &ctx->xin->flow, ctx->wc, &ctx->stack)) { | |
6705 | xlate_report_subfield(ctx, &a->subfield); | |
6706 | } else { | |
6707 | xlate_report_error(ctx, "stack underflow"); | |
6708 | } | |
6709 | } | |
6710 | ||
6711 | /* Restore translation context data that was stored earlier. */ | |
6712 | static void | |
6713 | xlate_ofpact_unroll_xlate(struct xlate_ctx *ctx, | |
6714 | const struct ofpact_unroll_xlate *a) | |
6715 | { | |
6716 | ctx->table_id = a->rule_table_id; | |
6717 | ctx->rule_cookie = a->rule_cookie; | |
6718 | xlate_report(ctx, OFT_THAW, "restored state: table=%"PRIu8", " | |
6719 | "cookie=%#"PRIx64, a->rule_table_id, a->rule_cookie); | |
6720 | } | |
6721 | ||
9583bc14 EJ |
6722 | static void |
6723 | do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
f5634764 KG |
6724 | struct xlate_ctx *ctx, bool is_last_action, |
6725 | bool group_bucket_action) | |
9583bc14 | 6726 | { |
49a73e0c | 6727 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 6728 | struct flow *flow = &ctx->xin->flow; |
9583bc14 EJ |
6729 | const struct ofpact *a; |
6730 | ||
f47ea021 JR |
6731 | /* dl_type already in the mask, not set below. */ |
6732 | ||
2d9b49dd BP |
6733 | if (!ofpacts_len) { |
6734 | xlate_report(ctx, OFT_ACTION, "drop"); | |
6735 | return; | |
6736 | } | |
6737 | ||
9583bc14 EJ |
6738 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
6739 | struct ofpact_controller *controller; | |
6740 | const struct ofpact_metadata *metadata; | |
b2dd70be JR |
6741 | const struct ofpact_set_field *set_field; |
6742 | const struct mf_field *mf; | |
feee58b9 AZ |
6743 | bool last = is_last_action && ofpact_last(a, ofpacts, ofpacts_len) |
6744 | && ctx->action_set.size; | |
9583bc14 | 6745 | |
fff1b9c0 JR |
6746 | if (ctx->error) { |
6747 | break; | |
6748 | } | |
6749 | ||
e12ec36b SH |
6750 | recirc_for_mpls(a, ctx); |
6751 | ||
e672ff9b JR |
6752 | if (ctx->exit) { |
6753 | /* Check if need to store the remaining actions for later | |
6754 | * execution. */ | |
1d361a81 BP |
6755 | if (ctx->freezing) { |
6756 | freeze_unroll_actions(a, ofpact_end(ofpacts, ofpacts_len), | |
e672ff9b JR |
6757 | ctx); |
6758 | } | |
6759 | break; | |
7bbdd84f SH |
6760 | } |
6761 | ||
2d9b49dd BP |
6762 | if (OVS_UNLIKELY(ctx->xin->trace)) { |
6763 | struct ds s = DS_EMPTY_INITIALIZER; | |
efefbcae BP |
6764 | struct ofpact_format_params fp = { .s = &s }; |
6765 | ofpacts_format(a, OFPACT_ALIGN(a->len), &fp); | |
2d9b49dd BP |
6766 | xlate_report(ctx, OFT_ACTION, "%s", ds_cstr(&s)); |
6767 | ds_destroy(&s); | |
6768 | } | |
6769 | ||
9583bc14 EJ |
6770 | switch (a->type) { |
6771 | case OFPACT_OUTPUT: | |
6772 | xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port, | |
8b496c72 | 6773 | ofpact_get_OUTPUT(a)->max_len, true, last, |
f5634764 | 6774 | false, group_bucket_action); |
9583bc14 EJ |
6775 | break; |
6776 | ||
7395c052 | 6777 | case OFPACT_GROUP: |
feee58b9 | 6778 | if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id, last)) { |
1d741d6d | 6779 | /* Group could not be found. */ |
db88b35c JR |
6780 | |
6781 | /* XXX: Terminates action list translation, but does not | |
6782 | * terminate the pipeline. */ | |
f4fb341b SH |
6783 | return; |
6784 | } | |
7395c052 NZ |
6785 | break; |
6786 | ||
9583bc14 EJ |
6787 | case OFPACT_CONTROLLER: |
6788 | controller = ofpact_get_CONTROLLER(a); | |
77ab5fd2 BP |
6789 | if (controller->pause) { |
6790 | ctx->pause = controller; | |
77ab5fd2 BP |
6791 | ctx_trigger_freeze(ctx); |
6792 | a = ofpact_next(a); | |
6793 | } else { | |
d39ec23d JP |
6794 | xlate_controller_action(ctx, controller->max_len, |
6795 | controller->reason, | |
6796 | controller->controller_id, | |
206ddb9a | 6797 | controller->provider_meter_id, |
d39ec23d JP |
6798 | controller->userdata, |
6799 | controller->userdata_len); | |
77ab5fd2 | 6800 | } |
9583bc14 EJ |
6801 | break; |
6802 | ||
6803 | case OFPACT_ENQUEUE: | |
16194afd DDP |
6804 | memset(&wc->masks.skb_priority, 0xff, |
6805 | sizeof wc->masks.skb_priority); | |
f5634764 KG |
6806 | xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a), last, |
6807 | group_bucket_action); | |
9583bc14 EJ |
6808 | break; |
6809 | ||
6810 | case OFPACT_SET_VLAN_VID: | |
f0fb825a EG |
6811 | wc->masks.vlans[0].tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
6812 | if (flow->vlans[0].tci & htons(VLAN_CFI) || | |
ca287d20 | 6813 | ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) { |
f0fb825a EG |
6814 | if (!flow->vlans[0].tpid) { |
6815 | flow->vlans[0].tpid = htons(ETH_TYPE_VLAN); | |
6816 | } | |
6817 | flow->vlans[0].tci &= ~htons(VLAN_VID_MASK); | |
6818 | flow->vlans[0].tci |= | |
6819 | (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid) | | |
6820 | htons(VLAN_CFI)); | |
ca287d20 | 6821 | } |
9583bc14 EJ |
6822 | break; |
6823 | ||
6824 | case OFPACT_SET_VLAN_PCP: | |
f0fb825a EG |
6825 | wc->masks.vlans[0].tci |= htons(VLAN_PCP_MASK | VLAN_CFI); |
6826 | if (flow->vlans[0].tci & htons(VLAN_CFI) || | |
ca287d20 | 6827 | ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) { |
f0fb825a EG |
6828 | if (!flow->vlans[0].tpid) { |
6829 | flow->vlans[0].tpid = htons(ETH_TYPE_VLAN); | |
6830 | } | |
6831 | flow->vlans[0].tci &= ~htons(VLAN_PCP_MASK); | |
6832 | flow->vlans[0].tci |= | |
6833 | htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp | |
6834 | << VLAN_PCP_SHIFT) | VLAN_CFI); | |
ca287d20 | 6835 | } |
9583bc14 EJ |
6836 | break; |
6837 | ||
6838 | case OFPACT_STRIP_VLAN: | |
f0fb825a | 6839 | flow_pop_vlan(flow, wc); |
9583bc14 EJ |
6840 | break; |
6841 | ||
6842 | case OFPACT_PUSH_VLAN: | |
f0fb825a EG |
6843 | flow_push_vlan_uninit(flow, wc); |
6844 | flow->vlans[0].tpid = ofpact_get_PUSH_VLAN(a)->ethertype; | |
6845 | flow->vlans[0].tci = htons(VLAN_CFI); | |
9583bc14 EJ |
6846 | break; |
6847 | ||
6848 | case OFPACT_SET_ETH_SRC: | |
74ff3298 JR |
6849 | WC_MASK_FIELD(wc, dl_src); |
6850 | flow->dl_src = ofpact_get_SET_ETH_SRC(a)->mac; | |
9583bc14 EJ |
6851 | break; |
6852 | ||
6853 | case OFPACT_SET_ETH_DST: | |
74ff3298 JR |
6854 | WC_MASK_FIELD(wc, dl_dst); |
6855 | flow->dl_dst = ofpact_get_SET_ETH_DST(a)->mac; | |
9583bc14 EJ |
6856 | break; |
6857 | ||
6858 | case OFPACT_SET_IPV4_SRC: | |
33bf9176 | 6859 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 6860 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); |
33bf9176 | 6861 | flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4; |
9583bc14 EJ |
6862 | } |
6863 | break; | |
6864 | ||
6865 | case OFPACT_SET_IPV4_DST: | |
33bf9176 | 6866 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 6867 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); |
33bf9176 | 6868 | flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4; |
9583bc14 EJ |
6869 | } |
6870 | break; | |
6871 | ||
04f01c24 BP |
6872 | case OFPACT_SET_IP_DSCP: |
6873 | if (is_ip_any(flow)) { | |
f47ea021 | 6874 | wc->masks.nw_tos |= IP_DSCP_MASK; |
33bf9176 | 6875 | flow->nw_tos &= ~IP_DSCP_MASK; |
04f01c24 | 6876 | flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp; |
9583bc14 EJ |
6877 | } |
6878 | break; | |
