]>
Commit | Line | Data |
---|---|---|
9bfe9334 | 1 | /* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016 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> |
a36de779 PS |
20 | #include <arpa/inet.h> |
21 | #include <net/if.h> | |
22 | #include <sys/socket.h> | |
23 | #include <netinet/in.h> | |
8449c4d6 | 24 | |
db7d4e46 | 25 | #include "bfd.h" |
9583bc14 EJ |
26 | #include "bitmap.h" |
27 | #include "bond.h" | |
28 | #include "bundle.h" | |
29 | #include "byte-order.h" | |
db7d4e46 | 30 | #include "cfm.h" |
9583bc14 EJ |
31 | #include "connmgr.h" |
32 | #include "coverage.h" | |
e14deea0 | 33 | #include "dp-packet.h" |
9583bc14 | 34 | #include "dpif.h" |
f7f1ea29 | 35 | #include "in-band.h" |
db7d4e46 | 36 | #include "lacp.h" |
9583bc14 EJ |
37 | #include "learn.h" |
38 | #include "mac-learning.h" | |
6d95c4e8 | 39 | #include "mcast-snooping.h" |
9583bc14 EJ |
40 | #include "multipath.h" |
41 | #include "netdev-vport.h" | |
42 | #include "netlink.h" | |
43 | #include "nx-match.h" | |
44 | #include "odp-execute.h" | |
9583bc14 | 45 | #include "ofproto/ofproto-dpif-ipfix.h" |
ec7ceaed | 46 | #include "ofproto/ofproto-dpif-mirror.h" |
60d02c72 | 47 | #include "ofproto/ofproto-dpif-monitor.h" |
9583bc14 EJ |
48 | #include "ofproto/ofproto-dpif-sflow.h" |
49 | #include "ofproto/ofproto-dpif.h" | |
6f00e29b | 50 | #include "ofproto/ofproto-provider.h" |
b598f214 BW |
51 | #include "openvswitch/dynamic-string.h" |
52 | #include "openvswitch/meta-flow.h" | |
53 | #include "openvswitch/list.h" | |
54 | #include "openvswitch/ofp-actions.h" | |
55 | #include "openvswitch/vlog.h" | |
56 | #include "ovs-lldp.h" | |
a36de779 | 57 | #include "ovs-router.h" |
b598f214 BW |
58 | #include "packets.h" |
59 | #include "tnl-neigh-cache.h" | |
a36de779 | 60 | #include "tnl-ports.h" |
9583bc14 | 61 | #include "tunnel.h" |
9583bc14 | 62 | |
46c88433 | 63 | COVERAGE_DEFINE(xlate_actions); |
0f032e95 | 64 | COVERAGE_DEFINE(xlate_actions_oversize); |
7d031d7e | 65 | COVERAGE_DEFINE(xlate_actions_too_many_output); |
9583bc14 EJ |
66 | |
67 | VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate); | |
68 | ||
8a553e9a EJ |
69 | /* Maximum depth of flow table recursion (due to resubmit actions) in a |
70 | * flow translation. */ | |
71 | #define MAX_RESUBMIT_RECURSION 64 | |
adcf00ba AZ |
72 | #define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in |
73 | internal table. */ | |
8a553e9a | 74 | |
98b07853 BP |
75 | /* Maximum number of resubmit actions in a flow translation, whether they are |
76 | * recursive or not. */ | |
77 | #define MAX_RESUBMITS (MAX_RESUBMIT_RECURSION * MAX_RESUBMIT_RECURSION) | |
78 | ||
46c88433 EJ |
79 | struct xbridge { |
80 | struct hmap_node hmap_node; /* Node in global 'xbridges' map. */ | |
81 | struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */ | |
82 | ||
ca6ba700 | 83 | struct ovs_list xbundles; /* Owned xbundles. */ |
46c88433 EJ |
84 | struct hmap xports; /* Indexed by ofp_port. */ |
85 | ||
86 | char *name; /* Name used in log messages. */ | |
89a8a7f0 | 87 | struct dpif *dpif; /* Datapath interface. */ |
46c88433 | 88 | struct mac_learning *ml; /* Mac learning handle. */ |
6d95c4e8 | 89 | struct mcast_snooping *ms; /* Multicast Snooping handle. */ |
46c88433 EJ |
90 | struct mbridge *mbridge; /* Mirroring. */ |
91 | struct dpif_sflow *sflow; /* SFlow handle, or null. */ | |
92 | struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */ | |
ce3955be | 93 | struct netflow *netflow; /* Netflow handle, or null. */ |
9d189a50 | 94 | struct stp *stp; /* STP or null if disabled. */ |
9efd308e | 95 | struct rstp *rstp; /* RSTP or null if disabled. */ |
46c88433 | 96 | |
46c88433 EJ |
97 | bool has_in_band; /* Bridge has in band control? */ |
98 | bool forward_bpdu; /* Bridge forwards STP BPDUs? */ | |
4b97b70d | 99 | |
b440dd8c JS |
100 | /* Datapath feature support. */ |
101 | struct dpif_backer_support support; | |
46c88433 EJ |
102 | }; |
103 | ||
104 | struct xbundle { | |
105 | struct hmap_node hmap_node; /* In global 'xbundles' map. */ | |
106 | struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */ | |
107 | ||
ca6ba700 | 108 | struct ovs_list list_node; /* In parent 'xbridges' list. */ |
46c88433 EJ |
109 | struct xbridge *xbridge; /* Parent xbridge. */ |
110 | ||
ca6ba700 | 111 | struct ovs_list xports; /* Contains "struct xport"s. */ |
46c88433 EJ |
112 | |
113 | char *name; /* Name used in log messages. */ | |
114 | struct bond *bond; /* Nonnull iff more than one port. */ | |
115 | struct lacp *lacp; /* LACP handle or null. */ | |
116 | ||
117 | enum port_vlan_mode vlan_mode; /* VLAN mode. */ | |
118 | int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */ | |
119 | unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1. | |
120 | * NULL if all VLANs are trunked. */ | |
121 | bool use_priority_tags; /* Use 802.1p tag for frames in VLAN 0? */ | |
122 | bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */ | |
123 | }; | |
124 | ||
125 | struct xport { | |
126 | struct hmap_node hmap_node; /* Node in global 'xports' map. */ | |
127 | struct ofport_dpif *ofport; /* Key in global 'xports map. */ | |
128 | ||
129 | struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */ | |
130 | ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */ | |
131 | ||
132 | odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */ | |
133 | ||
ca6ba700 | 134 | struct ovs_list bundle_node; /* In parent xbundle (if it exists). */ |
46c88433 EJ |
135 | struct xbundle *xbundle; /* Parent xbundle or null. */ |
136 | ||
137 | struct netdev *netdev; /* 'ofport''s netdev. */ | |
138 | ||
139 | struct xbridge *xbridge; /* Parent bridge. */ | |
140 | struct xport *peer; /* Patch port peer or null. */ | |
141 | ||
142 | enum ofputil_port_config config; /* OpenFlow port configuration. */ | |
dd8cd4b4 | 143 | enum ofputil_port_state state; /* OpenFlow port state. */ |
92cf817b | 144 | int stp_port_no; /* STP port number or -1 if not in use. */ |
f025bcb7 | 145 | struct rstp_port *rstp_port; /* RSTP port or null. */ |
46c88433 | 146 | |
55954f6e EJ |
147 | struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */ |
148 | ||
46c88433 EJ |
149 | bool may_enable; /* May be enabled in bonds. */ |
150 | bool is_tunnel; /* Is a tunnel port. */ | |
151 | ||
152 | struct cfm *cfm; /* CFM handle or null. */ | |
153 | struct bfd *bfd; /* BFD handle or null. */ | |
0477baa9 | 154 | struct lldp *lldp; /* LLDP handle or null. */ |
46c88433 EJ |
155 | }; |
156 | ||
4d0acc70 EJ |
157 | struct xlate_ctx { |
158 | struct xlate_in *xin; | |
159 | struct xlate_out *xout; | |
160 | ||
46c88433 | 161 | const struct xbridge *xbridge; |
4d0acc70 | 162 | |
621b8064 | 163 | /* Flow tables version at the beginning of the translation. */ |
18721c4a | 164 | cls_version_t tables_version; |
621b8064 | 165 | |
4d0acc70 EJ |
166 | /* Flow at the last commit. */ |
167 | struct flow base_flow; | |
168 | ||
169 | /* Tunnel IP destination address as received. This is stored separately | |
170 | * as the base_flow.tunnel is cleared on init to reflect the datapath | |
171 | * behavior. Used to make sure not to send tunneled output to ourselves, | |
172 | * which might lead to an infinite loop. This could happen easily | |
173 | * if a tunnel is marked as 'ip_remote=flow', and the flow does not | |
174 | * actually set the tun_dst field. */ | |
e4d3706c | 175 | struct in6_addr orig_tunnel_ipv6_dst; |
4d0acc70 EJ |
176 | |
177 | /* Stack for the push and pop actions. Each stack element is of type | |
178 | * "union mf_subvalue". */ | |
4d0acc70 EJ |
179 | struct ofpbuf stack; |
180 | ||
181 | /* The rule that we are currently translating, or NULL. */ | |
182 | struct rule_dpif *rule; | |
183 | ||
49a73e0c BP |
184 | /* Flow translation populates this with wildcards relevant in translation. |
185 | * When 'xin->wc' is nonnull, this is the same pointer. When 'xin->wc' is | |
186 | * null, this is a pointer to uninitialized scratch memory. This allows | |
187 | * code to blindly write to 'ctx->wc' without worrying about whether the | |
188 | * caller really wants wildcards. */ | |
189 | struct flow_wildcards *wc; | |
190 | ||
1520ef4f BP |
191 | /* Output buffer for datapath actions. When 'xin->odp_actions' is nonnull, |
192 | * this is the same pointer. When 'xin->odp_actions' is null, this points | |
193 | * to a scratch ofpbuf. This allows code to add actions to | |
194 | * 'ctx->odp_actions' without worrying about whether the caller really | |
195 | * wants actions. */ | |
196 | struct ofpbuf *odp_actions; | |
197 | ||
98b07853 | 198 | /* Resubmit statistics, via xlate_table_action(). */ |
9a9b001a | 199 | int indentation; /* Current resubmit nesting depth. */ |
98b07853 | 200 | int resubmits; /* Total number of resubmits. */ |
5a070238 | 201 | bool in_group; /* Currently translating ofgroup, if true. */ |
029ca940 | 202 | bool in_action_set; /* Currently translating action_set, if true. */ |
98b07853 | 203 | |
4d0acc70 | 204 | uint8_t table_id; /* OpenFlow table ID where flow was found. */ |
8b1e5560 JR |
205 | ovs_be64 rule_cookie; /* Cookie of the rule being translated. */ |
206 | uint32_t orig_skb_priority; /* Priority when packet arrived. */ | |
4d0acc70 | 207 | uint32_t sflow_n_outputs; /* Number of output ports. */ |
4e022ec0 | 208 | odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */ |
2031ef97 | 209 | ofp_port_t nf_output_iface; /* Output interface index for NetFlow. */ |
4d0acc70 | 210 | bool exit; /* No further actions should be processed. */ |
3d6151f3 | 211 | mirror_mask_t mirrors; /* Bitmap of associated mirrors. */ |
7fdb60a7 | 212 | |
1d361a81 BP |
213 | /* Freezing Translation |
214 | * ==================== | |
e672ff9b | 215 | * |
1d361a81 BP |
216 | * At some point during translation, the code may recognize the need to halt |
217 | * and checkpoint the translation in a way that it can be restarted again | |
218 | * later. We call the checkpointing process "freezing" and the restarting | |
219 | * process "thawing". | |
e672ff9b | 220 | * |
1d361a81 | 221 | * The use cases for freezing are: |
e672ff9b | 222 | * |
1d361a81 BP |
223 | * - "Recirculation", where the translation process discovers that it |
224 | * doesn't have enough information to complete translation without | |
225 | * actually executing the actions that have already been translated, | |
226 | * which provides the additionally needed information. In these | |
227 | * situations, translation freezes translation and assigns the frozen | |
228 | * data a unique "recirculation ID", which it associates with the data | |
229 | * in a table in userspace (see ofproto-dpif-rid.h). It also adds a | |
230 | * OVS_ACTION_ATTR_RECIRC action specifying that ID to the datapath | |
231 | * actions. When a packet hits that action, the datapath looks its | |
232 | * flow up again using the ID. If there's a miss, it comes back to | |
233 | * userspace, which find the recirculation table entry for the ID, | |
234 | * thaws the associated frozen data, and continues translation from | |
235 | * that point given the additional information that is now known. | |
e672ff9b | 236 | * |
1d361a81 BP |
237 | * The archetypal example is MPLS. As MPLS is implemented in |
238 | * OpenFlow, the protocol that follows the last MPLS label becomes | |
239 | * known only when that label is popped by an OpenFlow action. That | |
240 | * means that Open vSwitch can't extract the headers beyond the MPLS | |
241 | * labels until the pop action is executed. Thus, at that point | |
242 | * translation uses the recirculation process to extract the headers | |
243 | * beyond the MPLS labels. | |
e672ff9b | 244 | * |
1d361a81 BP |
245 | * (OVS also uses OVS_ACTION_ATTR_RECIRC to implement hashing for |
246 | * output to bonds. OVS pre-populates all the datapath flows for bond | |
247 | * output in the datapath, though, which means that the elaborate | |
248 | * process of coming back to userspace for a second round of | |
249 | * translation isn't needed, and so bonds don't follow the above | |
250 | * process.) | |
e672ff9b | 251 | * |
77ab5fd2 BP |
252 | * - "Continuation". A continuation is a way for an OpenFlow controller |
253 | * to interpose on a packet's traversal of the OpenFlow tables. When | |
254 | * the translation process encounters a "controller" action with the | |
255 | * "pause" flag, it freezes translation, serializes the frozen data, | |
256 | * and sends it to an OpenFlow controller. The controller then | |
257 | * examines and possibly modifies the frozen data and eventually sends | |
258 | * it back to the switch, which thaws it and continues translation. | |
e672ff9b | 259 | * |
1d361a81 BP |
260 | * The main problem of freezing translation is preserving state, so that |
261 | * when the translation is thawed later it resumes from where it left off, | |
262 | * without disruption. In particular, actions must be preserved as follows: | |
263 | * | |
264 | * - If we're freezing because an action needed more information, the | |
265 | * action that prompted it. | |
266 | * | |
267 | * - Any actions remaining to be translated within the current flow. | |
268 | * | |
269 | * - If translation was frozen within a NXAST_RESUBMIT, then any actions | |
270 | * following the resubmit action. Resubmit actions can be nested, so | |
271 | * this has to go all the way up the control stack. | |
e672ff9b JR |
272 | * |
273 | * - The OpenFlow 1.1+ action set. | |
274 | * | |
275 | * State that actions and flow table lookups can depend on, such as the | |
276 | * following, must also be preserved: | |
277 | * | |
278 | * - Metadata fields (input port, registers, OF1.1+ metadata, ...). | |
279 | * | |
1d361a81 | 280 | * - The stack used by NXAST_STACK_PUSH and NXAST_STACK_POP actions. |
e672ff9b JR |
281 | * |
282 | * - The table ID and cookie of the flow being translated at each level | |
1d361a81 BP |
283 | * of the control stack, because these can become visible through |
284 | * OFPAT_CONTROLLER actions (and other ways). | |
e672ff9b JR |
285 | * |
286 | * Translation allows for the control of this state preservation via these | |
1d361a81 BP |
287 | * members. When a need to freeze translation is identified, the |
288 | * translation process: | |
e672ff9b | 289 | * |
1d361a81 | 290 | * 1. Sets 'freezing' to true. |
e672ff9b JR |
291 | * |
292 | * 2. Sets 'exit' to true to tell later steps that we're exiting from the | |
293 | * translation process. | |
294 | * | |
1d361a81 BP |
295 | * 3. Adds an OFPACT_UNROLL_XLATE action to 'frozen_actions', and points |
296 | * frozen_actions.header to the action to make it easy to find it later. | |
297 | * This action holds the current table ID and cookie so that they can be | |
298 | * restored during a post-recirculation upcall translation. | |
e672ff9b JR |
299 | * |
300 | * 4. Adds the action that prompted recirculation and any actions following | |
1d361a81 | 301 | * it within the same flow to 'frozen_actions', so that they can be |
8a5fb3b4 | 302 | * executed during a post-recirculation upcall translation. |
e672ff9b JR |
303 | * |
304 | * 5. Returns. | |
305 | * | |
306 | * 6. The action that prompted recirculation might be nested in a stack of | |
307 | * nested "resubmit"s that have actions remaining. Each of these notices | |
1d361a81 BP |
308 | * that we're exiting and freezing and responds by adding more |
309 | * OFPACT_UNROLL_XLATE actions to 'frozen_actions', as necessary, | |
310 | * followed by any actions that were yet unprocessed. | |
e672ff9b | 311 | * |
1d361a81 BP |
312 | * If we're freezing because of recirculation, the caller generates a |
313 | * recirculation ID and associates all the state produced by this process | |
314 | * with it. For post-recirculation upcall translation, the caller passes it | |
315 | * back in for the new translation to execute. The process yielded a set of | |
316 | * ofpacts that can be translated directly, so it is not much of a special | |
317 | * case at that point. | |
e672ff9b | 318 | */ |
1d361a81 BP |
319 | bool freezing; |
320 | struct ofpbuf frozen_actions; | |
77ab5fd2 | 321 | const struct ofpact_controller *pause; |
e672ff9b | 322 | |
07659514 JS |
323 | /* True if conntrack has been performed on this packet during processing |
324 | * on the current bridge. This is used to determine whether conntrack | |
1d361a81 | 325 | * state from the datapath should be honored after thawing. */ |
07659514 JS |
326 | bool conntracked; |
327 | ||
9ac0aada JR |
328 | /* Pointer to an embedded NAT action in a conntrack action, or NULL. */ |
329 | struct ofpact_nat *ct_nat_action; | |
330 | ||
7fdb60a7 SH |
331 | /* OpenFlow 1.1+ action set. |
332 | * | |
333 | * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS. | |
334 | * When translation is otherwise complete, ofpacts_execute_action_set() | |
335 | * converts it to a set of "struct ofpact"s that can be translated into | |
ed9c9e3e | 336 | * datapath actions. */ |
c61f3870 | 337 | bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ |
7fdb60a7 | 338 | struct ofpbuf action_set; /* Action set. */ |
fff1b9c0 JR |
339 | |
340 | enum xlate_error error; /* Translation failed. */ | |
4d0acc70 EJ |
341 | }; |
342 | ||
fff1b9c0 JR |
343 | const char *xlate_strerror(enum xlate_error error) |
344 | { | |
345 | switch (error) { | |
346 | case XLATE_OK: | |
347 | return "OK"; | |
348 | case XLATE_BRIDGE_NOT_FOUND: | |
349 | return "Bridge not found"; | |
350 | case XLATE_RECURSION_TOO_DEEP: | |
351 | return "Recursion too deep"; | |
352 | case XLATE_TOO_MANY_RESUBMITS: | |
353 | return "Too many resubmits"; | |
354 | case XLATE_STACK_TOO_DEEP: | |
355 | return "Stack too deep"; | |
356 | case XLATE_NO_RECIRCULATION_CONTEXT: | |
357 | return "No recirculation context"; | |
358 | case XLATE_RECIRCULATION_CONFLICT: | |
359 | return "Recirculation conflict"; | |
360 | case XLATE_TOO_MANY_MPLS_LABELS: | |
361 | return "Too many MPLS labels"; | |
362 | } | |
363 | return "Unknown error"; | |
364 | } | |
365 | ||
ed9c9e3e | 366 | static void xlate_action_set(struct xlate_ctx *ctx); |
704bb0bf | 367 | static void xlate_commit_actions(struct xlate_ctx *ctx); |
ed9c9e3e | 368 | |
1d741d6d | 369 | static void |
1d361a81 | 370 | ctx_trigger_freeze(struct xlate_ctx *ctx) |
1d741d6d JR |
371 | { |
372 | ctx->exit = true; | |
1d361a81 | 373 | ctx->freezing = true; |
1d741d6d JR |
374 | } |
375 | ||
376 | static bool | |
1d361a81 | 377 | ctx_first_frozen_action(const struct xlate_ctx *ctx) |
1d741d6d | 378 | { |
1d361a81 | 379 | return !ctx->frozen_actions.size; |
e672ff9b JR |
380 | } |
381 | ||
3293cb85 | 382 | static void |
1d361a81 | 383 | ctx_cancel_freeze(struct xlate_ctx *ctx) |
3293cb85 | 384 | { |
1d361a81 BP |
385 | if (ctx->freezing) { |
386 | ctx->freezing = false; | |
387 | ofpbuf_clear(&ctx->frozen_actions); | |
388 | ctx->frozen_actions.header = NULL; | |
3293cb85 BP |
389 | } |
390 | } | |
391 | ||
77ab5fd2 | 392 | static void finish_freezing(struct xlate_ctx *ctx); |
e672ff9b | 393 | |
9583bc14 EJ |
394 | /* A controller may use OFPP_NONE as the ingress port to indicate that |
395 | * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for | |
396 | * when an input bundle is needed for validation (e.g., mirroring or | |
397 | * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have | |
3548d242 BP |
398 | * any 'port' structs, so care must be taken when dealing with it. */ |
399 | static struct xbundle ofpp_none_bundle = { | |
400 | .name = "OFPP_NONE", | |
401 | .vlan_mode = PORT_VLAN_TRUNK | |
402 | }; | |
9583bc14 | 403 | |
55954f6e EJ |
404 | /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from |
405 | * 'priority' (the datapath's term for QoS queue) to the dscp bits which all | |
406 | * traffic egressing the 'ofport' with that priority should be marked with. */ | |
407 | struct skb_priority_to_dscp { | |
408 | struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */ | |
409 | uint32_t skb_priority; /* Priority of this queue (see struct flow). */ | |
410 | ||
411 | uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */ | |
412 | }; | |
413 | ||
b256dc52 JS |
414 | enum xc_type { |
415 | XC_RULE, | |
416 | XC_BOND, | |
417 | XC_NETDEV, | |
418 | XC_NETFLOW, | |
419 | XC_MIRROR, | |
420 | XC_LEARN, | |
421 | XC_NORMAL, | |
422 | XC_FIN_TIMEOUT, | |
1e684d7d | 423 | XC_GROUP, |
53902038 | 424 | XC_TNL_NEIGH, |
b256dc52 JS |
425 | }; |
426 | ||
427 | /* xlate_cache entries hold enough information to perform the side effects of | |
428 | * xlate_actions() for a rule, without needing to perform rule translation | |
429 | * from scratch. The primary usage of these is to submit statistics to objects | |
430 | * that a flow relates to, although they may be used for other effects as well | |
431 | * (for instance, refreshing hard timeouts for learned flows). */ | |
432 | struct xc_entry { | |
433 | enum xc_type type; | |
434 | union { | |
435 | struct rule_dpif *rule; | |
436 | struct { | |
437 | struct netdev *tx; | |
438 | struct netdev *rx; | |
439 | struct bfd *bfd; | |
440 | } dev; | |
441 | struct { | |
442 | struct netflow *netflow; | |
443 | struct flow *flow; | |
444 | ofp_port_t iface; | |
445 | } nf; | |
446 | struct { | |
447 | struct mbridge *mbridge; | |
448 | mirror_mask_t mirrors; | |
449 | } mirror; | |
450 | struct { | |
451 | struct bond *bond; | |
452 | struct flow *flow; | |
453 | uint16_t vid; | |
454 | } bond; | |
455 | struct { | |
4165b5e0 JS |
456 | struct ofproto_dpif *ofproto; |
457 | struct ofputil_flow_mod *fm; | |
458 | struct ofpbuf *ofpacts; | |
b256dc52 JS |
459 | } learn; |
460 | struct { | |
461 | struct ofproto_dpif *ofproto; | |
462 | struct flow *flow; | |
463 | int vlan; | |
464 | } normal; | |
465 | struct { | |
466 | struct rule_dpif *rule; | |
467 | uint16_t idle; | |
468 | uint16_t hard; | |
469 | } fin; | |
1e684d7d RW |
470 | struct { |
471 | struct group_dpif *group; | |
472 | struct ofputil_bucket *bucket; | |
473 | } group; | |
a36de779 PS |
474 | struct { |
475 | char br_name[IFNAMSIZ]; | |
0b411137 | 476 | struct in6_addr d_ipv6; |
53902038 | 477 | } tnl_neigh_cache; |
b256dc52 JS |
478 | } u; |
479 | }; | |
480 | ||
6c38bdc0 HH |
481 | #define XC_ENTRY_FOR_EACH(ENTRY, ENTRIES, XCACHE) \ |
482 | ENTRIES = XCACHE->entries; \ | |
483 | for (ENTRY = ofpbuf_try_pull(&ENTRIES, sizeof *ENTRY); \ | |
484 | ENTRY; \ | |
485 | ENTRY = ofpbuf_try_pull(&ENTRIES, sizeof *ENTRY)) | |
b256dc52 JS |
486 | |
487 | struct xlate_cache { | |
488 | struct ofpbuf entries; | |
489 | }; | |
490 | ||
84f0f298 RW |
491 | /* Xlate config contains hash maps of all bridges, bundles and ports. |
492 | * Xcfgp contains the pointer to the current xlate configuration. | |
493 | * When the main thread needs to change the configuration, it copies xcfgp to | |
494 | * new_xcfg and edits new_xcfg. This enables the use of RCU locking which | |
495 | * does not block handler and revalidator threads. */ | |
496 | struct xlate_cfg { | |
497 | struct hmap xbridges; | |
498 | struct hmap xbundles; | |
499 | struct hmap xports; | |
500 | }; | |
b1b72f2d | 501 | static OVSRCU_TYPE(struct xlate_cfg *) xcfgp = OVSRCU_INITIALIZER(NULL); |
f439f23b | 502 | static struct xlate_cfg *new_xcfg = NULL; |
46c88433 EJ |
503 | |
504 | static bool may_receive(const struct xport *, struct xlate_ctx *); | |
9583bc14 EJ |
505 | static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
506 | struct xlate_ctx *); | |
adcf00ba | 507 | static void xlate_normal(struct xlate_ctx *); |
c1b3756c BP |
508 | static inline void xlate_report(struct xlate_ctx *, const char *, ...) |
509 | OVS_PRINTF_FORMAT(2, 3); | |
6d328fa2 SH |
510 | static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port, |
511 | uint8_t table_id, bool may_packet_in, | |
512 | bool honor_table_miss); | |
46c88433 EJ |
513 | static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn); |
514 | static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid); | |
515 | static void output_normal(struct xlate_ctx *, const struct xbundle *, | |
9583bc14 | 516 | uint16_t vlan); |
e93ef1c7 JR |
517 | |
518 | /* Optional bond recirculation parameter to compose_output_action(). */ | |
519 | struct xlate_bond_recirc { | |
520 | uint32_t recirc_id; /* !0 Use recirculation instead of output. */ | |
521 | uint8_t hash_alg; /* !0 Compute hash for recirc before. */ | |
522 | uint32_t hash_basis; /* Compute hash for recirc before. */ | |
523 | }; | |
524 | ||
525 | static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port, | |
526 | const struct xlate_bond_recirc *xr); | |
9583bc14 | 527 | |
84f0f298 RW |
528 | static struct xbridge *xbridge_lookup(struct xlate_cfg *, |
529 | const struct ofproto_dpif *); | |
290835f9 BP |
530 | static struct xbridge *xbridge_lookup_by_uuid(struct xlate_cfg *, |
531 | const struct uuid *); | |
84f0f298 RW |
532 | static struct xbundle *xbundle_lookup(struct xlate_cfg *, |
533 | const struct ofbundle *); | |
534 | static struct xport *xport_lookup(struct xlate_cfg *, | |
535 | const struct ofport_dpif *); | |
46c88433 | 536 | static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port); |
55954f6e EJ |
537 | static struct skb_priority_to_dscp *get_skb_priority(const struct xport *, |
538 | uint32_t skb_priority); | |
539 | static void clear_skb_priorities(struct xport *); | |
16194afd | 540 | static size_t count_skb_priorities(const struct xport *); |
55954f6e EJ |
541 | static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, |
542 | uint8_t *dscp); | |
46c88433 | 543 | |
b256dc52 JS |
544 | static struct xc_entry *xlate_cache_add_entry(struct xlate_cache *xc, |
545 | enum xc_type type); | |
84f0f298 RW |
546 | static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *); |
547 | static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *); | |
548 | static void xlate_xport_init(struct xlate_cfg *, struct xport *); | |
9efd308e | 549 | static void xlate_xbridge_set(struct xbridge *, struct dpif *, |
9efd308e DV |
550 | const struct mac_learning *, struct stp *, |
551 | struct rstp *, const struct mcast_snooping *, | |
552 | const struct mbridge *, | |
553 | const struct dpif_sflow *, | |
554 | const struct dpif_ipfix *, | |
2f47cdf4 | 555 | const struct netflow *, |
84f0f298 | 556 | bool forward_bpdu, bool has_in_band, |
b440dd8c | 557 | const struct dpif_backer_support *); |
84f0f298 RW |
558 | static void xlate_xbundle_set(struct xbundle *xbundle, |
559 | enum port_vlan_mode vlan_mode, int vlan, | |
560 | unsigned long *trunks, bool use_priority_tags, | |
561 | const struct bond *bond, const struct lacp *lacp, | |
562 | bool floodable); | |
563 | static void xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
564 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 DF |
565 | const struct bfd *bfd, const struct lldp *lldp, |
566 | int stp_port_no, const struct rstp_port *rstp_port, | |
84f0f298 RW |
567 | enum ofputil_port_config config, |
568 | enum ofputil_port_state state, bool is_tunnel, | |
569 | bool may_enable); | |
570 | static void xlate_xbridge_remove(struct xlate_cfg *, struct xbridge *); | |
571 | static void xlate_xbundle_remove(struct xlate_cfg *, struct xbundle *); | |
572 | static void xlate_xport_remove(struct xlate_cfg *, struct xport *); | |
573 | static void xlate_xbridge_copy(struct xbridge *); | |
574 | static void xlate_xbundle_copy(struct xbridge *, struct xbundle *); | |
575 | static void xlate_xport_copy(struct xbridge *, struct xbundle *, | |
576 | struct xport *); | |
577 | static void xlate_xcfg_free(struct xlate_cfg *); | |
b256dc52 | 578 | |
34dd0d78 | 579 | static inline void |
c1b3756c | 580 | xlate_report(struct xlate_ctx *ctx, const char *format, ...) |
34dd0d78 JR |
581 | { |
582 | if (OVS_UNLIKELY(ctx->xin->report_hook)) { | |
c1b3756c BP |
583 | va_list args; |
584 | ||
585 | va_start(args, format); | |
9a9b001a | 586 | ctx->xin->report_hook(ctx->xin, ctx->indentation, format, args); |
c1b3756c | 587 | va_end(args); |
34dd0d78 JR |
588 | } |
589 | } | |
84f0f298 | 590 | |
fff1b9c0 JR |
591 | static struct vlog_rate_limit error_report_rl = VLOG_RATE_LIMIT_INIT(1, 5); |
592 | ||
593 | #define XLATE_REPORT_ERROR(CTX, ...) \ | |
594 | do { \ | |
595 | if (OVS_UNLIKELY((CTX)->xin->report_hook)) { \ | |
596 | xlate_report(CTX, __VA_ARGS__); \ | |
597 | } else { \ | |
598 | VLOG_ERR_RL(&error_report_rl, __VA_ARGS__); \ | |
599 | } \ | |
600 | } while (0) | |
601 | ||
d6bef3cc BP |
602 | static inline void |
603 | xlate_report_actions(struct xlate_ctx *ctx, const char *title, | |
604 | const struct ofpact *ofpacts, size_t ofpacts_len) | |
605 | { | |
606 | if (OVS_UNLIKELY(ctx->xin->report_hook)) { | |
607 | struct ds s = DS_EMPTY_INITIALIZER; | |
608 | ofpacts_format(ofpacts, ofpacts_len, &s); | |
609 | xlate_report(ctx, "%s: %s", title, ds_cstr(&s)); | |
610 | ds_destroy(&s); | |
611 | } | |
612 | } | |
613 | ||
84f0f298 RW |
614 | static void |
615 | xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
616 | { | |
417e7e66 | 617 | ovs_list_init(&xbridge->xbundles); |
84f0f298 RW |
618 | hmap_init(&xbridge->xports); |
619 | hmap_insert(&xcfg->xbridges, &xbridge->hmap_node, | |
620 | hash_pointer(xbridge->ofproto, 0)); | |
621 | } | |
622 | ||
623 | static void | |
