]>
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 | 69 | /* Maximum depth of flow table recursion (due to resubmit actions) in a |
790c5d26 BP |
70 | * flow translation. |
71 | * | |
72 | * The goal of limiting the depth of resubmits is to ensure that flow | |
73 | * translation eventually terminates. Only resubmits to the same table or an | |
74 | * earlier table count against the maximum depth. This is because resubmits to | |
75 | * strictly monotonically increasing table IDs will eventually terminate, since | |
76 | * any OpenFlow switch has a finite number of tables. OpenFlow tables are most | |
77 | * commonly traversed in numerically increasing order, so this limit has little | |
78 | * effect on conventionally designed OpenFlow pipelines. | |
79 | * | |
80 | * Outputs to patch ports and to groups also count against the depth limit. */ | |
81 | #define MAX_DEPTH 64 | |
8a553e9a | 82 | |
98b07853 BP |
83 | /* Maximum number of resubmit actions in a flow translation, whether they are |
84 | * recursive or not. */ | |
790c5d26 | 85 | #define MAX_RESUBMITS (MAX_DEPTH * MAX_DEPTH) |
98b07853 | 86 | |
46c88433 EJ |
87 | struct xbridge { |
88 | struct hmap_node hmap_node; /* Node in global 'xbridges' map. */ | |
89 | struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */ | |
90 | ||
ca6ba700 | 91 | struct ovs_list xbundles; /* Owned xbundles. */ |
46c88433 EJ |
92 | struct hmap xports; /* Indexed by ofp_port. */ |
93 | ||
94 | char *name; /* Name used in log messages. */ | |
89a8a7f0 | 95 | struct dpif *dpif; /* Datapath interface. */ |
46c88433 | 96 | struct mac_learning *ml; /* Mac learning handle. */ |
6d95c4e8 | 97 | struct mcast_snooping *ms; /* Multicast Snooping handle. */ |
46c88433 EJ |
98 | struct mbridge *mbridge; /* Mirroring. */ |
99 | struct dpif_sflow *sflow; /* SFlow handle, or null. */ | |
100 | struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */ | |
ce3955be | 101 | struct netflow *netflow; /* Netflow handle, or null. */ |
9d189a50 | 102 | struct stp *stp; /* STP or null if disabled. */ |
9efd308e | 103 | struct rstp *rstp; /* RSTP or null if disabled. */ |
46c88433 | 104 | |
46c88433 EJ |
105 | bool has_in_band; /* Bridge has in band control? */ |
106 | bool forward_bpdu; /* Bridge forwards STP BPDUs? */ | |
4b97b70d | 107 | |
b440dd8c JS |
108 | /* Datapath feature support. */ |
109 | struct dpif_backer_support support; | |
46c88433 EJ |
110 | }; |
111 | ||
112 | struct xbundle { | |
113 | struct hmap_node hmap_node; /* In global 'xbundles' map. */ | |
114 | struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */ | |
115 | ||
ca6ba700 | 116 | struct ovs_list list_node; /* In parent 'xbridges' list. */ |
46c88433 EJ |
117 | struct xbridge *xbridge; /* Parent xbridge. */ |
118 | ||
ca6ba700 | 119 | struct ovs_list xports; /* Contains "struct xport"s. */ |
46c88433 EJ |
120 | |
121 | char *name; /* Name used in log messages. */ | |
122 | struct bond *bond; /* Nonnull iff more than one port. */ | |
123 | struct lacp *lacp; /* LACP handle or null. */ | |
124 | ||
125 | enum port_vlan_mode vlan_mode; /* VLAN mode. */ | |
126 | int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */ | |
127 | unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1. | |
128 | * NULL if all VLANs are trunked. */ | |
129 | bool use_priority_tags; /* Use 802.1p tag for frames in VLAN 0? */ | |
130 | bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */ | |
131 | }; | |
132 | ||
133 | struct xport { | |
134 | struct hmap_node hmap_node; /* Node in global 'xports' map. */ | |
135 | struct ofport_dpif *ofport; /* Key in global 'xports map. */ | |
136 | ||
137 | struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */ | |
138 | ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */ | |
139 | ||
140 | odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */ | |
141 | ||
ca6ba700 | 142 | struct ovs_list bundle_node; /* In parent xbundle (if it exists). */ |
46c88433 EJ |
143 | struct xbundle *xbundle; /* Parent xbundle or null. */ |
144 | ||
145 | struct netdev *netdev; /* 'ofport''s netdev. */ | |
146 | ||
147 | struct xbridge *xbridge; /* Parent bridge. */ | |
148 | struct xport *peer; /* Patch port peer or null. */ | |
149 | ||
150 | enum ofputil_port_config config; /* OpenFlow port configuration. */ | |
dd8cd4b4 | 151 | enum ofputil_port_state state; /* OpenFlow port state. */ |
92cf817b | 152 | int stp_port_no; /* STP port number or -1 if not in use. */ |
f025bcb7 | 153 | struct rstp_port *rstp_port; /* RSTP port or null. */ |
46c88433 | 154 | |
55954f6e EJ |
155 | struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */ |
156 | ||
46c88433 EJ |
157 | bool may_enable; /* May be enabled in bonds. */ |
158 | bool is_tunnel; /* Is a tunnel port. */ | |
159 | ||
160 | struct cfm *cfm; /* CFM handle or null. */ | |
161 | struct bfd *bfd; /* BFD handle or null. */ | |
0477baa9 | 162 | struct lldp *lldp; /* LLDP handle or null. */ |
46c88433 EJ |
163 | }; |
164 | ||
4d0acc70 EJ |
165 | struct xlate_ctx { |
166 | struct xlate_in *xin; | |
167 | struct xlate_out *xout; | |
168 | ||
46c88433 | 169 | const struct xbridge *xbridge; |
4d0acc70 | 170 | |
621b8064 | 171 | /* Flow tables version at the beginning of the translation. */ |
18721c4a | 172 | cls_version_t tables_version; |
621b8064 | 173 | |
4d0acc70 EJ |
174 | /* Flow at the last commit. */ |
175 | struct flow base_flow; | |
176 | ||
177 | /* Tunnel IP destination address as received. This is stored separately | |
178 | * as the base_flow.tunnel is cleared on init to reflect the datapath | |
179 | * behavior. Used to make sure not to send tunneled output to ourselves, | |