6879 | ||
ff14eb7a JR |
6880 | case OFPACT_SET_IP_ECN: |
6881 | if (is_ip_any(flow)) { | |
6882 | wc->masks.nw_tos |= IP_ECN_MASK; | |
6883 | flow->nw_tos &= ~IP_ECN_MASK; | |
6884 | flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn; | |
6885 | } | |
6886 | break; | |
6887 | ||
0c20dbe4 JR |
6888 | case OFPACT_SET_IP_TTL: |
6889 | if (is_ip_any(flow)) { | |
6890 | wc->masks.nw_ttl = 0xff; | |
6891 | flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl; | |
6892 | } | |
6893 | break; | |
6894 | ||
9583bc14 | 6895 | case OFPACT_SET_L4_SRC_PORT: |
b8778a0d | 6896 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
6897 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
6898 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
33bf9176 | 6899 | flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port); |
9583bc14 EJ |
6900 | } |
6901 | break; | |
6902 | ||
6903 | case OFPACT_SET_L4_DST_PORT: | |
b8778a0d | 6904 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
6905 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
6906 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
33bf9176 | 6907 | flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port); |
9583bc14 EJ |
6908 | } |
6909 | break; | |
6910 | ||
6911 | case OFPACT_RESUBMIT: | |
8bf009bf JR |
6912 | /* Freezing complicates resubmit. Some action in the flow |
6913 | * entry found by resubmit might trigger freezing. If that | |
6914 | * happens, then we do not want to execute the resubmit again after | |
6915 | * during thawing, so we want to skip back to the head of the loop | |
6916 | * to avoid that, only adding any actions that follow the resubmit | |
6917 | * to the frozen actions. | |
6b1c5734 | 6918 | */ |
feee58b9 | 6919 | xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a), last); |
6b1c5734 | 6920 | continue; |
9583bc14 EJ |
6921 | |
6922 | case OFPACT_SET_TUNNEL: | |
33bf9176 | 6923 | flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id); |
9583bc14 EJ |
6924 | break; |
6925 | ||
6926 | case OFPACT_SET_QUEUE: | |
16194afd DDP |
6927 | memset(&wc->masks.skb_priority, 0xff, |
6928 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
6929 | xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id); |
6930 | break; | |
6931 | ||
6932 | case OFPACT_POP_QUEUE: | |
16194afd DDP |
6933 | memset(&wc->masks.skb_priority, 0xff, |
6934 | sizeof wc->masks.skb_priority); | |
2d9b49dd BP |
6935 | if (flow->skb_priority != ctx->orig_skb_priority) { |
6936 | flow->skb_priority = ctx->orig_skb_priority; | |
6937 | xlate_report(ctx, OFT_DETAIL, "queue = %#"PRIx32, | |
6938 | flow->skb_priority); | |
6939 | } | |
9583bc14 EJ |
6940 | break; |
6941 | ||
6942 | case OFPACT_REG_MOVE: | |
2d9b49dd | 6943 | xlate_ofpact_reg_move(ctx, ofpact_get_REG_MOVE(a)); |
9583bc14 EJ |
6944 | break; |
6945 | ||
b2dd70be JR |
6946 | case OFPACT_SET_FIELD: |
6947 | set_field = ofpact_get_SET_FIELD(a); | |
6948 | mf = set_field->field; | |
b2dd70be | 6949 | |
aff49b8c JR |
6950 | /* Set the field only if the packet actually has it. */ |
6951 | if (mf_are_prereqs_ok(mf, flow, wc)) { | |
128684a6 JR |
6952 | mf_mask_field_masked(mf, ofpact_set_field_mask(set_field), wc); |
6953 | mf_set_flow_value_masked(mf, set_field->value, | |
6954 | ofpact_set_field_mask(set_field), | |
6955 | flow); | |
2d9b49dd BP |
6956 | } else { |
6957 | xlate_report(ctx, OFT_WARN, | |
6958 | "unmet prerequisites for %s, set_field ignored", | |
6959 | mf->name); | |
6960 | ||
b8778a0d | 6961 | } |
b2dd70be JR |
6962 | break; |
6963 | ||
9583bc14 | 6964 | case OFPACT_STACK_PUSH: |
33bf9176 BP |
6965 | nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc, |
6966 | &ctx->stack); | |
9583bc14 EJ |
6967 | break; |
6968 | ||
6969 | case OFPACT_STACK_POP: | |
2d9b49dd | 6970 | xlate_ofpact_stack_pop(ctx, ofpact_get_STACK_POP(a)); |
9583bc14 EJ |
6971 | break; |
6972 | ||
6973 | case OFPACT_PUSH_MPLS: | |
8bfd0fda | 6974 | compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)); |
9583bc14 EJ |
6975 | break; |
6976 | ||
6977 | case OFPACT_POP_MPLS: | |
8bfd0fda | 6978 | compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype); |
9583bc14 EJ |
6979 | break; |
6980 | ||
097d4939 | 6981 | case OFPACT_SET_MPLS_LABEL: |
8bfd0fda BP |
6982 | compose_set_mpls_label_action( |
6983 | ctx, ofpact_get_SET_MPLS_LABEL(a)->label); | |
1d741d6d | 6984 | break; |
097d4939 JR |
6985 | |
6986 | case OFPACT_SET_MPLS_TC: | |
8bfd0fda | 6987 | compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc); |
097d4939 JR |
6988 | break; |
6989 | ||
9583bc14 | 6990 | case OFPACT_SET_MPLS_TTL: |
8bfd0fda | 6991 | compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl); |
9583bc14 EJ |
6992 | break; |
6993 | ||
6994 | case OFPACT_DEC_MPLS_TTL: | |
9cfef3d0 | 6995 | if (compose_dec_mpls_ttl_action(ctx)) { |
ad3efdcb | 6996 | return; |
9583bc14 EJ |
6997 | } |
6998 | break; | |
6999 | ||
491e05c2 YY |
7000 | case OFPACT_DEC_NSH_TTL: |
7001 | if (compose_dec_nsh_ttl_action(ctx)) { | |
7002 | return; | |
7003 | } | |
7004 | break; | |
7005 | ||
9583bc14 | 7006 | case OFPACT_DEC_TTL: |
f74e7df7 | 7007 | wc->masks.nw_ttl = 0xff; |
9583bc14 | 7008 | if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) { |
ad3efdcb | 7009 | return; |
9583bc14 EJ |
7010 | } |
7011 | break; | |
7012 | ||
7013 | case OFPACT_NOTE: | |
7014 | /* Nothing to do. */ | |
7015 | break; | |
7016 | ||
7017 | case OFPACT_MULTIPATH: | |
33bf9176 | 7018 | multipath_execute(ofpact_get_MULTIPATH(a), flow, wc); |
2d9b49dd | 7019 | xlate_report_subfield(ctx, &ofpact_get_MULTIPATH(a)->dst); |
9583bc14 EJ |
7020 | break; |
7021 | ||
7022 | case OFPACT_BUNDLE: | |
f5634764 KG |
7023 | xlate_bundle_action(ctx, ofpact_get_BUNDLE(a), last, |
7024 | group_bucket_action); | |
9583bc14 EJ |
7025 | break; |
7026 | ||
7027 | case OFPACT_OUTPUT_REG: | |
f5634764 KG |
7028 | xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a), last, |
7029 | group_bucket_action); | |
9583bc14 EJ |
7030 | break; |
7031 | ||
aaca4fe0 WT |
7032 | case OFPACT_OUTPUT_TRUNC: |
7033 | xlate_output_trunc_action(ctx, ofpact_get_OUTPUT_TRUNC(a)->port, | |
f5634764 KG |
7034 | ofpact_get_OUTPUT_TRUNC(a)->max_len, last, |
7035 | group_bucket_action); | |
aaca4fe0 WT |
7036 | break; |
7037 | ||
9583bc14 EJ |
7038 | case OFPACT_LEARN: |
7039 | xlate_learn_action(ctx, ofpact_get_LEARN(a)); | |
7040 | break; | |
7041 | ||
2d9b49dd | 7042 | case OFPACT_CONJUNCTION: |
afc3987b BP |
7043 | /* A flow with a "conjunction" action represents part of a special |
7044 | * kind of "set membership match". Such a flow should not actually | |
7045 | * get executed, but it could via, say, a "packet-out", even though | |
7046 | * that wouldn't be useful. Log it to help debugging. */ | |
2d9b49dd | 7047 | xlate_report_error(ctx, "executing no-op conjunction action"); |
18080541 BP |
7048 | break; |
7049 | ||
9583bc14 EJ |
7050 | case OFPACT_EXIT: |
7051 | ctx->exit = true; | |
7052 | break; | |
7053 | ||
2d9b49dd BP |
7054 | case OFPACT_UNROLL_XLATE: |
7055 | xlate_ofpact_unroll_xlate(ctx, ofpact_get_UNROLL_XLATE(a)); | |