624 | xlate_xbundle_init(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
625 | { | |
417e7e66 BW |
626 | ovs_list_init(&xbundle->xports); |
627 | ovs_list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); | |
84f0f298 RW |
628 | hmap_insert(&xcfg->xbundles, &xbundle->hmap_node, |
629 | hash_pointer(xbundle->ofbundle, 0)); | |
630 | } | |
631 | ||
632 | static void | |
633 | xlate_xport_init(struct xlate_cfg *xcfg, struct xport *xport) | |
634 | { | |
635 | hmap_init(&xport->skb_priorities); | |
636 | hmap_insert(&xcfg->xports, &xport->hmap_node, | |
637 | hash_pointer(xport->ofport, 0)); | |
638 | hmap_insert(&xport->xbridge->xports, &xport->ofp_node, | |
639 | hash_ofp_port(xport->ofp_port)); | |
640 | } | |
641 | ||
642 | static void | |
643 | xlate_xbridge_set(struct xbridge *xbridge, | |
644 | struct dpif *dpif, | |
ec89fc6f | 645 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 646 | struct rstp *rstp, const struct mcast_snooping *ms, |
ec89fc6f | 647 | const struct mbridge *mbridge, |
46c88433 | 648 | const struct dpif_sflow *sflow, |
ce3955be | 649 | const struct dpif_ipfix *ipfix, |
2f47cdf4 | 650 | const struct netflow *netflow, |
4b97b70d | 651 | bool forward_bpdu, bool has_in_band, |
b440dd8c | 652 | const struct dpif_backer_support *support) |
46c88433 | 653 | { |
46c88433 EJ |
654 | if (xbridge->ml != ml) { |
655 | mac_learning_unref(xbridge->ml); | |
656 | xbridge->ml = mac_learning_ref(ml); | |
657 | } | |
658 | ||
6d95c4e8 FL |
659 | if (xbridge->ms != ms) { |
660 | mcast_snooping_unref(xbridge->ms); | |
661 | xbridge->ms = mcast_snooping_ref(ms); | |
662 | } | |
663 | ||
46c88433 EJ |
664 | if (xbridge->mbridge != mbridge) { |
665 | mbridge_unref(xbridge->mbridge); | |
666 | xbridge->mbridge = mbridge_ref(mbridge); | |
667 | } | |
668 | ||
669 | if (xbridge->sflow != sflow) { | |
670 | dpif_sflow_unref(xbridge->sflow); | |
671 | xbridge->sflow = dpif_sflow_ref(sflow); | |
672 | } | |
673 | ||
674 | if (xbridge->ipfix != ipfix) { | |
675 | dpif_ipfix_unref(xbridge->ipfix); | |
676 | xbridge->ipfix = dpif_ipfix_ref(ipfix); | |
677 | } | |
678 | ||
9d189a50 EJ |
679 | if (xbridge->stp != stp) { |
680 | stp_unref(xbridge->stp); | |
681 | xbridge->stp = stp_ref(stp); | |
682 | } | |
683 | ||
9efd308e DV |
684 | if (xbridge->rstp != rstp) { |
685 | rstp_unref(xbridge->rstp); | |
686 | xbridge->rstp = rstp_ref(rstp); | |
687 | } | |
688 | ||
ce3955be EJ |
689 | if (xbridge->netflow != netflow) { |
690 | netflow_unref(xbridge->netflow); | |
691 | xbridge->netflow = netflow_ref(netflow); | |
692 | } | |
693 | ||
89a8a7f0 | 694 | xbridge->dpif = dpif; |
46c88433 EJ |
695 | xbridge->forward_bpdu = forward_bpdu; |
696 | xbridge->has_in_band = has_in_band; | |
b440dd8c | 697 | xbridge->support = *support; |
46c88433 EJ |
698 | } |
699 | ||
84f0f298 RW |
700 | static void |
701 | xlate_xbundle_set(struct xbundle *xbundle, | |
702 | enum port_vlan_mode vlan_mode, int vlan, | |
703 | unsigned long *trunks, bool use_priority_tags, | |
704 | const struct bond *bond, const struct lacp *lacp, | |
705 | bool floodable) | |
706 | { | |
707 | ovs_assert(xbundle->xbridge); | |
708 | ||
709 | xbundle->vlan_mode = vlan_mode; | |
710 | xbundle->vlan = vlan; | |
711 | xbundle->trunks = trunks; | |
712 | xbundle->use_priority_tags = use_priority_tags; | |
713 | xbundle->floodable = floodable; | |
714 | ||
715 | if (xbundle->bond != bond) { | |
716 | bond_unref(xbundle->bond); | |
717 | xbundle->bond = bond_ref(bond); | |
718 | } | |
719 | ||
720 | if (xbundle->lacp != lacp) { | |
721 | lacp_unref(xbundle->lacp); | |
722 | xbundle->lacp = lacp_ref(lacp); | |
723 | } | |
724 | } | |
725 | ||
726 | static void | |
727 | xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
728 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 | 729 | const struct bfd *bfd, const struct lldp *lldp, int stp_port_no, |
f025bcb7 | 730 | const struct rstp_port* rstp_port, |
84f0f298 RW |
731 | enum ofputil_port_config config, enum ofputil_port_state state, |
732 | bool is_tunnel, bool may_enable) | |
733 | { | |
734 | xport->config = config; | |
735 | xport->state = state; | |
736 | xport->stp_port_no = stp_port_no; | |
737 | xport->is_tunnel = is_tunnel; | |
738 | xport->may_enable = may_enable; | |
739 | xport->odp_port = odp_port; | |
740 | ||
f025bcb7 JR |
741 | if (xport->rstp_port != rstp_port) { |
742 | rstp_port_unref(xport->rstp_port); | |
743 | xport->rstp_port = rstp_port_ref(rstp_port); | |
744 | } | |
745 | ||
84f0f298 RW |
746 | if (xport->cfm != cfm) { |
747 | cfm_unref(xport->cfm); | |
748 | xport->cfm = cfm_ref(cfm); | |
749 | } | |
750 | ||
751 | if (xport->bfd != bfd) { | |
752 | bfd_unref(xport->bfd); | |
753 | xport->bfd = bfd_ref(bfd); | |
754 | } | |
755 | ||
0477baa9 DF |
756 | if (xport->lldp != lldp) { |
757 | lldp_unref(xport->lldp); | |
758 | xport->lldp = lldp_ref(lldp); | |
759 | } | |
760 | ||
84f0f298 RW |
761 | if (xport->netdev != netdev) { |
762 | netdev_close(xport->netdev); | |
763 | xport->netdev = netdev_ref(netdev); | |
764 | } | |
765 | } | |
766 | ||
767 | static void | |
768 | xlate_xbridge_copy(struct xbridge *xbridge) | |
769 | { | |
770 | struct xbundle *xbundle; | |
771 | struct xport *xport; | |
772 | struct xbridge *new_xbridge = xzalloc(sizeof *xbridge); | |
773 | new_xbridge->ofproto = xbridge->ofproto; | |
774 | new_xbridge->name = xstrdup(xbridge->name); | |
775 | xlate_xbridge_init(new_xcfg, new_xbridge); | |
776 | ||
777 | xlate_xbridge_set(new_xbridge, | |
34dd0d78 | 778 | xbridge->dpif, xbridge->ml, xbridge->stp, |
9efd308e DV |
779 | xbridge->rstp, xbridge->ms, xbridge->mbridge, |
780 | xbridge->sflow, xbridge->ipfix, xbridge->netflow, | |
b440dd8c JS |
781 | xbridge->forward_bpdu, xbridge->has_in_band, |
782 | &xbridge->support); | |
84f0f298 RW |
783 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
784 | xlate_xbundle_copy(new_xbridge, xbundle); | |
785 | } | |
786 | ||
787 | /* Copy xports which are not part of a xbundle */ | |
788 | HMAP_FOR_EACH (xport, ofp_node, &xbridge->xports) { | |
789 | if (!xport->xbundle) { | |
790 | xlate_xport_copy(new_xbridge, NULL, xport); | |
791 | } | |
792 | } | |
793 | } | |
794 | ||
795 | static void | |
796 | xlate_xbundle_copy(struct xbridge *xbridge, struct xbundle *xbundle) | |
797 | { | |
798 | struct xport *xport; | |
799 | struct xbundle *new_xbundle = xzalloc(sizeof *xbundle); | |
800 | new_xbundle->ofbundle = xbundle->ofbundle; | |
801 | new_xbundle->xbridge = xbridge; | |
802 | new_xbundle->name = xstrdup(xbundle->name); | |
803 | xlate_xbundle_init(new_xcfg, new_xbundle); | |
804 | ||
805 | xlate_xbundle_set(new_xbundle, xbundle->vlan_mode, | |
806 | xbundle->vlan, xbundle->trunks, | |
807 | xbundle->use_priority_tags, xbundle->bond, xbundle->lacp, | |
808 | xbundle->floodable); | |
809 | LIST_FOR_EACH (xport, bundle_node, &xbundle->xports) { | |
810 | xlate_xport_copy(xbridge, new_xbundle, xport); | |
811 | } | |
812 | } | |
813 | ||
814 | static void | |
815 | xlate_xport_copy(struct xbridge *xbridge, struct xbundle *xbundle, | |
816 | struct xport *xport) | |
817 | { | |
818 | struct skb_priority_to_dscp *pdscp, *new_pdscp; | |
819 | struct xport *new_xport = xzalloc(sizeof *xport); | |
820 | new_xport->ofport = xport->ofport; | |
821 | new_xport->ofp_port = xport->ofp_port; | |
822 | new_xport->xbridge = xbridge; | |
823 | xlate_xport_init(new_xcfg, new_xport); | |
824 | ||
825 | xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm, | |
0477baa9 DF |
826 | xport->bfd, xport->lldp, xport->stp_port_no, |
827 | xport->rstp_port, xport->config, xport->state, | |
828 | xport->is_tunnel, xport->may_enable); | |
84f0f298 RW |
829 | |
830 | if (xport->peer) { | |
831 | struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport); | |
832 | if (peer) { | |
833 | new_xport->peer = peer; | |
834 | new_xport->peer->peer = new_xport; | |
835 | } | |
836 | } | |
837 | ||
838 | if (xbundle) { | |
839 | new_xport->xbundle = xbundle; | |
417e7e66 | 840 | ovs_list_insert(&new_xport->xbundle->xports, &new_xport->bundle_node); |
84f0f298 RW |
841 | } |
842 | ||
843 | HMAP_FOR_EACH (pdscp, hmap_node, &xport->skb_priorities) { | |
844 | new_pdscp = xmalloc(sizeof *pdscp); | |
845 | new_pdscp->skb_priority = pdscp->skb_priority; | |
846 | new_pdscp->dscp = pdscp->dscp; | |
847 | hmap_insert(&new_xport->skb_priorities, &new_pdscp->hmap_node, | |
848 | hash_int(new_pdscp->skb_priority, 0)); | |
849 | } | |
850 | } | |
851 | ||
852 | /* Sets the current xlate configuration to new_xcfg and frees the old xlate | |
853 | * configuration in xcfgp. | |
854 | * | |
855 | * This needs to be called after editing the xlate configuration. | |
856 | * | |
857 | * Functions that edit the new xlate configuration are | |
858 | * xlate_<ofport/bundle/ofport>_set and xlate_<ofport/bundle/ofport>_remove. | |
859 | * | |
860 | * A sample workflow: | |
861 | * | |
862 | * xlate_txn_start(); | |
863 | * ... | |
864 | * edit_xlate_configuration(); | |
865 | * ... | |
866 | * xlate_txn_commit(); */ | |
46c88433 | 867 | void |
84f0f298 RW |
868 | xlate_txn_commit(void) |
869 | { | |
870 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
871 | ||
872 | ovsrcu_set(&xcfgp, new_xcfg); | |
40a9c4c2 AW |
873 | ovsrcu_synchronize(); |
874 | xlate_xcfg_free(xcfg); | |
84f0f298 RW |
875 | new_xcfg = NULL; |
876 | } | |
877 | ||
878 | /* Copies the current xlate configuration in xcfgp to new_xcfg. | |
879 | * | |
880 | * This needs to be called prior to editing the xlate configuration. */ | |
881 | void | |
882 | xlate_txn_start(void) | |
883 | { | |
884 | struct xbridge *xbridge; | |
885 | struct xlate_cfg *xcfg; | |
886 | ||
887 | ovs_assert(!new_xcfg); | |
888 | ||
889 | new_xcfg = xmalloc(sizeof *new_xcfg); | |
890 | hmap_init(&new_xcfg->xbridges); | |
891 | hmap_init(&new_xcfg->xbundles); | |
892 | hmap_init(&new_xcfg->xports); | |
893 | ||
894 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
895 | if (!xcfg) { | |
896 | return; | |
897 | } | |
898 | ||
899 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
900 | xlate_xbridge_copy(xbridge); | |
901 | } | |
902 | } | |
903 | ||
904 | ||
905 | static void | |
906 | xlate_xcfg_free(struct xlate_cfg *xcfg) | |
907 | { | |
908 | struct xbridge *xbridge, *next_xbridge; | |
909 | ||
910 | if (!xcfg) { | |
911 | return; | |
912 | } | |
913 | ||
914 | HMAP_FOR_EACH_SAFE (xbridge, next_xbridge, hmap_node, &xcfg->xbridges) { | |
915 | xlate_xbridge_remove(xcfg, xbridge); | |
916 | } | |
917 | ||
918 | hmap_destroy(&xcfg->xbridges); | |
919 | hmap_destroy(&xcfg->xbundles); | |
920 | hmap_destroy(&xcfg->xports); | |
921 | free(xcfg); | |
922 | } | |
923 | ||
924 | void | |
925 | xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, | |
34dd0d78 | 926 | struct dpif *dpif, |
84f0f298 | 927 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 928 | struct rstp *rstp, const struct mcast_snooping *ms, |
84f0f298 RW |
929 | const struct mbridge *mbridge, |
930 | const struct dpif_sflow *sflow, | |
931 | const struct dpif_ipfix *ipfix, | |
2f47cdf4 | 932 | const struct netflow *netflow, |
b440dd8c JS |
933 | bool forward_bpdu, bool has_in_band, |
934 | const struct dpif_backer_support *support) | |
84f0f298 RW |
935 | { |
936 | struct xbridge *xbridge; | |
937 | ||
938 | ovs_assert(new_xcfg); | |
939 | ||
940 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
941 | if (!xbridge) { | |
942 | xbridge = xzalloc(sizeof *xbridge); | |
943 | xbridge->ofproto = ofproto; | |
944 | ||
945 | xlate_xbridge_init(new_xcfg, xbridge); | |
946 | } | |
947 | ||
948 | free(xbridge->name); | |
949 | xbridge->name = xstrdup(name); | |
950 | ||
34dd0d78 | 951 | xlate_xbridge_set(xbridge, dpif, ml, stp, rstp, ms, mbridge, sflow, ipfix, |
b440dd8c | 952 | netflow, forward_bpdu, has_in_band, support); |
84f0f298 RW |
953 | } |
954 | ||
955 | static void | |
956 | xlate_xbridge_remove(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
46c88433 | 957 | { |
46c88433 EJ |
958 | struct xbundle *xbundle, *next_xbundle; |
959 | struct xport *xport, *next_xport; | |
960 | ||
961 | if (!xbridge) { | |
962 | return; | |
963 | } | |
964 | ||
965 | HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) { | |
84f0f298 | 966 | xlate_xport_remove(xcfg, xport); |
46c88433 EJ |
967 | } |
968 | ||
969 | LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) { | |
84f0f298 | 970 | xlate_xbundle_remove(xcfg, xbundle); |
46c88433 EJ |
971 | } |
972 | ||
84f0f298 | 973 | hmap_remove(&xcfg->xbridges, &xbridge->hmap_node); |
795cc5c1 | 974 | mac_learning_unref(xbridge->ml); |
6d95c4e8 | 975 | mcast_snooping_unref(xbridge->ms); |
795cc5c1 EJ |
976 | mbridge_unref(xbridge->mbridge); |
977 | dpif_sflow_unref(xbridge->sflow); | |
978 | dpif_ipfix_unref(xbridge->ipfix); | |
979 | stp_unref(xbridge->stp); | |
9efd308e | 980 | rstp_unref(xbridge->rstp); |
795cc5c1 | 981 | hmap_destroy(&xbridge->xports); |
46c88433 EJ |
982 | free(xbridge->name); |
983 | free(xbridge); | |
984 | } | |
985 | ||
84f0f298 RW |
986 | void |
987 | xlate_remove_ofproto(struct ofproto_dpif *ofproto) | |
988 | { | |
989 | struct xbridge *xbridge; | |
990 | ||
991 | ovs_assert(new_xcfg); | |
992 | ||
993 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
994 | xlate_xbridge_remove(new_xcfg, xbridge); | |
995 | } | |
996 | ||
46c88433 EJ |
997 | void |
998 | xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
999 | const char *name, enum port_vlan_mode vlan_mode, int vlan, | |
1000 | unsigned long *trunks, bool use_priority_tags, | |
1001 | const struct bond *bond, const struct lacp *lacp, | |
1002 | bool floodable) | |
1003 | { | |
84f0f298 | 1004 | struct xbundle *xbundle; |
46c88433 | 1005 | |
84f0f298 RW |
1006 | ovs_assert(new_xcfg); |
1007 | ||
1008 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
46c88433 EJ |
1009 | if (!xbundle) { |
1010 | xbundle = xzalloc(sizeof *xbundle); | |
1011 | xbundle->ofbundle = ofbundle; | |
84f0f298 | 1012 | xbundle->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 1013 | |
84f0f298 | 1014 | xlate_xbundle_init(new_xcfg, xbundle); |
46c88433 EJ |
1015 | } |
1016 | ||
46c88433 EJ |
1017 | free(xbundle->name); |
1018 | xbundle->name = xstrdup(name); | |
1019 | ||
84f0f298 RW |
1020 | xlate_xbundle_set(xbundle, vlan_mode, vlan, trunks, |
1021 | use_priority_tags, bond, lacp, floodable); | |
46c88433 EJ |
1022 | } |
1023 | ||
84f0f298 RW |
1024 | static void |
1025 | xlate_xbundle_remove(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
46c88433 | 1026 | { |
5f03c983 | 1027 | struct xport *xport; |
46c88433 EJ |
1028 | |
1029 | if (!xbundle) { | |
1030 | return; | |
1031 | } | |
1032 | ||
5f03c983 | 1033 | LIST_FOR_EACH_POP (xport, bundle_node, &xbundle->xports) { |
46c88433 EJ |
1034 | xport->xbundle = NULL; |
1035 | } | |
1036 | ||
84f0f298 | 1037 | hmap_remove(&xcfg->xbundles, &xbundle->hmap_node); |
417e7e66 | 1038 | ovs_list_remove(&xbundle->list_node); |
46c88433 EJ |
1039 | bond_unref(xbundle->bond); |
1040 | lacp_unref(xbundle->lacp); | |
1041 | free(xbundle->name); | |
1042 | free(xbundle); | |
1043 | } | |
1044 | ||
84f0f298 RW |
1045 | void |
1046 | xlate_bundle_remove(struct ofbundle *ofbundle) | |
1047 | { | |
1048 | struct xbundle *xbundle; | |
1049 | ||
1050 | ovs_assert(new_xcfg); | |
1051 | ||
1052 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
1053 | xlate_xbundle_remove(new_xcfg, xbundle); | |
1054 | } | |
1055 | ||
46c88433 EJ |
1056 | void |
1057 | xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
1058 | struct ofport_dpif *ofport, ofp_port_t ofp_port, | |
1059 | odp_port_t odp_port, const struct netdev *netdev, | |
1060 | const struct cfm *cfm, const struct bfd *bfd, | |
0477baa9 DF |
1061 | const struct lldp *lldp, struct ofport_dpif *peer, |
1062 | int stp_port_no, const struct rstp_port *rstp_port, | |
55954f6e | 1063 | const struct ofproto_port_queue *qdscp_list, size_t n_qdscp, |
dd8cd4b4 SH |
1064 | enum ofputil_port_config config, |
1065 | enum ofputil_port_state state, bool is_tunnel, | |
9d189a50 | 1066 | bool may_enable) |
46c88433 | 1067 | { |
55954f6e | 1068 | size_t i; |
84f0f298 RW |
1069 | struct xport *xport; |
1070 | ||
1071 | ovs_assert(new_xcfg); | |
46c88433 | 1072 | |
84f0f298 | 1073 | xport = xport_lookup(new_xcfg, ofport); |
46c88433 EJ |
1074 | if (!xport) { |
1075 | xport = xzalloc(sizeof *xport); | |
1076 | xport->ofport = ofport; | |
84f0f298 | 1077 | xport->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 EJ |
1078 | xport->ofp_port = ofp_port; |
1079 | ||
84f0f298 | 1080 | xlate_xport_init(new_xcfg, xport); |
46c88433 EJ |
1081 | } |
1082 | ||
1083 | ovs_assert(xport->ofp_port == ofp_port); | |
1084 | ||
0477baa9 DF |
1085 | xlate_xport_set(xport, odp_port, netdev, cfm, bfd, lldp, |
1086 | stp_port_no, rstp_port, config, state, is_tunnel, | |
1087 | may_enable); | |
46c88433 EJ |
1088 | |
1089 | if (xport->peer) { | |
1090 | xport->peer->peer = NULL; | |
1091 | } | |
84f0f298 | 1092 | xport->peer = xport_lookup(new_xcfg, peer); |
46c88433 EJ |
1093 | if (xport->peer) { |
1094 | xport->peer->peer = xport; | |
1095 | } | |
1096 | ||
1097 | if (xport->xbundle) { | |
417e7e66 | 1098 | ovs_list_remove(&xport->bundle_node); |
46c88433 | 1099 | } |
84f0f298 | 1100 | xport->xbundle = xbundle_lookup(new_xcfg, ofbundle); |
46c88433 | 1101 | if (xport->xbundle) { |
417e7e66 | 1102 | ovs_list_insert(&xport->xbundle->xports, &xport->bundle_node); |
46c88433 | 1103 | } |
55954f6e EJ |
1104 | |
1105 | clear_skb_priorities(xport); | |
1106 | for (i = 0; i < n_qdscp; i++) { | |
1107 | struct skb_priority_to_dscp *pdscp; | |
1108 | uint32_t skb_priority; | |
1109 | ||
89a8a7f0 EJ |
1110 | if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue, |
1111 | &skb_priority)) { | |
55954f6e EJ |
1112 | continue; |
1113 | } | |
1114 | ||
1115 | pdscp = xmalloc(sizeof *pdscp); | |
1116 | pdscp->skb_priority = skb_priority; | |
1117 | pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK; | |
1118 | hmap_insert(&xport->skb_priorities, &pdscp->hmap_node, | |
1119 | hash_int(pdscp->skb_priority, 0)); | |
1120 | } | |
46c88433 EJ |
1121 | } |
1122 | ||
84f0f298 RW |
1123 | static void |
1124 | xlate_xport_remove(struct xlate_cfg *xcfg, struct xport *xport) | |
46c88433 | 1125 | { |
46c88433 EJ |
1126 | if (!xport) { |
1127 | return; | |
1128 | } | |
1129 | ||
1130 | if (xport->peer) { | |
1131 | xport->peer->peer = NULL; | |
1132 | xport->peer = NULL; | |
1133 | } | |
1134 | ||
e621a12d | 1135 | if (xport->xbundle) { |
417e7e66 | 1136 | ovs_list_remove(&xport->bundle_node); |
e621a12d EJ |
1137 | } |
1138 | ||
55954f6e EJ |
1139 | clear_skb_priorities(xport); |
1140 | hmap_destroy(&xport->skb_priorities); | |
1141 | ||
84f0f298 | 1142 | hmap_remove(&xcfg->xports, &xport->hmap_node); |
46c88433 EJ |
1143 | hmap_remove(&xport->xbridge->xports, &xport->ofp_node); |
1144 | ||
1145 | netdev_close(xport->netdev); | |
f025bcb7 | 1146 | rstp_port_unref(xport->rstp_port); |
46c88433 EJ |
1147 | cfm_unref(xport->cfm); |
1148 | bfd_unref(xport->bfd); | |
0477baa9 | 1149 | lldp_unref(xport->lldp); |
46c88433 EJ |
1150 | free(xport); |
1151 | } | |
1152 | ||
84f0f298 RW |
1153 | void |
1154 | xlate_ofport_remove(struct ofport_dpif *ofport) | |
1155 | { | |
1156 | struct xport *xport; | |
1157 | ||
1158 | ovs_assert(new_xcfg); | |
1159 | ||
1160 | xport = xport_lookup(new_xcfg, ofport); | |
1161 | xlate_xport_remove(new_xcfg, xport); | |
1162 | } | |
1163 | ||
ef377a58 JR |
1164 | static struct ofproto_dpif * |
1165 | xlate_lookup_ofproto_(const struct dpif_backer *backer, const struct flow *flow, | |
1166 | ofp_port_t *ofp_in_port, const struct xport **xportp) | |
1167 | { | |
e672ff9b | 1168 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
ef377a58 | 1169 | const struct xport *xport; |
f9038ef6 | 1170 | |
e672ff9b JR |
1171 | xport = xport_lookup(xcfg, tnl_port_should_receive(flow) |
1172 | ? tnl_port_receive(flow) | |
1173 | : odp_port_to_ofport(backer, flow->in_port.odp_port)); | |
1174 | if (OVS_UNLIKELY(!xport)) { | |
1175 | return NULL; | |
ef377a58 | 1176 | } |
e672ff9b | 1177 | *xportp = xport; |
f9038ef6 | 1178 | if (ofp_in_port) { |
e672ff9b | 1179 | *ofp_in_port = xport->ofp_port; |
f9038ef6 | 1180 | } |
e672ff9b | 1181 | return xport->xbridge->ofproto; |
ef377a58 JR |
1182 | } |
1183 | ||
1184 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively) | |
1185 | * returns the corresponding struct ofproto_dpif and OpenFlow port number. */ | |
1186 | struct ofproto_dpif * | |
1187 | xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow, | |
1188 | ofp_port_t *ofp_in_port) | |
1189 | { | |
1190 | const struct xport *xport; | |
1191 | ||
1192 | return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); | |
1193 | } | |
1194 | ||
cc377352 | 1195 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively), |
ef377a58 | 1196 | * optionally populates 'ofproto' with the ofproto_dpif, 'ofp_in_port' with the |
cc377352 | 1197 | * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate |
dcc2c6cd JR |
1198 | * handles for those protocols if they're enabled. Caller may use the returned |
1199 | * pointers until quiescing, for longer term use additional references must | |
1200 | * be taken. | |
8449c4d6 | 1201 | * |
f9038ef6 | 1202 | * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto. |
ef377a58 | 1203 | */ |
8449c4d6 | 1204 | int |
5c476ea3 JR |
1205 | xlate_lookup(const struct dpif_backer *backer, const struct flow *flow, |
1206 | struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix, | |
1207 | struct dpif_sflow **sflow, struct netflow **netflow, | |
1208 | ofp_port_t *ofp_in_port) | |
8449c4d6 | 1209 | { |
ef377a58 | 1210 | struct ofproto_dpif *ofproto; |
84f0f298 | 1211 | const struct xport *xport; |
8449c4d6 | 1212 | |
ef377a58 | 1213 | ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); |
8449c4d6 | 1214 | |
f9038ef6 | 1215 | if (!ofproto) { |
cc377352 | 1216 | return ENODEV; |
8449c4d6 | 1217 | } |
8449c4d6 | 1218 | |
ef377a58 JR |
1219 | if (ofprotop) { |
1220 | *ofprotop = ofproto; | |
8449c4d6 EJ |
1221 | } |
1222 | ||
1dfdb9b3 | 1223 | if (ipfix) { |
f9038ef6 | 1224 | *ipfix = xport ? xport->xbridge->ipfix : NULL; |
1dfdb9b3 EJ |
1225 | } |
1226 | ||
1227 | if (sflow) { | |
f9038ef6 | 1228 | *sflow = xport ? xport->xbridge->sflow : NULL; |
1dfdb9b3 EJ |
1229 | } |
1230 | ||
1231 | if (netflow) { | |
f9038ef6 | 1232 | *netflow = xport ? xport->xbridge->netflow : NULL; |
1dfdb9b3 | 1233 | } |
f9038ef6 | 1234 | |
cc377352 | 1235 | return 0; |
8449c4d6 EJ |
1236 | } |
1237 | ||
46c88433 | 1238 | static struct xbridge * |
84f0f298 | 1239 | xbridge_lookup(struct xlate_cfg *xcfg, const struct ofproto_dpif *ofproto) |
46c88433 | 1240 | { |
84f0f298 | 1241 | struct hmap *xbridges; |
46c88433 EJ |
1242 | struct xbridge *xbridge; |
1243 | ||
84f0f298 | 1244 | if (!ofproto || !xcfg) { |
5e6af486 EJ |
1245 | return NULL; |
1246 | } | |
1247 | ||
84f0f298 RW |
1248 | xbridges = &xcfg->xbridges; |
1249 | ||
46c88433 | 1250 | HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0), |
84f0f298 | 1251 | xbridges) { |
46c88433 EJ |
1252 | if (xbridge->ofproto == ofproto) { |
1253 | return xbridge; | |
1254 | } | |
1255 | } | |
1256 | return NULL; | |
1257 | } | |
1258 | ||
290835f9 BP |
1259 | static struct xbridge * |
1260 | xbridge_lookup_by_uuid(struct xlate_cfg *xcfg, const struct uuid *uuid) | |
1261 | { | |
1262 | struct xbridge *xbridge; | |
1263 | ||
1264 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
1265 | if (uuid_equals(ofproto_dpif_get_uuid(xbridge->ofproto), uuid)) { | |
1266 | return xbridge; | |
1267 | } | |
1268 | } | |
1269 | return NULL; | |
1270 | } | |
1271 | ||
46c88433 | 1272 | static struct xbundle * |
84f0f298 | 1273 | xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle) |
46c88433 | 1274 | { |
84f0f298 | 1275 | struct hmap *xbundles; |
46c88433 EJ |
1276 | struct xbundle *xbundle; |
1277 | ||
84f0f298 | 1278 | if (!ofbundle || !xcfg) { |
5e6af486 EJ |
1279 | return NULL; |
1280 | } | |
1281 | ||
84f0f298 RW |
1282 | xbundles = &xcfg->xbundles; |
1283 | ||
46c88433 | 1284 | HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), |
84f0f298 | 1285 | xbundles) { |
46c88433 EJ |
1286 | if (xbundle->ofbundle == ofbundle) { |
1287 | return xbundle; | |
1288 | } | |
1289 | } | |
1290 | return NULL; | |
1291 | } | |
1292 | ||
1293 | static struct xport * | |
84f0f298 | 1294 | xport_lookup(struct xlate_cfg *xcfg, const struct ofport_dpif *ofport) |
46c88433 | 1295 | { |
84f0f298 | 1296 | struct hmap *xports; |
46c88433 EJ |
1297 | struct xport *xport; |
1298 | ||
84f0f298 | 1299 | if (!ofport || !xcfg) { |
5e6af486 EJ |
1300 | return NULL; |
1301 | } | |
1302 | ||
84f0f298 RW |
1303 | xports = &xcfg->xports; |
1304 | ||
46c88433 | 1305 | HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0), |
84f0f298 | 1306 | xports) { |
46c88433 EJ |
1307 | if (xport->ofport == ofport) { |
1308 | return xport; | |
1309 | } | |
1310 | } | |
1311 | return NULL; | |
1312 | } | |
1313 | ||
40085e56 EJ |
1314 | static struct stp_port * |
1315 | xport_get_stp_port(const struct xport *xport) | |
1316 | { | |
92cf817b | 1317 | return xport->xbridge->stp && xport->stp_port_no != -1 |
40085e56 EJ |
1318 | ? stp_get_port(xport->xbridge->stp, xport->stp_port_no) |
1319 | : NULL; | |
1320 | } | |
9d189a50 | 1321 | |
0d1cee12 | 1322 | static bool |
9d189a50 EJ |
1323 | xport_stp_learn_state(const struct xport *xport) |
1324 | { | |
40085e56 | 1325 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1326 | return sp |
1327 | ? stp_learn_in_state(stp_port_get_state(sp)) | |
1328 | : true; | |
9d189a50 EJ |
1329 | } |
1330 | ||
1331 | static bool | |
1332 | xport_stp_forward_state(const struct xport *xport) | |
1333 | { | |
40085e56 | 1334 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1335 | return sp |
1336 | ? stp_forward_in_state(stp_port_get_state(sp)) | |
1337 | : true; | |
9d189a50 EJ |
1338 | } |
1339 | ||
0d1cee12 | 1340 | static bool |
bacdb85a | 1341 | xport_stp_should_forward_bpdu(const struct xport *xport) |
0d1cee12 K |
1342 | { |
1343 | struct stp_port *sp = xport_get_stp_port(xport); | |
bacdb85a | 1344 | return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED); |
0d1cee12 K |
1345 | } |
1346 | ||
9d189a50 EJ |
1347 | /* Returns true if STP should process 'flow'. Sets fields in 'wc' that |
1348 | * were used to make the determination.*/ | |
1349 | static bool | |
1350 | stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc) | |
1351 | { | |
bbbca389 | 1352 | /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */ |
9d189a50 | 1353 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
bbbca389 | 1354 | return is_stp(flow); |
9d189a50 EJ |
1355 | } |
1356 | ||
1357 | static void | |
cf62fa4c | 1358 | stp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9d189a50 | 1359 | { |
40085e56 | 1360 | struct stp_port *sp = xport_get_stp_port(xport); |
cf62fa4c PS |
1361 | struct dp_packet payload = *packet; |
1362 | struct eth_header *eth = dp_packet_data(&payload); | |
9d189a50 EJ |
1363 | |
1364 | /* Sink packets on ports that have STP disabled when the bridge has | |
1365 | * STP enabled. */ | |
1366 | if (!sp || stp_port_get_state(sp) == STP_DISABLED) { | |
1367 | return; | |
1368 | } | |
1369 | ||
1370 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1371 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1372 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9d189a50 EJ |
1373 | } |
1374 | ||
cf62fa4c PS |
1375 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1376 | stp_received_bpdu(sp, dp_packet_data(&payload), dp_packet_size(&payload)); | |
9d189a50 EJ |
1377 | } |
1378 | } | |
1379 | ||
f025bcb7 JR |
1380 | static enum rstp_state |
1381 | xport_get_rstp_port_state(const struct xport *xport) | |
9efd308e | 1382 | { |
f025bcb7 JR |
1383 | return xport->rstp_port |
1384 | ? rstp_port_get_state(xport->rstp_port) | |
1385 | : RSTP_DISABLED; | |
9efd308e DV |
1386 | } |
1387 | ||
1388 | static bool | |
1389 | xport_rstp_learn_state(const struct xport *xport) | |
1390 | { | |
4b5f1996 DV |
1391 | return xport->xbridge->rstp && xport->rstp_port |
1392 | ? rstp_learn_in_state(xport_get_rstp_port_state(xport)) | |
1393 | : true; | |
9efd308e DV |
1394 | } |
1395 | ||
1396 | static bool | |
1397 | xport_rstp_forward_state(const struct xport *xport) | |
1398 | { | |
4b5f1996 DV |
1399 | return xport->xbridge->rstp && xport->rstp_port |
1400 | ? rstp_forward_in_state(xport_get_rstp_port_state(xport)) | |
1401 | : true; | |
9efd308e DV |
1402 | } |
1403 | ||
1404 | static bool | |
1405 | xport_rstp_should_manage_bpdu(const struct xport *xport) | |
1406 | { | |
f025bcb7 | 1407 | return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport)); |
9efd308e DV |
1408 | } |
1409 | ||
1410 | static void | |
cf62fa4c | 1411 | rstp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9efd308e | 1412 | { |
cf62fa4c PS |
1413 | struct dp_packet payload = *packet; |
1414 | struct eth_header *eth = dp_packet_data(&payload); | |
9efd308e | 1415 | |
f025bcb7 JR |
1416 | /* Sink packets on ports that have no RSTP. */ |
1417 | if (!xport->rstp_port) { | |
9efd308e DV |
1418 | return; |
1419 | } | |
1420 | ||