180 | * which might lead to an infinite loop. This could happen easily | |
181 | * if a tunnel is marked as 'ip_remote=flow', and the flow does not | |
182 | * actually set the tun_dst field. */ | |
e4d3706c | 183 | struct in6_addr orig_tunnel_ipv6_dst; |
4d0acc70 EJ |
184 | |
185 | /* Stack for the push and pop actions. Each stack element is of type | |
186 | * "union mf_subvalue". */ | |
4d0acc70 EJ |
187 | struct ofpbuf stack; |
188 | ||
189 | /* The rule that we are currently translating, or NULL. */ | |
190 | struct rule_dpif *rule; | |
191 | ||
49a73e0c BP |
192 | /* Flow translation populates this with wildcards relevant in translation. |
193 | * When 'xin->wc' is nonnull, this is the same pointer. When 'xin->wc' is | |
c0e638aa | 194 | * null, this is a pointer to a temporary buffer. */ |
49a73e0c BP |
195 | struct flow_wildcards *wc; |
196 | ||
1520ef4f BP |
197 | /* Output buffer for datapath actions. When 'xin->odp_actions' is nonnull, |
198 | * this is the same pointer. When 'xin->odp_actions' is null, this points | |
199 | * to a scratch ofpbuf. This allows code to add actions to | |
200 | * 'ctx->odp_actions' without worrying about whether the caller really | |
201 | * wants actions. */ | |
202 | struct ofpbuf *odp_actions; | |
203 | ||
790c5d26 BP |
204 | /* Statistics maintained by xlate_table_action(). |
205 | * | |
206 | * 'indentation' is the nesting level for resubmits. It is used to indent | |
207 | * the output of resubmit_hook (e.g. for the "ofproto/trace" feature). | |
208 | * | |
209 | * The other statistics limit the amount of work that a single flow | |
210 | * translation can perform. The goal of the first of these, 'depth', is | |
211 | * primarily to prevent translation from performing an infinite amount of | |
212 | * work. It counts the current depth of nested "resubmit"s (and a few | |
213 | * other activities); when a resubmit returns, it decreases. Resubmits to | |
214 | * tables in strictly monotonically increasing order don't contribute to | |
215 | * 'depth' because they cannot cause a flow translation to take an infinite | |
216 | * amount of time (because the number of tables is finite). Translation | |
217 | * aborts when 'depth' exceeds MAX_DEPTH. | |
218 | * | |
219 | * 'resubmits', on the other hand, prevents flow translation from | |
220 | * performing an extraordinarily large while still finite amount of work. | |
221 | * It counts the total number of resubmits (and a few other activities) | |
222 | * that have been executed. Returning from a resubmit does not affect this | |
223 | * counter. Thus, this limits the amount of work that a particular | |
224 | * translation can perform. Translation aborts when 'resubmits' exceeds | |
225 | * MAX_RESUBMITS (which is much larger than MAX_DEPTH). | |
226 | */ | |
227 | int indentation; /* Indentation level for resubmit_hook. */ | |
228 | int depth; /* Current resubmit nesting depth. */ | |
98b07853 | 229 | int resubmits; /* Total number of resubmits. */ |
5a070238 | 230 | bool in_group; /* Currently translating ofgroup, if true. */ |
029ca940 | 231 | bool in_action_set; /* Currently translating action_set, if true. */ |
98b07853 | 232 | |
4d0acc70 | 233 | uint8_t table_id; /* OpenFlow table ID where flow was found. */ |
8b1e5560 JR |
234 | ovs_be64 rule_cookie; /* Cookie of the rule being translated. */ |
235 | uint32_t orig_skb_priority; /* Priority when packet arrived. */ | |
4d0acc70 | 236 | uint32_t sflow_n_outputs; /* Number of output ports. */ |
4e022ec0 | 237 | odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */ |
2031ef97 | 238 | ofp_port_t nf_output_iface; /* Output interface index for NetFlow. */ |
4d0acc70 | 239 | bool exit; /* No further actions should be processed. */ |
3d6151f3 | 240 | mirror_mask_t mirrors; /* Bitmap of associated mirrors. */ |
7fdb60a7 | 241 | |
1d361a81 BP |
242 | /* Freezing Translation |
243 | * ==================== | |
e672ff9b | 244 | * |
1d361a81 BP |
245 | * At some point during translation, the code may recognize the need to halt |
246 | * and checkpoint the translation in a way that it can be restarted again | |
247 | * later. We call the checkpointing process "freezing" and the restarting | |
248 | * process "thawing". | |
e672ff9b | 249 | * |
1d361a81 | 250 | * The use cases for freezing are: |
e672ff9b | 251 | * |
1d361a81 BP |
252 | * - "Recirculation", where the translation process discovers that it |
253 | * doesn't have enough information to complete translation without | |
254 | * actually executing the actions that have already been translated, | |
255 | * which provides the additionally needed information. In these | |
256 | * situations, translation freezes translation and assigns the frozen | |
257 | * data a unique "recirculation ID", which it associates with the data | |
258 | * in a table in userspace (see ofproto-dpif-rid.h). It also adds a | |
259 | * OVS_ACTION_ATTR_RECIRC action specifying that ID to the datapath | |
260 | * actions. When a packet hits that action, the datapath looks its | |
261 | * flow up again using the ID. If there's a miss, it comes back to | |
262 | * userspace, which find the recirculation table entry for the ID, | |
263 | * thaws the associated frozen data, and continues translation from | |
264 | * that point given the additional information that is now known. | |
e672ff9b | 265 | * |
1d361a81 BP |
266 | * The archetypal example is MPLS. As MPLS is implemented in |
267 | * OpenFlow, the protocol that follows the last MPLS label becomes | |
268 | * known only when that label is popped by an OpenFlow action. That | |
269 | * means that Open vSwitch can't extract the headers beyond the MPLS | |