e672ff9b | 7056 | break; |
2d9b49dd | 7057 | |
9583bc14 | 7058 | case OFPACT_FIN_TIMEOUT: |
33bf9176 | 7059 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
9583bc14 EJ |
7060 | xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a)); |
7061 | break; | |
7062 | ||
81f71381 YHW |
7063 | case OFPACT_DELETE_FIELD: |
7064 | xlate_delete_field(ctx, flow, ofpact_get_DELETE_FIELD(a)); | |
7065 | break; | |
7066 | ||
9583bc14 | 7067 | case OFPACT_CLEAR_ACTIONS: |
2d9b49dd | 7068 | xlate_report_action_set(ctx, "was"); |
7fdb60a7 | 7069 | ofpbuf_clear(&ctx->action_set); |
c61f3870 BP |
7070 | ctx->xin->flow.actset_output = OFPP_UNSET; |
7071 | ctx->action_set_has_group = false; | |
7fdb60a7 SH |
7072 | break; |
7073 | ||
7074 | case OFPACT_WRITE_ACTIONS: | |
7e7e8dbb | 7075 | xlate_write_actions(ctx, ofpact_get_WRITE_ACTIONS(a)); |
2d9b49dd | 7076 | xlate_report_action_set(ctx, "is"); |
9583bc14 EJ |
7077 | break; |
7078 | ||
7079 | case OFPACT_WRITE_METADATA: | |
7080 | metadata = ofpact_get_WRITE_METADATA(a); | |
33bf9176 BP |
7081 | flow->metadata &= ~metadata->mask; |
7082 | flow->metadata |= metadata->metadata & metadata->mask; | |
9583bc14 EJ |
7083 | break; |
7084 | ||
638a19b0 | 7085 | case OFPACT_METER: |
076caa2f | 7086 | xlate_meter_action(ctx, ofpact_get_METER(a)); |
638a19b0 JR |
7087 | break; |
7088 | ||
9583bc14 | 7089 | case OFPACT_GOTO_TABLE: { |
9583bc14 | 7090 | struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a); |
9583bc14 | 7091 | |
9167fc1a JR |
7092 | ovs_assert(ctx->table_id < ogt->table_id); |
7093 | ||
4468099e | 7094 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
96c3a6e5 AZ |
7095 | ogt->table_id, true, true, false, last, |
7096 | do_xlate_actions); | |
9583bc14 EJ |
7097 | break; |
7098 | } | |
7099 | ||
7100 | case OFPACT_SAMPLE: | |
7101 | xlate_sample_action(ctx, ofpact_get_SAMPLE(a)); | |
7102 | break; | |
d4abaff5 | 7103 | |
7ae62a67 | 7104 | case OFPACT_CLONE: |
feee58b9 | 7105 | compose_clone(ctx, ofpact_get_CLONE(a), last); |
7ae62a67 WT |
7106 | break; |
7107 | ||
f839892a JS |
7108 | case OFPACT_ENCAP: |
7109 | xlate_generic_encap_action(ctx, ofpact_get_ENCAP(a)); | |
7110 | break; | |
7111 | ||
7112 | case OFPACT_DECAP: { | |
7113 | bool recirc_needed = | |
7114 | xlate_generic_decap_action(ctx, ofpact_get_DECAP(a)); | |
7115 | if (!ctx->error && recirc_needed) { | |
7116 | /* Recirculate for parsing of inner packet. */ | |
7117 | ctx_trigger_freeze(ctx); | |
7118 | /* Then continue with next action. */ | |
7119 | a = ofpact_next(a); | |
7120 | } | |
7121 | break; | |
7122 | } | |
7123 | ||
07659514 | 7124 | case OFPACT_CT: |
feee58b9 | 7125 | compose_conntrack_action(ctx, ofpact_get_CT(a), last); |
07659514 JS |
7126 | break; |
7127 | ||
72fe7578 | 7128 | case OFPACT_CT_CLEAR: |
1fe178d2 | 7129 | compose_ct_clear_action(ctx); |
72fe7578 BP |
7130 | break; |
7131 | ||
9ac0aada JR |
7132 | case OFPACT_NAT: |
7133 | /* This will be processed by compose_conntrack_action(). */ | |
7134 | ctx->ct_nat_action = ofpact_get_NAT(a); | |
7135 | break; | |
7136 | ||
d4abaff5 | 7137 | case OFPACT_DEBUG_RECIRC: |
1d361a81 | 7138 | ctx_trigger_freeze(ctx); |
d4abaff5 BP |
7139 | a = ofpact_next(a); |
7140 | break; | |
a934a3dd JP |
7141 | |
7142 | case OFPACT_DEBUG_SLOW: | |
7143 | ctx->xout->slow |= SLOW_ACTION; | |
7144 | break; | |
5b34f8fc NS |
7145 | |
7146 | case OFPACT_CHECK_PKT_LARGER: { | |
7147 | if (last) { | |
7148 | /* If this is last action, then there is no need to | |
7149 | * translate the action. */ | |
7150 | break; | |
7151 | } | |
7152 | const struct ofpact *remaining_acts = ofpact_next(a); | |
7153 | size_t remaining_acts_len = ofpact_remaining_len(remaining_acts, | |
7154 | ofpacts, | |
7155 | ofpacts_len); | |
7156 | xlate_check_pkt_larger(ctx, ofpact_get_CHECK_PKT_LARGER(a), | |
7157 | remaining_acts, remaining_acts_len); | |
7158 | break; | |
7159 | } | |
9583bc14 | 7160 | } |
1d741d6d JR |
7161 | |
7162 | /* Check if need to store this and the remaining actions for later | |
7163 | * execution. */ | |
1d361a81 BP |
7164 | if (!ctx->error && ctx->exit && ctx_first_frozen_action(ctx)) { |
7165 | freeze_unroll_actions(a, ofpact_end(ofpacts, ofpacts_len), ctx); | |
1d741d6d JR |
7166 | break; |
7167 | } | |
9583bc14 | 7168 | } |
9583bc14 EJ |
7169 | } |
7170 | ||
7171 | void | |
7172 | xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, | |
1f4a8933 JR |
7173 | ovs_version_t version, const struct flow *flow, |
7174 | ofp_port_t in_port, struct rule_dpif *rule, uint16_t tcp_flags, | |
1520ef4f BP |
7175 | const struct dp_packet *packet, struct flow_wildcards *wc, |
7176 | struct ofpbuf *odp_actions) | |
9583bc14 EJ |
7177 | { |
7178 | xin->ofproto = ofproto; | |
1f4a8933 | 7179 | xin->tables_version = version; |
9583bc14 | 7180 | xin->flow = *flow; |
8d8ab6c2 | 7181 | xin->upcall_flow = flow; |
cc377352 | 7182 | xin->flow.in_port.ofp_port = in_port; |
c61f3870 | 7183 | xin->flow.actset_output = OFPP_UNSET; |
9583bc14 | 7184 | xin->packet = packet; |
df70a773 | 7185 | xin->allow_side_effects = packet != NULL; |
9583bc14 | 7186 | xin->rule = rule; |
b256dc52 | 7187 | xin->xcache = NULL; |
9583bc14 EJ |
7188 | xin->ofpacts = NULL; |
7189 | xin->ofpacts_len = 0; | |
7190 | xin->tcp_flags = tcp_flags; | |
2d9b49dd | 7191 | xin->trace = NULL; |
9583bc14 | 7192 | xin->resubmit_stats = NULL; |
790c5d26 | 7193 | xin->depth = 0; |
cdd42eda | 7194 | xin->resubmits = 0; |
49a73e0c | 7195 | xin->wc = wc; |
1520ef4f | 7196 | xin->odp_actions = odp_actions; |
331c07ac | 7197 | xin->in_packet_out = false; |
e6bc8e74 | 7198 | xin->recirc_queue = NULL; |
00135b86 | 7199 | xin->xport_uuid = UUID_ZERO; |
e672ff9b JR |
7200 | |
7201 | /* Do recirc lookup. */ | |
1d361a81 | 7202 | xin->frozen_state = NULL; |
29b1ea3f BP |
7203 | if (flow->recirc_id) { |
7204 | const struct recirc_id_node *node | |
7205 | = recirc_id_node_find(flow->recirc_id); | |
7206 | if (node) { | |
1d361a81 | 7207 | xin->frozen_state = &node->state; |
29b1ea3f BP |
7208 | } |
7209 | } | |
9583bc14 EJ |
7210 | } |
7211 | ||
7212 | void | |
7213 | xlate_out_uninit(struct xlate_out *xout) | |
7214 | { | |
e672ff9b | 7215 | if (xout) { |
fbf5d6ec | 7216 | recirc_refs_unref(&xout->recircs); |
9583bc14 EJ |
7217 | } |
7218 | } | |
9583bc14 | 7219 | \f |
55954f6e EJ |
7220 | static struct skb_priority_to_dscp * |
7221 | get_skb_priority(const struct xport *xport, uint32_t skb_priority) | |
7222 | { | |
7223 | struct skb_priority_to_dscp *pdscp; | |
7224 | uint32_t hash; | |
7225 | ||
7226 | hash = hash_int(skb_priority, 0); | |
7227 | HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) { | |
7228 | if (pdscp->skb_priority == skb_priority) { | |
7229 | return pdscp; | |
7230 | } | |
7231 | } | |
7232 | return NULL; | |
7233 | } | |
7234 | ||
7235 | static bool | |
7236 | dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority, | |
7237 | uint8_t *dscp) | |
7238 | { | |
7239 | struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority); | |
7240 | *dscp = pdscp ? pdscp->dscp : 0; | |
7241 | return pdscp != NULL; | |