1421 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1422 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1423 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9efd308e DV |
1424 | } |
1425 | ||
cf62fa4c PS |
1426 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1427 | rstp_port_received_bpdu(xport->rstp_port, dp_packet_data(&payload), | |
1428 | dp_packet_size(&payload)); | |
9efd308e DV |
1429 | } |
1430 | } | |
1431 | ||
46c88433 EJ |
1432 | static struct xport * |
1433 | get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1434 | { | |
1435 | struct xport *xport; | |
1436 | ||
1437 | HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port), | |
1438 | &xbridge->xports) { | |
1439 | if (xport->ofp_port == ofp_port) { | |
1440 | return xport; | |
1441 | } | |
1442 | } | |
1443 | return NULL; | |
1444 | } | |
1445 | ||
1446 | static odp_port_t | |
1447 | ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1448 | { | |
1449 | const struct xport *xport = get_ofp_port(xbridge, ofp_port); | |
1450 | return xport ? xport->odp_port : ODPP_NONE; | |
1451 | } | |
1452 | ||
dd8cd4b4 SH |
1453 | static bool |
1454 | odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port) | |
1455 | { | |
086fa873 BP |
1456 | struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1457 | return xport && xport->may_enable; | |
dd8cd4b4 SH |
1458 | } |
1459 | ||
1e684d7d | 1460 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1461 | group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *, |
1462 | int depth); | |
1463 | ||
1464 | static bool | |
1465 | group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) | |
1466 | { | |
1467 | struct group_dpif *group; | |
dd8cd4b4 | 1468 | |
dc25893e AZ |
1469 | if (group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group)) { |
1470 | struct ofputil_bucket *bucket; | |
dd8cd4b4 | 1471 | |
dc25893e AZ |
1472 | bucket = group_first_live_bucket(ctx, group, depth); |
1473 | group_dpif_unref(group); | |
1474 | return bucket == NULL; | |
1475 | } | |
dd8cd4b4 | 1476 | |
dc25893e | 1477 | return false; |
dd8cd4b4 SH |
1478 | } |
1479 | ||
1480 | #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ | |
1481 | ||
1482 | static bool | |
1483 | bucket_is_alive(const struct xlate_ctx *ctx, | |
1e684d7d | 1484 | struct ofputil_bucket *bucket, int depth) |
dd8cd4b4 SH |
1485 | { |
1486 | if (depth >= MAX_LIVENESS_RECURSION) { | |
1487 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
1488 | ||
1489 | VLOG_WARN_RL(&rl, "bucket chaining exceeded %d links", | |
1490 | MAX_LIVENESS_RECURSION); | |
1491 | return false; | |
1492 | } | |
1493 | ||
fdb1999b AZ |
1494 | return (!ofputil_bucket_has_liveness(bucket) |
1495 | || (bucket->watch_port != OFPP_ANY | |
1496 | && odp_port_is_alive(ctx, bucket->watch_port)) | |
1497 | || (bucket->watch_group != OFPG_ANY | |
1498 | && group_is_alive(ctx, bucket->watch_group, depth + 1))); | |
dd8cd4b4 SH |
1499 | } |
1500 | ||
1e684d7d | 1501 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1502 | group_first_live_bucket(const struct xlate_ctx *ctx, |
1503 | const struct group_dpif *group, int depth) | |
1504 | { | |
1505 | struct ofputil_bucket *bucket; | |
ca6ba700 | 1506 | const struct ovs_list *buckets; |
dd8cd4b4 SH |
1507 | |
1508 | group_dpif_get_buckets(group, &buckets); | |
1509 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1510 | if (bucket_is_alive(ctx, bucket, depth)) { | |
1511 | return bucket; | |
1512 | } | |
1513 | } | |
1514 | ||
1515 | return NULL; | |
1516 | } | |
1517 | ||
1e684d7d | 1518 | static struct ofputil_bucket * |
fe7e5749 SH |
1519 | group_best_live_bucket(const struct xlate_ctx *ctx, |
1520 | const struct group_dpif *group, | |
1521 | uint32_t basis) | |
1522 | { | |
1e684d7d | 1523 | struct ofputil_bucket *best_bucket = NULL; |
fe7e5749 SH |
1524 | uint32_t best_score = 0; |
1525 | int i = 0; | |
1526 | ||
1e684d7d | 1527 | struct ofputil_bucket *bucket; |
ca6ba700 | 1528 | const struct ovs_list *buckets; |
fe7e5749 SH |
1529 | |
1530 | group_dpif_get_buckets(group, &buckets); | |
1531 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1532 | if (bucket_is_alive(ctx, bucket, 0)) { | |
7cb279c2 | 1533 | uint32_t score = (hash_int(i, basis) & 0xffff) * bucket->weight; |
fe7e5749 SH |
1534 | if (score >= best_score) { |
1535 | best_bucket = bucket; | |
1536 | best_score = score; | |
1537 | } | |
1538 | } | |
1539 | i++; | |
1540 | } | |
1541 | ||
1542 | return best_bucket; | |
1543 | } | |
1544 | ||
9583bc14 | 1545 | static bool |
46c88433 | 1546 | xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan) |
9583bc14 EJ |
1547 | { |
1548 | return (bundle->vlan_mode != PORT_VLAN_ACCESS | |
1549 | && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan))); | |
1550 | } | |
1551 | ||
1552 | static bool | |
46c88433 EJ |
1553 | xbundle_includes_vlan(const struct xbundle *xbundle, uint16_t vlan) |
1554 | { | |
1555 | return vlan == xbundle->vlan || xbundle_trunks_vlan(xbundle, vlan); | |
1556 | } | |
1557 | ||
1558 | static mirror_mask_t | |
1559 | xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle) | |
1560 | { | |
1561 | return xbundle != &ofpp_none_bundle | |
1562 | ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle) | |
1563 | : 0; | |
1564 | } | |
1565 | ||
1566 | static mirror_mask_t | |
1567 | xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1568 | { |
46c88433 EJ |
1569 | return xbundle != &ofpp_none_bundle |
1570 | ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle) | |
1571 | : 0; | |
9583bc14 EJ |
1572 | } |
1573 | ||
46c88433 EJ |
1574 | static mirror_mask_t |
1575 | xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1576 | { |
46c88433 EJ |
1577 | return xbundle != &ofpp_none_bundle |
1578 | ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle) | |
1579 | : 0; | |
1580 | } | |
1581 | ||
1582 | static struct xbundle * | |
1583 | lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port, | |
1584 | bool warn, struct xport **in_xportp) | |
1585 | { | |
1586 | struct xport *xport; | |
9583bc14 EJ |
1587 | |
1588 | /* Find the port and bundle for the received packet. */ | |
46c88433 EJ |
1589 | xport = get_ofp_port(xbridge, in_port); |
1590 | if (in_xportp) { | |
1591 | *in_xportp = xport; | |
9583bc14 | 1592 | } |
46c88433 EJ |
1593 | if (xport && xport->xbundle) { |
1594 | return xport->xbundle; | |
9583bc14 EJ |
1595 | } |
1596 | ||
6362203b YT |
1597 | /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+), |
1598 | * which a controller may use as the ingress port for traffic that | |
1599 | * it is sourcing. */ | |
1600 | if (in_port == OFPP_CONTROLLER || in_port == OFPP_NONE) { | |
9583bc14 EJ |
1601 | return &ofpp_none_bundle; |
1602 | } | |
1603 | ||
1604 | /* Odd. A few possible reasons here: | |
1605 | * | |
1606 | * - We deleted a port but there are still a few packets queued up | |
1607 | * from it. | |
1608 | * | |
1609 | * - Someone externally added a port (e.g. "ovs-dpctl add-if") that | |
1610 | * we don't know about. | |
1611 | * | |
1612 | * - The ofproto client didn't configure the port as part of a bundle. | |
1613 | * This is particularly likely to happen if a packet was received on the | |
1614 | * port after it was created, but before the client had a chance to | |
1615 | * configure its bundle. | |
1616 | */ | |
1617 | if (warn) { | |
1618 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1619 | ||
1620 | VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown " | |
46c88433 | 1621 | "port %"PRIu16, xbridge->name, in_port); |
9583bc14 EJ |
1622 | } |
1623 | return NULL; | |
1624 | } | |
1625 | ||
faa624b4 BP |
1626 | /* Mirrors the packet represented by 'ctx' to appropriate mirror destinations, |
1627 | * given the packet is ingressing or egressing on 'xbundle', which has ingress | |
1628 | * or egress (as appropriate) mirrors 'mirrors'. */ | |
9583bc14 | 1629 | static void |
7efbc3b7 BP |
1630 | mirror_packet(struct xlate_ctx *ctx, struct xbundle *xbundle, |
1631 | mirror_mask_t mirrors) | |
9583bc14 | 1632 | { |
faa624b4 BP |
1633 | /* Figure out what VLAN the packet is in (because mirrors can select |
1634 | * packets on basis of VLAN). */ | |
7efbc3b7 BP |
1635 | bool warn = ctx->xin->packet != NULL; |
1636 | uint16_t vid = vlan_tci_to_vid(ctx->xin->flow.vlan_tci); | |
1637 | if (!input_vid_is_valid(vid, xbundle, warn)) { | |
9583bc14 EJ |
1638 | return; |
1639 | } | |
7efbc3b7 | 1640 | uint16_t vlan = input_vid_to_vlan(xbundle, vid); |
9583bc14 | 1641 | |
7efbc3b7 | 1642 | const struct xbridge *xbridge = ctx->xbridge; |
9583bc14 | 1643 | |
7efbc3b7 BP |
1644 | /* Don't mirror to destinations that we've already mirrored to. */ |
1645 | mirrors &= ~ctx->mirrors; | |
9583bc14 EJ |
1646 | if (!mirrors) { |
1647 | return; | |
1648 | } | |
1649 | ||
7efbc3b7 BP |
1650 | if (ctx->xin->resubmit_stats) { |
1651 | mirror_update_stats(xbridge->mbridge, mirrors, | |
1652 | ctx->xin->resubmit_stats->n_packets, | |
1653 | ctx->xin->resubmit_stats->n_bytes); | |
1654 | } | |
1655 | if (ctx->xin->xcache) { | |
1656 | struct xc_entry *entry; | |
1657 | ||
1658 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_MIRROR); | |
1659 | entry->u.mirror.mbridge = mbridge_ref(xbridge->mbridge); | |
1660 | entry->u.mirror.mirrors = mirrors; | |
1661 | } | |
9583bc14 | 1662 | |
faa624b4 BP |
1663 | /* 'mirrors' is a bit-mask of candidates for mirroring. Iterate as long as |
1664 | * some candidates remain. */ | |
9583bc14 | 1665 | while (mirrors) { |
7efbc3b7 | 1666 | const unsigned long *vlans; |
ec7ceaed EJ |
1667 | mirror_mask_t dup_mirrors; |
1668 | struct ofbundle *out; | |
ec7ceaed EJ |
1669 | int out_vlan; |
1670 | ||
faa624b4 | 1671 | /* Get the details of the mirror represented by the rightmost 1-bit. */ |
7efbc3b7 BP |
1672 | bool has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors), |
1673 | &vlans, &dup_mirrors, &out, &out_vlan); | |
ec7ceaed EJ |
1674 | ovs_assert(has_mirror); |
1675 | ||
faa624b4 BP |
1676 | /* If this mirror selects on the basis of VLAN, and it does not select |
1677 | * 'vlan', then discard this mirror and go on to the next one. */ | |
ec7ceaed | 1678 | if (vlans) { |
49a73e0c | 1679 | ctx->wc->masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK); |
9583bc14 | 1680 | } |
7efbc3b7 | 1681 | if (vlans && !bitmap_is_set(vlans, vlan)) { |
9583bc14 EJ |
1682 | mirrors = zero_rightmost_1bit(mirrors); |
1683 | continue; | |
1684 | } | |
1685 | ||
faa624b4 BP |
1686 | /* Record the mirror, and the mirrors that output to the same |
1687 | * destination, so that we don't mirror to them again. This must be | |
1688 | * done now to ensure that output_normal(), below, doesn't recursively | |
1689 | * output to the same mirrors. */ | |
3d6151f3 | 1690 | ctx->mirrors |= dup_mirrors; |
faa624b4 BP |
1691 | |
1692 | /* Send the packet to the mirror. */ | |
ec7ceaed | 1693 | if (out) { |
84f0f298 RW |
1694 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
1695 | struct xbundle *out_xbundle = xbundle_lookup(xcfg, out); | |
46c88433 EJ |
1696 | if (out_xbundle) { |
1697 | output_normal(ctx, out_xbundle, vlan); | |
1698 | } | |
ec7ceaed | 1699 | } else if (vlan != out_vlan |
7efbc3b7 | 1700 | && !eth_addr_is_reserved(ctx->xin->flow.dl_dst)) { |
46c88433 | 1701 | struct xbundle *xbundle; |
9583bc14 | 1702 | |
46c88433 EJ |
1703 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
1704 | if (xbundle_includes_vlan(xbundle, out_vlan) | |
1705 | && !xbundle_mirror_out(xbridge, xbundle)) { | |
1706 | output_normal(ctx, xbundle, out_vlan); | |
9583bc14 EJ |
1707 | } |
1708 | } | |
1709 | } | |
faa624b4 BP |
1710 | |
1711 | /* output_normal() could have recursively output (to different | |
1712 | * mirrors), so make sure that we don't send duplicates. */ | |
1713 | mirrors &= ~ctx->mirrors; | |
9583bc14 EJ |
1714 | } |
1715 | } | |
1716 | ||
7efbc3b7 BP |
1717 | static void |
1718 | mirror_ingress_packet(struct xlate_ctx *ctx) | |
1719 | { | |
1720 | if (mbridge_has_mirrors(ctx->xbridge->mbridge)) { | |
1721 | bool warn = ctx->xin->packet != NULL; | |
1722 | struct xbundle *xbundle = lookup_input_bundle( | |
1723 | ctx->xbridge, ctx->xin->flow.in_port.ofp_port, warn, NULL); | |
1724 | if (xbundle) { | |
1725 | mirror_packet(ctx, xbundle, | |
1726 | xbundle_mirror_src(ctx->xbridge, xbundle)); | |
1727 | } | |
1728 | } | |
1729 | } | |
1730 | ||
9583bc14 | 1731 | /* Given 'vid', the VID obtained from the 802.1Q header that was received as |
46c88433 | 1732 | * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle', |
9583bc14 EJ |
1733 | * the bundle on which the packet was received, returns the VLAN to which the |
1734 | * packet belongs. | |
1735 | * | |
1736 | * Both 'vid' and the return value are in the range 0...4095. */ | |
1737 | static uint16_t | |
46c88433 | 1738 | input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid) |
9583bc14 | 1739 | { |
46c88433 | 1740 | switch (in_xbundle->vlan_mode) { |
9583bc14 | 1741 | case PORT_VLAN_ACCESS: |
46c88433 | 1742 | return in_xbundle->vlan; |
9583bc14 EJ |
1743 | break; |
1744 | ||
1745 | case PORT_VLAN_TRUNK: | |
1746 | return vid; | |
1747 | ||
1748 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1749 | case PORT_VLAN_NATIVE_TAGGED: | |
46c88433 | 1750 | return vid ? vid : in_xbundle->vlan; |
9583bc14 EJ |
1751 | |
1752 | default: | |
428b2edd | 1753 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1754 | } |
1755 | } | |
1756 | ||
46c88433 | 1757 | /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'. |
9583bc14 EJ |
1758 | * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs |
1759 | * a warning. | |
1760 | * | |
1761 | * 'vid' should be the VID obtained from the 802.1Q header that was received as | |
1762 | * part of a packet (specify 0 if there was no 802.1Q header), in the range | |
1763 | * 0...4095. */ | |
1764 | static bool | |
46c88433 | 1765 | input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn) |
9583bc14 EJ |
1766 | { |
1767 | /* Allow any VID on the OFPP_NONE port. */ | |
46c88433 | 1768 | if (in_xbundle == &ofpp_none_bundle) { |
9583bc14 EJ |
1769 | return true; |
1770 | } | |
1771 | ||
46c88433 | 1772 | switch (in_xbundle->vlan_mode) { |
9583bc14 EJ |
1773 | case PORT_VLAN_ACCESS: |
1774 | if (vid) { | |
1775 | if (warn) { | |
1776 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1777 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" tagged " |
9583bc14 | 1778 | "packet received on port %s configured as VLAN " |
46c88433 EJ |
1779 | "%"PRIu16" access port", vid, in_xbundle->name, |
1780 | in_xbundle->vlan); | |
9583bc14 EJ |
1781 | } |
1782 | return false; | |
1783 | } | |
1784 | return true; | |
1785 | ||
1786 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1787 | case PORT_VLAN_NATIVE_TAGGED: | |
1788 | if (!vid) { | |
1789 | /* Port must always carry its native VLAN. */ | |
1790 | return true; | |
1791 | } | |
1792 | /* Fall through. */ | |
1793 | case PORT_VLAN_TRUNK: | |
46c88433 | 1794 | if (!xbundle_includes_vlan(in_xbundle, vid)) { |
9583bc14 EJ |
1795 | if (warn) { |
1796 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1797 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" packet " |
9583bc14 | 1798 | "received on port %s not configured for trunking " |
46c88433 | 1799 | "VLAN %"PRIu16, vid, in_xbundle->name, vid); |
9583bc14 EJ |
1800 | } |
1801 | return false; | |
1802 | } | |
1803 | return true; | |
1804 | ||
1805 | default: | |
428b2edd | 1806 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1807 | } |
1808 | ||
1809 | } | |
1810 | ||
1811 | /* Given 'vlan', the VLAN that a packet belongs to, and | |
46c88433 | 1812 | * 'out_xbundle', a bundle on which the packet is to be output, returns the VID |
9583bc14 EJ |
1813 | * that should be included in the 802.1Q header. (If the return value is 0, |
1814 | * then the 802.1Q header should only be included in the packet if there is a | |
1815 | * nonzero PCP.) | |
1816 | * | |
1817 | * Both 'vlan' and the return value are in the range 0...4095. */ | |
1818 | static uint16_t | |
46c88433 | 1819 | output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan) |
9583bc14 | 1820 | { |
46c88433 | 1821 | switch (out_xbundle->vlan_mode) { |
9583bc14 EJ |
1822 | case PORT_VLAN_ACCESS: |
1823 | return 0; | |
1824 | ||
1825 | case PORT_VLAN_TRUNK: | |
1826 | case PORT_VLAN_NATIVE_TAGGED: | |
1827 | return vlan; | |
1828 | ||
1829 | case PORT_VLAN_NATIVE_UNTAGGED: | |
46c88433 | 1830 | return vlan == out_xbundle->vlan ? 0 : vlan; |
9583bc14 EJ |
1831 | |
1832 | default: | |
428b2edd | 1833 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1834 | } |
1835 | } | |
1836 | ||
1837 | static void | |
46c88433 | 1838 | output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, |
9583bc14 EJ |
1839 | uint16_t vlan) |
1840 | { | |
33bf9176 | 1841 | ovs_be16 *flow_tci = &ctx->xin->flow.vlan_tci; |
9583bc14 EJ |
1842 | uint16_t vid; |
1843 | ovs_be16 tci, old_tci; | |
46c88433 | 1844 | struct xport *xport; |
e93ef1c7 JR |
1845 | struct xlate_bond_recirc xr; |
1846 | bool use_recirc = false; | |
9583bc14 | 1847 | |
46c88433 | 1848 | vid = output_vlan_to_vid(out_xbundle, vlan); |
417e7e66 | 1849 | if (ovs_list_is_empty(&out_xbundle->xports)) { |
46c88433 EJ |
1850 | /* Partially configured bundle with no slaves. Drop the packet. */ |
1851 | return; | |
1852 | } else if (!out_xbundle->bond) { | |
417e7e66 | 1853 | xport = CONTAINER_OF(ovs_list_front(&out_xbundle->xports), struct xport, |
46c88433 | 1854 | bundle_node); |
9583bc14 | 1855 | } else { |
84f0f298 | 1856 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
49a73e0c | 1857 | struct flow_wildcards *wc = ctx->wc; |
84f0f298 | 1858 | struct ofport_dpif *ofport; |
adcf00ba | 1859 | |
2494ccd7 | 1860 | if (ctx->xbridge->support.odp.recirc) { |
e93ef1c7 JR |
1861 | use_recirc = bond_may_recirc( |
1862 | out_xbundle->bond, &xr.recirc_id, &xr.hash_basis); | |
adcf00ba | 1863 | |
e93ef1c7 | 1864 | if (use_recirc) { |
adcf00ba | 1865 | /* Only TCP mode uses recirculation. */ |
e93ef1c7 | 1866 | xr.hash_alg = OVS_HASH_ALG_L4; |
adcf00ba | 1867 | bond_update_post_recirc_rules(out_xbundle->bond, false); |
54ecb5a2 AZ |
1868 | |
1869 | /* Recirculation does not require unmasking hash fields. */ | |
1870 | wc = NULL; | |
adcf00ba AZ |
1871 | } |
1872 | } | |
46c88433 | 1873 | |
54ecb5a2 AZ |
1874 | ofport = bond_choose_output_slave(out_xbundle->bond, |
1875 | &ctx->xin->flow, wc, vid); | |
84f0f298 | 1876 | xport = xport_lookup(xcfg, ofport); |
46c88433 EJ |
1877 | |
1878 | if (!xport) { | |
9583bc14 EJ |
1879 | /* No slaves enabled, so drop packet. */ |
1880 | return; | |
1881 | } | |
d6fc5f57 | 1882 | |
e93ef1c7 | 1883 | /* If use_recirc is set, the main thread will handle stats |
b256dc52 | 1884 | * accounting for this bond. */ |
e93ef1c7 | 1885 | if (!use_recirc) { |
b256dc52 JS |
1886 | if (ctx->xin->resubmit_stats) { |
1887 | bond_account(out_xbundle->bond, &ctx->xin->flow, vid, | |
1888 | ctx->xin->resubmit_stats->n_bytes); | |
1889 | } | |
1890 | if (ctx->xin->xcache) { | |
1891 | struct xc_entry *entry; | |
1892 | struct flow *flow; | |
1893 | ||
1894 | flow = &ctx->xin->flow; | |
1895 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND); | |
1896 | entry->u.bond.bond = bond_ref(out_xbundle->bond); | |
1897 | entry->u.bond.flow = xmemdup(flow, sizeof *flow); | |
1898 | entry->u.bond.vid = vid; | |
1899 | } | |
d6fc5f57 | 1900 | } |
9583bc14 EJ |
1901 | } |
1902 | ||
33bf9176 | 1903 | old_tci = *flow_tci; |
9583bc14 | 1904 | tci = htons(vid); |
46c88433 | 1905 | if (tci || out_xbundle->use_priority_tags) { |
33bf9176 | 1906 | tci |= *flow_tci & htons(VLAN_PCP_MASK); |
9583bc14 EJ |
1907 | if (tci) { |
1908 | tci |= htons(VLAN_CFI); | |
1909 | } | |
1910 | } | |
33bf9176 | 1911 | *flow_tci = tci; |
9583bc14 | 1912 | |
e93ef1c7 | 1913 | compose_output_action(ctx, xport->ofp_port, use_recirc ? &xr : NULL); |
33bf9176 | 1914 | *flow_tci = old_tci; |
9583bc14 EJ |
1915 | } |
1916 | ||
1917 | /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after | |
1918 | * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to | |
1919 | * indicate this; newer upstream kernels use gratuitous ARP requests. */ | |
1920 | static bool | |
1921 | is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) | |
1922 | { | |
1923 | if (flow->dl_type != htons(ETH_TYPE_ARP)) { | |
1924 | return false; | |
1925 | } | |
1926 | ||
1927 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1928 | if (!eth_addr_is_broadcast(flow->dl_dst)) { | |
1929 | return false; | |
1930 | } | |
1931 | ||
1932 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
1933 | if (flow->nw_proto == ARP_OP_REPLY) { | |
1934 | return true; | |
1935 | } else if (flow->nw_proto == ARP_OP_REQUEST) { | |
1936 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
1937 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
1938 | ||
1939 | return flow->nw_src == flow->nw_dst; | |
1940 | } else { | |
1941 | return false; | |
1942 | } | |
1943 | } | |
1944 | ||
ff69c24a FL |
1945 | /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or |
1946 | * dropped. Returns true if they may be forwarded, false if they should be | |
1947 | * dropped. | |
1948 | * | |
1949 | * 'in_port' must be the xport that corresponds to flow->in_port. | |
1950 | * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). | |
1951 | * | |
1952 | * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as | |
1953 | * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as | |
1954 | * checked by input_vid_is_valid(). | |
1955 | * | |
1956 | * May also add tags to '*tags', although the current implementation only does | |
1957 | * so in one special case. | |
1958 | */ | |
1959 | static bool | |
1960 | is_admissible(struct xlate_ctx *ctx, struct xport *in_port, | |
1961 | uint16_t vlan) | |
1962 | { | |
1963 | struct xbundle *in_xbundle = in_port->xbundle; | |
1964 | const struct xbridge *xbridge = ctx->xbridge; | |
1965 | struct flow *flow = &ctx->xin->flow; | |
1966 | ||
1967 | /* Drop frames for reserved multicast addresses | |
1968 | * only if forward_bpdu option is absent. */ | |
1969 | if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { | |
1970 | xlate_report(ctx, "packet has reserved destination MAC, dropping"); | |
1971 | return false; | |
1972 | } | |
1973 | ||
1974 | if (in_xbundle->bond) { | |
1975 | struct mac_entry *mac; | |
1976 | ||
1977 | switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, | |
1978 | flow->dl_dst)) { | |
1979 | case BV_ACCEPT: | |
1980 | break; | |
1981 | ||
1982 | case BV_DROP: | |
1983 | xlate_report(ctx, "bonding refused admissibility, dropping"); | |
1984 | return false; | |
1985 | ||
1986 | case BV_DROP_IF_MOVED: | |
1987 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
1988 | mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan); | |
9d078ec2 BP |
1989 | if (mac |
1990 | && mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle | |
49a73e0c | 1991 | && (!is_gratuitous_arp(flow, ctx->wc) |
9d078ec2 | 1992 | || mac_entry_is_grat_arp_locked(mac))) { |
ff69c24a FL |
1993 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
1994 | xlate_report(ctx, "SLB bond thinks this packet looped back, " | |
1995 | "dropping"); | |
1996 | return false; | |
1997 | } | |
1998 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
1999 | break; | |
2000 | } | |
2001 | } | |
2002 | ||
2003 | return true; | |
2004 | } | |
2005 | ||
ee047520 BP |
2006 | /* Checks whether a MAC learning update is necessary for MAC learning table |
2007 | * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in | |
2008 | * 'vlan'. | |
2009 | * | |
2010 | * Most packets processed through the MAC learning table do not actually | |
2011 | * change it in any way. This function requires only a read lock on the MAC | |
2012 | * learning table, so it is much cheaper in this common case. | |
2013 | * | |
2014 | * Keep the code here synchronized with that in update_learning_table__() | |
2015 | * below. */ | |
2016 | static bool | |
2017 | is_mac_learning_update_needed(const struct mac_learning *ml, | |
2018 | const struct flow *flow, | |
2019 | struct flow_wildcards *wc, | |
2020 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 2021 | OVS_REQ_RDLOCK(ml->rwlock) |
9583bc14 EJ |
2022 | { |
2023 | struct mac_entry *mac; | |
2024 | ||
ee047520 BP |
2025 | if (!mac_learning_may_learn(ml, flow->dl_src, vlan)) { |
2026 | return false; | |
2027 | } | |
2028 | ||
2029 | mac = mac_learning_lookup(ml, flow->dl_src, vlan); | |
2030 | if (!mac || mac_entry_age(ml, mac)) { | |
2031 | return true; | |
9583bc14 EJ |
2032 | } |
2033 | ||
ee047520 BP |
2034 | if (is_gratuitous_arp(flow, wc)) { |
2035 | /* We don't want to learn from gratuitous ARP packets that are | |
2036 | * reflected back over bond slaves so we lock the learning table. */ | |
2037 | if (!in_xbundle->bond) { | |
2038 | return true; | |
2039 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
2040 | return false; | |
2041 | } | |
2042 | } | |
2043 | ||
9d078ec2 | 2044 | return mac_entry_get_port(ml, mac) != in_xbundle->ofbundle; |
ee047520 BP |
2045 | } |
2046 | ||
2047 | ||
2048 | /* Updates MAC learning table 'ml' given that a packet matching 'flow' was | |
2049 | * received on 'in_xbundle' in 'vlan'. | |
2050 | * | |
2051 | * This code repeats all the checks in is_mac_learning_update_needed() because | |
2052 | * the lock was released between there and here and thus the MAC learning state | |
2053 | * could have changed. | |
2054 | * | |
2055 | * Keep the code here synchronized with that in is_mac_learning_update_needed() | |
2056 | * above. */ | |
2057 | static void | |
2058 | update_learning_table__(const struct xbridge *xbridge, | |
2059 | const struct flow *flow, struct flow_wildcards *wc, | |
2060 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 2061 | OVS_REQ_WRLOCK(xbridge->ml->rwlock) |
ee047520 BP |
2062 | { |
2063 | struct mac_entry *mac; | |
2064 | ||
46c88433 | 2065 | if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) { |
ee047520 | 2066 | return; |
9583bc14 EJ |
2067 | } |
2068 | ||
46c88433 | 2069 | mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan); |
9583bc14 EJ |
2070 | if (is_gratuitous_arp(flow, wc)) { |
2071 | /* We don't want to learn from gratuitous ARP packets that are | |
2072 | * reflected back over bond slaves so we lock the learning table. */ | |
46c88433 | 2073 | if (!in_xbundle->bond) { |
9583bc14 EJ |
2074 | mac_entry_set_grat_arp_lock(mac); |
2075 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
ee047520 | 2076 | return; |
9583bc14 EJ |
2077 | } |
2078 | } | |
2079 | ||
9d078ec2 | 2080 | if (mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle) { |
9583bc14 EJ |
2081 | /* The log messages here could actually be useful in debugging, |
2082 | * so keep the rate limit relatively high. */ | |
2083 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
ee047520 | 2084 | |
9583bc14 EJ |
2085 | VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is " |
2086 | "on port %s in VLAN %d", | |
46c88433 EJ |
2087 | xbridge->name, ETH_ADDR_ARGS(flow->dl_src), |
2088 | in_xbundle->name, vlan); | |
9583bc14 | 2089 | |
9d078ec2 | 2090 | mac_entry_set_port(xbridge->ml, mac, in_xbundle->ofbundle); |
9583bc14 | 2091 | } |
ee047520 BP |
2092 | } |
2093 | ||
2094 | static void | |
2095 | update_learning_table(const struct xbridge *xbridge, | |
2096 | const struct flow *flow, struct flow_wildcards *wc, | |
2097 | int vlan, struct xbundle *in_xbundle) | |
2098 | { | |
2099 | bool need_update; | |
2100 | ||
2101 | /* Don't learn the OFPP_NONE port. */ | |
2102 | if (in_xbundle == &ofpp_none_bundle) { | |
2103 | return; | |
2104 | } | |
2105 | ||
2106 | /* First try the common case: no change to MAC learning table. */ | |
2107 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
2108 | need_update = is_mac_learning_update_needed(xbridge->ml, flow, wc, vlan, | |
2109 | in_xbundle); | |
509c0149 | 2110 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
ee047520 BP |
2111 | |
2112 | if (need_update) { | |
2113 | /* Slow path: MAC learning table might need an update. */ | |
2114 | ovs_rwlock_wrlock(&xbridge->ml->rwlock); | |
2115 | update_learning_table__(xbridge, flow, wc, vlan, in_xbundle); | |
2116 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
2117 | } | |
9583bc14 EJ |
2118 | } |
2119 | ||
86e2dcdd FL |
2120 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2121 | * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */ | |
2122 | static void | |
06994f87 TLSC |
2123 | update_mcast_snooping_table4__(const struct xbridge *xbridge, |
2124 | const struct flow *flow, | |
2125 | struct mcast_snooping *ms, int vlan, | |
2126 | struct xbundle *in_xbundle, | |
2127 | const struct dp_packet *packet) | |
86e2dcdd FL |
2128 | OVS_REQ_WRLOCK(ms->rwlock) |
2129 | { | |
2130 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
e3102e42 | 2131 | int count; |
06994f87 | 2132 | ovs_be32 ip4 = flow->igmp_group_ip4; |
86e2dcdd FL |
2133 | |
2134 | switch (ntohs(flow->tp_src)) { | |
2135 | case IGMP_HOST_MEMBERSHIP_REPORT: | |
2136 | case IGMPV2_HOST_MEMBERSHIP_REPORT: | |
964a4d5f | 2137 | if (mcast_snooping_add_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
86e2dcdd FL |
2138 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping learned that " |
2139 | IP_FMT" is on port %s in VLAN %d", | |