270 | * labels until the pop action is executed. Thus, at that point | |
271 | * translation uses the recirculation process to extract the headers | |
272 | * beyond the MPLS labels. | |
e672ff9b | 273 | * |
1d361a81 BP |
274 | * (OVS also uses OVS_ACTION_ATTR_RECIRC to implement hashing for |
275 | * output to bonds. OVS pre-populates all the datapath flows for bond | |
276 | * output in the datapath, though, which means that the elaborate | |
277 | * process of coming back to userspace for a second round of | |
278 | * translation isn't needed, and so bonds don't follow the above | |
279 | * process.) | |
e672ff9b | 280 | * |
77ab5fd2 BP |
281 | * - "Continuation". A continuation is a way for an OpenFlow controller |
282 | * to interpose on a packet's traversal of the OpenFlow tables. When | |
283 | * the translation process encounters a "controller" action with the | |
284 | * "pause" flag, it freezes translation, serializes the frozen data, | |
285 | * and sends it to an OpenFlow controller. The controller then | |
286 | * examines and possibly modifies the frozen data and eventually sends | |
287 | * it back to the switch, which thaws it and continues translation. | |
e672ff9b | 288 | * |
1d361a81 BP |
289 | * The main problem of freezing translation is preserving state, so that |
290 | * when the translation is thawed later it resumes from where it left off, | |
291 | * without disruption. In particular, actions must be preserved as follows: | |
292 | * | |
293 | * - If we're freezing because an action needed more information, the | |
294 | * action that prompted it. | |
295 | * | |
296 | * - Any actions remaining to be translated within the current flow. | |
297 | * | |
298 | * - If translation was frozen within a NXAST_RESUBMIT, then any actions | |
299 | * following the resubmit action. Resubmit actions can be nested, so | |
300 | * this has to go all the way up the control stack. | |
e672ff9b JR |
301 | * |
302 | * - The OpenFlow 1.1+ action set. | |
303 | * | |
304 | * State that actions and flow table lookups can depend on, such as the | |
305 | * following, must also be preserved: | |
306 | * | |
307 | * - Metadata fields (input port, registers, OF1.1+ metadata, ...). | |
308 | * | |
1d361a81 | 309 | * - The stack used by NXAST_STACK_PUSH and NXAST_STACK_POP actions. |
e672ff9b JR |
310 | * |
311 | * - The table ID and cookie of the flow being translated at each level | |
1d361a81 BP |
312 | * of the control stack, because these can become visible through |
313 | * OFPAT_CONTROLLER actions (and other ways). | |
e672ff9b JR |
314 | * |
315 | * Translation allows for the control of this state preservation via these | |
1d361a81 BP |
316 | * members. When a need to freeze translation is identified, the |
317 | * translation process: | |
e672ff9b | 318 | * |
1d361a81 | 319 | * 1. Sets 'freezing' to true. |
e672ff9b JR |
320 | * |
321 | * 2. Sets 'exit' to true to tell later steps that we're exiting from the | |
322 | * translation process. | |
323 | * | |
1d361a81 BP |
324 | * 3. Adds an OFPACT_UNROLL_XLATE action to 'frozen_actions', and points |
325 | * frozen_actions.header to the action to make it easy to find it later. | |
326 | * This action holds the current table ID and cookie so that they can be | |
327 | * restored during a post-recirculation upcall translation. | |
e672ff9b JR |
328 | * |
329 | * 4. Adds the action that prompted recirculation and any actions following | |
1d361a81 | 330 | * it within the same flow to 'frozen_actions', so that they can be |
8a5fb3b4 | 331 | * executed during a post-recirculation upcall translation. |
e672ff9b JR |
332 | * |
333 | * 5. Returns. | |
334 | * | |
335 | * 6. The action that prompted recirculation might be nested in a stack of | |
336 | * nested "resubmit"s that have actions remaining. Each of these notices | |
1d361a81 BP |
337 | * that we're exiting and freezing and responds by adding more |
338 | * OFPACT_UNROLL_XLATE actions to 'frozen_actions', as necessary, | |
339 | * followed by any actions that were yet unprocessed. | |
e672ff9b | 340 | * |
1d361a81 BP |
341 | * If we're freezing because of recirculation, the caller generates a |
342 | * recirculation ID and associates all the state produced by this process | |
343 | * with it. For post-recirculation upcall translation, the caller passes it | |
344 | * back in for the new translation to execute. The process yielded a set of | |
345 | * ofpacts that can be translated directly, so it is not much of a special | |
346 | * case at that point. | |
e672ff9b | 347 | */ |
1d361a81 BP |
348 | bool freezing; |
349 | struct ofpbuf frozen_actions; | |
77ab5fd2 | 350 | const struct ofpact_controller *pause; |
e672ff9b | 351 | |
e12ec36b SH |
352 | /* True if a packet was but is no longer MPLS (due to an MPLS pop action). |
353 | * This is a trigger for recirculation in cases where translating an action | |
354 | * or looking up a flow requires access to the fields of the packet after | |
355 | * the MPLS label stack that was originally present. */ | |
356 | bool was_mpls; | |
357 | ||
07659514 JS |
358 | /* True if conntrack has been performed on this packet during processing |
359 | * on the current bridge. This is used to determine whether conntrack | |
1d361a81 | 360 | * state from the datapath should be honored after thawing. */ |
07659514 JS |
361 | bool conntracked; |
362 | ||
9ac0aada JR |
363 | /* Pointer to an embedded NAT action in a conntrack action, or NULL. */ |
364 | struct ofpact_nat *ct_nat_action; | |
365 | ||
7fdb60a7 SH |
366 | /* OpenFlow 1.1+ action set. |
367 | * | |