7242 | } | |
7243 | ||
16194afd DDP |
7244 | static size_t |
7245 | count_skb_priorities(const struct xport *xport) | |
7246 | { | |
7247 | return hmap_count(&xport->skb_priorities); | |
7248 | } | |
7249 | ||
55954f6e EJ |
7250 | static void |
7251 | clear_skb_priorities(struct xport *xport) | |
7252 | { | |
4ec3d7c7 | 7253 | struct skb_priority_to_dscp *pdscp; |
55954f6e | 7254 | |
4ec3d7c7 | 7255 | HMAP_FOR_EACH_POP (pdscp, hmap_node, &xport->skb_priorities) { |
55954f6e EJ |
7256 | free(pdscp); |
7257 | } | |
7258 | } | |
7259 | ||
ce4a6b76 BP |
7260 | static bool |
7261 | actions_output_to_local_port(const struct xlate_ctx *ctx) | |
7262 | { | |
46c88433 | 7263 | odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL); |
ce4a6b76 BP |
7264 | const struct nlattr *a; |
7265 | unsigned int left; | |
7266 | ||
1520ef4f BP |
7267 | NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->odp_actions->data, |
7268 | ctx->odp_actions->size) { | |
ce4a6b76 BP |
7269 | if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT |
7270 | && nl_attr_get_odp_port(a) == local_odp_port) { | |
7271 | return true; | |
7272 | } | |
7273 | } | |
7274 | return false; | |
7275 | } | |
9583bc14 | 7276 | |
5e2a6702 | 7277 | #if defined(__linux__) |
7d031d7e BP |
7278 | /* Returns the maximum number of packets that the Linux kernel is willing to |
7279 | * queue up internally to certain kinds of software-implemented ports, or the | |
7280 | * default (and rarely modified) value if it cannot be determined. */ | |
7281 | static int | |
7282 | netdev_max_backlog(void) | |
7283 | { | |
7284 | static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; | |
7285 | static int max_backlog = 1000; /* The normal default value. */ | |
7286 | ||
7287 | if (ovsthread_once_start(&once)) { | |
7288 | static const char filename[] = "/proc/sys/net/core/netdev_max_backlog"; | |
7289 | FILE *stream; | |
7290 | int n; | |
7291 | ||
7292 | stream = fopen(filename, "r"); | |
7293 | if (!stream) { | |
120c348f | 7294 | VLOG_INFO("%s: open failed (%s)", filename, ovs_strerror(errno)); |
7d031d7e BP |
7295 | } else { |
7296 | if (fscanf(stream, "%d", &n) != 1) { | |
7297 | VLOG_WARN("%s: read error", filename); | |
7298 | } else if (n <= 100) { | |
7299 | VLOG_WARN("%s: unexpectedly small value %d", filename, n); | |
7300 | } else { | |
7301 | max_backlog = n; | |
7302 | } | |
7303 | fclose(stream); | |
7304 | } | |
7305 | ovsthread_once_done(&once); | |
7306 | ||
7307 | VLOG_DBG("%s: using %d max_backlog", filename, max_backlog); | |
7308 | } | |
7309 | ||
7310 | return max_backlog; | |
7311 | } | |
7312 | ||
7313 | /* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in | |
7314 | * 'odp_actions'. */ | |
7315 | static int | |
7316 | count_output_actions(const struct ofpbuf *odp_actions) | |
7317 | { | |
7318 | const struct nlattr *a; | |
7319 | size_t left; | |
7320 | int n = 0; | |
7321 | ||
6fd6ed71 | 7322 | NL_ATTR_FOR_EACH_UNSAFE (a, left, odp_actions->data, odp_actions->size) { |
9df65060 VDA |
7323 | if ((a->nla_type == OVS_ACTION_ATTR_OUTPUT) || |
7324 | (a->nla_type == OVS_ACTION_ATTR_LB_OUTPUT)) { | |
7d031d7e BP |
7325 | n++; |
7326 | } | |
7327 | } | |
7328 | return n; | |
7329 | } | |
5e2a6702 | 7330 | #endif /* defined(__linux__) */ |
7d031d7e BP |
7331 | |
7332 | /* Returns true if 'odp_actions' contains more output actions than the datapath | |
7333 | * can reliably handle in one go. On Linux, this is the value of the | |
7334 | * net.core.netdev_max_backlog sysctl, which limits the maximum number of | |
7335 | * packets that the kernel is willing to queue up for processing while the | |
7336 | * datapath is processing a set of actions. */ | |
7337 | static bool | |
5e2a6702 | 7338 | too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) |
7d031d7e BP |
7339 | { |
7340 | #ifdef __linux__ | |
6fd6ed71 | 7341 | return (odp_actions->size / NL_A_U32_SIZE > netdev_max_backlog() |
7d031d7e BP |
7342 | && count_output_actions(odp_actions) > netdev_max_backlog()); |
7343 | #else | |
7344 | /* OSes other than Linux might have similar limits, but we don't know how | |
7345 | * to determine them.*/ | |
7346 | return false; | |
7347 | #endif | |
7348 | } | |
7349 | ||
234c3da9 BP |
7350 | static void |
7351 | xlate_wc_init(struct xlate_ctx *ctx) | |
7352 | { | |
7353 | flow_wildcards_init_catchall(ctx->wc); | |
7354 | ||
7355 | /* Some fields we consider to always be examined. */ | |
3d4b2e6e | 7356 | WC_MASK_FIELD(ctx->wc, packet_type); |
5e2e998a | 7357 | WC_MASK_FIELD(ctx->wc, in_port); |
29b5c0c3 | 7358 | WC_MASK_FIELD(ctx->wc, dl_type); |
234c3da9 | 7359 | if (is_ip_any(&ctx->xin->flow)) { |
5e2e998a | 7360 | WC_MASK_FIELD_MASK(ctx->wc, nw_frag, FLOW_NW_FRAG_MASK); |
234c3da9 BP |
7361 | } |
7362 | ||
7363 | if (ctx->xbridge->support.odp.recirc) { | |
7364 | /* Always exactly match recirc_id when datapath supports | |
7365 | * recirculation. */ | |
5e2e998a | 7366 | WC_MASK_FIELD(ctx->wc, recirc_id); |
234c3da9 BP |
7367 | } |
7368 | ||
7369 | if (ctx->xbridge->netflow) { | |
7370 | netflow_mask_wc(&ctx->xin->flow, ctx->wc); | |
7371 | } | |
7372 | ||
7373 | tnl_wc_init(&ctx->xin->flow, ctx->wc); | |
7374 | } | |
7375 | ||
7376 | static void | |
7377 | xlate_wc_finish(struct xlate_ctx *ctx) | |
7378 | { | |
f0fb825a EG |
7379 | int i; |
7380 | ||
234c3da9 BP |
7381 | /* Clear the metadata and register wildcard masks, because we won't |
7382 | * use non-header fields as part of the cache. */ | |
7383 | flow_wildcards_clear_non_packet_fields(ctx->wc); | |
7384 | ||
29b5c0c3 BP |
7385 | /* Wildcard Ethernet address fields if the original packet type was not |
7386 | * Ethernet. | |
7387 | * | |
7388 | * (The Ethertype field is used even when the original packet type is not | |
7389 | * Ethernet.) */ | |
beb75a40 JS |
7390 | if (ctx->xin->upcall_flow->packet_type != htonl(PT_ETH)) { |
7391 | ctx->wc->masks.dl_dst = eth_addr_zero; | |
7392 | ctx->wc->masks.dl_src = eth_addr_zero; | |
7393 | } | |
7394 | ||
234c3da9 BP |
7395 | /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow |
7396 | * uses the low 8 bits of the 16-bit tp_src and tp_dst members to | |
7397 | * represent these fields. The datapath interface, on the other hand, | |
7398 | * represents them with just 8 bits each. This means that if the high | |
7399 | * 8 bits of the masks for these fields somehow become set, then they | |
7400 | * will get chopped off by a round trip through the datapath, and | |
7401 | * revalidation will spot that as an inconsistency and delete the flow. | |
7402 | * Avoid the problem here by making sure that only the low 8 bits of | |
7403 | * either field can be unwildcarded for ICMP. | |
7404 | */ | |
a75636c8 | 7405 | if (is_icmpv4(&ctx->xin->flow, NULL) || is_icmpv6(&ctx->xin->flow, NULL)) { |
234c3da9 BP |
7406 | ctx->wc->masks.tp_src &= htons(UINT8_MAX); |
7407 | ctx->wc->masks.tp_dst &= htons(UINT8_MAX); | |
7408 | } | |
7409 | /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */ | |
f0fb825a EG |
7410 | for (i = 0; i < FLOW_MAX_VLAN_HEADERS; i++) { |
7411 | if (ctx->wc->masks.vlans[i].tci) { | |