2140 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
2141 | } | |
2142 | break; | |
2143 | case IGMP_HOST_LEAVE_MESSAGE: | |
964a4d5f | 2144 | if (mcast_snooping_leave_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
86e2dcdd FL |
2145 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping leaving " |
2146 | IP_FMT" is on port %s in VLAN %d", | |
2147 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
2148 | } | |
2149 | break; | |
2150 | case IGMP_HOST_MEMBERSHIP_QUERY: | |
2151 | if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan, | |
2152 | in_xbundle->ofbundle)) { | |
2153 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query from " | |
2154 | IP_FMT" is on port %s in VLAN %d", | |
2155 | xbridge->name, IP_ARGS(flow->nw_src), | |
2156 | in_xbundle->name, vlan); | |
2157 | } | |
2158 | break; | |
e3102e42 TLSC |
2159 | case IGMPV3_HOST_MEMBERSHIP_REPORT: |
2160 | if ((count = mcast_snooping_add_report(ms, packet, vlan, | |
2161 | in_xbundle->ofbundle))) { | |
2162 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping processed %d " | |
2163 | "addresses on port %s in VLAN %d", | |
2164 | xbridge->name, count, in_xbundle->name, vlan); | |
2165 | } | |
2166 | break; | |
86e2dcdd FL |
2167 | } |
2168 | } | |
2169 | ||
06994f87 TLSC |
2170 | static void |
2171 | update_mcast_snooping_table6__(const struct xbridge *xbridge, | |
2172 | const struct flow *flow, | |
2173 | struct mcast_snooping *ms, int vlan, | |
2174 | struct xbundle *in_xbundle, | |
2175 | const struct dp_packet *packet) | |
2176 | OVS_REQ_WRLOCK(ms->rwlock) | |
2177 | { | |
2178 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
2179 | int count; | |
2180 | ||
2181 | switch (ntohs(flow->tp_src)) { | |
2182 | case MLD_QUERY: | |
2183 | if (!ipv6_addr_equals(&flow->ipv6_src, &in6addr_any) | |
2184 | && mcast_snooping_add_mrouter(ms, vlan, in_xbundle->ofbundle)) { | |
2185 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query on port %s" | |
2186 | "in VLAN %d", | |
2187 | xbridge->name, in_xbundle->name, vlan); | |
2188 | } | |
2189 | break; | |
2190 | case MLD_REPORT: | |
2191 | case MLD_DONE: | |
2192 | case MLD2_REPORT: | |
2193 | count = mcast_snooping_add_mld(ms, packet, vlan, in_xbundle->ofbundle); | |
2194 | if (count) { | |
2195 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping processed %d " | |
2196 | "addresses on port %s in VLAN %d", | |
2197 | xbridge->name, count, in_xbundle->name, vlan); | |
2198 | } | |
2199 | break; | |
2200 | } | |
2201 | } | |
2202 | ||
86e2dcdd FL |
2203 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2204 | * was received on 'in_xbundle' in 'vlan'. */ | |
2205 | static void | |
2206 | update_mcast_snooping_table(const struct xbridge *xbridge, | |
2207 | const struct flow *flow, int vlan, | |
e3102e42 TLSC |
2208 | struct xbundle *in_xbundle, |
2209 | const struct dp_packet *packet) | |
86e2dcdd FL |
2210 | { |
2211 | struct mcast_snooping *ms = xbridge->ms; | |
2212 | struct xlate_cfg *xcfg; | |
2213 | struct xbundle *mcast_xbundle; | |
f4ae6e23 | 2214 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2215 | |
2216 | /* Don't learn the OFPP_NONE port. */ | |
2217 | if (in_xbundle == &ofpp_none_bundle) { | |
2218 | return; | |
2219 | } | |
2220 | ||
2221 | /* Don't learn from flood ports */ | |
2222 | mcast_xbundle = NULL; | |
2223 | ovs_rwlock_wrlock(&ms->rwlock); | |
2224 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2225 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2226 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2227 | if (mcast_xbundle == in_xbundle) { | |
2228 | break; | |
2229 | } | |
2230 | } | |
2231 | ||
2232 | if (!mcast_xbundle || mcast_xbundle != in_xbundle) { | |
06994f87 TLSC |
2233 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2234 | update_mcast_snooping_table4__(xbridge, flow, ms, vlan, | |
2235 | in_xbundle, packet); | |
2236 | } else { | |
2237 | update_mcast_snooping_table6__(xbridge, flow, ms, vlan, | |
2238 | in_xbundle, packet); | |
2239 | } | |
86e2dcdd FL |
2240 | } |
2241 | ovs_rwlock_unlock(&ms->rwlock); | |
2242 | } | |
2243 | ||
2244 | /* send the packet to ports having the multicast group learned */ | |
2245 | static void | |
2246 | xlate_normal_mcast_send_group(struct xlate_ctx *ctx, | |
2247 | struct mcast_snooping *ms OVS_UNUSED, | |
2248 | struct mcast_group *grp, | |
2249 | struct xbundle *in_xbundle, uint16_t vlan) | |
2250 | OVS_REQ_RDLOCK(ms->rwlock) | |
2251 | { | |
2252 | struct xlate_cfg *xcfg; | |
2253 | struct mcast_group_bundle *b; | |
2254 | struct xbundle *mcast_xbundle; | |
2255 | ||
2256 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2257 | LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) { | |
2258 | mcast_xbundle = xbundle_lookup(xcfg, b->port); | |
2259 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2260 | xlate_report(ctx, "forwarding to mcast group port"); | |
2261 | output_normal(ctx, mcast_xbundle, vlan); | |
2262 | } else if (!mcast_xbundle) { | |
2263 | xlate_report(ctx, "mcast group port is unknown, dropping"); | |
2264 | } else { | |
2265 | xlate_report(ctx, "mcast group port is input port, dropping"); | |
2266 | } | |
2267 | } | |
2268 | } | |
2269 | ||
2270 | /* send the packet to ports connected to multicast routers */ | |
2271 | static void | |
2272 | xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx, | |
2273 | struct mcast_snooping *ms, | |
2274 | struct xbundle *in_xbundle, uint16_t vlan) | |
2275 | OVS_REQ_RDLOCK(ms->rwlock) | |
2276 | { | |
2277 | struct xlate_cfg *xcfg; | |
2278 | struct mcast_mrouter_bundle *mrouter; | |
2279 | struct xbundle *mcast_xbundle; | |
2280 | ||
2281 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2282 | LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) { | |
2283 | mcast_xbundle = xbundle_lookup(xcfg, mrouter->port); | |
2284 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2285 | xlate_report(ctx, "forwarding to mcast router port"); | |
2286 | output_normal(ctx, mcast_xbundle, vlan); | |
2287 | } else if (!mcast_xbundle) { | |
2288 | xlate_report(ctx, "mcast router port is unknown, dropping"); | |
2289 | } else { | |
2290 | xlate_report(ctx, "mcast router port is input port, dropping"); | |
2291 | } | |
2292 | } | |
2293 | } | |
2294 | ||
2295 | /* send the packet to ports flagged to be flooded */ | |
2296 | static void | |
2297 | xlate_normal_mcast_send_fports(struct xlate_ctx *ctx, | |
2298 | struct mcast_snooping *ms, | |
2299 | struct xbundle *in_xbundle, uint16_t vlan) | |
2300 | OVS_REQ_RDLOCK(ms->rwlock) | |
2301 | { | |
2302 | struct xlate_cfg *xcfg; | |
f4ae6e23 | 2303 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2304 | struct xbundle *mcast_xbundle; |
2305 | ||
2306 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2307 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2308 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2309 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2310 | xlate_report(ctx, "forwarding to mcast flood port"); | |
2311 | output_normal(ctx, mcast_xbundle, vlan); | |
2312 | } else if (!mcast_xbundle) { | |
2313 | xlate_report(ctx, "mcast flood port is unknown, dropping"); | |
2314 | } else { | |
2315 | xlate_report(ctx, "mcast flood port is input port, dropping"); | |
2316 | } | |
2317 | } | |
2318 | } | |
2319 | ||
8e04a33f FL |
2320 | /* forward the Reports to configured ports */ |
2321 | static void | |
2322 | xlate_normal_mcast_send_rports(struct xlate_ctx *ctx, | |
2323 | struct mcast_snooping *ms, | |
2324 | struct xbundle *in_xbundle, uint16_t vlan) | |
2325 | OVS_REQ_RDLOCK(ms->rwlock) | |
2326 | { | |
2327 | struct xlate_cfg *xcfg; | |
2328 | struct mcast_port_bundle *rport; | |
2329 | struct xbundle *mcast_xbundle; | |
2330 | ||
2331 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2332 | LIST_FOR_EACH(rport, node, &ms->rport_list) { | |
2333 | mcast_xbundle = xbundle_lookup(xcfg, rport->port); | |
2334 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2335 | xlate_report(ctx, "forwarding Report to mcast flagged port"); | |
2336 | output_normal(ctx, mcast_xbundle, vlan); | |
2337 | } else if (!mcast_xbundle) { | |
2338 | xlate_report(ctx, "mcast port is unknown, dropping the Report"); | |
2339 | } else { | |
2340 | xlate_report(ctx, "mcast port is input port, dropping the Report"); | |
2341 | } | |
2342 | } | |
2343 | } | |
2344 | ||
682800a4 FL |
2345 | static void |
2346 | xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, | |
2347 | uint16_t vlan) | |
2348 | { | |
2349 | struct xbundle *xbundle; | |
2350 | ||
2351 | LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { | |
2352 | if (xbundle != in_xbundle | |
2353 | && xbundle_includes_vlan(xbundle, vlan) | |
2354 | && xbundle->floodable | |
2355 | && !xbundle_mirror_out(ctx->xbridge, xbundle)) { | |
2356 | output_normal(ctx, xbundle, vlan); | |
2357 | } | |
2358 | } | |
2031ef97 | 2359 | ctx->nf_output_iface = NF_OUT_FLOOD; |
682800a4 FL |
2360 | } |
2361 | ||
9583bc14 EJ |
2362 | static void |
2363 | xlate_normal(struct xlate_ctx *ctx) | |
2364 | { | |
49a73e0c | 2365 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 2366 | struct flow *flow = &ctx->xin->flow; |
46c88433 EJ |
2367 | struct xbundle *in_xbundle; |
2368 | struct xport *in_port; | |
9583bc14 | 2369 | struct mac_entry *mac; |
d6d5bbc9 | 2370 | void *mac_port; |
9583bc14 EJ |
2371 | uint16_t vlan; |
2372 | uint16_t vid; | |
2373 | ||
33bf9176 BP |
2374 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
2375 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1dd35f8a | 2376 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
9583bc14 | 2377 | |
46c88433 EJ |
2378 | in_xbundle = lookup_input_bundle(ctx->xbridge, flow->in_port.ofp_port, |
2379 | ctx->xin->packet != NULL, &in_port); | |
2380 | if (!in_xbundle) { | |
9583bc14 EJ |
2381 | xlate_report(ctx, "no input bundle, dropping"); |
2382 | return; | |
2383 | } | |
2384 | ||
2385 | /* Drop malformed frames. */ | |
33bf9176 BP |
2386 | if (flow->dl_type == htons(ETH_TYPE_VLAN) && |
2387 | !(flow->vlan_tci & htons(VLAN_CFI))) { | |
9583bc14 EJ |
2388 | if (ctx->xin->packet != NULL) { |
2389 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2390 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet with partial " | |
2391 | "VLAN tag received on port %s", | |
46c88433 | 2392 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2393 | } |
2394 | xlate_report(ctx, "partial VLAN tag, dropping"); | |
2395 | return; | |
2396 | } | |
2397 | ||
2398 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 2399 | if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) { |
9583bc14 EJ |
2400 | if (ctx->xin->packet != NULL) { |
2401 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2402 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
2403 | "%s, which is reserved exclusively for mirroring", | |
46c88433 | 2404 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2405 | } |
2406 | xlate_report(ctx, "input port is mirror output port, dropping"); | |
2407 | return; | |
2408 | } | |
2409 | ||
2410 | /* Check VLAN. */ | |
33bf9176 | 2411 | vid = vlan_tci_to_vid(flow->vlan_tci); |
46c88433 | 2412 | if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { |
9583bc14 EJ |
2413 | xlate_report(ctx, "disallowed VLAN VID for this input port, dropping"); |
2414 | return; | |
2415 | } | |
46c88433 | 2416 | vlan = input_vid_to_vlan(in_xbundle, vid); |
9583bc14 EJ |
2417 | |
2418 | /* Check other admissibility requirements. */ | |
2419 | if (in_port && !is_admissible(ctx, in_port, vlan)) { | |
2420 | return; | |
2421 | } | |
2422 | ||
2423 | /* Learn source MAC. */ | |
2424 | if (ctx->xin->may_learn) { | |
46c88433 | 2425 | update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle); |
9583bc14 | 2426 | } |
b256dc52 JS |
2427 | if (ctx->xin->xcache) { |
2428 | struct xc_entry *entry; | |
2429 | ||
2430 | /* Save enough info to update mac learning table later. */ | |
2431 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL); | |
9edf6b48 | 2432 | entry->u.normal.ofproto = ctx->xbridge->ofproto; |
b256dc52 JS |
2433 | entry->u.normal.flow = xmemdup(flow, sizeof *flow); |
2434 | entry->u.normal.vlan = vlan; | |
2435 | } | |
9583bc14 EJ |
2436 | |
2437 | /* Determine output bundle. */ | |
86e2dcdd FL |
2438 | if (mcast_snooping_enabled(ctx->xbridge->ms) |
2439 | && !eth_addr_is_broadcast(flow->dl_dst) | |
2440 | && eth_addr_is_multicast(flow->dl_dst) | |
06994f87 | 2441 | && is_ip_any(flow)) { |
86e2dcdd | 2442 | struct mcast_snooping *ms = ctx->xbridge->ms; |
06994f87 | 2443 | struct mcast_group *grp = NULL; |
86e2dcdd | 2444 | |
06994f87 | 2445 | if (is_igmp(flow)) { |
d29f137b TLSC |
2446 | if (mcast_snooping_is_membership(flow->tp_src) || |
2447 | mcast_snooping_is_query(flow->tp_src)) { | |
f7c5f6aa | 2448 | if (ctx->xin->may_learn && ctx->xin->packet) { |
86e2dcdd | 2449 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, |
e3102e42 | 2450 | in_xbundle, ctx->xin->packet); |
d29f137b TLSC |
2451 | } |
2452 | /* | |
2453 | * IGMP packets need to take the slow path, in order to be | |
2454 | * processed for mdb updates. That will prevent expires | |
2455 | * firing off even after hosts have sent reports. | |
2456 | */ | |
2457 | ctx->xout->slow |= SLOW_ACTION; | |
86e2dcdd | 2458 | } |
d6d5bbc9 | 2459 | |
86e2dcdd FL |
2460 | if (mcast_snooping_is_membership(flow->tp_src)) { |
2461 | ovs_rwlock_rdlock(&ms->rwlock); | |
2462 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
8e04a33f FL |
2463 | /* RFC4541: section 2.1.1, item 1: A snooping switch should |
2464 | * forward IGMP Membership Reports only to those ports where | |
2465 | * multicast routers are attached. Alternatively stated: a | |
2466 | * snooping switch should not forward IGMP Membership Reports | |
2467 | * to ports on which only hosts are attached. | |
2468 | * An administrative control may be provided to override this | |
2469 | * restriction, allowing the report messages to be flooded to | |
2470 | * other ports. */ | |
2471 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
86e2dcdd FL |
2472 | ovs_rwlock_unlock(&ms->rwlock); |
2473 | } else { | |
2474 | xlate_report(ctx, "multicast traffic, flooding"); | |
2475 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2476 | } | |
2477 | return; | |
06994f87 TLSC |
2478 | } else if (is_mld(flow)) { |
2479 | ctx->xout->slow |= SLOW_ACTION; | |
f7c5f6aa | 2480 | if (ctx->xin->may_learn && ctx->xin->packet) { |
06994f87 TLSC |
2481 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, |
2482 | in_xbundle, ctx->xin->packet); | |
2483 | } | |
2484 | if (is_mld_report(flow)) { | |
2485 | ovs_rwlock_rdlock(&ms->rwlock); | |
2486 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2487 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
2488 | ovs_rwlock_unlock(&ms->rwlock); | |
2489 | } else { | |
2490 | xlate_report(ctx, "MLD query, flooding"); | |
2491 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2492 | } | |
86e2dcdd | 2493 | } else { |
06994f87 TLSC |
2494 | if ((flow->dl_type == htons(ETH_TYPE_IP) |
2495 | && ip_is_local_multicast(flow->nw_dst)) | |
2496 | || (flow->dl_type == htons(ETH_TYPE_IPV6) | |
2497 | && ipv6_is_all_hosts(&flow->ipv6_dst))) { | |
86e2dcdd FL |
2498 | /* RFC4541: section 2.1.2, item 2: Packets with a dst IP |
2499 | * address in the 224.0.0.x range which are not IGMP must | |
2500 | * be forwarded on all ports */ | |
2501 | xlate_report(ctx, "RFC4541: section 2.1.2, item 2, flooding"); | |
2502 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2503 | return; | |
2504 | } | |
2505 | } | |
2506 | ||
2507 | /* forwarding to group base ports */ | |
2508 | ovs_rwlock_rdlock(&ms->rwlock); | |
06994f87 TLSC |
2509 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2510 | grp = mcast_snooping_lookup4(ms, flow->nw_dst, vlan); | |
2511 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
2512 | grp = mcast_snooping_lookup(ms, &flow->ipv6_dst, vlan); | |
2513 | } | |
86e2dcdd FL |
2514 | if (grp) { |
2515 | xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, vlan); | |
2516 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2517 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
9583bc14 | 2518 | } else { |
86e2dcdd FL |
2519 | if (mcast_snooping_flood_unreg(ms)) { |
2520 | xlate_report(ctx, "unregistered multicast, flooding"); | |
2521 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2522 | } else { | |
2523 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2524 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2525 | } | |
9583bc14 | 2526 | } |
86e2dcdd | 2527 | ovs_rwlock_unlock(&ms->rwlock); |
9583bc14 | 2528 | } else { |
86e2dcdd FL |
2529 | ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); |
2530 | mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); | |
9d078ec2 | 2531 | mac_port = mac ? mac_entry_get_port(ctx->xbridge->ml, mac) : NULL; |
86e2dcdd FL |
2532 | ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); |
2533 | ||
2534 | if (mac_port) { | |
2535 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2536 | struct xbundle *mac_xbundle = xbundle_lookup(xcfg, mac_port); | |
2537 | if (mac_xbundle && mac_xbundle != in_xbundle) { | |
2538 | xlate_report(ctx, "forwarding to learned port"); | |
2539 | output_normal(ctx, mac_xbundle, vlan); | |
2540 | } else if (!mac_xbundle) { | |
2541 | xlate_report(ctx, "learned port is unknown, dropping"); | |
2542 | } else { | |
2543 | xlate_report(ctx, "learned port is input port, dropping"); | |
2544 | } | |
2545 | } else { | |
2546 | xlate_report(ctx, "no learned MAC for destination, flooding"); | |
2547 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2548 | } | |
9583bc14 EJ |
2549 | } |
2550 | } | |
2551 | ||
a6092018 BP |
2552 | /* Appends a "sample" action for sFlow or IPFIX to 'ctx->odp_actions'. The |
2553 | * 'probability' is the number of packets out of UINT32_MAX to sample. The | |
2554 | * 'cookie' (of length 'cookie_size' bytes) is passed back in the callback for | |
2555 | * each sampled packet. 'tunnel_out_port', if not ODPP_NONE, is added as the | |
2556 | * OVS_USERSPACE_ATTR_EGRESS_TUN_PORT attribute. If 'include_actions', an | |
2557 | * OVS_USERSPACE_ATTR_ACTIONS attribute is added. | |
9583bc14 EJ |
2558 | */ |
2559 | static size_t | |
a6092018 | 2560 | compose_sample_action(struct xlate_ctx *ctx, |
9583bc14 EJ |
2561 | const uint32_t probability, |
2562 | const union user_action_cookie *cookie, | |
8b7ea2d4 | 2563 | const size_t cookie_size, |
7321bda3 NM |
2564 | const odp_port_t tunnel_out_port, |
2565 | bool include_actions) | |
9583bc14 | 2566 | { |
a6092018 BP |
2567 | size_t sample_offset = nl_msg_start_nested(ctx->odp_actions, |
2568 | OVS_ACTION_ATTR_SAMPLE); | |
9583bc14 | 2569 | |
a6092018 | 2570 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability); |
9583bc14 | 2571 | |
a6092018 BP |
2572 | size_t actions_offset = nl_msg_start_nested(ctx->odp_actions, |
2573 | OVS_SAMPLE_ATTR_ACTIONS); | |
9583bc14 | 2574 | |
a6092018 BP |
2575 | odp_port_t odp_port = ofp_port_to_odp_port( |
2576 | ctx->xbridge, ctx->xin->flow.in_port.ofp_port); | |
2577 | uint32_t pid = dpif_port_get_pid(ctx->xbridge->dpif, odp_port, | |
2578 | flow_hash_5tuple(&ctx->xin->flow, 0)); | |
2579 | int cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, | |
2580 | tunnel_out_port, | |
2581 | include_actions, | |
2582 | ctx->odp_actions); | |
89a8a7f0 | 2583 | |
a6092018 BP |
2584 | nl_msg_end_nested(ctx->odp_actions, actions_offset); |
2585 | nl_msg_end_nested(ctx->odp_actions, sample_offset); | |
9583bc14 | 2586 | |
9583bc14 EJ |
2587 | return cookie_offset; |
2588 | } | |
2589 | ||
a6092018 BP |
2590 | /* If sFLow is not enabled, returns 0 without doing anything. |
2591 | * | |
2592 | * If sFlow is enabled, appends a template "sample" action to the ODP actions | |
2593 | * in 'ctx'. This action is a template because some of the information needed | |
2594 | * to fill it out is not available until flow translation is complete. In this | |
2595 | * case, this functions returns an offset, which is always nonzero, to pass | |
2596 | * later to fix_sflow_action() to fill in the rest of the template. */ | |
9583bc14 | 2597 | static size_t |
a6092018 | 2598 | compose_sflow_action(struct xlate_ctx *ctx) |
9583bc14 | 2599 | { |
a6092018 BP |
2600 | struct dpif_sflow *sflow = ctx->xbridge->sflow; |
2601 | if (!sflow || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { | |
9583bc14 EJ |
2602 | return 0; |
2603 | } | |
2604 | ||
a6092018 BP |
2605 | union user_action_cookie cookie = { .type = USER_ACTION_COOKIE_SFLOW }; |
2606 | return compose_sample_action(ctx, dpif_sflow_get_probability(sflow), | |
7321bda3 NM |
2607 | &cookie, sizeof cookie.sflow, ODPP_NONE, |
2608 | true); | |
9583bc14 EJ |
2609 | } |
2610 | ||
a6092018 BP |
2611 | /* If IPFIX is enabled, this appends a "sample" action to implement IPFIX to |
2612 | * 'ctx->odp_actions'. */ | |
9583bc14 | 2613 | static void |
a6092018 | 2614 | compose_ipfix_action(struct xlate_ctx *ctx, odp_port_t output_odp_port) |
9583bc14 | 2615 | { |
a6092018 | 2616 | struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; |
8b7ea2d4 | 2617 | odp_port_t tunnel_out_port = ODPP_NONE; |
9583bc14 | 2618 | |
a6092018 | 2619 | if (!ipfix || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
2620 | return; |
2621 | } | |
2622 | ||
8b7ea2d4 WZ |
2623 | /* For input case, output_odp_port is ODPP_NONE, which is an invalid port |
2624 | * number. */ | |
2625 | if (output_odp_port == ODPP_NONE && | |
a6092018 | 2626 | !dpif_ipfix_get_bridge_exporter_input_sampling(ipfix)) { |
8b7ea2d4 WZ |
2627 | return; |
2628 | } | |
2629 | ||
2630 | /* For output case, output_odp_port is valid*/ | |
2631 | if (output_odp_port != ODPP_NONE) { | |
a6092018 | 2632 | if (!dpif_ipfix_get_bridge_exporter_output_sampling(ipfix)) { |
8b7ea2d4 WZ |
2633 | return; |
2634 | } | |
2635 | /* If tunnel sampling is enabled, put an additional option attribute: | |
2636 | * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT | |
2637 | */ | |
a6092018 BP |
2638 | if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(ipfix) && |
2639 | dpif_ipfix_get_tunnel_port(ipfix, output_odp_port) ) { | |
8b7ea2d4 WZ |
2640 | tunnel_out_port = output_odp_port; |
2641 | } | |
2642 | } | |
2643 | ||
a6092018 BP |
2644 | union user_action_cookie cookie = { |
2645 | .ipfix = { | |
2646 | .type = USER_ACTION_COOKIE_IPFIX, | |
2647 | .output_odp_port = output_odp_port, | |
2648 | } | |
2649 | }; | |
2650 | compose_sample_action(ctx, | |
2651 | dpif_ipfix_get_bridge_exporter_probability(ipfix), | |
7321bda3 NM |
2652 | &cookie, sizeof cookie.ipfix, tunnel_out_port, |
2653 | false); | |
9583bc14 EJ |
2654 | } |
2655 | ||
a6092018 BP |
2656 | /* Fix "sample" action according to data collected while composing ODP actions, |
2657 | * as described in compose_sflow_action(). | |
2658 | * | |
2659 | * 'user_cookie_offset' must be the offset returned by add_sflow_action(). */ | |
9583bc14 | 2660 | static void |
a6092018 | 2661 | fix_sflow_action(struct xlate_ctx *ctx, unsigned int user_cookie_offset) |
9583bc14 EJ |
2662 | { |
2663 | const struct flow *base = &ctx->base_flow; | |
2664 | union user_action_cookie *cookie; | |
2665 | ||
a6092018 | 2666 | cookie = ofpbuf_at(ctx->odp_actions, user_cookie_offset, |
9583bc14 EJ |
2667 | sizeof cookie->sflow); |
2668 | ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); | |
2669 | ||
a6092018 BP |
2670 | cookie->type = USER_ACTION_COOKIE_SFLOW; |
2671 | cookie->sflow.vlan_tci = base->vlan_tci; | |
2672 | ||
2673 | /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output | |
2674 | * port information") for the interpretation of cookie->output. */ | |
2675 | switch (ctx->sflow_n_outputs) { | |
2676 | case 0: | |
2677 | /* 0x40000000 | 256 means "packet dropped for unknown reason". */ | |
2678 | cookie->sflow.output = 0x40000000 | 256; | |
2679 | break; | |
2680 | ||
2681 | case 1: | |
2682 | cookie->sflow.output = dpif_sflow_odp_port_to_ifindex( | |
2683 | ctx->xbridge->sflow, ctx->sflow_odp_port); | |
2684 | if (cookie->sflow.output) { | |
2685 | break; | |
2686 | } | |
2687 | /* Fall through. */ | |
2688 | default: | |
2689 | /* 0x80000000 means "multiple output ports. */ | |
2690 | cookie->sflow.output = 0x80000000 | ctx->sflow_n_outputs; | |
2691 | break; | |
2692 | } | |
9583bc14 EJ |
2693 | } |
2694 | ||
515793d5 BP |
2695 | static bool |
2696 | process_special(struct xlate_ctx *ctx, const struct xport *xport) | |
db7d4e46 | 2697 | { |
515793d5 | 2698 | const struct flow *flow = &ctx->xin->flow; |
49a73e0c | 2699 | struct flow_wildcards *wc = ctx->wc; |
46c88433 | 2700 | const struct xbridge *xbridge = ctx->xbridge; |
515793d5 BP |
2701 | const struct dp_packet *packet = ctx->xin->packet; |
2702 | enum slow_path_reason slow; | |
642dc74d | 2703 | |
46c88433 | 2704 | if (!xport) { |
515793d5 | 2705 | slow = 0; |
46c88433 | 2706 | } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) { |
db7d4e46 | 2707 | if (packet) { |
46c88433 | 2708 | cfm_process_heartbeat(xport->cfm, packet); |
db7d4e46 | 2709 | } |
515793d5 | 2710 | slow = SLOW_CFM; |
fab52e16 | 2711 | } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) { |
db7d4e46 | 2712 | if (packet) { |
46c88433 | 2713 | bfd_process_packet(xport->bfd, flow, packet); |
60d02c72 AW |
2714 | /* If POLL received, immediately sends FINAL back. */ |
2715 | if (bfd_should_send_packet(xport->bfd)) { | |
6d308b28 | 2716 | ofproto_dpif_monitor_port_send_soon(xport->ofport); |
60d02c72 | 2717 | } |
db7d4e46 | 2718 | } |
515793d5 | 2719 | slow = SLOW_BFD; |
46c88433 | 2720 | } else if (xport->xbundle && xport->xbundle->lacp |
db7d4e46 JP |
2721 | && flow->dl_type == htons(ETH_TYPE_LACP)) { |
2722 | if (packet) { | |
46c88433 | 2723 | lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet); |
db7d4e46 | 2724 | } |
515793d5 | 2725 | slow = SLOW_LACP; |
9efd308e DV |
2726 | } else if ((xbridge->stp || xbridge->rstp) && |
2727 | stp_should_process_flow(flow, wc)) { | |
db7d4e46 | 2728 | if (packet) { |
f025bcb7 JR |
2729 | xbridge->stp |
2730 | ? stp_process_packet(xport, packet) | |
2731 | : rstp_process_packet(xport, packet); | |
db7d4e46 | 2732 | } |
515793d5 | 2733 | slow = SLOW_STP; |
19aef6ef | 2734 | } else if (xport->lldp && lldp_should_process_flow(xport->lldp, flow)) { |
0477baa9 DF |
2735 | if (packet) { |
2736 | lldp_process_packet(xport->lldp, packet); | |
2737 | } | |
515793d5 | 2738 | slow = SLOW_LLDP; |
db7d4e46 | 2739 | } else { |
515793d5 BP |
2740 | slow = 0; |
2741 | } | |
2742 | ||
2743 | if (slow) { | |
2744 | ctx->xout->slow |= slow; | |
2745 | return true; | |
2746 | } else { | |
2747 | return false; | |
db7d4e46 JP |
2748 | } |
2749 | } | |
2750 | ||
a36de779 PS |
2751 | static int |
2752 | tnl_route_lookup_flow(const struct flow *oflow, | |
a8704b50 PS |
2753 | struct in6_addr *ip, struct in6_addr *src, |
2754 | struct xport **out_port) | |
a36de779 PS |
2755 | { |
2756 | char out_dev[IFNAMSIZ]; | |
2757 | struct xbridge *xbridge; | |
2758 | struct xlate_cfg *xcfg; | |
c2b878e0 TLSC |
2759 | struct in6_addr gw; |
2760 | struct in6_addr dst; | |
a36de779 | 2761 | |
c2b878e0 | 2762 | dst = flow_tnl_dst(&oflow->tunnel); |
a8704b50 | 2763 | if (!ovs_router_lookup(&dst, out_dev, src, &gw)) { |
a36de779 PS |
2764 | return -ENOENT; |
2765 | } | |
2766 | ||
c2b878e0 TLSC |
2767 | if (ipv6_addr_is_set(&gw) && |
2768 | (!IN6_IS_ADDR_V4MAPPED(&gw) || in6_addr_get_mapped_ipv4(&gw))) { | |
a36de779 PS |
2769 | *ip = gw; |
2770 | } else { | |
c2b878e0 | 2771 | *ip = dst; |
a36de779 PS |
2772 | } |
2773 | ||
2774 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2775 | ovs_assert(xcfg); | |
2776 | ||
2777 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
2778 | if (!strncmp(xbridge->name, out_dev, IFNAMSIZ)) { | |
2779 | struct xport *port; | |
2780 | ||
2781 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
2782 | if (!strncmp(netdev_get_name(port->netdev), out_dev, IFNAMSIZ)) { | |
2783 | *out_port = port; | |
2784 | return 0; | |
2785 | } | |
2786 | } | |
2787 | } | |
2788 | } | |
2789 | return -ENOENT; | |
2790 | } | |
2791 | ||
2792 | static int | |
cdd42eda JG |
2793 | compose_table_xlate(struct xlate_ctx *ctx, const struct xport *out_dev, |
2794 | struct dp_packet *packet) | |
a36de779 | 2795 | { |
cdd42eda | 2796 | struct xbridge *xbridge = out_dev->xbridge; |
a36de779 PS |
2797 | struct ofpact_output output; |
2798 | struct flow flow; | |
2799 | ||
2800 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); | |
cf62fa4c | 2801 | flow_extract(packet, &flow); |
cdd42eda JG |
2802 | flow.in_port.ofp_port = out_dev->ofp_port; |
2803 | output.port = OFPP_TABLE; | |
a36de779 PS |
2804 | output.max_len = 0; |
2805 | ||
cdd42eda JG |
2806 | return ofproto_dpif_execute_actions__(xbridge->ofproto, &flow, NULL, |
2807 | &output.ofpact, sizeof output, | |
9a9b001a | 2808 | ctx->indentation, ctx->resubmits, packet); |
a36de779 PS |
2809 | } |
2810 | ||
c2b878e0 TLSC |
2811 | static void |
2812 | tnl_send_nd_request(struct xlate_ctx *ctx, const struct xport *out_dev, | |
2813 | const struct eth_addr eth_src, | |
2814 | struct in6_addr * ipv6_src, struct in6_addr * ipv6_dst) | |
2815 | { | |