368 | * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS. | |
369 | * When translation is otherwise complete, ofpacts_execute_action_set() | |
370 | * converts it to a set of "struct ofpact"s that can be translated into | |
ed9c9e3e | 371 | * datapath actions. */ |
c61f3870 | 372 | bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ |
7fdb60a7 | 373 | struct ofpbuf action_set; /* Action set. */ |
fff1b9c0 JR |
374 | |
375 | enum xlate_error error; /* Translation failed. */ | |
4d0acc70 EJ |
376 | }; |
377 | ||
fff1b9c0 JR |
378 | const char *xlate_strerror(enum xlate_error error) |
379 | { | |
380 | switch (error) { | |
381 | case XLATE_OK: | |
382 | return "OK"; | |
383 | case XLATE_BRIDGE_NOT_FOUND: | |
384 | return "Bridge not found"; | |
385 | case XLATE_RECURSION_TOO_DEEP: | |
386 | return "Recursion too deep"; | |
387 | case XLATE_TOO_MANY_RESUBMITS: | |
388 | return "Too many resubmits"; | |
389 | case XLATE_STACK_TOO_DEEP: | |
390 | return "Stack too deep"; | |
391 | case XLATE_NO_RECIRCULATION_CONTEXT: | |
392 | return "No recirculation context"; | |
393 | case XLATE_RECIRCULATION_CONFLICT: | |
394 | return "Recirculation conflict"; | |
395 | case XLATE_TOO_MANY_MPLS_LABELS: | |
396 | return "Too many MPLS labels"; | |
397 | } | |
398 | return "Unknown error"; | |
399 | } | |
400 | ||
ed9c9e3e | 401 | static void xlate_action_set(struct xlate_ctx *ctx); |
704bb0bf | 402 | static void xlate_commit_actions(struct xlate_ctx *ctx); |
ed9c9e3e | 403 | |
1d741d6d | 404 | static void |
1d361a81 | 405 | ctx_trigger_freeze(struct xlate_ctx *ctx) |
1d741d6d JR |
406 | { |
407 | ctx->exit = true; | |
1d361a81 | 408 | ctx->freezing = true; |
1d741d6d JR |
409 | } |
410 | ||
411 | static bool | |
1d361a81 | 412 | ctx_first_frozen_action(const struct xlate_ctx *ctx) |
1d741d6d | 413 | { |
1d361a81 | 414 | return !ctx->frozen_actions.size; |
e672ff9b JR |
415 | } |
416 | ||
3293cb85 | 417 | static void |
1d361a81 | 418 | ctx_cancel_freeze(struct xlate_ctx *ctx) |
3293cb85 | 419 | { |
1d361a81 BP |
420 | if (ctx->freezing) { |
421 | ctx->freezing = false; | |
422 | ofpbuf_clear(&ctx->frozen_actions); | |
423 | ctx->frozen_actions.header = NULL; | |
3293cb85 BP |
424 | } |
425 | } | |
426 | ||
77ab5fd2 | 427 | static void finish_freezing(struct xlate_ctx *ctx); |
e672ff9b | 428 | |
9583bc14 EJ |
429 | /* A controller may use OFPP_NONE as the ingress port to indicate that |
430 | * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for | |
431 | * when an input bundle is needed for validation (e.g., mirroring or | |
432 | * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have | |
3548d242 BP |
433 | * any 'port' structs, so care must be taken when dealing with it. */ |
434 | static struct xbundle ofpp_none_bundle = { | |
435 | .name = "OFPP_NONE", | |
436 | .vlan_mode = PORT_VLAN_TRUNK | |
437 | }; | |
9583bc14 | 438 | |
55954f6e EJ |
439 | /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from |
440 | * 'priority' (the datapath's term for QoS queue) to the dscp bits which all | |
441 | * traffic egressing the 'ofport' with that priority should be marked with. */ | |
442 | struct skb_priority_to_dscp { | |
443 | struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */ | |
444 | uint32_t skb_priority; /* Priority of this queue (see struct flow). */ | |
445 | ||
446 | uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */ | |
447 | }; | |
448 | ||
b256dc52 JS |
449 | enum xc_type { |
450 | XC_RULE, | |
451 | XC_BOND, | |
452 | XC_NETDEV, | |
453 | XC_NETFLOW, | |
454 | XC_MIRROR, | |
455 | XC_LEARN, | |
456 | XC_NORMAL, | |
457 | XC_FIN_TIMEOUT, | |
1e684d7d | 458 | XC_GROUP, |
53902038 | 459 | XC_TNL_NEIGH, |
b256dc52 JS |
460 | }; |
461 | ||
462 | /* xlate_cache entries hold enough information to perform the side effects of | |
463 | * xlate_actions() for a rule, without needing to perform rule translation | |
464 | * from scratch. The primary usage of these is to submit statistics to objects | |
465 | * that a flow relates to, although they may be used for other effects as well | |
466 | * (for instance, refreshing hard timeouts for learned flows). */ | |
467 | struct xc_entry { | |
468 | enum xc_type type; | |
469 | union { | |
470 | struct rule_dpif *rule; | |
471 | struct { | |
472 | struct netdev *tx; | |
473 | struct netdev *rx; | |
474 | struct bfd *bfd; | |
475 | } dev; | |
476 | struct { | |
477 | struct netflow *netflow; | |
478 | struct flow *flow; | |
479 | ofp_port_t iface; | |
480 | } nf; | |
481 | struct { | |
482 | struct mbridge *mbridge; | |
483 | mirror_mask_t mirrors; | |
484 | } mirror; | |
485 | struct { | |
486 | struct bond *bond; | |
487 | struct flow *flow; | |
488 | uint16_t vid; | |
489 | } bond; | |
490 | struct { | |
4165b5e0 JS |
491 | struct ofproto_dpif *ofproto; |
492 | struct ofputil_flow_mod *fm; | |
493 | struct ofpbuf *ofpacts; | |
b256dc52 JS |
494 | } learn; |
495 | struct { | |
496 | struct ofproto_dpif *ofproto; | |
497 | struct flow *flow; | |
498 | int vlan; | |
499 | } normal; | |
500 | struct { | |
501 | struct rule_dpif *rule; | |
502 | uint16_t idle; | |
503 | uint16_t hard; | |
504 | } fin; | |
1e684d7d RW |
505 | struct { |
506 | struct group_dpif *group; | |
507 | struct ofputil_bucket *bucket; | |
508 | } group; | |
a36de779 PS |
509 | struct { |
510 | char br_name[IFNAMSIZ]; | |
0b411137 | 511 | struct in6_addr d_ipv6; |
53902038 | 512 | } tnl_neigh_cache; |
b256dc52 JS |
513 | } u; |
514 | }; | |
515 | ||
6c38bdc0 HH |
516 | #define XC_ENTRY_FOR_EACH(ENTRY, ENTRIES, XCACHE) \ |
517 | ENTRIES = XCACHE->entries; \ | |
518 | for (ENTRY = ofpbuf_try_pull(&ENTRIES, sizeof *ENTRY); \ | |
519 | ENTRY; \ | |