7412 | ctx->wc->masks.vlans[i].tci |= htons(VLAN_CFI); | |
7413 | } | |
234c3da9 | 7414 | } |
4a7ab326 DDP |
7415 | |
7416 | /* The classifier might return masks that match on tp_src and tp_dst even | |
7417 | * for later fragments. This happens because there might be flows that | |
7418 | * match on tp_src or tp_dst without matching on the frag bits, because | |
7419 | * it is not a prerequisite for OpenFlow. Since it is a prerequisite for | |
7420 | * datapath flows and since tp_src and tp_dst are always going to be 0, | |
7421 | * wildcard the fields here. */ | |
7422 | if (ctx->xin->flow.nw_frag & FLOW_NW_FRAG_LATER) { | |
7423 | ctx->wc->masks.tp_src = 0; | |
7424 | ctx->wc->masks.tp_dst = 0; | |
7425 | } | |
736a8111 VDA |
7426 | |
7427 | /* Clear flow wildcard bits for fields which are not present | |
7428 | * in the original packet header. These wildcards may get set | |
7429 | * due to push/set_field actions. This results into frequent | |
7430 | * invalidation of datapath flows by revalidator thread. */ | |
7431 | ||
7432 | /* Clear mpls label wc bits if original packet is non-mpls. */ | |
7433 | if (!eth_type_mpls(ctx->xin->upcall_flow->dl_type)) { | |
7434 | for (i = 0; i < FLOW_MAX_MPLS_LABELS; i++) { | |
7435 | ctx->wc->masks.mpls_lse[i] = 0; | |
7436 | } | |
7437 | } | |
7438 | /* Clear vlan header wc bits if original packet does not have | |
7439 | * vlan header. */ | |
7440 | for (i = 0; i < FLOW_MAX_VLAN_HEADERS; i++) { | |
7441 | if (!eth_type_vlan(ctx->xin->upcall_flow->vlans[i].tpid)) { | |
7442 | ctx->wc->masks.vlans[i].tpid = 0; | |
7443 | ctx->wc->masks.vlans[i].tci = 0; | |
7444 | } | |
7445 | } | |
234c3da9 BP |
7446 | } |
7447 | ||
e672ff9b JR |
7448 | /* Translates the flow, actions, or rule in 'xin' into datapath actions in |
7449 | * 'xout'. | |
56450a41 | 7450 | * The caller must take responsibility for eventually freeing 'xout', with |
fff1b9c0 JR |
7451 | * xlate_out_uninit(). |
7452 | * Returns 'XLATE_OK' if translation was successful. In case of an error an | |
7453 | * empty set of actions will be returned in 'xin->odp_actions' (if non-NULL), | |
7454 | * so that most callers may ignore the return value and transparently install a | |
7455 | * drop flow when the translation fails. */ | |
7456 | enum xlate_error | |
84f0f298 | 7457 | xlate_actions(struct xlate_in *xin, struct xlate_out *xout) |
9583bc14 | 7458 | { |
e467ea42 BP |
7459 | *xout = (struct xlate_out) { |
7460 | .slow = 0, | |
fbf5d6ec | 7461 | .recircs = RECIRC_REFS_EMPTY_INITIALIZER, |
e467ea42 BP |
7462 | }; |
7463 | ||
84f0f298 | 7464 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
bb00fdef BP |
7465 | struct xbridge *xbridge = xbridge_lookup(xcfg, xin->ofproto); |
7466 | if (!xbridge) { | |
fff1b9c0 | 7467 | return XLATE_BRIDGE_NOT_FOUND; |
bb00fdef BP |
7468 | } |
7469 | ||
33bf9176 BP |
7470 | struct flow *flow = &xin->flow; |
7471 | ||
84cf3c1f | 7472 | uint8_t stack_stub[1024]; |
bb00fdef | 7473 | uint64_t action_set_stub[1024 / 8]; |
1d361a81 | 7474 | uint64_t frozen_actions_stub[1024 / 8]; |
1520ef4f BP |
7475 | uint64_t actions_stub[256 / 8]; |
7476 | struct ofpbuf scratch_actions = OFPBUF_STUB_INITIALIZER(actions_stub); | |
bb00fdef BP |
7477 | struct xlate_ctx ctx = { |
7478 | .xin = xin, | |
7479 | .xout = xout, | |
7480 | .base_flow = *flow, | |
c2b878e0 | 7481 | .orig_tunnel_ipv6_dst = flow_tnl_dst(&flow->tunnel), |
0506f184 | 7482 | .xcfg = xcfg, |
bb00fdef BP |
7483 | .xbridge = xbridge, |
7484 | .stack = OFPBUF_STUB_INITIALIZER(stack_stub), | |
7485 | .rule = xin->rule, | |
c0e638aa BP |
7486 | .wc = (xin->wc |
7487 | ? xin->wc | |
f36efd90 | 7488 | : &(struct flow_wildcards) { .masks = { .dl_type = 0 } }), |
1520ef4f | 7489 | .odp_actions = xin->odp_actions ? xin->odp_actions : &scratch_actions, |
bb00fdef | 7490 | |
790c5d26 | 7491 | .depth = xin->depth, |
cdd42eda | 7492 | .resubmits = xin->resubmits, |
bb00fdef | 7493 | .in_action_set = false, |
331c07ac | 7494 | .in_packet_out = xin->in_packet_out, |
f839892a | 7495 | .pending_encap = false, |
88ec1e0a | 7496 | .pending_decap = false, |
1fc11c59 | 7497 | .encap_data = NULL, |
bb00fdef BP |
7498 | |
7499 | .table_id = 0, | |
7500 | .rule_cookie = OVS_BE64_MAX, | |
7501 | .orig_skb_priority = flow->skb_priority, | |
7502 | .sflow_n_outputs = 0, | |
7503 | .sflow_odp_port = 0, | |
2031ef97 | 7504 | .nf_output_iface = NF_OUT_DROP, |
bb00fdef | 7505 | .exit = false, |
fff1b9c0 | 7506 | .error = XLATE_OK, |
3d6151f3 | 7507 | .mirrors = 0, |
bb00fdef | 7508 | |
1d361a81 | 7509 | .freezing = false, |
53cc166a | 7510 | .recirc_update_dp_hash = false, |
1d361a81 | 7511 | .frozen_actions = OFPBUF_STUB_INITIALIZER(frozen_actions_stub), |
77ab5fd2 | 7512 | .pause = NULL, |
bb00fdef | 7513 | |
e12ec36b | 7514 | .was_mpls = false, |
07659514 | 7515 | .conntracked = false, |
bb00fdef | 7516 | |
9ac0aada JR |
7517 | .ct_nat_action = NULL, |
7518 | ||
bb00fdef BP |
7519 | .action_set_has_group = false, |
7520 | .action_set = OFPBUF_STUB_INITIALIZER(action_set_stub), | |
7521 | }; | |
865ca6cf BP |
7522 | |
7523 | /* 'base_flow' reflects the packet as it came in, but we need it to reflect | |
42deb67d PS |
7524 | * the packet as the datapath will treat it for output actions. Our |
7525 | * datapath doesn't retain tunneling information without us re-setting | |
7526 | * it, so clear the tunnel data. | |
865ca6cf | 7527 | */ |
42deb67d | 7528 | |
bb00fdef | 7529 | memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); |
865ca6cf | 7530 | |
1520ef4f | 7531 | ofpbuf_reserve(ctx.odp_actions, NL_A_U32_SIZE); |
c0e638aa | 7532 | xlate_wc_init(&ctx); |
bb00fdef | 7533 | |
46c88433 | 7534 | COVERAGE_INC(xlate_actions); |
9583bc14 | 7535 | |
2d9b49dd BP |
7536 | xin->trace = xlate_report(&ctx, OFT_BRIDGE, "bridge(\"%s\")", |
7537 | xbridge->name); | |
1d361a81 BP |
7538 | if (xin->frozen_state) { |
7539 | const struct frozen_state *state = xin->frozen_state; | |
e672ff9b | 7540 | |
2d9b49dd BP |
7541 | struct ovs_list *old_trace = xin->trace; |
7542 | xin->trace = xlate_report(&ctx, OFT_THAW, "thaw"); | |
d6bef3cc | 7543 | |
e672ff9b | 7544 | if (xin->ofpacts_len > 0 || ctx.rule) { |
2d9b49dd BP |
7545 | xlate_report_error(&ctx, "Recirculation conflict (%s)!", |
7546 | xin->ofpacts_len ? "actions" : "rule"); | |
fff1b9c0 | 7547 | ctx.error = XLATE_RECIRCULATION_CONFLICT; |
1520ef4f | 7548 | goto exit; |
e672ff9b JR |
7549 | } |
7550 | ||
7551 | /* Set the bridge for post-recirculation processing if needed. */ | |
07a3cd5c | 7552 | if (!uuid_equals(&ctx.xbridge->ofproto->uuid, &state->ofproto_uuid)) { |
2082425c | 7553 | const struct xbridge *new_bridge |
290835f9 | 7554 | = xbridge_lookup_by_uuid(xcfg, &state->ofproto_uuid); |
e672ff9b JR |
7555 | |
7556 | if (OVS_UNLIKELY(!new_bridge)) { | |
7557 | /* Drop the packet if the bridge cannot be found. */ | |
2d9b49dd | 7558 | xlate_report_error(&ctx, "Frozen bridge no longer exists."); |
fff1b9c0 | 7559 | ctx.error = XLATE_BRIDGE_NOT_FOUND; |
2d9b49dd | 7560 | xin->trace = old_trace; |
1520ef4f | 7561 | goto exit; |