2816 | struct dp_packet packet; | |
2817 | ||
2818 | dp_packet_init(&packet, 0); | |
2819 | compose_nd(&packet, eth_src, ipv6_src, ipv6_dst); | |
2820 | compose_table_xlate(ctx, out_dev, &packet); | |
2821 | dp_packet_uninit(&packet); | |
2822 | } | |
2823 | ||
a36de779 | 2824 | static void |
cdd42eda | 2825 | tnl_send_arp_request(struct xlate_ctx *ctx, const struct xport *out_dev, |
74ff3298 | 2826 | const struct eth_addr eth_src, |
a36de779 PS |
2827 | ovs_be32 ip_src, ovs_be32 ip_dst) |
2828 | { | |
cf62fa4c | 2829 | struct dp_packet packet; |
a36de779 | 2830 | |
cf62fa4c | 2831 | dp_packet_init(&packet, 0); |
eb0b295e BP |
2832 | compose_arp(&packet, ARP_OP_REQUEST, |
2833 | eth_src, eth_addr_zero, true, ip_src, ip_dst); | |
a36de779 | 2834 | |
cdd42eda | 2835 | compose_table_xlate(ctx, out_dev, &packet); |
cf62fa4c | 2836 | dp_packet_uninit(&packet); |
a36de779 PS |
2837 | } |
2838 | ||
2839 | static int | |
81de18ec | 2840 | build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport, |
a36de779 PS |
2841 | const struct flow *flow, odp_port_t tunnel_odp_port) |
2842 | { | |
2843 | struct ovs_action_push_tnl tnl_push_data; | |
2844 | struct xport *out_dev = NULL; | |
c2b878e0 TLSC |
2845 | ovs_be32 s_ip = 0, d_ip = 0; |
2846 | struct in6_addr s_ip6 = in6addr_any; | |
2847 | struct in6_addr d_ip6 = in6addr_any; | |
74ff3298 JR |
2848 | struct eth_addr smac; |
2849 | struct eth_addr dmac; | |
a36de779 | 2850 | int err; |
c2b878e0 TLSC |
2851 | char buf_sip6[INET6_ADDRSTRLEN]; |
2852 | char buf_dip6[INET6_ADDRSTRLEN]; | |
a36de779 | 2853 | |
a8704b50 | 2854 | err = tnl_route_lookup_flow(flow, &d_ip6, &s_ip6, &out_dev); |
a36de779 | 2855 | if (err) { |
81de18ec | 2856 | xlate_report(ctx, "native tunnel routing failed"); |
a36de779 PS |
2857 | return err; |
2858 | } | |
c2b878e0 TLSC |
2859 | |
2860 | xlate_report(ctx, "tunneling to %s via %s", | |
2861 | ipv6_string_mapped(buf_dip6, &d_ip6), | |
2862 | netdev_get_name(out_dev->netdev)); | |
a36de779 PS |
2863 | |
2864 | /* Use mac addr of bridge port of the peer. */ | |
74ff3298 | 2865 | err = netdev_get_etheraddr(out_dev->netdev, &smac); |
a36de779 | 2866 | if (err) { |
81de18ec | 2867 | xlate_report(ctx, "tunnel output device lacks Ethernet address"); |
a36de779 PS |
2868 | return err; |
2869 | } | |
2870 | ||
c2b878e0 TLSC |
2871 | d_ip = in6_addr_get_mapped_ipv4(&d_ip6); |
2872 | if (d_ip) { | |
a8704b50 | 2873 | s_ip = in6_addr_get_mapped_ipv4(&s_ip6); |
a36de779 PS |
2874 | } |
2875 | ||
c2b878e0 | 2876 | err = tnl_neigh_lookup(out_dev->xbridge->name, &d_ip6, &dmac); |
a36de779 | 2877 | if (err) { |
c2b878e0 TLSC |
2878 | xlate_report(ctx, "neighbor cache miss for %s on bridge %s, " |
2879 | "sending %s request", | |
2880 | buf_dip6, out_dev->xbridge->name, d_ip ? "ARP" : "ND"); | |
2881 | if (d_ip) { | |
2882 | tnl_send_arp_request(ctx, out_dev, smac, s_ip, d_ip); | |
2883 | } else { | |
2884 | tnl_send_nd_request(ctx, out_dev, smac, &s_ip6, &d_ip6); | |
2885 | } | |
a36de779 PS |
2886 | return err; |
2887 | } | |
c2b878e0 | 2888 | |
a36de779 PS |
2889 | if (ctx->xin->xcache) { |
2890 | struct xc_entry *entry; | |
2891 | ||
53902038 TLSC |
2892 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_NEIGH); |
2893 | ovs_strlcpy(entry->u.tnl_neigh_cache.br_name, out_dev->xbridge->name, | |
2894 | sizeof entry->u.tnl_neigh_cache.br_name); | |
c2b878e0 | 2895 | entry->u.tnl_neigh_cache.d_ipv6 = d_ip6; |
a36de779 | 2896 | } |
81de18ec | 2897 | |
c2b878e0 TLSC |
2898 | xlate_report(ctx, "tunneling from "ETH_ADDR_FMT" %s" |
2899 | " to "ETH_ADDR_FMT" %s", | |
2900 | ETH_ADDR_ARGS(smac), ipv6_string_mapped(buf_sip6, &s_ip6), | |
2901 | ETH_ADDR_ARGS(dmac), buf_dip6); | |
2902 | ||
a36de779 | 2903 | err = tnl_port_build_header(xport->ofport, flow, |
370e373b | 2904 | dmac, smac, &s_ip6, &tnl_push_data); |
a36de779 PS |
2905 | if (err) { |
2906 | return err; | |
2907 | } | |
2908 | tnl_push_data.tnl_port = odp_to_u32(tunnel_odp_port); | |
2909 | tnl_push_data.out_port = odp_to_u32(out_dev->odp_port); | |
1520ef4f | 2910 | odp_put_tnl_push_action(ctx->odp_actions, &tnl_push_data); |
a36de779 PS |
2911 | return 0; |
2912 | } | |
2913 | ||
704bb0bf JS |
2914 | static void |
2915 | xlate_commit_actions(struct xlate_ctx *ctx) | |
2916 | { | |
2917 | bool use_masked = ctx->xbridge->support.masked_set_action; | |
2918 | ||
2919 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, | |
2920 | ctx->odp_actions, ctx->wc, | |
2921 | use_masked); | |
2922 | } | |
2923 | ||
07659514 JS |
2924 | static void |
2925 | clear_conntrack(struct flow *flow) | |
2926 | { | |
2927 | flow->ct_state = 0; | |
2928 | flow->ct_zone = 0; | |
8e53fe8c | 2929 | flow->ct_mark = 0; |
9daf2348 | 2930 | memset(&flow->ct_label, 0, sizeof flow->ct_label); |
07659514 JS |
2931 | } |
2932 | ||
9583bc14 | 2933 | static void |
4e022ec0 | 2934 | compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
e93ef1c7 | 2935 | const struct xlate_bond_recirc *xr, bool check_stp) |
9583bc14 | 2936 | { |
46c88433 | 2937 | const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
49a73e0c | 2938 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 2939 | struct flow *flow = &ctx->xin->flow; |
a36de779 | 2940 | struct flow_tnl flow_tnl; |
9583bc14 | 2941 | ovs_be16 flow_vlan_tci; |
1362e248 | 2942 | uint32_t flow_pkt_mark; |
9583bc14 | 2943 | uint8_t flow_nw_tos; |
4e022ec0 | 2944 | odp_port_t out_port, odp_port; |
a36de779 | 2945 | bool tnl_push_pop_send = false; |
ca077186 | 2946 | uint8_t dscp; |
9583bc14 EJ |
2947 | |
2948 | /* If 'struct flow' gets additional metadata, we'll need to zero it out | |
2949 | * before traversing a patch port. */ | |
ffe4c74f | 2950 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 35); |
a36de779 | 2951 | memset(&flow_tnl, 0, sizeof flow_tnl); |
9583bc14 | 2952 | |
46c88433 | 2953 | if (!xport) { |
9583bc14 EJ |
2954 | xlate_report(ctx, "Nonexistent output port"); |
2955 | return; | |
46c88433 | 2956 | } else if (xport->config & OFPUTIL_PC_NO_FWD) { |
9583bc14 EJ |
2957 | xlate_report(ctx, "OFPPC_NO_FWD set, skipping output"); |
2958 | return; | |
0d1cee12 | 2959 | } else if (check_stp) { |
bbbca389 | 2960 | if (is_stp(&ctx->base_flow)) { |
9efd308e DV |
2961 | if (!xport_stp_should_forward_bpdu(xport) && |
2962 | !xport_rstp_should_manage_bpdu(xport)) { | |
2963 | if (ctx->xbridge->stp != NULL) { | |
2964 | xlate_report(ctx, "STP not in listening state, " | |
2965 | "skipping bpdu output"); | |
2966 | } else if (ctx->xbridge->rstp != NULL) { | |
2967 | xlate_report(ctx, "RSTP not managing BPDU in this state, " | |
2968 | "skipping bpdu output"); | |
2969 | } | |
0d1cee12 K |
2970 | return; |
2971 | } | |
9efd308e DV |
2972 | } else if (!xport_stp_forward_state(xport) || |
2973 | !xport_rstp_forward_state(xport)) { | |
2974 | if (ctx->xbridge->stp != NULL) { | |
2975 | xlate_report(ctx, "STP not in forwarding state, " | |
2976 | "skipping output"); | |
2977 | } else if (ctx->xbridge->rstp != NULL) { | |
2978 | xlate_report(ctx, "RSTP not in forwarding state, " | |
2979 | "skipping output"); | |
2980 | } | |
0d1cee12 K |
2981 | return; |
2982 | } | |
9583bc14 EJ |
2983 | } |
2984 | ||
46c88433 EJ |
2985 | if (xport->peer) { |
2986 | const struct xport *peer = xport->peer; | |
9583bc14 | 2987 | struct flow old_flow = ctx->xin->flow; |
07659514 | 2988 | bool old_conntrack = ctx->conntracked; |
18721c4a | 2989 | cls_version_t old_version = ctx->tables_version; |
1774d762 JR |
2990 | struct ofpbuf old_stack = ctx->stack; |
2991 | union mf_subvalue new_stack[1024 / sizeof(union mf_subvalue)]; | |
ed9c9e3e JR |
2992 | struct ofpbuf old_action_set = ctx->action_set; |
2993 | uint64_t actset_stub[1024 / 8]; | |
9583bc14 | 2994 | |
1774d762 | 2995 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); |
ed9c9e3e | 2996 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); |
46c88433 EJ |
2997 | ctx->xbridge = peer->xbridge; |
2998 | flow->in_port.ofp_port = peer->ofp_port; | |
33bf9176 BP |
2999 | flow->metadata = htonll(0); |
3000 | memset(&flow->tunnel, 0, sizeof flow->tunnel); | |
3001 | memset(flow->regs, 0, sizeof flow->regs); | |
c61f3870 | 3002 | flow->actset_output = OFPP_UNSET; |
07659514 JS |
3003 | ctx->conntracked = false; |
3004 | clear_conntrack(flow); | |
9583bc14 | 3005 | |
621b8064 JR |
3006 | /* The bridge is now known so obtain its table version. */ |
3007 | ctx->tables_version | |
3008 | = ofproto_dpif_get_tables_version(ctx->xbridge->ofproto); | |
3009 | ||
515793d5 | 3010 | if (!process_special(ctx, peer) && may_receive(peer, ctx)) { |
9efd308e | 3011 | if (xport_stp_forward_state(peer) && xport_rstp_forward_state(peer)) { |
e672ff9b | 3012 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); |
1d361a81 | 3013 | if (!ctx->freezing) { |
8a5fb3b4 | 3014 | xlate_action_set(ctx); |
e672ff9b | 3015 | } |
1d361a81 | 3016 | if (ctx->freezing) { |
77ab5fd2 | 3017 | finish_freezing(ctx); |
ed9c9e3e | 3018 | } |
9583bc14 | 3019 | } else { |
9efd308e DV |
3020 | /* Forwarding is disabled by STP and RSTP. Let OFPP_NORMAL and |
3021 | * the learning action look at the packet, then drop it. */ | |
9583bc14 | 3022 | struct flow old_base_flow = ctx->base_flow; |
1520ef4f | 3023 | size_t old_size = ctx->odp_actions->size; |
3d6151f3 | 3024 | mirror_mask_t old_mirrors = ctx->mirrors; |
f3d5b473 | 3025 | |
e672ff9b | 3026 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); |
3d6151f3 | 3027 | ctx->mirrors = old_mirrors; |
9583bc14 | 3028 | ctx->base_flow = old_base_flow; |
1520ef4f | 3029 | ctx->odp_actions->size = old_size; |
e672ff9b | 3030 | |
1d361a81 BP |
3031 | /* Undo changes that may have been done for freezing. */ |
3032 | ctx_cancel_freeze(ctx); | |
9583bc14 EJ |
3033 | } |
3034 | } | |
3035 | ||
3036 | ctx->xin->flow = old_flow; | |
832554e3 | 3037 | ctx->xbridge = xport->xbridge; |
ed9c9e3e JR |
3038 | ofpbuf_uninit(&ctx->action_set); |
3039 | ctx->action_set = old_action_set; | |
1774d762 JR |
3040 | ofpbuf_uninit(&ctx->stack); |
3041 | ctx->stack = old_stack; | |
9583bc14 | 3042 | |
621b8064 JR |
3043 | /* Restore calling bridge's lookup version. */ |
3044 | ctx->tables_version = old_version; | |
3045 | ||
07659514 JS |
3046 | /* The peer bridge's conntrack execution should have no effect on the |
3047 | * original bridge. */ | |
3048 | ctx->conntracked = old_conntrack; | |
3049 | ||
f3d5b473 JR |
3050 | /* The fact that the peer bridge exits (for any reason) does not mean |
3051 | * that the original bridge should exit. Specifically, if the peer | |
1d361a81 BP |
3052 | * bridge freezes translation, the original bridge must continue |
3053 | * processing with the original, not the frozen packet! */ | |
f3d5b473 JR |
3054 | ctx->exit = false; |
3055 | ||
fff1b9c0 JR |
3056 | /* Peer bridge errors do not propagate back. */ |
3057 | ctx->error = XLATE_OK; | |
3058 | ||
9583bc14 | 3059 | if (ctx->xin->resubmit_stats) { |
46c88433 EJ |
3060 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
3061 | netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats); | |
a1aeea86 AW |
3062 | if (peer->bfd) { |
3063 | bfd_account_rx(peer->bfd, ctx->xin->resubmit_stats); | |
3064 | } | |
9583bc14 | 3065 | } |
b256dc52 JS |
3066 | if (ctx->xin->xcache) { |
3067 | struct xc_entry *entry; | |
3068 | ||
3069 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
3070 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
3071 | entry->u.dev.rx = netdev_ref(peer->netdev); | |
3072 | entry->u.dev.bfd = bfd_ref(peer->bfd); | |
3073 | } | |
9583bc14 EJ |
3074 | return; |
3075 | } | |
3076 | ||
33bf9176 | 3077 | flow_vlan_tci = flow->vlan_tci; |
1362e248 | 3078 | flow_pkt_mark = flow->pkt_mark; |
33bf9176 | 3079 | flow_nw_tos = flow->nw_tos; |
9583bc14 | 3080 | |
16194afd DDP |
3081 | if (count_skb_priorities(xport)) { |
3082 | memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority); | |
3083 | if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { | |
3084 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
3085 | flow->nw_tos &= ~IP_DSCP_MASK; | |
3086 | flow->nw_tos |= dscp; | |
3087 | } | |
9583bc14 EJ |
3088 | } |
3089 | ||
46c88433 | 3090 | if (xport->is_tunnel) { |
c2b878e0 | 3091 | struct in6_addr dst; |
9583bc14 EJ |
3092 | /* Save tunnel metadata so that changes made due to |
3093 | * the Logical (tunnel) Port are not visible for any further | |
3094 | * matches, while explicit set actions on tunnel metadata are. | |
3095 | */ | |
a36de779 | 3096 | flow_tnl = flow->tunnel; |
49a73e0c | 3097 | odp_port = tnl_port_send(xport->ofport, flow, ctx->wc); |
4e022ec0 | 3098 | if (odp_port == ODPP_NONE) { |
9583bc14 EJ |
3099 | xlate_report(ctx, "Tunneling decided against output"); |
3100 | goto out; /* restore flow_nw_tos */ | |
3101 | } | |
c2b878e0 TLSC |
3102 | dst = flow_tnl_dst(&flow->tunnel); |
3103 | if (ipv6_addr_equals(&dst, &ctx->orig_tunnel_ipv6_dst)) { | |
9583bc14 EJ |
3104 | xlate_report(ctx, "Not tunneling to our own address"); |
3105 | goto out; /* restore flow_nw_tos */ | |
3106 | } | |
3107 | if (ctx->xin->resubmit_stats) { | |
46c88433 | 3108 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
9583bc14 | 3109 | } |
b256dc52 JS |
3110 | if (ctx->xin->xcache) { |
3111 | struct xc_entry *entry; | |
3112 | ||
3113 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
3114 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
3115 | } | |
9583bc14 | 3116 | out_port = odp_port; |
a36de779 | 3117 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
81de18ec | 3118 | xlate_report(ctx, "output to native tunnel"); |
a36de779 PS |
3119 | tnl_push_pop_send = true; |
3120 | } else { | |
81de18ec | 3121 | xlate_report(ctx, "output to kernel tunnel"); |
1520ef4f | 3122 | commit_odp_tunnel_action(flow, &ctx->base_flow, ctx->odp_actions); |
a36de779 PS |
3123 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ |
3124 | } | |
9583bc14 | 3125 | } else { |
46c88433 | 3126 | odp_port = xport->odp_port; |
7614e5d0 | 3127 | out_port = odp_port; |
9583bc14 | 3128 | } |
9583bc14 | 3129 | |
4e022ec0 | 3130 | if (out_port != ODPP_NONE) { |
704bb0bf | 3131 | xlate_commit_actions(ctx); |
adcf00ba | 3132 | |
e93ef1c7 | 3133 | if (xr) { |
347bf289 | 3134 | struct ovs_action_hash *act_hash; |
adcf00ba | 3135 | |
347bf289 | 3136 | /* Hash action. */ |
1520ef4f | 3137 | act_hash = nl_msg_put_unspec_uninit(ctx->odp_actions, |
347bf289 AZ |
3138 | OVS_ACTION_ATTR_HASH, |
3139 | sizeof *act_hash); | |
3140 | act_hash->hash_alg = xr->hash_alg; | |
62ac1f20 | 3141 | act_hash->hash_basis = xr->hash_basis; |
347bf289 AZ |
3142 | |
3143 | /* Recirc action. */ | |
1520ef4f | 3144 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, |
347bf289 | 3145 | xr->recirc_id); |
adcf00ba | 3146 | } else { |
a36de779 PS |
3147 | |
3148 | if (tnl_push_pop_send) { | |
3149 | build_tunnel_send(ctx, xport, flow, odp_port); | |
3150 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ | |
3151 | } else { | |
3152 | odp_port_t odp_tnl_port = ODPP_NONE; | |
3153 | ||
3154 | /* XXX: Write better Filter for tunnel port. We can use inport | |
3155 | * int tunnel-port flow to avoid these checks completely. */ | |
3156 | if (ofp_port == OFPP_LOCAL && | |
3157 | ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { | |
3158 | ||
3159 | odp_tnl_port = tnl_port_map_lookup(flow, wc); | |
3160 | } | |
3161 | ||
3162 | if (odp_tnl_port != ODPP_NONE) { | |
1520ef4f | 3163 | nl_msg_put_odp_port(ctx->odp_actions, |
a36de779 PS |
3164 | OVS_ACTION_ATTR_TUNNEL_POP, |
3165 | odp_tnl_port); | |
3166 | } else { | |
3167 | /* Tunnel push-pop action is not compatible with | |
3168 | * IPFIX action. */ | |
a6092018 | 3169 | compose_ipfix_action(ctx, out_port); |
1520ef4f | 3170 | nl_msg_put_odp_port(ctx->odp_actions, |
a36de779 PS |
3171 | OVS_ACTION_ATTR_OUTPUT, |
3172 | out_port); | |
3173 | } | |
3174 | } | |
adcf00ba | 3175 | } |
9583bc14 | 3176 | |
6cbbf4fa EJ |
3177 | ctx->sflow_odp_port = odp_port; |
3178 | ctx->sflow_n_outputs++; | |
2031ef97 | 3179 | ctx->nf_output_iface = ofp_port; |
6cbbf4fa EJ |
3180 | } |
3181 | ||
7efbc3b7 BP |
3182 | if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { |
3183 | mirror_packet(ctx, xport->xbundle, | |
3184 | xbundle_mirror_dst(xport->xbundle->xbridge, | |
3185 | xport->xbundle)); | |
3186 | } | |
3187 | ||
6cbbf4fa | 3188 | out: |
9583bc14 | 3189 | /* Restore flow */ |
33bf9176 | 3190 | flow->vlan_tci = flow_vlan_tci; |
1362e248 | 3191 | flow->pkt_mark = flow_pkt_mark; |
33bf9176 | 3192 | flow->nw_tos = flow_nw_tos; |
9583bc14 EJ |
3193 | } |
3194 | ||
3195 | static void | |
e93ef1c7 JR |
3196 | compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
3197 | const struct xlate_bond_recirc *xr) | |
9583bc14 | 3198 | { |
e93ef1c7 | 3199 | compose_output_action__(ctx, ofp_port, xr, true); |
9583bc14 EJ |
3200 | } |
3201 | ||
bb61b33d BP |
3202 | static void |
3203 | xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule) | |
bb61b33d BP |
3204 | { |
3205 | struct rule_dpif *old_rule = ctx->rule; | |
8b1e5560 | 3206 | ovs_be64 old_cookie = ctx->rule_cookie; |
dc723c44 | 3207 | const struct rule_actions *actions; |
bb61b33d BP |
3208 | |
3209 | if (ctx->xin->resubmit_stats) { | |
70742c7f | 3210 | rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats); |
bb61b33d BP |
3211 | } |
3212 | ||
98b07853 | 3213 | ctx->resubmits++; |
9a9b001a | 3214 | ctx->indentation++; |
bb61b33d | 3215 | ctx->rule = rule; |
8b1e5560 | 3216 | ctx->rule_cookie = rule_dpif_get_flow_cookie(rule); |
6f00e29b BP |
3217 | actions = rule_dpif_get_actions(rule); |
3218 | do_xlate_actions(actions->ofpacts, actions->ofpacts_len, ctx); | |
8b1e5560 | 3219 | ctx->rule_cookie = old_cookie; |
bb61b33d | 3220 | ctx->rule = old_rule; |
9a9b001a | 3221 | ctx->indentation--; |
bb61b33d BP |
3222 | } |
3223 | ||
bd3240ba SH |
3224 | static bool |
3225 | xlate_resubmit_resource_check(struct xlate_ctx *ctx) | |
9583bc14 | 3226 | { |
9a9b001a | 3227 | if (ctx->indentation >= MAX_RESUBMIT_RECURSION + MAX_INTERNAL_RESUBMITS) { |
fff1b9c0 JR |
3228 | XLATE_REPORT_ERROR(ctx, "resubmit actions recursed over %d times", |
3229 | MAX_RESUBMIT_RECURSION); | |
3230 | ctx->error = XLATE_RECURSION_TOO_DEEP; | |
adcf00ba | 3231 | } else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) { |
fff1b9c0 JR |
3232 | XLATE_REPORT_ERROR(ctx, "over %d resubmit actions", MAX_RESUBMITS); |
3233 | ctx->error = XLATE_TOO_MANY_RESUBMITS; | |
1520ef4f | 3234 | } else if (ctx->odp_actions->size > UINT16_MAX) { |
fff1b9c0 JR |
3235 | XLATE_REPORT_ERROR(ctx, "resubmits yielded over 64 kB of actions"); |
3236 | /* NOT an error, as we'll be slow-pathing the flow in this case? */ | |
3237 | ctx->exit = true; /* XXX: translation still terminated! */ | |
6fd6ed71 | 3238 | } else if (ctx->stack.size >= 65536) { |
fff1b9c0 JR |
3239 | XLATE_REPORT_ERROR(ctx, "resubmits yielded over 64 kB of stack"); |
3240 | ctx->error = XLATE_STACK_TOO_DEEP; | |
98b07853 | 3241 | } else { |
bd3240ba SH |
3242 | return true; |
3243 | } | |
3244 | ||
3245 | return false; | |
3246 | } | |
3247 | ||
3248 | static void | |
6d328fa2 SH |
3249 | xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, |
3250 | bool may_packet_in, bool honor_table_miss) | |
bd3240ba SH |
3251 | { |
3252 | if (xlate_resubmit_resource_check(ctx)) { | |
9583bc14 | 3253 | uint8_t old_table_id = ctx->table_id; |
3f207910 | 3254 | struct rule_dpif *rule; |
9583bc14 EJ |
3255 | |
3256 | ctx->table_id = table_id; | |
3257 | ||
34dd0d78 | 3258 | rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, |
621b8064 | 3259 | ctx->tables_version, |
49a73e0c | 3260 | &ctx->xin->flow, ctx->xin->wc, |
34dd0d78 JR |
3261 | ctx->xin->resubmit_stats, |
3262 | &ctx->table_id, in_port, | |
3263 | may_packet_in, honor_table_miss); | |
9583bc14 | 3264 | |
a8c31348 | 3265 | if (OVS_UNLIKELY(ctx->xin->resubmit_hook)) { |
9a9b001a | 3266 | ctx->xin->resubmit_hook(ctx->xin, rule, ctx->indentation + 1); |
ad3efdcb EJ |
3267 | } |
3268 | ||
a2143702 | 3269 | if (rule) { |
83709dfa JR |
3270 | /* Fill in the cache entry here instead of xlate_recursively |
3271 | * to make the reference counting more explicit. We take a | |
3272 | * reference in the lookups above if we are going to cache the | |
3273 | * rule. */ | |
3274 | if (ctx->xin->xcache) { | |
3275 | struct xc_entry *entry; | |
3276 | ||
3277 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE); | |
3278 | entry->u.rule = rule; | |
1e1e1d19 | 3279 | rule_dpif_ref(rule); |
83709dfa | 3280 | } |
bb61b33d | 3281 | xlate_recursively(ctx, rule); |
ad3efdcb EJ |
3282 | } |
3283 | ||
9583bc14 | 3284 | ctx->table_id = old_table_id; |
98b07853 | 3285 | return; |
9583bc14 EJ |
3286 | } |
3287 | } | |
3288 | ||
f4fb341b | 3289 | static void |
1e684d7d RW |
3290 | xlate_group_stats(struct xlate_ctx *ctx, struct group_dpif *group, |
3291 | struct ofputil_bucket *bucket) | |
3292 | { | |
3293 | if (ctx->xin->resubmit_stats) { | |
3294 | group_dpif_credit_stats(group, bucket, ctx->xin->resubmit_stats); | |
3295 | } | |
3296 | if (ctx->xin->xcache) { | |
3297 | struct xc_entry *entry; | |
3298 | ||
3299 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_GROUP); | |
3300 | entry->u.group.group = group_dpif_ref(group); | |
3301 | entry->u.group.bucket = bucket; | |
3302 | } | |
3303 | } | |
3304 | ||
3305 | static void | |
3306 | xlate_group_bucket(struct xlate_ctx *ctx, struct ofputil_bucket *bucket) | |
f4fb341b SH |
3307 | { |
3308 | uint64_t action_list_stub[1024 / 8]; | |
0a2869d5 BP |
3309 | struct ofpbuf action_list = OFPBUF_STUB_INITIALIZER(action_list_stub); |
3310 | struct ofpbuf action_set = ofpbuf_const_initializer(bucket->ofpacts, | |
3311 | bucket->ofpacts_len); | |
5b09e569 | 3312 | struct flow old_flow = ctx->xin->flow; |
f4fb341b | 3313 | |
f4fb341b | 3314 | ofpacts_execute_action_set(&action_list, &action_set); |
9a9b001a | 3315 | ctx->indentation++; |
6fd6ed71 | 3316 | do_xlate_actions(action_list.data, action_list.size, ctx); |
9a9b001a | 3317 | ctx->indentation--; |
f4fb341b | 3318 | |
f4fb341b | 3319 | ofpbuf_uninit(&action_list); |
5b09e569 | 3320 | |
77ab5fd2 | 3321 | /* Check if need to freeze. */ |
1d361a81 | 3322 | if (ctx->freezing) { |
77ab5fd2 | 3323 | finish_freezing(ctx); |
e672ff9b JR |
3324 | } |
3325 | ||
5b09e569 JR |
3326 | /* Roll back flow to previous state. |
3327 | * This is equivalent to cloning the packet for each bucket. | |
3328 | * | |
3329 | * As a side effect any subsequently applied actions will | |
3330 | * also effectively be applied to a clone of the packet taken | |
3331 | * just before applying the all or indirect group. | |
3332 | * | |
3333 | * Note that group buckets are action sets, hence they cannot modify the | |
3334 | * main action set. Also any stack actions are ignored when executing an | |
3335 | * action set, so group buckets cannot change the stack either. | |
3336 | * However, we do allow resubmit actions in group buckets, which could | |
3337 | * break the above assumptions. It is up to the controller to not mess up | |
3338 | * with the action_set and stack in the tables resubmitted to from | |
3339 | * group buckets. */ | |
3340 | ctx->xin->flow = old_flow; | |
3341 | ||
5b09e569 JR |
3342 | /* The fact that the group bucket exits (for any reason) does not mean that |
3343 | * the translation after the group action should exit. Specifically, if | |
1d361a81 BP |
3344 | * the group bucket freezes translation, the actions after the group action |
3345 | * must continue processing with the original, not the frozen packet! */ | |
5b09e569 | 3346 | ctx->exit = false; |
f4fb341b SH |
3347 | } |
3348 | ||
3349 | static void | |
3350 | xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3351 | { | |
1e684d7d | 3352 | struct ofputil_bucket *bucket; |
ca6ba700 | 3353 | const struct ovs_list *buckets; |
f4fb341b SH |
3354 | |
3355 | group_dpif_get_buckets(group, &buckets); | |
3356 | ||
3357 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
3358 | xlate_group_bucket(ctx, bucket); | |
f4fb341b | 3359 | } |
1e684d7d | 3360 | xlate_group_stats(ctx, group, NULL); |
f4fb341b SH |
3361 | } |
3362 | ||
dd8cd4b4 SH |
3363 | static void |
3364 | xlate_ff_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3365 | { | |
1e684d7d | 3366 | struct ofputil_bucket *bucket; |
dd8cd4b4 SH |
3367 | |
3368 | bucket = group_first_live_bucket(ctx, group, 0); | |
3369 | if (bucket) { | |
3370 | xlate_group_bucket(ctx, bucket); | |
1e684d7d | 3371 | xlate_group_stats(ctx, group, bucket); |
dd8cd4b4 SH |
3372 | } |
3373 | } | |
3374 | ||
fe7e5749 | 3375 | static void |
7565c3e4 | 3376 | xlate_default_select_group(struct xlate_ctx *ctx, struct group_dpif *group) |
fe7e5749 | 3377 | { |
49a73e0c | 3378 | struct flow_wildcards *wc = ctx->wc; |
1e684d7d | 3379 | struct ofputil_bucket *bucket; |
fe7e5749 SH |
3380 | uint32_t basis; |
3381 | ||
1d1aae0b | 3382 | basis = flow_hash_symmetric_l4(&ctx->xin->flow, 0); |
80e3509d | 3383 | flow_mask_hash_fields(&ctx->xin->flow, wc, NX_HASH_FIELDS_SYMMETRIC_L4); |
fe7e5749 SH |
3384 | bucket = group_best_live_bucket(ctx, group, basis); |
3385 | if (bucket) { | |
fe7e5749 | 3386 | xlate_group_bucket(ctx, bucket); |
1e684d7d | 3387 | xlate_group_stats(ctx, group, bucket); |
fe7e5749 SH |
3388 | } |
3389 | } | |
3390 | ||
0c4b9393 SH |
3391 | static void |
3392 | xlate_hash_fields_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3393 | { | |
3394 | struct mf_bitmap hash_fields = MF_BITMAP_INITIALIZER; | |
0c4b9393 SH |
3395 | const struct field_array *fields; |
3396 | struct ofputil_bucket *bucket; | |
3397 | uint32_t basis; | |
3398 | int i; | |
3399 | ||
3400 | fields = group_dpif_get_fields(group); | |
3401 | basis = hash_uint64(group_dpif_get_selection_method_param(group)); | |
3402 | ||
3403 | /* Determine which fields to hash */ | |
3404 | for (i = 0; i < MFF_N_IDS; i++) { | |
3405 | if (bitmap_is_set(fields->used.bm, i)) { | |
3406 | const struct mf_field *mf; | |
3407 | ||
3408 | /* If the field is already present in 'hash_fields' then | |
3409 | * this loop has already checked that it and its pre-requisites | |
3410 | * are present in the flow and its pre-requisites have | |
3411 | * already been added to 'hash_fields'. There is nothing more | |
3412 | * to do here and as an optimisation the loop can continue. */ | |
3413 | if (bitmap_is_set(hash_fields.bm, i)) { | |
3414 | continue; | |
3415 | } | |
3416 | ||
3417 | mf = mf_from_id(i); | |
3418 | ||
3419 | /* Only hash a field if it and its pre-requisites are present | |
3420 | * in the flow. */ | |
3421 | if (!mf_are_prereqs_ok(mf, &ctx->xin->flow)) { | |
3422 | continue; | |
3423 | } | |
3424 | ||
3425 | /* Hash both the field and its pre-requisites */ | |
3426 | mf_bitmap_set_field_and_prereqs(mf, &hash_fields); | |
3427 | } | |
3428 | } | |
3429 | ||
3430 | /* Hash the fields */ | |
3431 | for (i = 0; i < MFF_N_IDS; i++) { | |
3432 | if (bitmap_is_set(hash_fields.bm, i)) { | |
3433 | const struct mf_field *mf = mf_from_id(i); | |
3434 | union mf_value value; | |
3435 | int j; | |
3436 | ||
3437 | mf_get_value(mf, &ctx->xin->flow, &value); | |
3438 | /* This seems inefficient but so does apply_mask() */ | |
3439 | for (j = 0; j < mf->n_bytes; j++) { | |
3440 | ((uint8_t *) &value)[j] &= ((uint8_t *) &fields->value[i])[j]; | |
3441 | } | |
3442 | basis = hash_bytes(&value, mf->n_bytes, basis); | |
3443 | ||
1cb20095 JG |
3444 | /* For tunnels, hash in whether the field is present. */ |
3445 | if (mf_is_tun_metadata(mf)) { | |
3446 | basis = hash_boolean(mf_is_set(mf, &ctx->xin->flow), basis); | |
3447 | } | |
3448 | ||
49a73e0c | 3449 | mf_mask_field(mf, &ctx->wc->masks); |
0c4b9393 SH |
3450 | } |
3451 | } | |
3452 | ||
3453 | bucket = group_best_live_bucket(ctx, group, basis); | |
3454 | if (bucket) { | |
3455 | xlate_group_bucket(ctx, bucket); | |
3456 | xlate_group_stats(ctx, group, bucket); | |
3457 | } | |
3458 | } | |
3459 | ||
7565c3e4 SH |
3460 | static void |
3461 | xlate_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3462 | { | |
3463 | const char *selection_method = group_dpif_get_selection_method(group); | |
3464 | ||
3465 | if (selection_method[0] == '\0') { | |
3466 | xlate_default_select_group(ctx, group); | |
0c4b9393 SH |
3467 | } else if (!strcasecmp("hash", selection_method)) { |
3468 | xlate_hash_fields_select_group(ctx, group); | |
7565c3e4 SH |
3469 | } else { |
3470 | /* Parsing of groups should ensure this never happens */ | |
3471 | OVS_NOT_REACHED(); | |
3472 | } | |
3473 | } | |
3474 | ||
f4fb341b SH |
3475 | static void |
3476 | xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group) | |
3477 | { | |
0eb48fe1 | 3478 | bool was_in_group = ctx->in_group; |
5a070238 BP |
3479 | ctx->in_group = true; |
3480 | ||
f4fb341b SH |
3481 | switch (group_dpif_get_type(group)) { |
3482 | case OFPGT11_ALL: | |
3483 | case OFPGT11_INDIRECT: | |
3484 | xlate_all_group(ctx, group); | |
3485 | break; | |
3486 | case OFPGT11_SELECT: | |
fe7e5749 | 3487 | xlate_select_group(ctx, group); |
f4fb341b | 3488 | break; |
dd8cd4b4 SH |
3489 | case OFPGT11_FF: |
3490 | xlate_ff_group(ctx, group); | |
3491 | break; | |
f4fb341b | 3492 | default: |
428b2edd | 3493 | OVS_NOT_REACHED(); |
f4fb341b | 3494 | } |
809c7548 | 3495 | group_dpif_unref(group); |
5a070238 | 3496 | |
0eb48fe1 | 3497 | ctx->in_group = was_in_group; |