520 | ENTRY = ofpbuf_try_pull(&ENTRIES, sizeof *ENTRY)) | |
b256dc52 JS |
521 | |
522 | struct xlate_cache { | |
523 | struct ofpbuf entries; | |
524 | }; | |
525 | ||
84f0f298 RW |
526 | /* Xlate config contains hash maps of all bridges, bundles and ports. |
527 | * Xcfgp contains the pointer to the current xlate configuration. | |
528 | * When the main thread needs to change the configuration, it copies xcfgp to | |
529 | * new_xcfg and edits new_xcfg. This enables the use of RCU locking which | |
530 | * does not block handler and revalidator threads. */ | |
531 | struct xlate_cfg { | |
532 | struct hmap xbridges; | |
533 | struct hmap xbundles; | |
534 | struct hmap xports; | |
535 | }; | |
b1b72f2d | 536 | static OVSRCU_TYPE(struct xlate_cfg *) xcfgp = OVSRCU_INITIALIZER(NULL); |
f439f23b | 537 | static struct xlate_cfg *new_xcfg = NULL; |
46c88433 EJ |
538 | |
539 | static bool may_receive(const struct xport *, struct xlate_ctx *); | |
9583bc14 EJ |
540 | static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
541 | struct xlate_ctx *); | |
adcf00ba | 542 | static void xlate_normal(struct xlate_ctx *); |
c1b3756c BP |
543 | static inline void xlate_report(struct xlate_ctx *, const char *, ...) |
544 | OVS_PRINTF_FORMAT(2, 3); | |
6d328fa2 SH |
545 | static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port, |
546 | uint8_t table_id, bool may_packet_in, | |
547 | bool honor_table_miss); | |
46c88433 EJ |
548 | static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn); |
549 | static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid); | |
550 | static void output_normal(struct xlate_ctx *, const struct xbundle *, | |
9583bc14 | 551 | uint16_t vlan); |
e93ef1c7 JR |
552 | |
553 | /* Optional bond recirculation parameter to compose_output_action(). */ | |
554 | struct xlate_bond_recirc { | |
555 | uint32_t recirc_id; /* !0 Use recirculation instead of output. */ | |
556 | uint8_t hash_alg; /* !0 Compute hash for recirc before. */ | |
557 | uint32_t hash_basis; /* Compute hash for recirc before. */ | |
558 | }; | |
559 | ||
560 | static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port, | |
561 | const struct xlate_bond_recirc *xr); | |
9583bc14 | 562 | |
84f0f298 RW |
563 | static struct xbridge *xbridge_lookup(struct xlate_cfg *, |
564 | const struct ofproto_dpif *); | |
290835f9 BP |
565 | static struct xbridge *xbridge_lookup_by_uuid(struct xlate_cfg *, |
566 | const struct uuid *); | |
84f0f298 RW |
567 | static struct xbundle *xbundle_lookup(struct xlate_cfg *, |
568 | const struct ofbundle *); | |
569 | static struct xport *xport_lookup(struct xlate_cfg *, | |
570 | const struct ofport_dpif *); | |
46c88433 | 571 | static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port); |
55954f6e EJ |
572 | static struct skb_priority_to_dscp *get_skb_priority(const struct xport *, |
573 | uint32_t skb_priority); | |
574 | static void clear_skb_priorities(struct xport *); | |
16194afd | 575 | static size_t count_skb_priorities(const struct xport *); |
55954f6e EJ |
576 | static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, |
577 | uint8_t *dscp); | |
46c88433 | 578 | |
b256dc52 JS |
579 | static struct xc_entry *xlate_cache_add_entry(struct xlate_cache *xc, |
580 | enum xc_type type); | |
84f0f298 RW |
581 | static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *); |
582 | static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *); | |
583 | static void xlate_xport_init(struct xlate_cfg *, struct xport *); | |
9efd308e | 584 | static void xlate_xbridge_set(struct xbridge *, struct dpif *, |
9efd308e DV |
585 | const struct mac_learning *, struct stp *, |
586 | struct rstp *, const struct mcast_snooping *, | |
587 | const struct mbridge *, | |
588 | const struct dpif_sflow *, | |
589 | const struct dpif_ipfix *, | |
2f47cdf4 | 590 | const struct netflow *, |
84f0f298 | 591 | bool forward_bpdu, bool has_in_band, |
b440dd8c | 592 | const struct dpif_backer_support *); |
84f0f298 RW |
593 | static void xlate_xbundle_set(struct xbundle *xbundle, |
594 | enum port_vlan_mode vlan_mode, int vlan, | |
595 | unsigned long *trunks, bool use_priority_tags, | |
596 | const struct bond *bond, const struct lacp *lacp, | |
597 | bool floodable); | |
598 | static void xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
599 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 DF |
600 | const struct bfd *bfd, const struct lldp *lldp, |
601 | int stp_port_no, const struct rstp_port *rstp_port, | |
84f0f298 RW |
602 | enum ofputil_port_config config, |
603 | enum ofputil_port_state state, bool is_tunnel, | |
604 | bool may_enable); | |
605 | static void xlate_xbridge_remove(struct xlate_cfg *, struct xbridge *); | |
606 | static void xlate_xbundle_remove(struct xlate_cfg *, struct xbundle *); | |
607 | static void xlate_xport_remove(struct xlate_cfg *, struct xport *); | |
608 | static void xlate_xbridge_copy(struct xbridge *); | |
609 | static void xlate_xbundle_copy(struct xbridge *, struct xbundle *); | |
610 | static void xlate_xport_copy(struct xbridge *, struct xbundle *, | |
611 | struct xport *); | |
612 | static void xlate_xcfg_free(struct xlate_cfg *); | |
b256dc52 | 613 | |
34dd0d78 | 614 | static inline void |
c1b3756c | 615 | xlate_report(struct xlate_ctx *ctx, const char *format, ...) |
34dd0d78 JR |
616 | { |
617 | if (OVS_UNLIKELY(ctx->xin->report_hook)) { | |
c1b3756c BP |
618 | va_list args; |
619 | ||
620 | va_start(args, format); | |
9a9b001a | 621 | ctx->xin->report_hook(ctx->xin, ctx->indentation, format, args); |
c1b3756c | 622 | va_end(args); |
34dd0d78 JR |
623 | } |
624 | } | |