e672ff9b JR |
7562 | } |
7563 | ctx.xbridge = new_bridge; | |
1f4a8933 JR |
7564 | /* The bridge is now known so obtain its table version. */ |
7565 | ctx.xin->tables_version | |
7566 | = ofproto_dpif_get_tables_version(ctx.xbridge->ofproto); | |
e672ff9b JR |
7567 | } |
7568 | ||
1d361a81 BP |
7569 | /* Set the thawed table id. Note: A table lookup is done only if there |
7570 | * are no frozen actions. */ | |
2082425c | 7571 | ctx.table_id = state->table_id; |
2d9b49dd BP |
7572 | xlate_report(&ctx, OFT_THAW, |
7573 | "Resuming from table %"PRIu8, ctx.table_id); | |
e672ff9b | 7574 | |
40b0fbd3 | 7575 | ctx.conntracked = state->conntracked; |
07659514 | 7576 | if (!state->conntracked) { |
72fe7578 | 7577 | clear_conntrack(&ctx); |
07659514 JS |
7578 | } |
7579 | ||
e672ff9b | 7580 | /* Restore pipeline metadata. May change flow's in_port and other |
1d361a81 | 7581 | * metadata to the values that existed when freezing was triggered. */ |
02f641e2 YS |
7582 | frozen_metadata_to_flow(&ctx.xbridge->ofproto->up, |
7583 | &state->metadata, flow); | |
e672ff9b JR |
7584 | |
7585 | /* Restore stack, if any. */ | |
2082425c | 7586 | if (state->stack) { |
84cf3c1f | 7587 | ofpbuf_put(&ctx.stack, state->stack, state->stack_size); |
e672ff9b JR |
7588 | } |
7589 | ||
29bae541 BP |
7590 | /* Restore mirror state. */ |
7591 | ctx.mirrors = state->mirrors; | |
7592 | ||
e672ff9b | 7593 | /* Restore action set, if any. */ |
2082425c | 7594 | if (state->action_set_len) { |
2d9b49dd | 7595 | xlate_report_actions(&ctx, OFT_THAW, "Restoring action set", |
417509fa | 7596 | state->action_set, state->action_set_len); |
d6bef3cc | 7597 | |
7e7e8dbb BP |
7598 | flow->actset_output = OFPP_UNSET; |
7599 | xlate_write_actions__(&ctx, state->action_set, | |
7600 | state->action_set_len); | |
e672ff9b JR |
7601 | } |
7602 | ||
1d361a81 BP |
7603 | /* Restore frozen actions. If there are no actions, processing will |
7604 | * start with a lookup in the table set above. */ | |
417509fa BP |
7605 | xin->ofpacts = state->ofpacts; |
7606 | xin->ofpacts_len = state->ofpacts_len; | |
7607 | if (state->ofpacts_len) { | |
2d9b49dd | 7608 | xlate_report_actions(&ctx, OFT_THAW, "Restoring actions", |
d6bef3cc | 7609 | xin->ofpacts, xin->ofpacts_len); |
e672ff9b | 7610 | } |
e672ff9b | 7611 | |
2d9b49dd BP |
7612 | xin->trace = old_trace; |
7613 | } else if (OVS_UNLIKELY(flow->recirc_id)) { | |
7614 | xlate_report_error(&ctx, | |
7615 | "Recirculation context not found for ID %"PRIx32, | |
7616 | flow->recirc_id); | |
fff1b9c0 | 7617 | ctx.error = XLATE_NO_RECIRCULATION_CONTEXT; |
1520ef4f | 7618 | goto exit; |
e672ff9b | 7619 | } |
9583bc14 | 7620 | |
8d8ab6c2 JG |
7621 | /* Tunnel metadata in udpif format must be normalized before translation. */ |
7622 | if (flow->tunnel.flags & FLOW_TNL_F_UDPIF) { | |
5b09d9f7 MS |
7623 | const struct tun_table *tun_tab = ofproto_get_tun_tab( |
7624 | &ctx.xbridge->ofproto->up); | |
8d8ab6c2 JG |
7625 | int err; |
7626 | ||
7627 | err = tun_metadata_from_geneve_udpif(tun_tab, &xin->upcall_flow->tunnel, | |
7628 | &xin->upcall_flow->tunnel, | |
7629 | &flow->tunnel); | |
7630 | if (err) { | |
2d9b49dd | 7631 | xlate_report_error(&ctx, "Invalid Geneve tunnel metadata"); |
8d8ab6c2 JG |
7632 | ctx.error = XLATE_INVALID_TUNNEL_METADATA; |
7633 | goto exit; | |
7634 | } | |
02f641e2 | 7635 | } else if (!flow->tunnel.metadata.tab) { |
8d8ab6c2 JG |
7636 | /* If the original flow did not come in on a tunnel, then it won't have |
7637 | * FLOW_TNL_F_UDPIF set. However, we still need to have a metadata | |
7638 | * table in case we generate tunnel actions. */ | |
5b09d9f7 MS |
7639 | flow->tunnel.metadata.tab = ofproto_get_tun_tab( |
7640 | &ctx.xbridge->ofproto->up); | |
8d8ab6c2 JG |
7641 | } |
7642 | ctx.wc->masks.tunnel.metadata.tab = flow->tunnel.metadata.tab; | |
7643 | ||
beb75a40 JS |
7644 | /* Get the proximate input port of the packet. (If xin->frozen_state, |
7645 | * flow->in_port is the ultimate input port of the packet.) */ | |
7646 | struct xport *in_port = get_ofp_port(xbridge, | |
7647 | ctx.base_flow.in_port.ofp_port); | |
00135b86 ZB |
7648 | if (in_port && !in_port->peer) { |
7649 | ctx.xin->xport_uuid = in_port->uuid; | |
7650 | } | |
beb75a40 | 7651 | |
875ab130 BP |
7652 | if (flow->packet_type != htonl(PT_ETH) && in_port && |
7653 | in_port->pt_mode == NETDEV_PT_LEGACY_L3 && ctx.table_id == 0) { | |
beb75a40 JS |
7654 | /* Add dummy Ethernet header to non-L2 packet if it's coming from a |
7655 | * L3 port. So all packets will be L2 packets for lookup. | |
7656 | * The dl_type has already been set from the packet_type. */ | |
7657 | flow->packet_type = htonl(PT_ETH); | |
7658 | flow->dl_src = eth_addr_zero; | |
7659 | flow->dl_dst = eth_addr_zero; | |
f839892a | 7660 | ctx.pending_encap = true; |
beb75a40 JS |
7661 | } |
7662 | ||
10c44245 | 7663 | if (!xin->ofpacts && !ctx.rule) { |
b2e89cc9 | 7664 | ctx.rule = rule_dpif_lookup_from_table( |
1f4a8933 | 7665 | ctx.xbridge->ofproto, ctx.xin->tables_version, flow, ctx.wc, |
1e1e1d19 | 7666 | ctx.xin->resubmit_stats, &ctx.table_id, |
a027899e | 7667 | flow->in_port.ofp_port, true, true, ctx.xin->xcache); |
10c44245 | 7668 | if (ctx.xin->resubmit_stats) { |
16441315 | 7669 | rule_dpif_credit_stats(ctx.rule, ctx.xin->resubmit_stats, false); |
10c44245 | 7670 | } |
b256dc52 JS |
7671 | if (ctx.xin->xcache) { |
7672 | struct xc_entry *entry; | |
7673 | ||
7674 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE); | |
901a517e | 7675 | entry->rule = ctx.rule; |
07a3cd5c | 7676 | ofproto_rule_ref(&ctx.rule->up); |
b256dc52 | 7677 | } |
a8c31348 | 7678 | |
2d9b49dd | 7679 | xlate_report_table(&ctx, ctx.rule, ctx.table_id); |
10c44245 | 7680 | } |
10c44245 | 7681 | |
1d361a81 BP |
7682 | /* Tunnel stats only for not-thawed packets. */ |
7683 | if (!xin->frozen_state && in_port && in_port->is_tunnel) { | |
b256dc52 JS |
7684 | if (ctx.xin->resubmit_stats) { |
7685 | netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats); | |
7686 | if (in_port->bfd) { | |
7687 | bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats); | |
7688 | } | |
7689 | } | |
7690 | if (ctx.xin->xcache) { | |
7691 | struct xc_entry *entry; | |
7692 | ||
7693 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV); | |
901a517e JR |
7694 | entry->dev.rx = netdev_ref(in_port->netdev); |
7695 | entry->dev.bfd = bfd_ref(in_port->bfd); | |
d6fc5f57 EJ |
7696 | } |
7697 | } | |
7698 | ||
1d361a81 | 7699 | if (!xin->frozen_state && process_special(&ctx, in_port)) { |
bef1403e BP |
7700 | /* process_special() did all the processing for this packet. |
7701 | * | |
1d361a81 BP |
7702 | * We do not perform special processing on thawed packets, since that |
7703 | * was done before they were frozen and should not be redone. */ | |
27a9c0e3 | 7704 | mirror_ingress_packet(&ctx); |
bef1403e BP |
7705 | } else if (in_port && in_port->xbundle |
7706 | && xbundle_mirror_out(xbridge, in_port->xbundle)) { | |
2d9b49dd BP |
7707 | xlate_report_error(&ctx, "dropping packet received on port " |
7708 | "%s, which is reserved exclusively for mirroring", | |