f4fb341b SH |
3498 | } |
3499 | ||
3500 | static bool | |
3501 | xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id) | |
3502 | { | |
0eb48fe1 | 3503 | if (xlate_resubmit_resource_check(ctx)) { |
f4fb341b SH |
3504 | struct group_dpif *group; |
3505 | bool got_group; | |
3506 | ||
3507 | got_group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group); | |
3508 | if (got_group) { | |
3509 | xlate_group_action__(ctx, group); | |
3510 | } else { | |
3511 | return true; | |
3512 | } | |
3513 | } | |
3514 | ||
3515 | return false; | |
3516 | } | |
3517 | ||
9583bc14 EJ |
3518 | static void |
3519 | xlate_ofpact_resubmit(struct xlate_ctx *ctx, | |
3520 | const struct ofpact_resubmit *resubmit) | |
3521 | { | |
4e022ec0 | 3522 | ofp_port_t in_port; |
9583bc14 | 3523 | uint8_t table_id; |
adcf00ba AZ |
3524 | bool may_packet_in = false; |
3525 | bool honor_table_miss = false; | |
3526 | ||
3527 | if (ctx->rule && rule_dpif_is_internal(ctx->rule)) { | |
3528 | /* Still allow missed packets to be sent to the controller | |
3529 | * if resubmitting from an internal table. */ | |
3530 | may_packet_in = true; | |
3531 | honor_table_miss = true; | |
3532 | } | |
9583bc14 EJ |
3533 | |
3534 | in_port = resubmit->in_port; | |
3535 | if (in_port == OFPP_IN_PORT) { | |
4e022ec0 | 3536 | in_port = ctx->xin->flow.in_port.ofp_port; |
9583bc14 EJ |
3537 | } |
3538 | ||
3539 | table_id = resubmit->table_id; | |
3540 | if (table_id == 255) { | |
3541 | table_id = ctx->table_id; | |
3542 | } | |
3543 | ||
adcf00ba AZ |
3544 | xlate_table_action(ctx, in_port, table_id, may_packet_in, |
3545 | honor_table_miss); | |
9583bc14 EJ |
3546 | } |
3547 | ||
3548 | static void | |
3549 | flood_packets(struct xlate_ctx *ctx, bool all) | |
3550 | { | |
46c88433 | 3551 | const struct xport *xport; |
9583bc14 | 3552 | |
46c88433 EJ |
3553 | HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) { |
3554 | if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3555 | continue; |
3556 | } | |
3557 | ||
3558 | if (all) { | |
e93ef1c7 | 3559 | compose_output_action__(ctx, xport->ofp_port, NULL, false); |
46c88433 | 3560 | } else if (!(xport->config & OFPUTIL_PC_NO_FLOOD)) { |
e93ef1c7 | 3561 | compose_output_action(ctx, xport->ofp_port, NULL); |
9583bc14 EJ |
3562 | } |
3563 | } | |
3564 | ||
2031ef97 | 3565 | ctx->nf_output_iface = NF_OUT_FLOOD; |
9583bc14 EJ |
3566 | } |
3567 | ||
3568 | static void | |
3569 | execute_controller_action(struct xlate_ctx *ctx, int len, | |
3570 | enum ofp_packet_in_reason reason, | |
bdcad671 BP |
3571 | uint16_t controller_id, |
3572 | const uint8_t *userdata, size_t userdata_len) | |
9583bc14 | 3573 | { |
e14deea0 | 3574 | struct dp_packet *packet; |
9583bc14 | 3575 | |
04594cd5 | 3576 | ctx->xout->slow |= SLOW_CONTROLLER; |
b476e2f2 | 3577 | xlate_commit_actions(ctx); |
9583bc14 EJ |
3578 | if (!ctx->xin->packet) { |
3579 | return; | |
3580 | } | |
3581 | ||
cf62fa4c | 3582 | packet = dp_packet_clone(ctx->xin->packet); |
9583bc14 | 3583 | |
41ccaa24 | 3584 | odp_execute_actions(NULL, &packet, 1, false, |
1520ef4f | 3585 | ctx->odp_actions->data, ctx->odp_actions->size, NULL); |
9583bc14 | 3586 | |
9bfe9334 BP |
3587 | /* A packet sent by an action in a table-miss rule is considered an |
3588 | * explicit table miss. OpenFlow before 1.3 doesn't have that concept so | |
3589 | * it will get translated back to OFPR_ACTION for those versions. */ | |
3590 | if (reason == OFPR_ACTION | |
3591 | && ctx->rule && rule_dpif_is_table_miss(ctx->rule)) { | |
3592 | reason = OFPR_EXPLICIT_MISS; | |
3593 | } | |
3594 | ||
3595 | size_t packet_len = dp_packet_size(packet); | |
0fb7792a | 3596 | |
a2b53dec BP |
3597 | struct ofproto_async_msg *am = xmalloc(sizeof *am); |
3598 | *am = (struct ofproto_async_msg) { | |
9bfe9334 | 3599 | .controller_id = controller_id, |
a2b53dec BP |
3600 | .oam = OAM_PACKET_IN, |
3601 | .pin = { | |
3602 | .up = { | |
77ab5fd2 BP |
3603 | .public = { |
3604 | .packet = dp_packet_steal_data(packet), | |
3605 | .packet_len = packet_len, | |
3606 | .reason = reason, | |
3607 | .table_id = ctx->table_id, | |
3608 | .cookie = ctx->rule_cookie, | |
3609 | .userdata = (userdata_len | |
3610 | ? xmemdup(userdata, userdata_len) | |
3611 | : NULL), | |
3612 | .userdata_len = userdata_len, | |
3613 | } | |
a2b53dec BP |
3614 | }, |
3615 | .max_len = len, | |
9bfe9334 | 3616 | }, |
9bfe9334 | 3617 | }; |
77ab5fd2 | 3618 | flow_get_metadata(&ctx->xin->flow, &am->pin.up.public.flow_metadata); |
9583bc14 | 3619 | |
a2b53dec | 3620 | ofproto_dpif_send_async_msg(ctx->xbridge->ofproto, am); |
e14deea0 | 3621 | dp_packet_delete(packet); |
9583bc14 EJ |
3622 | } |
3623 | ||
7bbdd84f | 3624 | static void |
77ab5fd2 | 3625 | emit_continuation(struct xlate_ctx *ctx, const struct frozen_state *state) |
7bbdd84f | 3626 | { |
77ab5fd2 BP |
3627 | struct ofproto_async_msg *am = xmalloc(sizeof *am); |
3628 | *am = (struct ofproto_async_msg) { | |
3629 | .controller_id = ctx->pause->controller_id, | |
3630 | .oam = OAM_PACKET_IN, | |
3631 | .pin = { | |
3632 | .up = { | |
3633 | .public = { | |
3634 | .userdata = xmemdup(ctx->pause->userdata, | |
3635 | ctx->pause->userdata_len), | |
3636 | .userdata_len = ctx->pause->userdata_len, | |
3637 | .packet = xmemdup(dp_packet_data(ctx->xin->packet), | |
3638 | dp_packet_size(ctx->xin->packet)), | |
3639 | .packet_len = dp_packet_size(ctx->xin->packet), | |
0b024e49 | 3640 | .reason = ctx->pause->reason, |
77ab5fd2 BP |
3641 | }, |
3642 | .bridge = *ofproto_dpif_get_uuid(ctx->xbridge->ofproto), | |
3643 | .stack = xmemdup(state->stack, | |
3644 | state->n_stack * sizeof *state->stack), | |
3645 | .n_stack = state->n_stack, | |
3646 | .mirrors = state->mirrors, | |
3647 | .conntracked = state->conntracked, | |
3648 | .actions = xmemdup(state->ofpacts, state->ofpacts_len), | |
3649 | .actions_len = state->ofpacts_len, | |
3650 | .action_set = xmemdup(state->action_set, | |
3651 | state->action_set_len), | |
3652 | .action_set_len = state->action_set_len, | |
3653 | }, | |
3654 | .max_len = UINT16_MAX, | |
3655 | }, | |
3656 | }; | |
3657 | flow_get_metadata(&ctx->xin->flow, &am->pin.up.public.flow_metadata); | |
3658 | ofproto_dpif_send_async_msg(ctx->xbridge->ofproto, am); | |
3659 | } | |
7bbdd84f | 3660 | |
77ab5fd2 BP |
3661 | static void |
3662 | finish_freezing__(struct xlate_ctx *ctx, uint8_t table) | |
3663 | { | |
1d361a81 | 3664 | ovs_assert(ctx->freezing); |
7bbdd84f | 3665 | |
1d361a81 | 3666 | struct frozen_state state = { |
07659514 | 3667 | .table_id = table, |
290835f9 | 3668 | .ofproto_uuid = *ofproto_dpif_get_uuid(ctx->xbridge->ofproto), |
5c1b2314 BP |
3669 | .stack = ctx->stack.data, |
3670 | .n_stack = ctx->stack.size / sizeof(union mf_subvalue), | |
29bae541 | 3671 | .mirrors = ctx->mirrors, |
07659514 | 3672 | .conntracked = ctx->conntracked, |
1d361a81 BP |
3673 | .ofpacts = ctx->frozen_actions.data, |
3674 | .ofpacts_len = ctx->frozen_actions.size, | |
417509fa | 3675 | .action_set = ctx->action_set.data, |
8a5fb3b4 | 3676 | .action_set_len = ctx->action_set.size, |
2082425c | 3677 | }; |
77ab5fd2 | 3678 | frozen_metadata_from_flow(&state.metadata, &ctx->xin->flow); |
2082425c | 3679 | |
77ab5fd2 BP |
3680 | if (ctx->pause) { |
3681 | if (ctx->xin->packet) { | |
3682 | emit_continuation(ctx, &state); | |
3683 | } | |
3684 | } else { | |
3685 | /* Allocate a unique recirc id for the given metadata state in the | |
3686 | * flow. An existing id, with a new reference to the corresponding | |
3687 | * recirculation context, will be returned if possible. | |
3688 | * The life-cycle of this recirc id is managed by associating it | |
3689 | * with the udpif key ('ukey') created for each new datapath flow. */ | |
3690 | uint32_t id = recirc_alloc_id_ctx(&state); | |
3691 | if (!id) { | |
3692 | XLATE_REPORT_ERROR(ctx, "Failed to allocate recirculation id"); | |
3693 | ctx->error = XLATE_NO_RECIRCULATION_CONTEXT; | |
3694 | return; | |
3695 | } | |
3696 | recirc_refs_add(&ctx->xout->recircs, id); | |
7bbdd84f | 3697 | |
77ab5fd2 BP |
3698 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, id); |
3699 | } | |
e672ff9b | 3700 | |
1d361a81 BP |
3701 | /* Undo changes done by freezing. */ |
3702 | ctx_cancel_freeze(ctx); | |
7bbdd84f SH |
3703 | } |
3704 | ||
1d361a81 | 3705 | /* Called only when we're freezing. */ |
07659514 | 3706 | static void |
77ab5fd2 | 3707 | finish_freezing(struct xlate_ctx *ctx) |
07659514 JS |
3708 | { |
3709 | xlate_commit_actions(ctx); | |
77ab5fd2 | 3710 | finish_freezing__(ctx, 0); |
07659514 JS |
3711 | } |
3712 | ||
e37b8437 JS |
3713 | /* Fork the pipeline here. The current packet will continue processing the |
3714 | * current action list. A clone of the current packet will recirculate, skip | |
3715 | * the remainder of the current action list and asynchronously resume pipeline | |
3716 | * processing in 'table' with the current metadata and action set. */ | |
3717 | static void | |
3718 | compose_recirculate_and_fork(struct xlate_ctx *ctx, uint8_t table) | |
3719 | { | |
1d361a81 | 3720 | ctx->freezing = true; |
77ab5fd2 | 3721 | finish_freezing__(ctx, table); |
e37b8437 JS |
3722 | } |
3723 | ||
8bfd0fda BP |
3724 | static void |
3725 | compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) | |
9583bc14 | 3726 | { |
33bf9176 | 3727 | struct flow *flow = &ctx->xin->flow; |
8bfd0fda | 3728 | int n; |
33bf9176 | 3729 | |
8bfd0fda | 3730 | ovs_assert(eth_type_mpls(mpls->ethertype)); |
b0a17866 | 3731 | |
49a73e0c | 3732 | n = flow_count_mpls_labels(flow, ctx->wc); |
8bfd0fda | 3733 | if (!n) { |
704bb0bf | 3734 | xlate_commit_actions(ctx); |
8bfd0fda BP |
3735 | } else if (n >= FLOW_MAX_MPLS_LABELS) { |
3736 | if (ctx->xin->packet != NULL) { | |
fff1b9c0 | 3737 | XLATE_REPORT_ERROR(ctx, "bridge %s: dropping packet on which an " |
8bfd0fda BP |
3738 | "MPLS push action can't be performed as it would " |
3739 | "have more MPLS LSEs than the %d supported.", | |
3740 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
9583bc14 | 3741 | } |
fff1b9c0 | 3742 | ctx->error = XLATE_TOO_MANY_MPLS_LABELS; |
8bfd0fda | 3743 | return; |
9583bc14 | 3744 | } |
b0a17866 | 3745 | |
49a73e0c | 3746 | flow_push_mpls(flow, n, mpls->ethertype, ctx->wc); |
9583bc14 EJ |
3747 | } |
3748 | ||
8bfd0fda | 3749 | static void |
9cfef3d0 | 3750 | compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) |
9583bc14 | 3751 | { |
8bfd0fda | 3752 | struct flow *flow = &ctx->xin->flow; |
49a73e0c | 3753 | int n = flow_count_mpls_labels(flow, ctx->wc); |
33bf9176 | 3754 | |
49a73e0c | 3755 | if (flow_pop_mpls(flow, n, eth_type, ctx->wc)) { |
8bf009bf JR |
3756 | if (!eth_type_mpls(eth_type) && ctx->xbridge->support.odp.recirc) { |
3757 | ctx_trigger_freeze(ctx); | |
7bbdd84f SH |
3758 | } |
3759 | } else if (n >= FLOW_MAX_MPLS_LABELS) { | |
8bfd0fda | 3760 | if (ctx->xin->packet != NULL) { |
fff1b9c0 | 3761 | XLATE_REPORT_ERROR(ctx, "bridge %s: dropping packet on which an " |
8bfd0fda BP |
3762 | "MPLS pop action can't be performed as it has " |
3763 | "more MPLS LSEs than the %d supported.", | |
3764 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
3765 | } | |
fff1b9c0 | 3766 | ctx->error = XLATE_TOO_MANY_MPLS_LABELS; |
1520ef4f | 3767 | ofpbuf_clear(ctx->odp_actions); |
9583bc14 EJ |
3768 | } |
3769 | } | |
3770 | ||
3771 | static bool | |
3772 | compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids) | |
3773 | { | |
33bf9176 BP |
3774 | struct flow *flow = &ctx->xin->flow; |
3775 | ||
3776 | if (!is_ip_any(flow)) { | |
9583bc14 EJ |
3777 | return false; |
3778 | } | |
3779 | ||
49a73e0c | 3780 | ctx->wc->masks.nw_ttl = 0xff; |
33bf9176 BP |
3781 | if (flow->nw_ttl > 1) { |
3782 | flow->nw_ttl--; | |
9583bc14 EJ |
3783 | return false; |
3784 | } else { | |
3785 | size_t i; | |
3786 | ||
3787 | for (i = 0; i < ids->n_controllers; i++) { | |
3788 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, | |
bdcad671 | 3789 | ids->cnt_ids[i], NULL, 0); |
9583bc14 EJ |
3790 | } |
3791 | ||
3792 | /* Stop processing for current table. */ | |
3793 | return true; | |
3794 | } | |
3795 | } | |
3796 | ||
8bfd0fda | 3797 | static void |
097d4939 JR |
3798 | compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label) |
3799 | { | |
8bfd0fda | 3800 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3801 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK); |
8bfd0fda | 3802 | set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label); |
097d4939 | 3803 | } |
097d4939 JR |
3804 | } |
3805 | ||
8bfd0fda | 3806 | static void |
097d4939 JR |
3807 | compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc) |
3808 | { | |
8bfd0fda | 3809 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3810 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TC_MASK); |
8bfd0fda | 3811 | set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc); |
097d4939 | 3812 | } |
097d4939 JR |
3813 | } |
3814 | ||
8bfd0fda | 3815 | static void |
9cfef3d0 | 3816 | compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl) |
9583bc14 | 3817 | { |
8bfd0fda | 3818 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3819 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda | 3820 | set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl); |
b0a17866 | 3821 | } |
9583bc14 EJ |
3822 | } |
3823 | ||
3824 | static bool | |
9cfef3d0 | 3825 | compose_dec_mpls_ttl_action(struct xlate_ctx *ctx) |
9583bc14 | 3826 | { |
33bf9176 | 3827 | struct flow *flow = &ctx->xin->flow; |
1dd35f8a | 3828 | |
8bfd0fda | 3829 | if (eth_type_mpls(flow->dl_type)) { |
22d38fca JR |
3830 | uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]); |
3831 | ||
49a73e0c | 3832 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda BP |
3833 | if (ttl > 1) { |
3834 | ttl--; | |
3835 | set_mpls_lse_ttl(&flow->mpls_lse[0], ttl); | |
3836 | return false; | |
3837 | } else { | |
bdcad671 BP |
3838 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0, |
3839 | NULL, 0); | |
8bfd0fda | 3840 | } |
9583bc14 | 3841 | } |
22d38fca JR |
3842 | |
3843 | /* Stop processing for current table. */ | |
3844 | return true; | |
9583bc14 EJ |
3845 | } |
3846 | ||
3847 | static void | |
3848 | xlate_output_action(struct xlate_ctx *ctx, | |
4e022ec0 | 3849 | ofp_port_t port, uint16_t max_len, bool may_packet_in) |
9583bc14 | 3850 | { |
2031ef97 | 3851 | ofp_port_t prev_nf_output_iface = ctx->nf_output_iface; |
9583bc14 | 3852 | |
2031ef97 | 3853 | ctx->nf_output_iface = NF_OUT_DROP; |
9583bc14 EJ |
3854 | |
3855 | switch (port) { | |
3856 | case OFPP_IN_PORT: | |
e93ef1c7 | 3857 | compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port, NULL); |
9583bc14 EJ |
3858 | break; |
3859 | case OFPP_TABLE: | |
4e022ec0 | 3860 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 3861 | 0, may_packet_in, true); |
9583bc14 EJ |
3862 | break; |
3863 | case OFPP_NORMAL: | |
3864 | xlate_normal(ctx); | |
3865 | break; | |
3866 | case OFPP_FLOOD: | |
3867 | flood_packets(ctx, false); | |
3868 | break; | |
3869 | case OFPP_ALL: | |
3870 | flood_packets(ctx, true); | |
3871 | break; | |
3872 | case OFPP_CONTROLLER: | |
3a11fd5b | 3873 | execute_controller_action(ctx, max_len, |
029ca940 SS |
3874 | (ctx->in_group ? OFPR_GROUP |
3875 | : ctx->in_action_set ? OFPR_ACTION_SET | |
3876 | : OFPR_ACTION), | |
bdcad671 | 3877 | 0, NULL, 0); |
9583bc14 EJ |
3878 | break; |
3879 | case OFPP_NONE: | |
3880 | break; | |
3881 | case OFPP_LOCAL: | |
3882 | default: | |
4e022ec0 | 3883 | if (port != ctx->xin->flow.in_port.ofp_port) { |
e93ef1c7 | 3884 | compose_output_action(ctx, port, NULL); |
9583bc14 EJ |
3885 | } else { |
3886 | xlate_report(ctx, "skipping output to input port"); | |
3887 | } | |
3888 | break; | |
3889 | } | |
3890 | ||
3891 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
2031ef97 BP |
3892 | ctx->nf_output_iface = NF_OUT_FLOOD; |
3893 | } else if (ctx->nf_output_iface == NF_OUT_DROP) { | |
3894 | ctx->nf_output_iface = prev_nf_output_iface; | |
9583bc14 | 3895 | } else if (prev_nf_output_iface != NF_OUT_DROP && |
2031ef97 BP |
3896 | ctx->nf_output_iface != NF_OUT_FLOOD) { |
3897 | ctx->nf_output_iface = NF_OUT_MULTI; | |
9583bc14 EJ |
3898 | } |
3899 | } | |
3900 | ||
3901 | static void | |
3902 | xlate_output_reg_action(struct xlate_ctx *ctx, | |
3903 | const struct ofpact_output_reg *or) | |
3904 | { | |
3905 | uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow); | |
3906 | if (port <= UINT16_MAX) { | |
3907 | union mf_subvalue value; | |
3908 | ||
3909 | memset(&value, 0xff, sizeof value); | |
49a73e0c | 3910 | mf_write_subfield_flow(&or->src, &value, &ctx->wc->masks); |
4e022ec0 AW |
3911 | xlate_output_action(ctx, u16_to_ofp(port), |
3912 | or->max_len, false); | |
9583bc14 EJ |
3913 | } |
3914 | } | |
3915 | ||
3916 | static void | |
3917 | xlate_enqueue_action(struct xlate_ctx *ctx, | |
3918 | const struct ofpact_enqueue *enqueue) | |
3919 | { | |
4e022ec0 | 3920 | ofp_port_t ofp_port = enqueue->port; |
9583bc14 EJ |
3921 | uint32_t queue_id = enqueue->queue; |
3922 | uint32_t flow_priority, priority; | |
3923 | int error; | |
3924 | ||
3925 | /* Translate queue to priority. */ | |
89a8a7f0 | 3926 | error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority); |
9583bc14 EJ |
3927 | if (error) { |
3928 | /* Fall back to ordinary output action. */ | |
3929 | xlate_output_action(ctx, enqueue->port, 0, false); | |
3930 | return; | |
3931 | } | |
3932 | ||
3933 | /* Check output port. */ | |
3934 | if (ofp_port == OFPP_IN_PORT) { | |
4e022ec0 AW |
3935 | ofp_port = ctx->xin->flow.in_port.ofp_port; |
3936 | } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3937 | return; |
3938 | } | |
3939 | ||
3940 | /* Add datapath actions. */ | |
3941 | flow_priority = ctx->xin->flow.skb_priority; | |
3942 | ctx->xin->flow.skb_priority = priority; | |
e93ef1c7 | 3943 | compose_output_action(ctx, ofp_port, NULL); |
9583bc14 EJ |
3944 | ctx->xin->flow.skb_priority = flow_priority; |
3945 | ||
3946 | /* Update NetFlow output port. */ | |
2031ef97 BP |
3947 | if (ctx->nf_output_iface == NF_OUT_DROP) { |
3948 | ctx->nf_output_iface = ofp_port; | |
3949 | } else if (ctx->nf_output_iface != NF_OUT_FLOOD) { | |
3950 | ctx->nf_output_iface = NF_OUT_MULTI; | |
9583bc14 EJ |
3951 | } |
3952 | } | |
3953 | ||
3954 | static void | |
3955 | xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id) | |
3956 | { | |
3957 | uint32_t skb_priority; | |
3958 | ||
89a8a7f0 | 3959 | if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) { |
9583bc14 EJ |
3960 | ctx->xin->flow.skb_priority = skb_priority; |
3961 | } else { | |
3962 | /* Couldn't translate queue to a priority. Nothing to do. A warning | |
3963 | * has already been logged. */ | |
3964 | } | |
3965 | } | |
3966 | ||
3967 | static bool | |
46c88433 | 3968 | slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_) |
9583bc14 | 3969 | { |
46c88433 EJ |
3970 | const struct xbridge *xbridge = xbridge_; |
3971 | struct xport *port; | |
9583bc14 EJ |
3972 | |
3973 | switch (ofp_port) { | |
3974 | case OFPP_IN_PORT: | |
3975 | case OFPP_TABLE: | |
3976 | case OFPP_NORMAL: | |
3977 | case OFPP_FLOOD: | |
3978 | case OFPP_ALL: | |
3979 | case OFPP_NONE: | |
3980 | return true; | |
3981 | case OFPP_CONTROLLER: /* Not supported by the bundle action. */ | |
3982 | return false; | |
3983 | default: | |
46c88433 | 3984 | port = get_ofp_port(xbridge, ofp_port); |
9583bc14 EJ |
3985 | return port ? port->may_enable : false; |
3986 | } | |
3987 | } | |
3988 | ||
3989 | static void | |
3990 | xlate_bundle_action(struct xlate_ctx *ctx, | |
3991 | const struct ofpact_bundle *bundle) | |
3992 | { | |
4e022ec0 | 3993 | ofp_port_t port; |
9583bc14 | 3994 | |
49a73e0c | 3995 | port = bundle_execute(bundle, &ctx->xin->flow, ctx->wc, slave_enabled_cb, |
46c88433 | 3996 | CONST_CAST(struct xbridge *, ctx->xbridge)); |
9583bc14 | 3997 | if (bundle->dst.field) { |
49a73e0c | 3998 | nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow, ctx->wc); |
9583bc14 EJ |
3999 | } else { |
4000 | xlate_output_action(ctx, port, 0, false); | |
4001 | } | |
4002 | } | |
4003 | ||
4004 | static void | |
4165b5e0 JS |
4005 | xlate_learn_action__(struct xlate_ctx *ctx, const struct ofpact_learn *learn, |
4006 | struct ofputil_flow_mod *fm, struct ofpbuf *ofpacts) | |
9583bc14 | 4007 | { |
4165b5e0 JS |
4008 | learn_execute(learn, &ctx->xin->flow, fm, ofpacts); |
4009 | if (ctx->xin->may_learn) { | |
4010 | ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm); | |
4011 | } | |
4012 | } | |
9583bc14 | 4013 | |
4165b5e0 JS |
4014 | static void |
4015 | xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) | |
4016 | { | |
49a73e0c | 4017 | learn_mask(learn, ctx->wc); |
9583bc14 | 4018 | |
b256dc52 JS |
4019 | if (ctx->xin->xcache) { |
4020 | struct xc_entry *entry; | |
4021 | ||
4022 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN); | |
4165b5e0 JS |
4023 | entry->u.learn.ofproto = ctx->xbridge->ofproto; |
4024 | entry->u.learn.fm = xmalloc(sizeof *entry->u.learn.fm); | |
4025 | entry->u.learn.ofpacts = ofpbuf_new(64); | |
4026 | xlate_learn_action__(ctx, learn, entry->u.learn.fm, | |
4027 | entry->u.learn.ofpacts); | |
4028 | } else if (ctx->xin->may_learn) { | |
4029 | uint64_t ofpacts_stub[1024 / 8]; | |
4030 | struct ofputil_flow_mod fm; | |
4031 | struct ofpbuf ofpacts; | |
4032 | ||
4033 | ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); | |
4034 | xlate_learn_action__(ctx, learn, &fm, &ofpacts); | |
4035 | ofpbuf_uninit(&ofpacts); | |
b256dc52 JS |
4036 | } |
4037 | } | |
4038 | ||
4039 | static void | |
4040 | xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags, | |
4041 | uint16_t idle_timeout, uint16_t hard_timeout) | |
4042 | { | |
4043 | if (tcp_flags & (TCP_FIN | TCP_RST)) { | |
4044 | rule_dpif_reduce_timeouts(rule, idle_timeout, hard_timeout); | |
4045 | } | |
9583bc14 EJ |
4046 | } |
4047 | ||
9583bc14 EJ |
4048 | static void |
4049 | xlate_fin_timeout(struct xlate_ctx *ctx, | |
4050 | const struct ofpact_fin_timeout *oft) | |
4051 | { | |
b256dc52 JS |
4052 | if (ctx->rule) { |
4053 | xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags, | |
4054 | oft->fin_idle_timeout, oft->fin_hard_timeout); | |
4055 | if (ctx->xin->xcache) { | |
4056 | struct xc_entry *entry; | |
4057 | ||
4058 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT); | |
83709dfa JR |
4059 | /* XC_RULE already holds a reference on the rule, none is taken |
4060 | * here. */ | |
b256dc52 JS |
4061 | entry->u.fin.rule = ctx->rule; |
4062 | entry->u.fin.idle = oft->fin_idle_timeout; | |
4063 | entry->u.fin.hard = oft->fin_hard_timeout; | |
b256dc52 | 4064 | } |
9583bc14 EJ |
4065 | } |
4066 | } | |
4067 | ||
4068 | static void | |
4069 | xlate_sample_action(struct xlate_ctx *ctx, | |
4070 | const struct ofpact_sample *os) | |
4071 | { | |
e824d78d JR |
4072 | /* Scale the probability from 16-bit to 32-bit while representing |
4073 | * the same percentage. */ | |
4074 | uint32_t probability = (os->probability << 16) | os->probability; | |
4075 | ||
b440dd8c | 4076 | if (!ctx->xbridge->support.variable_length_userdata) { |
e824d78d JR |
4077 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); |
4078 | ||
4079 | VLOG_ERR_RL(&rl, "ignoring NXAST_SAMPLE action because datapath " | |
4080 | "lacks support (needs Linux 3.10+ or kernel module from " | |
4081 | "OVS 1.11+)"); | |
4082 | return; | |
4083 | } | |
4084 | ||
704bb0bf | 4085 | xlate_commit_actions(ctx); |
e824d78d | 4086 | |
a6092018 BP |
4087 | union user_action_cookie cookie = { |
4088 | .flow_sample = { | |
4089 | .type = USER_ACTION_COOKIE_FLOW_SAMPLE, | |
4090 | .probability = os->probability, | |
4091 | .collector_set_id = os->collector_set_id, | |
4092 | .obs_domain_id = os->obs_domain_id, | |
4093 | .obs_point_id = os->obs_point_id, | |
4094 | } | |
4095 | }; | |
4096 | compose_sample_action(ctx, probability, &cookie, sizeof cookie.flow_sample, | |
4097 | ODPP_NONE, false); | |
9583bc14 EJ |
4098 | } |
4099 | ||
4100 | static bool | |
46c88433 | 4101 | may_receive(const struct xport *xport, struct xlate_ctx *ctx) |
9583bc14 | 4102 | { |
bbbca389 | 4103 | if (xport->config & (is_stp(&ctx->xin->flow) |
46c88433 EJ |
4104 | ? OFPUTIL_PC_NO_RECV_STP |
4105 | : OFPUTIL_PC_NO_RECV)) { | |
9583bc14 EJ |
4106 | return false; |
4107 | } | |
4108 | ||
4109 | /* Only drop packets here if both forwarding and learning are | |
4110 | * disabled. If just learning is enabled, we need to have | |
4111 | * OFPP_NORMAL and the learning action have a look at the packet | |
4112 | * before we can drop it. */ | |
9efd308e DV |
4113 | if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) || |
4114 | (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) { | |
9583bc14 EJ |
4115 | return false; |
4116 | } | |
4117 | ||
4118 | return true; | |
4119 | } | |
4120 | ||
7fdb60a7 | 4121 | static void |
7e7e8dbb BP |
4122 | xlate_write_actions__(struct xlate_ctx *ctx, |
4123 | const struct ofpact *ofpacts, size_t ofpacts_len) | |
7fdb60a7 | 4124 | { |
c61f3870 BP |
4125 | /* Maintain actset_output depending on the contents of the action set: |
4126 | * | |
4127 | * - OFPP_UNSET, if there is no "output" action. | |
4128 | * | |
4129 | * - The output port, if there is an "output" action and no "group" | |
4130 | * action. | |
4131 | * | |
4132 | * - OFPP_UNSET, if there is a "group" action. | |
4133 | */ | |
4134 | if (!ctx->action_set_has_group) { | |
7e7e8dbb BP |
4135 | const struct ofpact *a; |
4136 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { | |
4137 | if (a->type == OFPACT_OUTPUT) { | |
4138 | ctx->xin->flow.actset_output = ofpact_get_OUTPUT(a)->port; | |
4139 | } else if (a->type == OFPACT_GROUP) { | |
c61f3870 BP |
4140 | ctx->xin->flow.actset_output = OFPP_UNSET; |
4141 | ctx->action_set_has_group = true; | |
9055ca9a | 4142 | break; |
c61f3870 BP |
4143 | } |
4144 | } | |
4145 | } | |
4146 | ||
7e7e8dbb BP |
4147 | ofpbuf_put(&ctx->action_set, ofpacts, ofpacts_len); |
4148 | } | |
4149 | ||
4150 | static void | |
4151 | xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact_nest *a) | |
4152 | { | |
4153 | xlate_write_actions__(ctx, a->actions, ofpact_nest_get_action_len(a)); | |
7fdb60a7 SH |
4154 | } |
4155 | ||
4156 | static void | |
4157 | xlate_action_set(struct xlate_ctx *ctx) | |
4158 | { | |
4159 | uint64_t action_list_stub[1024 / 64]; | |
4160 | struct ofpbuf action_list; | |
4161 | ||
029ca940 | 4162 | ctx->in_action_set = true; |
7fdb60a7 SH |
4163 | ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub); |
4164 | ofpacts_execute_action_set(&action_list, &ctx->action_set); | |
ed9c9e3e JR |
4165 | /* Clear the action set, as it is not needed any more. */ |
4166 | ofpbuf_clear(&ctx->action_set); | |
6fd6ed71 | 4167 | do_xlate_actions(action_list.data, action_list.size, ctx); |
029ca940 | 4168 | ctx->in_action_set = false; |
7fdb60a7 SH |
4169 | ofpbuf_uninit(&action_list); |
4170 | } | |
4171 | ||
e672ff9b | 4172 | static void |
1d361a81 | 4173 | freeze_put_unroll_xlate(struct xlate_ctx *ctx) |
e672ff9b | 4174 | { |
1d361a81 | 4175 | struct ofpact_unroll_xlate *unroll = ctx->frozen_actions.header; |
e672ff9b JR |
4176 | |
4177 | /* Restore the table_id and rule cookie for a potential PACKET | |
4178 | * IN if needed. */ | |
4179 | if (!unroll || | |
4180 | (ctx->table_id != unroll->rule_table_id | |
4181 | || ctx->rule_cookie != unroll->rule_cookie)) { | |
1d361a81 | 4182 | unroll = ofpact_put_UNROLL_XLATE(&ctx->frozen_actions); |
e672ff9b JR |
4183 | unroll->rule_table_id = ctx->table_id; |
4184 | unroll->rule_cookie = ctx->rule_cookie; | |
1d361a81 | 4185 | ctx->frozen_actions.header = unroll; |
e672ff9b JR |
4186 | } |
4187 | } | |
4188 | ||
4189 | ||
1d361a81 BP |
4190 | /* Copy actions 'a' through 'end' to ctx->frozen_actions, which will be |
4191 | * executed after thawing. Inserts an UNROLL_XLATE action, if none is already | |
4192 | * present, before any action that may depend on the current table ID or flow | |
4193 | * cookie. */ | |
e672ff9b | 4194 | static void |
1d361a81 | 4195 | freeze_unroll_actions(const struct ofpact *a, const struct ofpact *end, |
e672ff9b JR |
4196 | struct xlate_ctx *ctx) |
4197 | { | |
c2b283b7 | 4198 | for (; a < end; a = ofpact_next(a)) { |
e672ff9b | 4199 | switch (a->type) { |
e672ff9b JR |
4200 | case OFPACT_OUTPUT_REG: |
4201 | case OFPACT_GROUP: | |
4202 | case OFPACT_OUTPUT: | |
4203 | case OFPACT_CONTROLLER: | |
4204 | case OFPACT_DEC_MPLS_TTL: | |
4205 | case OFPACT_DEC_TTL: | |
83a31283 BP |
4206 | /* These actions may generate asynchronous messages, which include |
4207 | * table ID and flow cookie information. */ | |
1d361a81 | 4208 | freeze_put_unroll_xlate(ctx); |
e672ff9b JR |
4209 | break; |
4210 | ||
83a31283 BP |
4211 | case OFPACT_RESUBMIT: |
4212 | if (ofpact_get_RESUBMIT(a)->table_id == 0xff) { | |
4213 | /* This resubmit action is relative to the current table, so we | |
4214 | * need to track what table that is.*/ | |
1d361a81 | 4215 | freeze_put_unroll_xlate(ctx); |
83a31283 BP |
4216 | } |
4217 | break; | |
4218 | ||
e672ff9b JR |
4219 | case OFPACT_SET_TUNNEL: |
4220 | case OFPACT_REG_MOVE: | |
4221 | case OFPACT_SET_FIELD: | |
4222 | case OFPACT_STACK_PUSH: | |
4223 | case OFPACT_STACK_POP: | |
4224 | case OFPACT_LEARN: | |
4225 | case OFPACT_WRITE_METADATA: | |