84f0f298 | 625 | |
fff1b9c0 JR |
626 | static struct vlog_rate_limit error_report_rl = VLOG_RATE_LIMIT_INIT(1, 5); |
627 | ||
628 | #define XLATE_REPORT_ERROR(CTX, ...) \ | |
629 | do { \ | |
630 | if (OVS_UNLIKELY((CTX)->xin->report_hook)) { \ | |
631 | xlate_report(CTX, __VA_ARGS__); \ | |
632 | } else { \ | |
633 | VLOG_ERR_RL(&error_report_rl, __VA_ARGS__); \ | |
634 | } \ | |
635 | } while (0) | |
636 | ||
d6bef3cc BP |
637 | static inline void |
638 | xlate_report_actions(struct xlate_ctx *ctx, const char *title, | |
639 | const struct ofpact *ofpacts, size_t ofpacts_len) | |
640 | { | |
641 | if (OVS_UNLIKELY(ctx->xin->report_hook)) { | |
642 | struct ds s = DS_EMPTY_INITIALIZER; | |
643 | ofpacts_format(ofpacts, ofpacts_len, &s); | |
644 | xlate_report(ctx, "%s: %s", title, ds_cstr(&s)); | |
645 | ds_destroy(&s); | |
646 | } | |
647 | } | |
648 | ||
84f0f298 RW |
649 | static void |
650 | xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
651 | { | |
417e7e66 | 652 | ovs_list_init(&xbridge->xbundles); |
84f0f298 RW |
653 | hmap_init(&xbridge->xports); |
654 | hmap_insert(&xcfg->xbridges, &xbridge->hmap_node, | |
655 | hash_pointer(xbridge->ofproto, 0)); | |
656 | } | |
657 | ||
658 | static void | |
659 | xlate_xbundle_init(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
660 | { | |
417e7e66 BW |
661 | ovs_list_init(&xbundle->xports); |
662 | ovs_list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); | |
84f0f298 RW |
663 | hmap_insert(&xcfg->xbundles, &xbundle->hmap_node, |
664 | hash_pointer(xbundle->ofbundle, 0)); | |
665 | } | |
666 | ||
667 | static void | |
668 | xlate_xport_init(struct xlate_cfg *xcfg, struct xport *xport) | |
669 | { | |
670 | hmap_init(&xport->skb_priorities); | |
671 | hmap_insert(&xcfg->xports, &xport->hmap_node, | |
672 | hash_pointer(xport->ofport, 0)); | |
673 | hmap_insert(&xport->xbridge->xports, &xport->ofp_node, | |
674 | hash_ofp_port(xport->ofp_port)); | |
675 | } | |
676 | ||
677 | static void | |
678 | xlate_xbridge_set(struct xbridge *xbridge, | |
679 | struct dpif *dpif, | |
ec89fc6f | 680 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 681 | struct rstp *rstp, const struct mcast_snooping *ms, |
ec89fc6f | 682 | const struct mbridge *mbridge, |
46c88433 | 683 | const struct dpif_sflow *sflow, |
ce3955be | 684 | const struct dpif_ipfix *ipfix, |
2f47cdf4 | 685 | const struct netflow *netflow, |
4b97b70d | 686 | bool forward_bpdu, bool has_in_band, |
b440dd8c | 687 | const struct dpif_backer_support *support) |
46c88433 | 688 | { |
46c88433 EJ |
689 | if (xbridge->ml != ml) { |
690 | mac_learning_unref(xbridge->ml); | |
691 | xbridge->ml = mac_learning_ref(ml); | |
692 | } | |
693 | ||
6d95c4e8 FL |
694 | if (xbridge->ms != ms) { |
695 | mcast_snooping_unref(xbridge->ms); | |
696 | xbridge->ms = mcast_snooping_ref(ms); | |
697 | } | |
698 | ||
46c88433 EJ |
699 | if (xbridge->mbridge != mbridge) { |
700 | mbridge_unref(xbridge->mbridge); | |
701 | xbridge->mbridge = mbridge_ref(mbridge); | |
702 | } | |
703 | ||
704 | if (xbridge->sflow != sflow) { | |
705 | dpif_sflow_unref(xbridge->sflow); | |
706 | xbridge->sflow = dpif_sflow_ref(sflow); | |
707 | } | |
708 | ||
709 | if (xbridge->ipfix != ipfix) { | |
710 | dpif_ipfix_unref(xbridge->ipfix); | |
711 | xbridge->ipfix = dpif_ipfix_ref(ipfix); | |
712 | } | |
713 | ||
9d189a50 EJ |
714 | if (xbridge->stp != stp) { |
715 | stp_unref(xbridge->stp); | |
716 | xbridge->stp = stp_ref(stp); | |
717 | } | |
718 | ||
9efd308e DV |
719 | if (xbridge->rstp != rstp) { |
720 | rstp_unref(xbridge->rstp); | |
721 | xbridge->rstp = rstp_ref(rstp); | |
722 | } | |
723 | ||
ce3955be EJ |
724 | if (xbridge->netflow != netflow) { |
725 | netflow_unref(xbridge->netflow); | |
726 | xbridge->netflow = netflow_ref(netflow); | |
727 | } | |
728 | ||
89a8a7f0 | 729 | xbridge->dpif = dpif; |
46c88433 EJ |
730 | xbridge->forward_bpdu = forward_bpdu; |
731 | xbridge->has_in_band = has_in_band; | |
b440dd8c | 732 | xbridge->support = *support; |
46c88433 EJ |
733 | } |
734 | ||
84f0f298 RW |
735 | static void |
736 | xlate_xbundle_set(struct xbundle *xbundle, | |
737 | enum port_vlan_mode vlan_mode, int vlan, | |
738 | unsigned long *trunks, bool use_priority_tags, | |
739 | const struct bond *bond, const struct lacp *lacp, | |
740 | bool floodable) | |
741 | { | |
742 | ovs_assert(xbundle->xbridge); | |
743 | ||
744 | xbundle->vlan_mode = vlan_mode; | |
745 | xbundle->vlan = vlan; | |
746 | xbundle->trunks = trunks; | |
747 | xbundle->use_priority_tags = use_priority_tags; | |
748 | xbundle->floodable = floodable; | |
749 | ||
750 | if (xbundle->bond != bond) { | |
751 | bond_unref(xbundle->bond); | |
752 | xbundle->bond = bond_ref(bond); | |
753 | } | |
754 | ||
755 | if (xbundle->lacp != lacp) { | |
756 | lacp_unref(xbundle->lacp); | |
757 | xbundle->lacp = lacp_ref(lacp); | |
758 | } | |
759 | } | |
760 | ||
761 | static void | |
762 | xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
763 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 | 764 | const struct bfd *bfd, const struct lldp *lldp, int stp_port_no, |
f025bcb7 | 765 | const struct rstp_port* rstp_port, |
84f0f298 RW |
766 | enum ofputil_port_config config, enum ofputil_port_state state, |
767 | bool is_tunnel, bool may_enable) | |
768 | { | |
769 | xport->config = config; | |
770 | xport->state = state; | |
771 | xport->stp_port_no = stp_port_no; | |
772 | xport->is_tunnel = is_tunnel; | |
773 | xport->may_enable = may_enable; | |
774 | xport->odp_port = odp_port; | |
775 | ||
f025bcb7 JR |
776 | if (xport->rstp_port != rstp_port) { |
777 | rstp_port_unref(xport->rstp_port); | |