7709 | in_port->xbundle->name); | |
bef1403e | 7710 | } else { |
1d361a81 | 7711 | /* Sampling is done on initial reception; don't redo after thawing. */ |
a6092018 | 7712 | unsigned int user_cookie_offset = 0; |
1d361a81 | 7713 | if (!xin->frozen_state) { |
a6092018 BP |
7714 | user_cookie_offset = compose_sflow_action(&ctx); |
7715 | compose_ipfix_action(&ctx, ODPP_NONE); | |
e672ff9b | 7716 | } |
0731abc5 | 7717 | size_t sample_actions_len = ctx.odp_actions->size; |
a13a0209 | 7718 | bool ecn_drop = !tnl_process_ecn(flow); |
9583bc14 | 7719 | |
a13a0209 | 7720 | if (!ecn_drop |
234c3da9 | 7721 | && (!in_port || may_receive(in_port, &ctx))) { |
1806291d BP |
7722 | const struct ofpact *ofpacts; |
7723 | size_t ofpacts_len; | |
7724 | ||
7725 | if (xin->ofpacts) { | |
7726 | ofpacts = xin->ofpacts; | |
7727 | ofpacts_len = xin->ofpacts_len; | |
7728 | } else if (ctx.rule) { | |
7729 | const struct rule_actions *actions | |
07a3cd5c | 7730 | = rule_get_actions(&ctx.rule->up); |
1806291d BP |
7731 | ofpacts = actions->ofpacts; |
7732 | ofpacts_len = actions->ofpacts_len; | |
07a3cd5c | 7733 | ctx.rule_cookie = ctx.rule->up.flow_cookie; |
1806291d BP |
7734 | } else { |
7735 | OVS_NOT_REACHED(); | |
7736 | } | |
7737 | ||
7efbc3b7 | 7738 | mirror_ingress_packet(&ctx); |
f5634764 | 7739 | do_xlate_actions(ofpacts, ofpacts_len, &ctx, true, false); |
fff1b9c0 JR |
7740 | if (ctx.error) { |
7741 | goto exit; | |
7742 | } | |
9583bc14 EJ |
7743 | |
7744 | /* We've let OFPP_NORMAL and the learning action look at the | |
1d361a81 | 7745 | * packet, so cancel all actions and freezing if forwarding is |
8a5fb3b4 | 7746 | * disabled. */ |
9efd308e DV |
7747 | if (in_port && (!xport_stp_forward_state(in_port) || |
7748 | !xport_rstp_forward_state(in_port))) { | |
1520ef4f | 7749 | ctx.odp_actions->size = sample_actions_len; |
1d361a81 | 7750 | ctx_cancel_freeze(&ctx); |
8a5fb3b4 | 7751 | ofpbuf_clear(&ctx.action_set); |
a13a0209 | 7752 | ctx.error = XLATE_FORWARDING_DISABLED; |
8a5fb3b4 BP |
7753 | } |
7754 | ||
1d361a81 | 7755 | if (!ctx.freezing) { |
8a5fb3b4 | 7756 | xlate_action_set(&ctx); |
e672ff9b | 7757 | } |
1d361a81 | 7758 | if (ctx.freezing) { |
77ab5fd2 | 7759 | finish_freezing(&ctx); |
9583bc14 | 7760 | } |
a13a0209 AT |
7761 | } else if (ecn_drop) { |
7762 | ctx.error = XLATE_CONGESTION_DROP; | |
9583bc14 EJ |
7763 | } |
7764 | ||
e672ff9b | 7765 | /* Output only fully processed packets. */ |
1d361a81 | 7766 | if (!ctx.freezing |
e672ff9b | 7767 | && xbridge->has_in_band |
ce4a6b76 BP |
7768 | && in_band_must_output_to_local_port(flow) |
7769 | && !actions_output_to_local_port(&ctx)) { | |
c9dc050d AT |
7770 | WC_MASK_FIELD(ctx.wc, nw_proto); |
7771 | WC_MASK_FIELD(ctx.wc, tp_src); | |
7772 | WC_MASK_FIELD(ctx.wc, tp_dst); | |
7773 | WC_MASK_FIELD(ctx.wc, dl_type); | |
b1986b08 BP |
7774 | xlate_report(&ctx, OFT_DETAIL, "outputting DHCP packet " |
7775 | "to local port for in-band control"); | |
11938578 | 7776 | compose_output_action(&ctx, OFPP_LOCAL, NULL, false, false); |
9583bc14 | 7777 | } |
aaa0fbae | 7778 | |
a6092018 BP |
7779 | if (user_cookie_offset) { |
7780 | fix_sflow_action(&ctx, user_cookie_offset); | |
e672ff9b | 7781 | } |
9583bc14 EJ |
7782 | } |
7783 | ||
1520ef4f | 7784 | if (nl_attr_oversized(ctx.odp_actions->size)) { |
542024c4 | 7785 | /* These datapath actions are too big for a Netlink attribute, so we |
0f032e95 BP |
7786 | * can't hand them to the kernel directly. dpif_execute() can execute |
7787 | * them one by one with help, so just mark the result as SLOW_ACTION to | |
7788 | * prevent the flow from being installed. */ | |
7789 | COVERAGE_INC(xlate_actions_oversize); | |
7790 | ctx.xout->slow |= SLOW_ACTION; | |
1520ef4f | 7791 | } else if (too_many_output_actions(ctx.odp_actions)) { |
7d031d7e BP |
7792 | COVERAGE_INC(xlate_actions_too_many_output); |
7793 | ctx.xout->slow |= SLOW_ACTION; | |
542024c4 BP |
7794 | } |
7795 | ||
64fb5f82 JP |
7796 | /* Update NetFlow for non-frozen traffic. */ |
7797 | if (xbridge->netflow && !xin->frozen_state) { | |
1806291d BP |
7798 | if (ctx.xin->resubmit_stats) { |
7799 | netflow_flow_update(xbridge->netflow, flow, | |
2031ef97 | 7800 | ctx.nf_output_iface, |
1806291d BP |
7801 | ctx.xin->resubmit_stats); |
7802 | } | |
7803 | if (ctx.xin->xcache) { | |
7804 | struct xc_entry *entry; | |
b256dc52 | 7805 | |
1806291d | 7806 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW); |
901a517e JR |
7807 | entry->nf.netflow = netflow_ref(xbridge->netflow); |
7808 | entry->nf.flow = xmemdup(flow, sizeof *flow); | |
7809 | entry->nf.iface = ctx.nf_output_iface; | |
d6fc5f57 EJ |
7810 | } |
7811 | } | |
7812 | ||
8d8ab6c2 JG |
7813 | /* Translate tunnel metadata masks to udpif format if necessary. */ |
7814 | if (xin->upcall_flow->tunnel.flags & FLOW_TNL_F_UDPIF) { | |
7815 | if (ctx.wc->masks.tunnel.metadata.present.map) { | |
7816 | const struct flow_tnl *upcall_tnl = &xin->upcall_flow->tunnel; | |
7817 | struct geneve_opt opts[TLV_TOT_OPT_SIZE / | |
7818 | sizeof(struct geneve_opt)]; | |
7819 | ||
7820 | tun_metadata_to_geneve_udpif_mask(&flow->tunnel, | |
7821 | &ctx.wc->masks.tunnel, | |
7822 | upcall_tnl->metadata.opts.gnv, | |
7823 | upcall_tnl->metadata.present.len, | |
7824 | opts); | |
7825 | memset(&ctx.wc->masks.tunnel.metadata, 0, | |
7826 | sizeof ctx.wc->masks.tunnel.metadata); | |
7827 | memcpy(&ctx.wc->masks.tunnel.metadata.opts.gnv, opts, | |
7828 | upcall_tnl->metadata.present.len); | |
7829 | } | |
7830 | ctx.wc->masks.tunnel.metadata.present.len = 0xff; | |
7831 | ctx.wc->masks.tunnel.metadata.tab = NULL; | |
7832 | ctx.wc->masks.tunnel.flags |= FLOW_TNL_F_UDPIF; | |
7833 | } else if (!xin->upcall_flow->tunnel.metadata.tab) { | |
7834 | /* If we didn't have options in UDPIF format and didn't have an existing | |
7835 | * metadata table, then it means that there were no options at all when | |
7836 | * we started processing and any wildcards we picked up were from | |
7837 | * action generation. Without options on the incoming packet, wildcards | |
7838 | * aren't meaningful. To avoid them possibly getting misinterpreted, | |
7839 | * just clear everything. */ | |
7840 | if (ctx.wc->masks.tunnel.metadata.present.map) { | |
7841 | memset(&ctx.wc->masks.tunnel.metadata, 0, | |
7842 | sizeof ctx.wc->masks.tunnel.metadata); | |
7843 | } else { | |
7844 | ctx.wc->masks.tunnel.metadata.tab = NULL; | |
7845 | } | |
7846 | } | |
7847 | ||
c0e638aa | 7848 | xlate_wc_finish(&ctx); |
1520ef4f BP |
7849 | |
7850 | exit: | |
8d8ab6c2 JG |
7851 | /* Reset the table to what it was when we came in. If we only fetched |
7852 | * it locally, then it has no meaning outside of flow translation. */ | |
7853 | flow->tunnel.metadata.tab = xin->upcall_flow->tunnel.metadata.tab; | |
7854 | ||
1520ef4f BP |
7855 | ofpbuf_uninit(&ctx.stack); |
7856 | ofpbuf_uninit(&ctx.action_set); | |
1d361a81 | 7857 | ofpbuf_uninit(&ctx.frozen_actions); |
1520ef4f | 7858 | ofpbuf_uninit(&scratch_actions); |
1fc11c59 | 7859 | ofpbuf_delete(ctx.encap_data); |
fff1b9c0 JR |
7860 | |