83a31283 | 4226 | case OFPACT_GOTO_TABLE: |
e672ff9b JR |
4227 | case OFPACT_ENQUEUE: |
4228 | case OFPACT_SET_VLAN_VID: | |
4229 | case OFPACT_SET_VLAN_PCP: | |
4230 | case OFPACT_STRIP_VLAN: | |
4231 | case OFPACT_PUSH_VLAN: | |
4232 | case OFPACT_SET_ETH_SRC: | |
4233 | case OFPACT_SET_ETH_DST: | |
4234 | case OFPACT_SET_IPV4_SRC: | |
4235 | case OFPACT_SET_IPV4_DST: | |
4236 | case OFPACT_SET_IP_DSCP: | |
4237 | case OFPACT_SET_IP_ECN: | |
4238 | case OFPACT_SET_IP_TTL: | |
4239 | case OFPACT_SET_L4_SRC_PORT: | |
4240 | case OFPACT_SET_L4_DST_PORT: | |
4241 | case OFPACT_SET_QUEUE: | |
4242 | case OFPACT_POP_QUEUE: | |
4243 | case OFPACT_PUSH_MPLS: | |
4244 | case OFPACT_POP_MPLS: | |
4245 | case OFPACT_SET_MPLS_LABEL: | |
4246 | case OFPACT_SET_MPLS_TC: | |
4247 | case OFPACT_SET_MPLS_TTL: | |
4248 | case OFPACT_MULTIPATH: | |
4249 | case OFPACT_BUNDLE: | |
4250 | case OFPACT_EXIT: | |
4251 | case OFPACT_UNROLL_XLATE: | |
4252 | case OFPACT_FIN_TIMEOUT: | |
4253 | case OFPACT_CLEAR_ACTIONS: | |
4254 | case OFPACT_WRITE_ACTIONS: | |
4255 | case OFPACT_METER: | |
4256 | case OFPACT_SAMPLE: | |
d4abaff5 | 4257 | case OFPACT_DEBUG_RECIRC: |
07659514 | 4258 | case OFPACT_CT: |
9ac0aada | 4259 | case OFPACT_NAT: |
83a31283 | 4260 | /* These may not generate PACKET INs. */ |
e672ff9b JR |
4261 | break; |
4262 | ||
e672ff9b JR |
4263 | case OFPACT_NOTE: |
4264 | case OFPACT_CONJUNCTION: | |
83a31283 | 4265 | /* These need not be copied for restoration. */ |
e672ff9b JR |
4266 | continue; |
4267 | } | |
4268 | /* Copy the action over. */ | |
1d361a81 | 4269 | ofpbuf_put(&ctx->frozen_actions, a, OFPACT_ALIGN(a->len)); |
e672ff9b JR |
4270 | } |
4271 | } | |
4272 | ||
8e53fe8c | 4273 | static void |
f2d105b5 JS |
4274 | put_ct_mark(const struct flow *flow, struct ofpbuf *odp_actions, |
4275 | struct flow_wildcards *wc) | |
8e53fe8c | 4276 | { |
2a754f4a JS |
4277 | if (wc->masks.ct_mark) { |
4278 | struct { | |
4279 | uint32_t key; | |
4280 | uint32_t mask; | |
4281 | } *odp_ct_mark; | |
4282 | ||
4283 | odp_ct_mark = nl_msg_put_unspec_uninit(odp_actions, OVS_CT_ATTR_MARK, | |
4284 | sizeof(*odp_ct_mark)); | |
4285 | odp_ct_mark->key = flow->ct_mark & wc->masks.ct_mark; | |
4286 | odp_ct_mark->mask = wc->masks.ct_mark; | |
8e53fe8c JS |
4287 | } |
4288 | } | |
4289 | ||
9daf2348 | 4290 | static void |
f2d105b5 JS |
4291 | put_ct_label(const struct flow *flow, struct ofpbuf *odp_actions, |
4292 | struct flow_wildcards *wc) | |
9daf2348 | 4293 | { |
f2d105b5 | 4294 | if (!ovs_u128_is_zero(&wc->masks.ct_label)) { |
9daf2348 JS |
4295 | struct { |
4296 | ovs_u128 key; | |
4297 | ovs_u128 mask; | |
4298 | } *odp_ct_label; | |
4299 | ||
4300 | odp_ct_label = nl_msg_put_unspec_uninit(odp_actions, | |
4301 | OVS_CT_ATTR_LABELS, | |
4302 | sizeof(*odp_ct_label)); | |
f2d105b5 | 4303 | odp_ct_label->key = ovs_u128_and(flow->ct_label, wc->masks.ct_label); |
9daf2348 JS |
4304 | odp_ct_label->mask = wc->masks.ct_label; |
4305 | } | |
4306 | } | |
4307 | ||
d787ad39 JS |
4308 | static void |
4309 | put_ct_helper(struct ofpbuf *odp_actions, struct ofpact_conntrack *ofc) | |
4310 | { | |
4311 | if (ofc->alg) { | |
4312 | if (ofc->alg == IPPORT_FTP) { | |
4313 | nl_msg_put_string(odp_actions, OVS_CT_ATTR_HELPER, "ftp"); | |
4314 | } else { | |
4315 | VLOG_WARN("Cannot serialize ct_helper %d\n", ofc->alg); | |
4316 | } | |
4317 | } | |
4318 | } | |
4319 | ||
9ac0aada JR |
4320 | static void |
4321 | put_ct_nat(struct xlate_ctx *ctx) | |
4322 | { | |
4323 | struct ofpact_nat *ofn = ctx->ct_nat_action; | |
4324 | size_t nat_offset; | |
4325 | ||
4326 | if (!ofn) { | |
4327 | return; | |
4328 | } | |
4329 | ||
4330 | nat_offset = nl_msg_start_nested(ctx->odp_actions, OVS_CT_ATTR_NAT); | |
4331 | if (ofn->flags & NX_NAT_F_SRC || ofn->flags & NX_NAT_F_DST) { | |
4332 | nl_msg_put_flag(ctx->odp_actions, ofn->flags & NX_NAT_F_SRC | |
4333 | ? OVS_NAT_ATTR_SRC : OVS_NAT_ATTR_DST); | |
4334 | if (ofn->flags & NX_NAT_F_PERSISTENT) { | |
4335 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PERSISTENT); | |
4336 | } | |
4337 | if (ofn->flags & NX_NAT_F_PROTO_HASH) { | |
4338 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_HASH); | |
4339 | } else if (ofn->flags & NX_NAT_F_PROTO_RANDOM) { | |
4340 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_RANDOM); | |
4341 | } | |
4342 | if (ofn->range_af == AF_INET) { | |
73e8bc23 | 4343 | nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, |
9ac0aada JR |
4344 | ofn->range.addr.ipv4.min); |
4345 | if (ofn->range.addr.ipv4.max && | |
73e8bc23 BP |
4346 | (ntohl(ofn->range.addr.ipv4.max) |
4347 | > ntohl(ofn->range.addr.ipv4.min))) { | |
4348 | nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, | |
4349 | ofn->range.addr.ipv4.max); | |
9ac0aada JR |
4350 | } |
4351 | } else if (ofn->range_af == AF_INET6) { | |
4352 | nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, | |
4353 | &ofn->range.addr.ipv6.min, | |
4354 | sizeof ofn->range.addr.ipv6.min); | |
4355 | if (!ipv6_mask_is_any(&ofn->range.addr.ipv6.max) && | |
4356 | memcmp(&ofn->range.addr.ipv6.max, &ofn->range.addr.ipv6.min, | |
4357 | sizeof ofn->range.addr.ipv6.max) > 0) { | |
4358 | nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, | |
4359 | &ofn->range.addr.ipv6.max, | |
4360 | sizeof ofn->range.addr.ipv6.max); | |
4361 | } | |
4362 | } | |
4363 | if (ofn->range_af != AF_UNSPEC && ofn->range.proto.min) { | |
4364 | nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MIN, | |
4365 | ofn->range.proto.min); | |
4366 | if (ofn->range.proto.max && | |
4367 | ofn->range.proto.max > ofn->range.proto.min) { | |
4368 | nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MAX, | |
4369 | ofn->range.proto.max); | |
4370 | } | |
4371 | } | |
4372 | } | |
4373 | nl_msg_end_nested(ctx->odp_actions, nat_offset); | |
4374 | } | |
4375 | ||
07659514 JS |
4376 | static void |
4377 | compose_conntrack_action(struct xlate_ctx *ctx, struct ofpact_conntrack *ofc) | |
4378 | { | |
9daf2348 | 4379 | ovs_u128 old_ct_label = ctx->base_flow.ct_label; |
f2d105b5 | 4380 | ovs_u128 old_ct_label_mask = ctx->wc->masks.ct_label; |
8e53fe8c | 4381 | uint32_t old_ct_mark = ctx->base_flow.ct_mark; |
f2d105b5 | 4382 | uint32_t old_ct_mark_mask = ctx->wc->masks.ct_mark; |
07659514 JS |
4383 | size_t ct_offset; |
4384 | uint16_t zone; | |
4385 | ||
4386 | /* Ensure that any prior actions are applied before composing the new | |
4387 | * conntrack action. */ | |
4388 | xlate_commit_actions(ctx); | |
4389 | ||
8e53fe8c | 4390 | /* Process nested actions first, to populate the key. */ |
9ac0aada | 4391 | ctx->ct_nat_action = NULL; |
f2d105b5 JS |
4392 | ctx->wc->masks.ct_mark = 0; |
4393 | ctx->wc->masks.ct_label.u64.hi = ctx->wc->masks.ct_label.u64.lo = 0; | |
8e53fe8c JS |
4394 | do_xlate_actions(ofc->actions, ofpact_ct_get_action_len(ofc), ctx); |
4395 | ||
07659514 JS |
4396 | if (ofc->zone_src.field) { |
4397 | zone = mf_get_subfield(&ofc->zone_src, &ctx->xin->flow); | |
4398 | } else { | |
4399 | zone = ofc->zone_imm; | |
4400 | } | |
4401 | ||
4402 | ct_offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CT); | |
4403 | if (ofc->flags & NX_CT_F_COMMIT) { | |
4404 | nl_msg_put_flag(ctx->odp_actions, OVS_CT_ATTR_COMMIT); | |
4405 | } | |
4406 | nl_msg_put_u16(ctx->odp_actions, OVS_CT_ATTR_ZONE, zone); | |
f2d105b5 JS |
4407 | put_ct_mark(&ctx->xin->flow, ctx->odp_actions, ctx->wc); |
4408 | put_ct_label(&ctx->xin->flow, ctx->odp_actions, ctx->wc); | |
d787ad39 | 4409 | put_ct_helper(ctx->odp_actions, ofc); |
9ac0aada JR |
4410 | put_ct_nat(ctx); |
4411 | ctx->ct_nat_action = NULL; | |
07659514 JS |
4412 | nl_msg_end_nested(ctx->odp_actions, ct_offset); |
4413 | ||
8e53fe8c JS |
4414 | /* Restore the original ct fields in the key. These should only be exposed |
4415 | * after recirculation to another table. */ | |
4416 | ctx->base_flow.ct_mark = old_ct_mark; | |
f2d105b5 | 4417 | ctx->wc->masks.ct_mark = old_ct_mark_mask; |
9daf2348 | 4418 | ctx->base_flow.ct_label = old_ct_label; |
f2d105b5 | 4419 | ctx->wc->masks.ct_label = old_ct_label_mask; |
8e53fe8c | 4420 | |
07659514 JS |
4421 | if (ofc->recirc_table == NX_CT_RECIRC_NONE) { |
4422 | /* If we do not recirculate as part of this action, hide the results of | |
4423 | * connection tracking from subsequent recirculations. */ | |
4424 | ctx->conntracked = false; | |
4425 | } else { | |
4426 | /* Use ct_* fields from datapath during recirculation upcall. */ | |
4427 | ctx->conntracked = true; | |
e37b8437 | 4428 | compose_recirculate_and_fork(ctx, ofc->recirc_table); |
07659514 JS |
4429 | } |
4430 | } | |
4431 | ||
9583bc14 EJ |
4432 | static void |
4433 | do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
4434 | struct xlate_ctx *ctx) | |
4435 | { | |
49a73e0c | 4436 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 4437 | struct flow *flow = &ctx->xin->flow; |
9583bc14 EJ |
4438 | const struct ofpact *a; |
4439 | ||
a36de779 | 4440 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
53902038 | 4441 | tnl_neigh_snoop(flow, wc, ctx->xbridge->name); |
a36de779 | 4442 | } |
f47ea021 JR |
4443 | /* dl_type already in the mask, not set below. */ |
4444 | ||
9583bc14 EJ |
4445 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
4446 | struct ofpact_controller *controller; | |
4447 | const struct ofpact_metadata *metadata; | |
b2dd70be JR |
4448 | const struct ofpact_set_field *set_field; |
4449 | const struct mf_field *mf; | |
9583bc14 | 4450 | |
fff1b9c0 JR |
4451 | if (ctx->error) { |
4452 | break; | |
4453 | } | |
4454 | ||
e672ff9b JR |
4455 | if (ctx->exit) { |
4456 | /* Check if need to store the remaining actions for later | |
4457 | * execution. */ | |
1d361a81 BP |
4458 | if (ctx->freezing) { |
4459 | freeze_unroll_actions(a, ofpact_end(ofpacts, ofpacts_len), | |
e672ff9b JR |
4460 | ctx); |
4461 | } | |
4462 | break; | |
7bbdd84f SH |
4463 | } |
4464 | ||
9583bc14 EJ |
4465 | switch (a->type) { |
4466 | case OFPACT_OUTPUT: | |
4467 | xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port, | |
4468 | ofpact_get_OUTPUT(a)->max_len, true); | |
4469 | break; | |
4470 | ||
7395c052 | 4471 | case OFPACT_GROUP: |
f4fb341b | 4472 | if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id)) { |
1d741d6d | 4473 | /* Group could not be found. */ |
f4fb341b SH |
4474 | return; |
4475 | } | |
7395c052 NZ |
4476 | break; |
4477 | ||
9583bc14 EJ |
4478 | case OFPACT_CONTROLLER: |
4479 | controller = ofpact_get_CONTROLLER(a); | |
77ab5fd2 BP |
4480 | if (controller->pause) { |
4481 | ctx->pause = controller; | |
4482 | ctx->xout->slow |= SLOW_CONTROLLER; | |
4483 | ctx_trigger_freeze(ctx); | |
4484 | a = ofpact_next(a); | |
4485 | } else { | |
4486 | execute_controller_action(ctx, controller->max_len, | |
4487 | controller->reason, | |
4488 | controller->controller_id, | |
4489 | controller->userdata, | |
4490 | controller->userdata_len); | |
4491 | } | |
9583bc14 EJ |
4492 | break; |
4493 | ||
4494 | case OFPACT_ENQUEUE: | |
16194afd DDP |
4495 | memset(&wc->masks.skb_priority, 0xff, |
4496 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4497 | xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a)); |
4498 | break; | |
4499 | ||
4500 | case OFPACT_SET_VLAN_VID: | |
f74e7df7 | 4501 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
ca287d20 JR |
4502 | if (flow->vlan_tci & htons(VLAN_CFI) || |
4503 | ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) { | |
4504 | flow->vlan_tci &= ~htons(VLAN_VID_MASK); | |
4505 | flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid) | |
4506 | | htons(VLAN_CFI)); | |
4507 | } | |
9583bc14 EJ |
4508 | break; |
4509 | ||
4510 | case OFPACT_SET_VLAN_PCP: | |
f74e7df7 | 4511 | wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI); |
ca287d20 JR |
4512 | if (flow->vlan_tci & htons(VLAN_CFI) || |
4513 | ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) { | |
4514 | flow->vlan_tci &= ~htons(VLAN_PCP_MASK); | |
4515 | flow->vlan_tci |= htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp | |
4516 | << VLAN_PCP_SHIFT) | VLAN_CFI); | |
4517 | } | |
9583bc14 EJ |
4518 | break; |
4519 | ||
4520 | case OFPACT_STRIP_VLAN: | |
f74e7df7 | 4521 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 4522 | flow->vlan_tci = htons(0); |
9583bc14 EJ |
4523 | break; |
4524 | ||
4525 | case OFPACT_PUSH_VLAN: | |
4526 | /* XXX 802.1AD(QinQ) */ | |
f74e7df7 | 4527 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 4528 | flow->vlan_tci = htons(VLAN_CFI); |
9583bc14 EJ |
4529 | break; |
4530 | ||
4531 | case OFPACT_SET_ETH_SRC: | |
74ff3298 JR |
4532 | WC_MASK_FIELD(wc, dl_src); |
4533 | flow->dl_src = ofpact_get_SET_ETH_SRC(a)->mac; | |
9583bc14 EJ |
4534 | break; |
4535 | ||
4536 | case OFPACT_SET_ETH_DST: | |
74ff3298 JR |
4537 | WC_MASK_FIELD(wc, dl_dst); |
4538 | flow->dl_dst = ofpact_get_SET_ETH_DST(a)->mac; | |
9583bc14 EJ |
4539 | break; |
4540 | ||
4541 | case OFPACT_SET_IPV4_SRC: | |
33bf9176 | 4542 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 4543 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); |
33bf9176 | 4544 | flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4; |
9583bc14 EJ |
4545 | } |
4546 | break; | |
4547 | ||
4548 | case OFPACT_SET_IPV4_DST: | |
33bf9176 | 4549 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 4550 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); |
33bf9176 | 4551 | flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4; |
9583bc14 EJ |
4552 | } |
4553 | break; | |
4554 | ||
04f01c24 BP |
4555 | case OFPACT_SET_IP_DSCP: |
4556 | if (is_ip_any(flow)) { | |
f47ea021 | 4557 | wc->masks.nw_tos |= IP_DSCP_MASK; |
33bf9176 | 4558 | flow->nw_tos &= ~IP_DSCP_MASK; |
04f01c24 | 4559 | flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp; |
9583bc14 EJ |
4560 | } |
4561 | break; | |
4562 | ||
ff14eb7a JR |
4563 | case OFPACT_SET_IP_ECN: |
4564 | if (is_ip_any(flow)) { | |
4565 | wc->masks.nw_tos |= IP_ECN_MASK; | |
4566 | flow->nw_tos &= ~IP_ECN_MASK; | |
4567 | flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn; | |
4568 | } | |
4569 | break; | |
4570 | ||
0c20dbe4 JR |
4571 | case OFPACT_SET_IP_TTL: |
4572 | if (is_ip_any(flow)) { | |
4573 | wc->masks.nw_ttl = 0xff; | |
4574 | flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl; | |
4575 | } | |
4576 | break; | |
4577 | ||
9583bc14 | 4578 | case OFPACT_SET_L4_SRC_PORT: |
b8778a0d | 4579 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
4580 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
4581 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
33bf9176 | 4582 | flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port); |
9583bc14 EJ |
4583 | } |
4584 | break; | |
4585 | ||
4586 | case OFPACT_SET_L4_DST_PORT: | |
b8778a0d | 4587 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
4588 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
4589 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
33bf9176 | 4590 | flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port); |
9583bc14 EJ |
4591 | } |
4592 | break; | |
4593 | ||
4594 | case OFPACT_RESUBMIT: | |
8bf009bf JR |
4595 | /* Freezing complicates resubmit. Some action in the flow |
4596 | * entry found by resubmit might trigger freezing. If that | |
4597 | * happens, then we do not want to execute the resubmit again after | |
4598 | * during thawing, so we want to skip back to the head of the loop | |
4599 | * to avoid that, only adding any actions that follow the resubmit | |
4600 | * to the frozen actions. | |
6b1c5734 | 4601 | */ |
9583bc14 | 4602 | xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a)); |
6b1c5734 | 4603 | continue; |
9583bc14 EJ |
4604 | |
4605 | case OFPACT_SET_TUNNEL: | |
33bf9176 | 4606 | flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id); |
9583bc14 EJ |
4607 | break; |
4608 | ||
4609 | case OFPACT_SET_QUEUE: | |
16194afd DDP |
4610 | memset(&wc->masks.skb_priority, 0xff, |
4611 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4612 | xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id); |
4613 | break; | |
4614 | ||
4615 | case OFPACT_POP_QUEUE: | |
16194afd DDP |
4616 | memset(&wc->masks.skb_priority, 0xff, |
4617 | sizeof wc->masks.skb_priority); | |
33bf9176 | 4618 | flow->skb_priority = ctx->orig_skb_priority; |
9583bc14 EJ |
4619 | break; |
4620 | ||
4621 | case OFPACT_REG_MOVE: | |
33bf9176 | 4622 | nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc); |
9583bc14 EJ |
4623 | break; |
4624 | ||
b2dd70be JR |
4625 | case OFPACT_SET_FIELD: |
4626 | set_field = ofpact_get_SET_FIELD(a); | |
4627 | mf = set_field->field; | |
b2dd70be JR |
4628 | |
4629 | /* Set field action only ever overwrites packet's outermost | |
4630 | * applicable header fields. Do nothing if no header exists. */ | |
4479846e JR |
4631 | if (mf->id == MFF_VLAN_VID) { |
4632 | wc->masks.vlan_tci |= htons(VLAN_CFI); | |
4633 | if (!(flow->vlan_tci & htons(VLAN_CFI))) { | |
4634 | break; | |
4635 | } | |
4636 | } else if ((mf->id == MFF_MPLS_LABEL || mf->id == MFF_MPLS_TC) | |
4637 | /* 'dl_type' is already unwildcarded. */ | |
4638 | && !eth_type_mpls(flow->dl_type)) { | |
4639 | break; | |
b2dd70be | 4640 | } |
67794d56 | 4641 | /* A flow may wildcard nw_frag. Do nothing if setting a transport |
b8778a0d | 4642 | * header field on a packet that does not have them. */ |
4d182934 | 4643 | mf_mask_field_and_prereqs__(mf, &set_field->mask, wc); |
b8778a0d JR |
4644 | if (mf_are_prereqs_ok(mf, flow)) { |
4645 | mf_set_flow_value_masked(mf, &set_field->value, | |
4646 | &set_field->mask, flow); | |
4647 | } | |
b2dd70be JR |
4648 | break; |
4649 | ||
9583bc14 | 4650 | case OFPACT_STACK_PUSH: |
33bf9176 BP |
4651 | nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc, |
4652 | &ctx->stack); | |
9583bc14 EJ |
4653 | break; |
4654 | ||
4655 | case OFPACT_STACK_POP: | |
f74e7df7 JP |
4656 | nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc, |
4657 | &ctx->stack); | |
9583bc14 EJ |
4658 | break; |
4659 | ||
4660 | case OFPACT_PUSH_MPLS: | |
8bfd0fda | 4661 | compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)); |
9583bc14 EJ |
4662 | break; |
4663 | ||
4664 | case OFPACT_POP_MPLS: | |
8bfd0fda | 4665 | compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype); |
9583bc14 EJ |
4666 | break; |
4667 | ||
097d4939 | 4668 | case OFPACT_SET_MPLS_LABEL: |
8bfd0fda BP |
4669 | compose_set_mpls_label_action( |
4670 | ctx, ofpact_get_SET_MPLS_LABEL(a)->label); | |
1d741d6d | 4671 | break; |
097d4939 JR |
4672 | |
4673 | case OFPACT_SET_MPLS_TC: | |
8bfd0fda | 4674 | compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc); |
097d4939 JR |
4675 | break; |
4676 | ||
9583bc14 | 4677 | case OFPACT_SET_MPLS_TTL: |
8bfd0fda | 4678 | compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl); |
9583bc14 EJ |
4679 | break; |
4680 | ||
4681 | case OFPACT_DEC_MPLS_TTL: | |
9cfef3d0 | 4682 | if (compose_dec_mpls_ttl_action(ctx)) { |
ad3efdcb | 4683 | return; |
9583bc14 EJ |
4684 | } |
4685 | break; | |
4686 | ||
4687 | case OFPACT_DEC_TTL: | |
f74e7df7 | 4688 | wc->masks.nw_ttl = 0xff; |
9583bc14 | 4689 | if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) { |
ad3efdcb | 4690 | return; |
9583bc14 EJ |
4691 | } |
4692 | break; | |
4693 | ||
4694 | case OFPACT_NOTE: | |
4695 | /* Nothing to do. */ | |
4696 | break; | |
4697 | ||
4698 | case OFPACT_MULTIPATH: | |
33bf9176 | 4699 | multipath_execute(ofpact_get_MULTIPATH(a), flow, wc); |
9583bc14 EJ |
4700 | break; |
4701 | ||
4702 | case OFPACT_BUNDLE: | |
9583bc14 EJ |
4703 | xlate_bundle_action(ctx, ofpact_get_BUNDLE(a)); |
4704 | break; | |
4705 | ||
4706 | case OFPACT_OUTPUT_REG: | |
4707 | xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a)); | |
4708 | break; | |
4709 | ||
4710 | case OFPACT_LEARN: | |
4711 | xlate_learn_action(ctx, ofpact_get_LEARN(a)); | |
4712 | break; | |
4713 | ||
afc3987b BP |
4714 | case OFPACT_CONJUNCTION: { |
4715 | /* A flow with a "conjunction" action represents part of a special | |
4716 | * kind of "set membership match". Such a flow should not actually | |
4717 | * get executed, but it could via, say, a "packet-out", even though | |
4718 | * that wouldn't be useful. Log it to help debugging. */ | |
4719 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4720 | VLOG_INFO_RL(&rl, "executing no-op conjunction action"); | |
18080541 | 4721 | break; |
afc3987b | 4722 | } |
18080541 | 4723 | |
9583bc14 EJ |
4724 | case OFPACT_EXIT: |
4725 | ctx->exit = true; | |
4726 | break; | |
4727 | ||
e672ff9b JR |
4728 | case OFPACT_UNROLL_XLATE: { |
4729 | struct ofpact_unroll_xlate *unroll = ofpact_get_UNROLL_XLATE(a); | |
4730 | ||
4731 | /* Restore translation context data that was stored earlier. */ | |
4732 | ctx->table_id = unroll->rule_table_id; | |
4733 | ctx->rule_cookie = unroll->rule_cookie; | |
4734 | break; | |
4735 | } | |
9583bc14 | 4736 | case OFPACT_FIN_TIMEOUT: |
33bf9176 | 4737 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
9583bc14 EJ |
4738 | xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a)); |
4739 | break; | |
4740 | ||
4741 | case OFPACT_CLEAR_ACTIONS: | |
7fdb60a7 | 4742 | ofpbuf_clear(&ctx->action_set); |
c61f3870 BP |
4743 | ctx->xin->flow.actset_output = OFPP_UNSET; |
4744 | ctx->action_set_has_group = false; | |
7fdb60a7 SH |
4745 | break; |
4746 | ||
4747 | case OFPACT_WRITE_ACTIONS: | |
7e7e8dbb | 4748 | xlate_write_actions(ctx, ofpact_get_WRITE_ACTIONS(a)); |
9583bc14 EJ |
4749 | break; |
4750 | ||
4751 | case OFPACT_WRITE_METADATA: | |
4752 | metadata = ofpact_get_WRITE_METADATA(a); | |
33bf9176 BP |
4753 | flow->metadata &= ~metadata->mask; |
4754 | flow->metadata |= metadata->metadata & metadata->mask; | |
9583bc14 EJ |
4755 | break; |
4756 | ||
638a19b0 JR |
4757 | case OFPACT_METER: |
4758 | /* Not implemented yet. */ | |
4759 | break; | |
4760 | ||
9583bc14 | 4761 | case OFPACT_GOTO_TABLE: { |
9583bc14 | 4762 | struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a); |
9583bc14 | 4763 | |
9167fc1a JR |
4764 | ovs_assert(ctx->table_id < ogt->table_id); |
4765 | ||
4468099e | 4766 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 4767 | ogt->table_id, true, true); |
9583bc14 EJ |
4768 | break; |
4769 | } | |
4770 | ||
4771 | case OFPACT_SAMPLE: | |
4772 | xlate_sample_action(ctx, ofpact_get_SAMPLE(a)); | |
4773 | break; | |
d4abaff5 | 4774 | |
07659514 | 4775 | case OFPACT_CT: |
07659514 JS |
4776 | compose_conntrack_action(ctx, ofpact_get_CT(a)); |
4777 | break; | |
4778 | ||
9ac0aada JR |
4779 | case OFPACT_NAT: |
4780 | /* This will be processed by compose_conntrack_action(). */ | |
4781 | ctx->ct_nat_action = ofpact_get_NAT(a); | |
4782 | break; | |
4783 | ||
d4abaff5 | 4784 | case OFPACT_DEBUG_RECIRC: |
1d361a81 | 4785 | ctx_trigger_freeze(ctx); |
d4abaff5 BP |
4786 | a = ofpact_next(a); |
4787 | break; | |
9583bc14 | 4788 | } |
1d741d6d JR |
4789 | |
4790 | /* Check if need to store this and the remaining actions for later | |
4791 | * execution. */ | |
1d361a81 BP |
4792 | if (!ctx->error && ctx->exit && ctx_first_frozen_action(ctx)) { |
4793 | freeze_unroll_actions(a, ofpact_end(ofpacts, ofpacts_len), ctx); | |
1d741d6d JR |
4794 | break; |
4795 | } | |
9583bc14 | 4796 | } |
9583bc14 EJ |
4797 | } |
4798 | ||
4799 | void | |
4800 | xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, | |
cc377352 EJ |
4801 | const struct flow *flow, ofp_port_t in_port, |
4802 | struct rule_dpif *rule, uint16_t tcp_flags, | |
1520ef4f BP |
4803 | const struct dp_packet *packet, struct flow_wildcards *wc, |
4804 | struct ofpbuf *odp_actions) | |
9583bc14 EJ |
4805 | { |
4806 | xin->ofproto = ofproto; | |
4807 | xin->flow = *flow; | |
cc377352 | 4808 | xin->flow.in_port.ofp_port = in_port; |
c61f3870 | 4809 | xin->flow.actset_output = OFPP_UNSET; |
9583bc14 EJ |
4810 | xin->packet = packet; |
4811 | xin->may_learn = packet != NULL; | |
4812 | xin->rule = rule; | |
b256dc52 | 4813 | xin->xcache = NULL; |
9583bc14 EJ |
4814 | xin->ofpacts = NULL; |
4815 | xin->ofpacts_len = 0; | |
4816 | xin->tcp_flags = tcp_flags; | |
4817 | xin->resubmit_hook = NULL; | |
4818 | xin->report_hook = NULL; | |
4819 | xin->resubmit_stats = NULL; | |
9a9b001a | 4820 | xin->indentation = 0; |
cdd42eda | 4821 | xin->resubmits = 0; |
49a73e0c | 4822 | xin->wc = wc; |
1520ef4f | 4823 | xin->odp_actions = odp_actions; |
e672ff9b JR |
4824 | |
4825 | /* Do recirc lookup. */ | |
1d361a81 | 4826 | xin->frozen_state = NULL; |
29b1ea3f BP |
4827 | if (flow->recirc_id) { |
4828 | const struct recirc_id_node *node | |
4829 | = recirc_id_node_find(flow->recirc_id); | |
4830 | if (node) { | |
1d361a81 | 4831 | xin->frozen_state = &node->state; |
29b1ea3f BP |
4832 | } |
4833 | } | |
9583bc14 EJ |
4834 | } |
4835 | ||
4836 | void | |
4837 | xlate_out_uninit(struct xlate_out *xout) | |
4838 | { | |
e672ff9b | 4839 | if (xout) { |
fbf5d6ec | 4840 | recirc_refs_unref(&xout->recircs); |
9583bc14 EJ |
4841 | } |
4842 | } | |
4843 | ||
4844 | /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts' | |
4845 | * into datapath actions, using 'ctx', and discards the datapath actions. */ | |
4846 | void | |
4847 | xlate_actions_for_side_effects(struct xlate_in *xin) | |
4848 | { | |
4849 | struct xlate_out xout; | |
fff1b9c0 JR |
4850 | enum xlate_error error; |
4851 | ||
4852 | error = xlate_actions(xin, &xout); | |
4853 | if (error) { | |
4854 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
4855 | ||
4856 | VLOG_WARN_RL(&rl, "xlate_actions failed (%s)!", xlate_strerror(error)); | |
4857 | } | |
9583bc14 | 4858 | |
9583bc14 EJ |
4859 | xlate_out_uninit(&xout); |
4860 | } | |
9583bc14 | 4861 | \f |
55954f6e EJ |
4862 | static struct skb_priority_to_dscp * |
4863 | get_skb_priority(const struct xport *xport, uint32_t skb_priority) | |
4864 | { | |
4865 | struct skb_priority_to_dscp *pdscp; | |
4866 | uint32_t hash; | |
4867 | ||
4868 | hash = hash_int(skb_priority, 0); | |
4869 | HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) { | |
4870 | if (pdscp->skb_priority == skb_priority) { | |
4871 | return pdscp; | |
4872 | } | |
4873 | } | |
4874 | return NULL; | |
4875 | } | |
4876 | ||
4877 | static bool | |
4878 | dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority, | |
4879 | uint8_t *dscp) | |
4880 | { | |
4881 | struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority); | |
4882 | *dscp = pdscp ? pdscp->dscp : 0; | |
4883 | return pdscp != NULL; | |
4884 | } | |
4885 | ||
16194afd DDP |
4886 | static size_t |
4887 | count_skb_priorities(const struct xport *xport) | |
4888 | { | |
4889 | return hmap_count(&xport->skb_priorities); | |
4890 | } | |
4891 | ||
55954f6e EJ |
4892 | static void |
4893 | clear_skb_priorities(struct xport *xport) | |
4894 | { | |
4ec3d7c7 | 4895 | struct skb_priority_to_dscp *pdscp; |
55954f6e | 4896 | |
4ec3d7c7 | 4897 | HMAP_FOR_EACH_POP (pdscp, hmap_node, &xport->skb_priorities) { |
55954f6e EJ |
4898 | free(pdscp); |
4899 | } | |
4900 | } | |
4901 | ||
ce4a6b76 BP |
4902 | static bool |
4903 | actions_output_to_local_port(const struct xlate_ctx *ctx) | |
4904 | { | |
46c88433 | 4905 | odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL); |
ce4a6b76 BP |
4906 | const struct nlattr *a; |
4907 | unsigned int left; | |
4908 | ||
1520ef4f BP |
4909 | NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->odp_actions->data, |
4910 | ctx->odp_actions->size) { | |
ce4a6b76 BP |
4911 | if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT |
4912 | && nl_attr_get_odp_port(a) == local_odp_port) { | |
4913 | return true; | |
4914 | } | |
4915 | } | |
4916 | return false; | |
4917 | } | |
9583bc14 | 4918 | |
5e2a6702 | 4919 | #if defined(__linux__) |
7d031d7e BP |
4920 | /* Returns the maximum number of packets that the Linux kernel is willing to |
4921 | * queue up internally to certain kinds of software-implemented ports, or the | |
4922 | * default (and rarely modified) value if it cannot be determined. */ | |
4923 | static int | |
4924 | netdev_max_backlog(void) | |
4925 | { | |
4926 | static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; | |
4927 | static int max_backlog = 1000; /* The normal default value. */ | |
4928 | ||
4929 | if (ovsthread_once_start(&once)) { | |
4930 | static const char filename[] = "/proc/sys/net/core/netdev_max_backlog"; | |
4931 | FILE *stream; | |
4932 | int n; | |
4933 | ||
4934 | stream = fopen(filename, "r"); | |
4935 | if (!stream) { | |
120c348f | 4936 | VLOG_INFO("%s: open failed (%s)", filename, ovs_strerror(errno)); |
7d031d7e BP |
4937 | } else { |
4938 | if (fscanf(stream, "%d", &n) != 1) { | |
4939 | VLOG_WARN("%s: read error", filename); | |
4940 | } else if (n <= 100) { | |
4941 | VLOG_WARN("%s: unexpectedly small value %d", filename, n); | |
4942 | } else { | |
4943 | max_backlog = n; | |
4944 | } | |
4945 | fclose(stream); | |
4946 | } | |
4947 | ovsthread_once_done(&once); | |
4948 | ||
4949 | VLOG_DBG("%s: using %d max_backlog", filename, max_backlog); | |
4950 | } | |
4951 | ||
4952 | return max_backlog; | |
4953 | } | |
4954 | ||
4955 | /* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in | |
4956 | * 'odp_actions'. */ | |
4957 | static int | |
4958 | count_output_actions(const struct ofpbuf *odp_actions) | |
4959 | { | |
4960 | const struct nlattr *a; | |
4961 | size_t left; | |
4962 | int n = 0; | |
4963 | ||