778 | xport->rstp_port = rstp_port_ref(rstp_port); | |
779 | } | |
780 | ||
84f0f298 RW |
781 | if (xport->cfm != cfm) { |
782 | cfm_unref(xport->cfm); | |
783 | xport->cfm = cfm_ref(cfm); | |
784 | } | |
785 | ||
786 | if (xport->bfd != bfd) { | |
787 | bfd_unref(xport->bfd); | |
788 | xport->bfd = bfd_ref(bfd); | |
789 | } | |
790 | ||
0477baa9 DF |
791 | if (xport->lldp != lldp) { |
792 | lldp_unref(xport->lldp); | |
793 | xport->lldp = lldp_ref(lldp); | |
794 | } | |
795 | ||
84f0f298 RW |
796 | if (xport->netdev != netdev) { |
797 | netdev_close(xport->netdev); | |
798 | xport->netdev = netdev_ref(netdev); | |
799 | } | |
800 | } | |
801 | ||
802 | static void | |
803 | xlate_xbridge_copy(struct xbridge *xbridge) | |
804 | { | |
805 | struct xbundle *xbundle; | |
806 | struct xport *xport; | |
807 | struct xbridge *new_xbridge = xzalloc(sizeof *xbridge); | |
808 | new_xbridge->ofproto = xbridge->ofproto; | |
809 | new_xbridge->name = xstrdup(xbridge->name); | |
810 | xlate_xbridge_init(new_xcfg, new_xbridge); | |
811 | ||
812 | xlate_xbridge_set(new_xbridge, | |
34dd0d78 | 813 | xbridge->dpif, xbridge->ml, xbridge->stp, |
9efd308e DV |
814 | xbridge->rstp, xbridge->ms, xbridge->mbridge, |
815 | xbridge->sflow, xbridge->ipfix, xbridge->netflow, | |
b440dd8c JS |
816 | xbridge->forward_bpdu, xbridge->has_in_band, |
817 | &xbridge->support); | |
84f0f298 RW |
818 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
819 | xlate_xbundle_copy(new_xbridge, xbundle); | |
820 | } | |
821 | ||
822 | /* Copy xports which are not part of a xbundle */ | |
823 | HMAP_FOR_EACH (xport, ofp_node, &xbridge->xports) { | |
824 | if (!xport->xbundle) { | |
825 | xlate_xport_copy(new_xbridge, NULL, xport); | |
826 | } | |
827 | } | |
828 | } | |
829 | ||
830 | static void | |
831 | xlate_xbundle_copy(struct xbridge *xbridge, struct xbundle *xbundle) | |
832 | { | |
833 | struct xport *xport; | |
834 | struct xbundle *new_xbundle = xzalloc(sizeof *xbundle); | |
835 | new_xbundle->ofbundle = xbundle->ofbundle; | |
836 | new_xbundle->xbridge = xbridge; | |
837 | new_xbundle->name = xstrdup(xbundle->name); | |
838 | xlate_xbundle_init(new_xcfg, new_xbundle); | |
839 | ||
840 | xlate_xbundle_set(new_xbundle, xbundle->vlan_mode, | |
841 | xbundle->vlan, xbundle->trunks, | |
842 | xbundle->use_priority_tags, xbundle->bond, xbundle->lacp, | |
843 | xbundle->floodable); | |
844 | LIST_FOR_EACH (xport, bundle_node, &xbundle->xports) { | |
845 | xlate_xport_copy(xbridge, new_xbundle, xport); | |
846 | } | |
847 | } | |
848 | ||
849 | static void | |
850 | xlate_xport_copy(struct xbridge *xbridge, struct xbundle *xbundle, | |
851 | struct xport *xport) | |
852 | { | |
853 | struct skb_priority_to_dscp *pdscp, *new_pdscp; | |
854 | struct xport *new_xport = xzalloc(sizeof *xport); | |
855 | new_xport->ofport = xport->ofport; | |
856 | new_xport->ofp_port = xport->ofp_port; | |
857 | new_xport->xbridge = xbridge; | |
858 | xlate_xport_init(new_xcfg, new_xport); | |
859 | ||
860 | xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm, | |
0477baa9 DF |
861 | xport->bfd, xport->lldp, xport->stp_port_no, |
862 | xport->rstp_port, xport->config, xport->state, | |
863 | xport->is_tunnel, xport->may_enable); | |
84f0f298 RW |
864 | |
865 | if (xport->peer) { | |
866 | struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport); | |
867 | if (peer) { | |
868 | new_xport->peer = peer; | |
869 | new_xport->peer->peer = new_xport; | |
870 | } | |
871 | } | |
872 | ||
873 | if (xbundle) { | |
874 | new_xport->xbundle = xbundle; | |
417e7e66 | 875 | ovs_list_insert(&new_xport->xbundle->xports, &new_xport->bundle_node); |
84f0f298 RW |
876 | } |
877 | ||
878 | HMAP_FOR_EACH (pdscp, hmap_node, &xport->skb_priorities) { | |
879 | new_pdscp = xmalloc(sizeof *pdscp); | |
880 | new_pdscp->skb_priority = pdscp->skb_priority; | |
881 | new_pdscp->dscp = pdscp->dscp; | |
882 | hmap_insert(&new_xport->skb_priorities, &new_pdscp->hmap_node, | |
883 | hash_int(new_pdscp->skb_priority, 0)); | |
884 | } | |
885 | } | |
886 | ||
887 | /* Sets the current xlate configuration to new_xcfg and frees the old xlate | |
888 | * configuration in xcfgp. | |
889 | * | |
890 | * This needs to be called after editing the xlate configuration. | |
891 | * | |
892 | * Functions that edit the new xlate configuration are | |
6cd20a22 | 893 | * xlate_<ofproto/bundle/ofport>_set and xlate_<ofproto/bundle/ofport>_remove. |
84f0f298 RW |
894 | * |
895 | * A sample workflow: | |
896 | * | |
897 | * xlate_txn_start(); | |
898 | * ... | |
899 | * edit_xlate_configuration(); | |
900 | * ... | |
901 | * xlate_txn_commit(); */ | |
46c88433 | 902 | void |
84f0f298 RW |
903 | xlate_txn_commit(void) |
904 | { | |
905 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
906 | ||
907 | ovsrcu_set(&xcfgp, new_xcfg); | |
40a9c4c2 AW |
908 | ovsrcu_synchronize(); |
909 | xlate_xcfg_free(xcfg); | |
84f0f298 RW |
910 | new_xcfg = NULL; |
911 | } | |
912 | ||
913 | /* Copies the current xlate configuration in xcfgp to new_xcfg. | |
914 | * | |
915 | * This needs to be called prior to editing the xlate configuration. */ | |
916 | void | |
917 | xlate_txn_start(void) | |
918 | { | |
919 | struct xbridge *xbridge; | |
920 | struct xlate_cfg *xcfg; | |
921 | ||
922 | ovs_assert(!new_xcfg); | |
923 | ||
924 | new_xcfg = xmalloc(sizeof *new_xcfg); | |
925 | hmap_init(&new_xcfg->xbridges); | |
926 | hmap_init(&new_xcfg->xbundles); | |
927 | hmap_init(&new_xcfg->xports); | |
928 | ||
929 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
930 | if (!xcfg) { | |
931 | return; | |
932 | } | |
933 | ||
934 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