7861 | /* Make sure we return a "drop flow" in case of an error. */ | |
7862 | if (ctx.error) { | |
7863 | xout->slow = 0; | |
7864 | if (xin->odp_actions) { | |
7865 | ofpbuf_clear(xin->odp_actions); | |
7866 | } | |
7867 | } | |
a13a0209 AT |
7868 | |
7869 | /* Install drop action if datapath supports explicit drop action. */ | |
7870 | if (xin->odp_actions && !xin->odp_actions->size && | |
7871 | ovs_explicit_drop_action_supported(ctx.xbridge->ofproto)) { | |
7872 | put_drop_action(xin->odp_actions, ctx.error); | |
7873 | } | |
7874 | ||
7875 | /* Since congestion drop and forwarding drop are not exactly | |
7876 | * translation error, we are resetting the translation error. | |
7877 | */ | |
7878 | if (ctx.error == XLATE_CONGESTION_DROP || | |
7879 | ctx.error == XLATE_FORWARDING_DISABLED) { | |
7880 | ctx.error = XLATE_OK; | |
7881 | } | |
7882 | ||
fff1b9c0 | 7883 | return ctx.error; |
91d6cd12 AW |
7884 | } |
7885 | ||
77ab5fd2 BP |
7886 | enum ofperr |
7887 | xlate_resume(struct ofproto_dpif *ofproto, | |
7888 | const struct ofputil_packet_in_private *pin, | |
7889 | struct ofpbuf *odp_actions, | |
2f355bff YHW |
7890 | enum slow_path_reason *slow, |
7891 | struct flow *flow, | |
7892 | struct xlate_cache *xcache) | |
77ab5fd2 BP |
7893 | { |
7894 | struct dp_packet packet; | |
4d617a87 BP |
7895 | dp_packet_use_const(&packet, pin->base.packet, |
7896 | pin->base.packet_len); | |
77ab5fd2 | 7897 | |
07706c50 | 7898 | pkt_metadata_from_flow(&packet.md, &pin->base.flow_metadata.flow); |
2f355bff | 7899 | flow_extract(&packet, flow); |
77ab5fd2 BP |
7900 | |
7901 | struct xlate_in xin; | |
1f4a8933 | 7902 | xlate_in_init(&xin, ofproto, ofproto_dpif_get_tables_version(ofproto), |
2f355bff | 7903 | flow, 0, NULL, ntohs(flow->tcp_flags), |
77ab5fd2 | 7904 | &packet, NULL, odp_actions); |
2f355bff | 7905 | xin.xcache = xcache; |
77ab5fd2 BP |
7906 | |
7907 | struct ofpact_note noop; | |
7908 | ofpact_init_NOTE(&noop); | |
7909 | noop.length = 0; | |
7910 | ||
7911 | bool any_actions = pin->actions_len > 0; | |
7912 | struct frozen_state state = { | |
7913 | .table_id = 0, /* Not the table where NXAST_PAUSE was executed. */ | |
7914 | .ofproto_uuid = pin->bridge, | |
7915 | .stack = pin->stack, | |
84cf3c1f | 7916 | .stack_size = pin->stack_size, |
77ab5fd2 BP |
7917 | .mirrors = pin->mirrors, |
7918 | .conntracked = pin->conntracked, | |
00135b86 | 7919 | .xport_uuid = UUID_ZERO, |
77ab5fd2 BP |
7920 | |
7921 | /* When there are no actions, xlate_actions() will search the flow | |
7922 | * table. We don't want it to do that (we want it to resume), so | |
7923 | * supply a no-op action if there aren't any. | |
7924 | * | |
7925 | * (We can't necessarily avoid translating actions entirely if there | |
7926 | * aren't any actions, because there might be some finishing-up to do | |
7927 | * at the end of the pipeline, and we don't check for those | |
7928 | * conditions.) */ | |
7929 | .ofpacts = any_actions ? pin->actions : &noop.ofpact, | |
7930 | .ofpacts_len = any_actions ? pin->actions_len : sizeof noop, | |
7931 | ||
7932 | .action_set = pin->action_set, | |
7933 | .action_set_len = pin->action_set_len, | |
7934 | }; | |
7935 | frozen_metadata_from_flow(&state.metadata, | |
4d617a87 | 7936 | &pin->base.flow_metadata.flow); |
77ab5fd2 BP |
7937 | xin.frozen_state = &state; |
7938 | ||
7939 | struct xlate_out xout; | |
7940 | enum xlate_error error = xlate_actions(&xin, &xout); | |
7941 | *slow = xout.slow; | |
7942 | xlate_out_uninit(&xout); | |
7943 | ||
7944 | /* xlate_actions() can generate a number of errors, but only | |
7945 | * XLATE_BRIDGE_NOT_FOUND really stands out to me as one that we should be | |
7946 | * sure to report over OpenFlow. The others could come up in packet-outs | |
7947 | * or regular flow translation and I don't think that it's going to be too | |
7948 | * useful to report them to the controller. */ | |
7949 | return error == XLATE_BRIDGE_NOT_FOUND ? OFPERR_NXR_STALE : 0; | |
7950 | } | |
7951 | ||
2eb79142 JG |
7952 | /* Sends 'packet' out 'ofport'. If 'port' is a tunnel and that tunnel type |
7953 | * supports a notion of an OAM flag, sets it if 'oam' is true. | |
91d6cd12 AW |
7954 | * May modify 'packet'. |
7955 | * Returns 0 if successful, otherwise a positive errno value. */ | |
7956 | int | |
2eb79142 JG |
7957 | xlate_send_packet(const struct ofport_dpif *ofport, bool oam, |
7958 | struct dp_packet *packet) | |
91d6cd12 | 7959 | { |
84f0f298 | 7960 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
91d6cd12 | 7961 | struct xport *xport; |
2eb79142 JG |
7962 | uint64_t ofpacts_stub[1024 / 8]; |
7963 | struct ofpbuf ofpacts; | |
91d6cd12 | 7964 | struct flow flow; |
91d6cd12 | 7965 | |
2eb79142 | 7966 | ofpbuf_use_stack(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); |
91d6cd12 | 7967 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ |
cf62fa4c | 7968 | flow_extract(packet, &flow); |
b5e7e61a | 7969 | flow.in_port.ofp_port = OFPP_NONE; |
91d6cd12 | 7970 | |
84f0f298 | 7971 | xport = xport_lookup(xcfg, ofport); |
91d6cd12 | 7972 | if (!xport) { |
02ea2703 | 7973 | return EINVAL; |
91d6cd12 | 7974 | } |
2eb79142 JG |
7975 | |
7976 | if (oam) { | |
71f21279 BP |
7977 | const ovs_be16 flag = htons(NX_TUN_FLAG_OAM); |
7978 | ofpact_put_set_field(&ofpacts, mf_from_id(MFF_TUN_FLAGS), | |
7979 | &flag, &flag); | |
2eb79142 JG |
7980 | } |
7981 | ||
7982 | ofpact_put_OUTPUT(&ofpacts)->port = xport->ofp_port; | |
e491a67a | 7983 | |
1f4a8933 JR |
7984 | /* Actions here are not referring to anything versionable (flow tables or |
7985 | * groups) so we don't need to worry about the version here. */ | |
7986 | return ofproto_dpif_execute_actions(xport->xbridge->ofproto, | |
7987 | OVS_VERSION_MAX, &flow, NULL, | |
2eb79142 | 7988 | ofpacts.data, ofpacts.size, packet); |
9583bc14 | 7989 | } |
b256dc52 | 7990 | |
901a517e | 7991 | void |
064799a1 JR |
7992 | xlate_mac_learning_update(const struct ofproto_dpif *ofproto, |
7993 | ofp_port_t in_port, struct eth_addr dl_src, | |
7994 | int vlan, bool is_grat_arp) | |
b256dc52 | 7995 | { |
84f0f298 | 7996 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
b256dc52 JS |
7997 | struct xbridge *xbridge; |
7998 | struct xbundle *xbundle; | |
b256dc52 | 7999 | |
84f0f298 | 8000 | xbridge = xbridge_lookup(xcfg, ofproto); |
b256dc52 JS |
8001 | if (!xbridge) { |
8002 | return; | |
8003 | } | |
8004 | ||
2d9b49dd | 8005 | xbundle = lookup_input_bundle__(xbridge, in_port, NULL); |
b256dc52 JS |
8006 | if (!xbundle) { |
8007 | return; | |
8008 | } | |
8009 | ||
2d9b49dd | 8010 | update_learning_table__(xbridge, xbundle, dl_src, vlan, is_grat_arp); |
b256dc52 | 8011 | } |
bef503e8 | 8012 | |
88186383 AZ |
8013 | void |
8014 | xlate_set_support(const struct ofproto_dpif *ofproto, | |
8015 | const struct dpif_backer_support *support) | |
8016 | { | |
8017 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
8018 | struct xbridge *xbridge = xbridge_lookup(xcfg, ofproto); | |
8019 | ||
8020 | if (xbridge) { | |
8021 | xbridge->support = *support; | |
8022 | } | |
8023 | } |