6fd6ed71 | 4964 | NL_ATTR_FOR_EACH_UNSAFE (a, left, odp_actions->data, odp_actions->size) { |
7d031d7e BP |
4965 | if (a->nla_type == OVS_ACTION_ATTR_OUTPUT) { |
4966 | n++; | |
4967 | } | |
4968 | } | |
4969 | return n; | |
4970 | } | |
5e2a6702 | 4971 | #endif /* defined(__linux__) */ |
7d031d7e BP |
4972 | |
4973 | /* Returns true if 'odp_actions' contains more output actions than the datapath | |
4974 | * can reliably handle in one go. On Linux, this is the value of the | |
4975 | * net.core.netdev_max_backlog sysctl, which limits the maximum number of | |
4976 | * packets that the kernel is willing to queue up for processing while the | |
4977 | * datapath is processing a set of actions. */ | |
4978 | static bool | |
5e2a6702 | 4979 | too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) |
7d031d7e BP |
4980 | { |
4981 | #ifdef __linux__ | |
6fd6ed71 | 4982 | return (odp_actions->size / NL_A_U32_SIZE > netdev_max_backlog() |
7d031d7e BP |
4983 | && count_output_actions(odp_actions) > netdev_max_backlog()); |
4984 | #else | |
4985 | /* OSes other than Linux might have similar limits, but we don't know how | |
4986 | * to determine them.*/ | |
4987 | return false; | |
4988 | #endif | |
4989 | } | |
4990 | ||
234c3da9 BP |
4991 | static void |
4992 | xlate_wc_init(struct xlate_ctx *ctx) | |
4993 | { | |
4994 | flow_wildcards_init_catchall(ctx->wc); | |
4995 | ||
4996 | /* Some fields we consider to always be examined. */ | |
5e2e998a JR |
4997 | WC_MASK_FIELD(ctx->wc, in_port); |
4998 | WC_MASK_FIELD(ctx->wc, dl_type); | |
234c3da9 | 4999 | if (is_ip_any(&ctx->xin->flow)) { |
5e2e998a | 5000 | WC_MASK_FIELD_MASK(ctx->wc, nw_frag, FLOW_NW_FRAG_MASK); |
234c3da9 BP |
5001 | } |
5002 | ||
5003 | if (ctx->xbridge->support.odp.recirc) { | |
5004 | /* Always exactly match recirc_id when datapath supports | |
5005 | * recirculation. */ | |
5e2e998a | 5006 | WC_MASK_FIELD(ctx->wc, recirc_id); |
234c3da9 BP |
5007 | } |
5008 | ||
5009 | if (ctx->xbridge->netflow) { | |
5010 | netflow_mask_wc(&ctx->xin->flow, ctx->wc); | |
5011 | } | |
5012 | ||
5013 | tnl_wc_init(&ctx->xin->flow, ctx->wc); | |
5014 | } | |
5015 | ||
5016 | static void | |
5017 | xlate_wc_finish(struct xlate_ctx *ctx) | |
5018 | { | |
5019 | /* Clear the metadata and register wildcard masks, because we won't | |
5020 | * use non-header fields as part of the cache. */ | |
5021 | flow_wildcards_clear_non_packet_fields(ctx->wc); | |
5022 | ||
5023 | /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow | |
5024 | * uses the low 8 bits of the 16-bit tp_src and tp_dst members to | |
5025 | * represent these fields. The datapath interface, on the other hand, | |
5026 | * represents them with just 8 bits each. This means that if the high | |
5027 | * 8 bits of the masks for these fields somehow become set, then they | |
5028 | * will get chopped off by a round trip through the datapath, and | |
5029 | * revalidation will spot that as an inconsistency and delete the flow. | |
5030 | * Avoid the problem here by making sure that only the low 8 bits of | |
5031 | * either field can be unwildcarded for ICMP. | |
5032 | */ | |
5033 | if (is_icmpv4(&ctx->xin->flow) || is_icmpv6(&ctx->xin->flow)) { | |
5034 | ctx->wc->masks.tp_src &= htons(UINT8_MAX); | |
5035 | ctx->wc->masks.tp_dst &= htons(UINT8_MAX); | |
5036 | } | |
5037 | /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */ | |
5038 | if (ctx->wc->masks.vlan_tci) { | |
5039 | ctx->wc->masks.vlan_tci |= htons(VLAN_CFI); | |
5040 | } | |
5041 | } | |
5042 | ||
e672ff9b JR |
5043 | /* Translates the flow, actions, or rule in 'xin' into datapath actions in |
5044 | * 'xout'. | |
56450a41 | 5045 | * The caller must take responsibility for eventually freeing 'xout', with |
fff1b9c0 JR |
5046 | * xlate_out_uninit(). |
5047 | * Returns 'XLATE_OK' if translation was successful. In case of an error an | |
5048 | * empty set of actions will be returned in 'xin->odp_actions' (if non-NULL), | |
5049 | * so that most callers may ignore the return value and transparently install a | |
5050 | * drop flow when the translation fails. */ | |
5051 | enum xlate_error | |
84f0f298 | 5052 | xlate_actions(struct xlate_in *xin, struct xlate_out *xout) |
9583bc14 | 5053 | { |
e467ea42 BP |
5054 | *xout = (struct xlate_out) { |
5055 | .slow = 0, | |
fbf5d6ec | 5056 | .recircs = RECIRC_REFS_EMPTY_INITIALIZER, |
e467ea42 BP |
5057 | }; |
5058 | ||
84f0f298 | 5059 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
bb00fdef BP |
5060 | struct xbridge *xbridge = xbridge_lookup(xcfg, xin->ofproto); |
5061 | if (!xbridge) { | |
fff1b9c0 | 5062 | return XLATE_BRIDGE_NOT_FOUND; |
bb00fdef BP |
5063 | } |
5064 | ||
33bf9176 BP |
5065 | struct flow *flow = &xin->flow; |
5066 | ||
bb00fdef BP |
5067 | union mf_subvalue stack_stub[1024 / sizeof(union mf_subvalue)]; |
5068 | uint64_t action_set_stub[1024 / 8]; | |
1d361a81 | 5069 | uint64_t frozen_actions_stub[1024 / 8]; |
49a73e0c | 5070 | struct flow_wildcards scratch_wc; |
1520ef4f BP |
5071 | uint64_t actions_stub[256 / 8]; |
5072 | struct ofpbuf scratch_actions = OFPBUF_STUB_INITIALIZER(actions_stub); | |
bb00fdef BP |
5073 | struct xlate_ctx ctx = { |
5074 | .xin = xin, | |
5075 | .xout = xout, | |
5076 | .base_flow = *flow, | |
c2b878e0 | 5077 | .orig_tunnel_ipv6_dst = flow_tnl_dst(&flow->tunnel), |
bb00fdef BP |
5078 | .xbridge = xbridge, |
5079 | .stack = OFPBUF_STUB_INITIALIZER(stack_stub), | |
5080 | .rule = xin->rule, | |
49a73e0c | 5081 | .wc = xin->wc ? xin->wc : &scratch_wc, |
1520ef4f | 5082 | .odp_actions = xin->odp_actions ? xin->odp_actions : &scratch_actions, |
bb00fdef | 5083 | |
9a9b001a | 5084 | .indentation = xin->indentation, |
cdd42eda | 5085 | .resubmits = xin->resubmits, |
bb00fdef BP |
5086 | .in_group = false, |
5087 | .in_action_set = false, | |
5088 | ||
5089 | .table_id = 0, | |
5090 | .rule_cookie = OVS_BE64_MAX, | |
5091 | .orig_skb_priority = flow->skb_priority, | |
5092 | .sflow_n_outputs = 0, | |
5093 | .sflow_odp_port = 0, | |
2031ef97 | 5094 | .nf_output_iface = NF_OUT_DROP, |
bb00fdef | 5095 | .exit = false, |
fff1b9c0 | 5096 | .error = XLATE_OK, |
3d6151f3 | 5097 | .mirrors = 0, |
bb00fdef | 5098 | |
1d361a81 BP |
5099 | .freezing = false, |
5100 | .frozen_actions = OFPBUF_STUB_INITIALIZER(frozen_actions_stub), | |
77ab5fd2 | 5101 | .pause = NULL, |
bb00fdef | 5102 | |
07659514 | 5103 | .conntracked = false, |
bb00fdef | 5104 | |
9ac0aada JR |
5105 | .ct_nat_action = NULL, |
5106 | ||
bb00fdef BP |
5107 | .action_set_has_group = false, |
5108 | .action_set = OFPBUF_STUB_INITIALIZER(action_set_stub), | |
5109 | }; | |
865ca6cf BP |
5110 | |
5111 | /* 'base_flow' reflects the packet as it came in, but we need it to reflect | |
42deb67d PS |
5112 | * the packet as the datapath will treat it for output actions. Our |
5113 | * datapath doesn't retain tunneling information without us re-setting | |
5114 | * it, so clear the tunnel data. | |
865ca6cf | 5115 | */ |
42deb67d | 5116 | |
bb00fdef | 5117 | memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); |
865ca6cf | 5118 | |
1520ef4f | 5119 | ofpbuf_reserve(ctx.odp_actions, NL_A_U32_SIZE); |
234c3da9 BP |
5120 | if (xin->wc) { |
5121 | xlate_wc_init(&ctx); | |
5122 | } | |
bb00fdef | 5123 | |
46c88433 | 5124 | COVERAGE_INC(xlate_actions); |
9583bc14 | 5125 | |
1d361a81 BP |
5126 | if (xin->frozen_state) { |
5127 | const struct frozen_state *state = xin->frozen_state; | |
e672ff9b | 5128 | |
1d361a81 | 5129 | xlate_report(&ctx, "Thawing frozen state:"); |
d6bef3cc | 5130 | |
e672ff9b JR |
5131 | if (xin->ofpacts_len > 0 || ctx.rule) { |
5132 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
d6bef3cc | 5133 | const char *conflict = xin->ofpacts_len ? "actions" : "rule"; |
e672ff9b | 5134 | |
d6bef3cc BP |
5135 | VLOG_WARN_RL(&rl, "Recirculation conflict (%s)!", conflict); |
5136 | xlate_report(&ctx, "- Recirculation conflict (%s)!", conflict); | |
fff1b9c0 | 5137 | ctx.error = XLATE_RECIRCULATION_CONFLICT; |
1520ef4f | 5138 | goto exit; |
e672ff9b JR |
5139 | } |
5140 | ||
5141 | /* Set the bridge for post-recirculation processing if needed. */ | |
290835f9 BP |
5142 | if (!uuid_equals(ofproto_dpif_get_uuid(ctx.xbridge->ofproto), |
5143 | &state->ofproto_uuid)) { | |
e672ff9b | 5144 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
2082425c | 5145 | const struct xbridge *new_bridge |
290835f9 | 5146 | = xbridge_lookup_by_uuid(xcfg, &state->ofproto_uuid); |
e672ff9b JR |
5147 | |
5148 | if (OVS_UNLIKELY(!new_bridge)) { | |
5149 | /* Drop the packet if the bridge cannot be found. */ | |
5150 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
1d361a81 BP |
5151 | VLOG_WARN_RL(&rl, "Frozen bridge no longer exists."); |
5152 | xlate_report(&ctx, "- Frozen bridge no longer exists."); | |
fff1b9c0 | 5153 | ctx.error = XLATE_BRIDGE_NOT_FOUND; |
1520ef4f | 5154 | goto exit; |
e672ff9b JR |
5155 | } |
5156 | ctx.xbridge = new_bridge; | |
5157 | } | |
5158 | ||
1d361a81 BP |
5159 | /* Set the thawed table id. Note: A table lookup is done only if there |
5160 | * are no frozen actions. */ | |
2082425c | 5161 | ctx.table_id = state->table_id; |
d6bef3cc | 5162 | xlate_report(&ctx, "- Resuming from table %"PRIu8, ctx.table_id); |
e672ff9b | 5163 | |
07659514 JS |
5164 | if (!state->conntracked) { |
5165 | clear_conntrack(flow); | |
5166 | } | |
5167 | ||
e672ff9b | 5168 | /* Restore pipeline metadata. May change flow's in_port and other |
1d361a81 BP |
5169 | * metadata to the values that existed when freezing was triggered. */ |
5170 | frozen_metadata_to_flow(&state->metadata, flow); | |
e672ff9b JR |
5171 | |
5172 | /* Restore stack, if any. */ | |
2082425c | 5173 | if (state->stack) { |
5c1b2314 BP |
5174 | ofpbuf_put(&ctx.stack, state->stack, |
5175 | state->n_stack * sizeof *state->stack); | |
e672ff9b JR |
5176 | } |
5177 | ||
29bae541 BP |
5178 | /* Restore mirror state. */ |
5179 | ctx.mirrors = state->mirrors; | |
5180 | ||
e672ff9b | 5181 | /* Restore action set, if any. */ |
2082425c | 5182 | if (state->action_set_len) { |
d6bef3cc | 5183 | xlate_report_actions(&ctx, "- Restoring action set", |
417509fa | 5184 | state->action_set, state->action_set_len); |
d6bef3cc | 5185 | |
7e7e8dbb BP |
5186 | flow->actset_output = OFPP_UNSET; |
5187 | xlate_write_actions__(&ctx, state->action_set, | |
5188 | state->action_set_len); | |
e672ff9b JR |
5189 | } |
5190 | ||
1d361a81 BP |
5191 | /* Restore frozen actions. If there are no actions, processing will |
5192 | * start with a lookup in the table set above. */ | |
417509fa BP |
5193 | xin->ofpacts = state->ofpacts; |
5194 | xin->ofpacts_len = state->ofpacts_len; | |
5195 | if (state->ofpacts_len) { | |
d6bef3cc BP |
5196 | xlate_report_actions(&ctx, "- Restoring actions", |
5197 | xin->ofpacts, xin->ofpacts_len); | |
e672ff9b JR |
5198 | } |
5199 | } else if (OVS_UNLIKELY(flow->recirc_id)) { | |
5200 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
5201 | ||
5202 | VLOG_WARN_RL(&rl, "Recirculation context not found for ID %"PRIx32, | |
5203 | flow->recirc_id); | |
fff1b9c0 | 5204 | ctx.error = XLATE_NO_RECIRCULATION_CONTEXT; |
1520ef4f | 5205 | goto exit; |
e672ff9b | 5206 | } |
621b8064 JR |
5207 | /* The bridge is now known so obtain its table version. */ |
5208 | ctx.tables_version = ofproto_dpif_get_tables_version(ctx.xbridge->ofproto); | |
9583bc14 | 5209 | |
10c44245 | 5210 | if (!xin->ofpacts && !ctx.rule) { |
b2e89cc9 BP |
5211 | ctx.rule = rule_dpif_lookup_from_table( |
5212 | ctx.xbridge->ofproto, ctx.tables_version, flow, xin->wc, | |
1e1e1d19 | 5213 | ctx.xin->resubmit_stats, &ctx.table_id, |
b2e89cc9 | 5214 | flow->in_port.ofp_port, true, true); |
10c44245 | 5215 | if (ctx.xin->resubmit_stats) { |
b2e89cc9 | 5216 | rule_dpif_credit_stats(ctx.rule, ctx.xin->resubmit_stats); |
10c44245 | 5217 | } |
b256dc52 JS |
5218 | if (ctx.xin->xcache) { |
5219 | struct xc_entry *entry; | |
5220 | ||
5221 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE); | |
b2e89cc9 | 5222 | entry->u.rule = ctx.rule; |
1e1e1d19 | 5223 | rule_dpif_ref(ctx.rule); |
b256dc52 | 5224 | } |
a8c31348 BP |
5225 | |
5226 | if (OVS_UNLIKELY(ctx.xin->resubmit_hook)) { | |
b2e89cc9 | 5227 | ctx.xin->resubmit_hook(ctx.xin, ctx.rule, 0); |
a8c31348 | 5228 | } |
10c44245 | 5229 | } |
10c44245 | 5230 | |
1d361a81 | 5231 | /* Get the proximate input port of the packet. (If xin->frozen_state, |
14d2b8b2 BP |
5232 | * flow->in_port is the ultimate input port of the packet.) */ |
5233 | struct xport *in_port = get_ofp_port(xbridge, | |
5234 | ctx.base_flow.in_port.ofp_port); | |
5235 | ||
1d361a81 BP |
5236 | /* Tunnel stats only for not-thawed packets. */ |
5237 | if (!xin->frozen_state && in_port && in_port->is_tunnel) { | |
b256dc52 JS |
5238 | if (ctx.xin->resubmit_stats) { |
5239 | netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats); | |
5240 | if (in_port->bfd) { | |
5241 | bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats); | |
5242 | } | |
5243 | } | |
5244 | if (ctx.xin->xcache) { | |
5245 | struct xc_entry *entry; | |
5246 | ||
5247 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV); | |
5248 | entry->u.dev.rx = netdev_ref(in_port->netdev); | |
5249 | entry->u.dev.bfd = bfd_ref(in_port->bfd); | |
d6fc5f57 EJ |
5250 | } |
5251 | } | |
5252 | ||
1d361a81 | 5253 | if (!xin->frozen_state && process_special(&ctx, in_port)) { |
bef1403e BP |
5254 | /* process_special() did all the processing for this packet. |
5255 | * | |
1d361a81 BP |
5256 | * We do not perform special processing on thawed packets, since that |
5257 | * was done before they were frozen and should not be redone. */ | |
bef1403e BP |
5258 | } else if (in_port && in_port->xbundle |
5259 | && xbundle_mirror_out(xbridge, in_port->xbundle)) { | |
5260 | if (ctx.xin->packet != NULL) { | |
5261 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
5262 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
5263 | "%s, which is reserved exclusively for mirroring", | |
5264 | ctx.xbridge->name, in_port->xbundle->name); | |
5265 | } | |
5266 | } else { | |
1d361a81 | 5267 | /* Sampling is done on initial reception; don't redo after thawing. */ |
a6092018 | 5268 | unsigned int user_cookie_offset = 0; |
1d361a81 | 5269 | if (!xin->frozen_state) { |
a6092018 BP |
5270 | user_cookie_offset = compose_sflow_action(&ctx); |
5271 | compose_ipfix_action(&ctx, ODPP_NONE); | |
e672ff9b | 5272 | } |
0731abc5 | 5273 | size_t sample_actions_len = ctx.odp_actions->size; |
9583bc14 | 5274 | |
234c3da9 BP |
5275 | if (tnl_process_ecn(flow) |
5276 | && (!in_port || may_receive(in_port, &ctx))) { | |
1806291d BP |
5277 | const struct ofpact *ofpacts; |
5278 | size_t ofpacts_len; | |
5279 | ||
5280 | if (xin->ofpacts) { | |
5281 | ofpacts = xin->ofpacts; | |
5282 | ofpacts_len = xin->ofpacts_len; | |
5283 | } else if (ctx.rule) { | |
5284 | const struct rule_actions *actions | |
5285 | = rule_dpif_get_actions(ctx.rule); | |
5286 | ofpacts = actions->ofpacts; | |
5287 | ofpacts_len = actions->ofpacts_len; | |
5288 | ctx.rule_cookie = rule_dpif_get_flow_cookie(ctx.rule); | |
5289 | } else { | |
5290 | OVS_NOT_REACHED(); | |
5291 | } | |
5292 | ||
7efbc3b7 | 5293 | mirror_ingress_packet(&ctx); |
9583bc14 | 5294 | do_xlate_actions(ofpacts, ofpacts_len, &ctx); |
fff1b9c0 JR |
5295 | if (ctx.error) { |
5296 | goto exit; | |
5297 | } | |
9583bc14 EJ |
5298 | |
5299 | /* We've let OFPP_NORMAL and the learning action look at the | |
1d361a81 | 5300 | * packet, so cancel all actions and freezing if forwarding is |
8a5fb3b4 | 5301 | * disabled. */ |
9efd308e DV |
5302 | if (in_port && (!xport_stp_forward_state(in_port) || |
5303 | !xport_rstp_forward_state(in_port))) { | |
1520ef4f | 5304 | ctx.odp_actions->size = sample_actions_len; |
1d361a81 | 5305 | ctx_cancel_freeze(&ctx); |
8a5fb3b4 BP |
5306 | ofpbuf_clear(&ctx.action_set); |
5307 | } | |
5308 | ||
1d361a81 | 5309 | if (!ctx.freezing) { |
8a5fb3b4 | 5310 | xlate_action_set(&ctx); |
e672ff9b | 5311 | } |
1d361a81 | 5312 | if (ctx.freezing) { |
77ab5fd2 | 5313 | finish_freezing(&ctx); |
9583bc14 EJ |
5314 | } |
5315 | } | |
5316 | ||
e672ff9b | 5317 | /* Output only fully processed packets. */ |
1d361a81 | 5318 | if (!ctx.freezing |
e672ff9b | 5319 | && xbridge->has_in_band |
ce4a6b76 BP |
5320 | && in_band_must_output_to_local_port(flow) |
5321 | && !actions_output_to_local_port(&ctx)) { | |
e93ef1c7 | 5322 | compose_output_action(&ctx, OFPP_LOCAL, NULL); |
9583bc14 | 5323 | } |
aaa0fbae | 5324 | |
a6092018 BP |
5325 | if (user_cookie_offset) { |
5326 | fix_sflow_action(&ctx, user_cookie_offset); | |
e672ff9b | 5327 | } |
9583bc14 EJ |
5328 | } |
5329 | ||
1520ef4f | 5330 | if (nl_attr_oversized(ctx.odp_actions->size)) { |
542024c4 | 5331 | /* These datapath actions are too big for a Netlink attribute, so we |
0f032e95 BP |
5332 | * can't hand them to the kernel directly. dpif_execute() can execute |
5333 | * them one by one with help, so just mark the result as SLOW_ACTION to | |
5334 | * prevent the flow from being installed. */ | |
5335 | COVERAGE_INC(xlate_actions_oversize); | |
5336 | ctx.xout->slow |= SLOW_ACTION; | |
1520ef4f | 5337 | } else if (too_many_output_actions(ctx.odp_actions)) { |
7d031d7e BP |
5338 | COVERAGE_INC(xlate_actions_too_many_output); |
5339 | ctx.xout->slow |= SLOW_ACTION; | |
542024c4 BP |
5340 | } |
5341 | ||
1d361a81 BP |
5342 | /* Do netflow only for packets on initial reception, that are not sent to |
5343 | * the controller. We consider packets sent to the controller to be part | |
5344 | * of the control plane rather than the data plane. */ | |
5345 | if (!xin->frozen_state | |
5346 | && xbridge->netflow | |
5347 | && !(xout->slow & SLOW_CONTROLLER)) { | |
1806291d BP |
5348 | if (ctx.xin->resubmit_stats) { |
5349 | netflow_flow_update(xbridge->netflow, flow, | |
2031ef97 | 5350 | ctx.nf_output_iface, |
1806291d BP |
5351 | ctx.xin->resubmit_stats); |
5352 | } | |
5353 | if (ctx.xin->xcache) { | |
5354 | struct xc_entry *entry; | |
b256dc52 | 5355 | |
1806291d BP |
5356 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW); |
5357 | entry->u.nf.netflow = netflow_ref(xbridge->netflow); | |
5358 | entry->u.nf.flow = xmemdup(flow, sizeof *flow); | |
2031ef97 | 5359 | entry->u.nf.iface = ctx.nf_output_iface; |
d6fc5f57 EJ |
5360 | } |
5361 | } | |
5362 | ||
49a73e0c | 5363 | if (xin->wc) { |
234c3da9 | 5364 | xlate_wc_finish(&ctx); |
c56fac1b | 5365 | } |
1520ef4f BP |
5366 | |
5367 | exit: | |
5368 | ofpbuf_uninit(&ctx.stack); | |
5369 | ofpbuf_uninit(&ctx.action_set); | |
1d361a81 | 5370 | ofpbuf_uninit(&ctx.frozen_actions); |
1520ef4f | 5371 | ofpbuf_uninit(&scratch_actions); |
fff1b9c0 JR |
5372 | |
5373 | /* Make sure we return a "drop flow" in case of an error. */ | |
5374 | if (ctx.error) { | |
5375 | xout->slow = 0; | |
5376 | if (xin->odp_actions) { | |
5377 | ofpbuf_clear(xin->odp_actions); | |
5378 | } | |
5379 | } | |
5380 | return ctx.error; | |
91d6cd12 AW |
5381 | } |
5382 | ||
77ab5fd2 BP |
5383 | enum ofperr |
5384 | xlate_resume(struct ofproto_dpif *ofproto, | |
5385 | const struct ofputil_packet_in_private *pin, | |
5386 | struct ofpbuf *odp_actions, | |
5387 | enum slow_path_reason *slow) | |
5388 | { | |
5389 | struct dp_packet packet; | |
5390 | dp_packet_use_const(&packet, pin->public.packet, | |
5391 | pin->public.packet_len); | |
5392 | ||
5393 | struct flow flow; | |
5394 | flow_extract(&packet, &flow); | |
5395 | ||
5396 | struct xlate_in xin; | |
5397 | xlate_in_init(&xin, ofproto, &flow, 0, NULL, ntohs(flow.tcp_flags), | |
5398 | &packet, NULL, odp_actions); | |
5399 | ||
5400 | struct ofpact_note noop; | |
5401 | ofpact_init_NOTE(&noop); | |
5402 | noop.length = 0; | |
5403 | ||
5404 | bool any_actions = pin->actions_len > 0; | |
5405 | struct frozen_state state = { | |
5406 | .table_id = 0, /* Not the table where NXAST_PAUSE was executed. */ | |
5407 | .ofproto_uuid = pin->bridge, | |
5408 | .stack = pin->stack, | |
5409 | .n_stack = pin->n_stack, | |
5410 | .mirrors = pin->mirrors, | |
5411 | .conntracked = pin->conntracked, | |
5412 | ||
5413 | /* When there are no actions, xlate_actions() will search the flow | |
5414 | * table. We don't want it to do that (we want it to resume), so | |
5415 | * supply a no-op action if there aren't any. | |
5416 | * | |
5417 | * (We can't necessarily avoid translating actions entirely if there | |
5418 | * aren't any actions, because there might be some finishing-up to do | |
5419 | * at the end of the pipeline, and we don't check for those | |
5420 | * conditions.) */ | |
5421 | .ofpacts = any_actions ? pin->actions : &noop.ofpact, | |
5422 | .ofpacts_len = any_actions ? pin->actions_len : sizeof noop, | |
5423 | ||
5424 | .action_set = pin->action_set, | |
5425 | .action_set_len = pin->action_set_len, | |
5426 | }; | |
5427 | frozen_metadata_from_flow(&state.metadata, | |
5428 | &pin->public.flow_metadata.flow); | |
5429 | xin.frozen_state = &state; | |
5430 | ||
5431 | struct xlate_out xout; | |
5432 | enum xlate_error error = xlate_actions(&xin, &xout); | |
5433 | *slow = xout.slow; | |
5434 | xlate_out_uninit(&xout); | |
5435 | ||
5436 | /* xlate_actions() can generate a number of errors, but only | |
5437 | * XLATE_BRIDGE_NOT_FOUND really stands out to me as one that we should be | |
5438 | * sure to report over OpenFlow. The others could come up in packet-outs | |
5439 | * or regular flow translation and I don't think that it's going to be too | |
5440 | * useful to report them to the controller. */ | |
5441 | return error == XLATE_BRIDGE_NOT_FOUND ? OFPERR_NXR_STALE : 0; | |
5442 | } | |
5443 | ||
91d6cd12 AW |
5444 | /* Sends 'packet' out 'ofport'. |
5445 | * May modify 'packet'. | |
5446 | * Returns 0 if successful, otherwise a positive errno value. */ | |
5447 | int | |
cf62fa4c | 5448 | xlate_send_packet(const struct ofport_dpif *ofport, struct dp_packet *packet) |
91d6cd12 | 5449 | { |
84f0f298 | 5450 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
91d6cd12 | 5451 | struct xport *xport; |
91d6cd12 | 5452 | struct ofpact_output output; |
91d6cd12 | 5453 | struct flow flow; |
91d6cd12 | 5454 | |
91d6cd12 AW |
5455 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); |
5456 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ | |
cf62fa4c | 5457 | flow_extract(packet, &flow); |
b5e7e61a | 5458 | flow.in_port.ofp_port = OFPP_NONE; |
91d6cd12 | 5459 | |
84f0f298 | 5460 | xport = xport_lookup(xcfg, ofport); |
91d6cd12 | 5461 | if (!xport) { |
02ea2703 | 5462 | return EINVAL; |
91d6cd12 | 5463 | } |
91d6cd12 AW |
5464 | output.port = xport->ofp_port; |
5465 | output.max_len = 0; | |
e491a67a YT |
5466 | |
5467 | return ofproto_dpif_execute_actions(xport->xbridge->ofproto, &flow, NULL, | |
5468 | &output.ofpact, sizeof output, | |
5469 | packet); | |
9583bc14 | 5470 | } |
b256dc52 JS |
5471 | |
5472 | struct xlate_cache * | |
5473 | xlate_cache_new(void) | |
5474 | { | |
5475 | struct xlate_cache *xcache = xmalloc(sizeof *xcache); | |
5476 | ||
5477 | ofpbuf_init(&xcache->entries, 512); | |
5478 | return xcache; | |
5479 | } | |
5480 | ||
5481 | static struct xc_entry * | |
5482 | xlate_cache_add_entry(struct xlate_cache *xcache, enum xc_type type) | |
5483 | { | |
5484 | struct xc_entry *entry; | |
5485 | ||
5486 | entry = ofpbuf_put_zeros(&xcache->entries, sizeof *entry); | |
5487 | entry->type = type; | |
5488 | ||
5489 | return entry; | |
5490 | } | |
5491 | ||
5492 | static void | |
5493 | xlate_cache_netdev(struct xc_entry *entry, const struct dpif_flow_stats *stats) | |
5494 | { | |
5495 | if (entry->u.dev.tx) { | |
5496 | netdev_vport_inc_tx(entry->u.dev.tx, stats); | |
5497 | } | |
5498 | if (entry->u.dev.rx) { | |
5499 | netdev_vport_inc_rx(entry->u.dev.rx, stats); | |
5500 | } | |
5501 | if (entry->u.dev.bfd) { | |
5502 | bfd_account_rx(entry->u.dev.bfd, stats); | |
5503 | } | |
5504 | } | |
5505 | ||
5506 | static void | |
5507 | xlate_cache_normal(struct ofproto_dpif *ofproto, struct flow *flow, int vlan) | |
5508 | { | |
84f0f298 | 5509 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
b256dc52 JS |
5510 | struct xbridge *xbridge; |
5511 | struct xbundle *xbundle; | |
5512 | struct flow_wildcards wc; | |
5513 | ||
84f0f298 | 5514 | xbridge = xbridge_lookup(xcfg, ofproto); |
b256dc52 JS |
5515 | if (!xbridge) { |
5516 | return; | |
5517 | } | |
5518 | ||
5519 | xbundle = lookup_input_bundle(xbridge, flow->in_port.ofp_port, false, | |
5520 | NULL); | |
5521 | if (!xbundle) { | |
5522 | return; | |
5523 | } | |
5524 | ||
5525 | update_learning_table(xbridge, flow, &wc, vlan, xbundle); | |
5526 | } | |
5527 | ||
5528 | /* Push stats and perform side effects of flow translation. */ | |
5529 | void | |
0725e747 | 5530 | xlate_push_stats(struct xlate_cache *xcache, |
b256dc52 JS |
5531 | const struct dpif_flow_stats *stats) |
5532 | { | |
5533 | struct xc_entry *entry; | |
5534 | struct ofpbuf entries = xcache->entries; | |
74ff3298 | 5535 | struct eth_addr dmac; |
b256dc52 | 5536 | |
bc388583 BP |
5537 | if (!stats->n_packets) { |
5538 | return; | |
5539 | } | |
5540 | ||
b256dc52 JS |
5541 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { |
5542 | switch (entry->type) { | |
5543 | case XC_RULE: | |
5544 | rule_dpif_credit_stats(entry->u.rule, stats); | |
5545 | break; | |
5546 | case XC_BOND: | |
5547 | bond_account(entry->u.bond.bond, entry->u.bond.flow, | |
5548 | entry->u.bond.vid, stats->n_bytes); | |
5549 | break; | |
5550 | case XC_NETDEV: | |
5551 | xlate_cache_netdev(entry, stats); | |
5552 | break; | |
5553 | case XC_NETFLOW: | |
5554 | netflow_flow_update(entry->u.nf.netflow, entry->u.nf.flow, | |
5555 | entry->u.nf.iface, stats); | |
5556 | break; | |
5557 | case XC_MIRROR: | |
5558 | mirror_update_stats(entry->u.mirror.mbridge, | |
5559 | entry->u.mirror.mirrors, | |
5560 | stats->n_packets, stats->n_bytes); | |
5561 | break; | |
5562 | case XC_LEARN: | |
0725e747 | 5563 | ofproto_dpif_flow_mod(entry->u.learn.ofproto, entry->u.learn.fm); |
b256dc52 JS |
5564 | break; |
5565 | case XC_NORMAL: | |
0725e747 BP |
5566 | xlate_cache_normal(entry->u.normal.ofproto, entry->u.normal.flow, |
5567 | entry->u.normal.vlan); | |
b256dc52 JS |
5568 | break; |
5569 | case XC_FIN_TIMEOUT: | |
5570 | xlate_fin_timeout__(entry->u.fin.rule, stats->tcp_flags, | |
5571 | entry->u.fin.idle, entry->u.fin.hard); | |
5572 | break; | |
1e684d7d RW |
5573 | case XC_GROUP: |
5574 | group_dpif_credit_stats(entry->u.group.group, entry->u.group.bucket, | |
5575 | stats); | |
5576 | break; | |
53902038 TLSC |
5577 | case XC_TNL_NEIGH: |
5578 | /* Lookup neighbor to avoid timeout. */ | |
0b411137 TLSC |
5579 | tnl_neigh_lookup(entry->u.tnl_neigh_cache.br_name, |
5580 | &entry->u.tnl_neigh_cache.d_ipv6, &dmac); | |
a36de779 | 5581 | break; |
b256dc52 JS |
5582 | default: |
5583 | OVS_NOT_REACHED(); | |
5584 | } | |
5585 | } | |
5586 | } | |
5587 | ||
5588 | static void | |
5589 | xlate_dev_unref(struct xc_entry *entry) | |
5590 | { | |
5591 | if (entry->u.dev.tx) { | |
5592 | netdev_close(entry->u.dev.tx); | |
5593 | } | |
5594 | if (entry->u.dev.rx) { | |
5595 | netdev_close(entry->u.dev.rx); | |
5596 | } | |
5597 | if (entry->u.dev.bfd) { | |
5598 | bfd_unref(entry->u.dev.bfd); | |
5599 | } | |
5600 | } | |
5601 | ||
5602 | static void | |
5603 | xlate_cache_clear_netflow(struct netflow *netflow, struct flow *flow) | |
5604 | { | |
b256dc52 JS |
5605 | netflow_flow_clear(netflow, flow); |
5606 | netflow_unref(netflow); | |
5607 | free(flow); | |
5608 | } | |
5609 | ||
5610 | void | |
5611 | xlate_cache_clear(struct xlate_cache *xcache) | |
5612 | { | |
5613 | struct xc_entry *entry; | |
5614 | struct ofpbuf entries; | |
5615 | ||
5616 | if (!xcache) { | |
5617 | return; | |
5618 | } | |
5619 | ||
5620 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { | |
5621 | switch (entry->type) { | |
5622 | case XC_RULE: | |
5623 | rule_dpif_unref(entry->u.rule); | |
5624 | break; | |
5625 | case XC_BOND: | |
5626 | free(entry->u.bond.flow); | |
5627 | bond_unref(entry->u.bond.bond); | |
5628 | break; | |
5629 | case XC_NETDEV: | |
5630 | xlate_dev_unref(entry); | |
5631 | break; | |
5632 | case XC_NETFLOW: | |
5633 | xlate_cache_clear_netflow(entry->u.nf.netflow, entry->u.nf.flow); | |
5634 | break; | |
5635 | case XC_MIRROR: | |
5636 | mbridge_unref(entry->u.mirror.mbridge); | |
5637 | break; | |
5638 | case XC_LEARN: | |
4165b5e0 JS |
5639 | free(entry->u.learn.fm); |
5640 | ofpbuf_delete(entry->u.learn.ofpacts); | |
b256dc52 JS |
5641 | break; |
5642 | case XC_NORMAL: | |
5643 | free(entry->u.normal.flow); | |
5644 | break; | |
5645 | case XC_FIN_TIMEOUT: | |
83709dfa JR |
5646 | /* 'u.fin.rule' is always already held as a XC_RULE, which |
5647 | * has already released it's reference above. */ | |
b256dc52 | 5648 | break; |
1e684d7d RW |
5649 | case XC_GROUP: |
5650 | group_dpif_unref(entry->u.group.group); | |
5651 | break; | |
53902038 | 5652 | case XC_TNL_NEIGH: |
a36de779 | 5653 | break; |
b256dc52 JS |
5654 | default: |
5655 | OVS_NOT_REACHED(); | |
5656 | } | |
5657 | } | |
5658 | ||
5659 | ofpbuf_clear(&xcache->entries); | |
5660 | } | |
5661 | ||
5662 | void | |
5663 | xlate_cache_delete(struct xlate_cache *xcache) | |
5664 | { | |
5665 | xlate_cache_clear(xcache); | |
5666 | ofpbuf_uninit(&xcache->entries); | |
5667 | free(xcache); | |
5668 | } |