935 | xlate_xbridge_copy(xbridge); | |
936 | } | |
937 | } | |
938 | ||
939 | ||
940 | static void | |
941 | xlate_xcfg_free(struct xlate_cfg *xcfg) | |
942 | { | |
943 | struct xbridge *xbridge, *next_xbridge; | |
944 | ||
945 | if (!xcfg) { | |
946 | return; | |
947 | } | |
948 | ||
949 | HMAP_FOR_EACH_SAFE (xbridge, next_xbridge, hmap_node, &xcfg->xbridges) { | |
950 | xlate_xbridge_remove(xcfg, xbridge); | |
951 | } | |
952 | ||
953 | hmap_destroy(&xcfg->xbridges); | |
954 | hmap_destroy(&xcfg->xbundles); | |
955 | hmap_destroy(&xcfg->xports); | |
956 | free(xcfg); | |
957 | } | |
958 | ||
959 | void | |
960 | xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, | |
34dd0d78 | 961 | struct dpif *dpif, |
84f0f298 | 962 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 963 | struct rstp *rstp, const struct mcast_snooping *ms, |
84f0f298 RW |
964 | const struct mbridge *mbridge, |
965 | const struct dpif_sflow *sflow, | |
966 | const struct dpif_ipfix *ipfix, | |
2f47cdf4 | 967 | const struct netflow *netflow, |
b440dd8c JS |
968 | bool forward_bpdu, bool has_in_band, |
969 | const struct dpif_backer_support *support) | |
84f0f298 RW |
970 | { |
971 | struct xbridge *xbridge; | |
972 | ||
973 | ovs_assert(new_xcfg); | |
974 | ||
975 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
976 | if (!xbridge) { | |
977 | xbridge = xzalloc(sizeof *xbridge); | |
978 | xbridge->ofproto = ofproto; | |
979 | ||
980 | xlate_xbridge_init(new_xcfg, xbridge); | |
981 | } | |
982 | ||
983 | free(xbridge->name); | |
984 | xbridge->name = xstrdup(name); | |
985 | ||
34dd0d78 | 986 | xlate_xbridge_set(xbridge, dpif, ml, stp, rstp, ms, mbridge, sflow, ipfix, |
b440dd8c | 987 | netflow, forward_bpdu, has_in_band, support); |
84f0f298 RW |
988 | } |
989 | ||
990 | static void | |
991 | xlate_xbridge_remove(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
46c88433 | 992 | { |
46c88433 EJ |
993 | struct xbundle *xbundle, *next_xbundle; |
994 | struct xport *xport, *next_xport; | |
995 | ||
996 | if (!xbridge) { | |
997 | return; | |
998 | } | |
999 | ||
1000 | HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) { | |
84f0f298 | 1001 | xlate_xport_remove(xcfg, xport); |
46c88433 EJ |
1002 | } |
1003 | ||
1004 | LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) { | |
84f0f298 | 1005 | xlate_xbundle_remove(xcfg, xbundle); |
46c88433 EJ |
1006 | } |
1007 | ||
84f0f298 | 1008 | hmap_remove(&xcfg->xbridges, &xbridge->hmap_node); |
795cc5c1 | 1009 | mac_learning_unref(xbridge->ml); |
6d95c4e8 | 1010 | mcast_snooping_unref(xbridge->ms); |
795cc5c1 EJ |
1011 | mbridge_unref(xbridge->mbridge); |
1012 | dpif_sflow_unref(xbridge->sflow); | |
1013 | dpif_ipfix_unref(xbridge->ipfix); | |
1014 | stp_unref(xbridge->stp); | |
9efd308e | 1015 | rstp_unref(xbridge->rstp); |
795cc5c1 | 1016 | hmap_destroy(&xbridge->xports); |
46c88433 EJ |
1017 | free(xbridge->name); |
1018 | free(xbridge); | |
1019 | } | |
1020 | ||
84f0f298 RW |
1021 | void |
1022 | xlate_remove_ofproto(struct ofproto_dpif *ofproto) | |
1023 | { | |
1024 | struct xbridge *xbridge; | |
1025 | ||
1026 | ovs_assert(new_xcfg); | |
1027 | ||
1028 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
1029 | xlate_xbridge_remove(new_xcfg, xbridge); | |
1030 | } | |
1031 | ||
46c88433 EJ |
1032 | void |
1033 | xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
1034 | const char *name, enum port_vlan_mode vlan_mode, int vlan, | |
1035 | unsigned long *trunks, bool use_priority_tags, | |
1036 | const struct bond *bond, const struct lacp *lacp, | |
1037 | bool floodable) | |
1038 | { | |
84f0f298 | 1039 | struct xbundle *xbundle; |
46c88433 | 1040 | |
84f0f298 RW |
1041 | ovs_assert(new_xcfg); |
1042 | ||
1043 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
46c88433 EJ |
1044 | if (!xbundle) { |
1045 | xbundle = xzalloc(sizeof *xbundle); | |
1046 | xbundle->ofbundle = ofbundle; | |
84f0f298 | 1047 | xbundle->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 1048 | |
84f0f298 | 1049 | xlate_xbundle_init(new_xcfg, xbundle); |
46c88433 EJ |
1050 | } |
1051 | ||
46c88433 EJ |
1052 | free(xbundle->name); |
1053 | xbundle->name = xstrdup(name); | |
1054 | ||
84f0f298 RW |
1055 | xlate_xbundle_set(xbundle, vlan_mode, vlan, trunks, |
1056 | use_priority_tags, bond, lacp, floodable); | |
46c88433 EJ |
1057 | } |
1058 | ||
84f0f298 RW |
1059 | static void |
1060 | xlate_xbundle_remove(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
46c88433 | 1061 | { |
5f03c983 | 1062 | struct xport *xport; |
46c88433 EJ |
1063 | |
1064 | if (!xbundle) { | |
1065 | return; | |
1066 | } | |
1067 | ||
5f03c983 | 1068 | LIST_FOR_EACH_POP (xport, bundle_node, &xbundle->xports) { |
46c88433 EJ |
1069 | xport->xbundle = NULL; |
1070 | } | |
1071 | ||
84f0f298 | 1072 | hmap_remove(&xcfg->xbundles, &xbundle->hmap_node); |
417e7e66 | 1073 | ovs_list_remove(&xbundle->list_node); |
46c88433 EJ |
1074 | bond_unref(xbundle->bond); |
1075 | lacp_unref(xbundle->lacp); | |
1076 | free(xbundle->name); | |
1077 | free(xbundle); | |
1078 | } | |
1079 | ||
84f0f298 RW |
1080 | void |
1081 | xlate_bundle_remove(struct ofbundle *ofbundle) | |
1082 | { | |
1083 | struct xbundle *xbundle; | |
1084 | ||
1085 | ovs_assert(new_xcfg); | |
1086 | ||
1087 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
1088 | xlate_xbundle_remove(new_xcfg, xbundle); | |
1089 | } | |
1090 | ||
46c88433 EJ |
1091 | void |
1092 | xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
1093 | struct ofport_dpif *ofport, ofp_port_t ofp_port, | |
1094 | odp_port_t odp_port, const struct netdev *netdev, | |
1095 | const struct cfm *cfm, const struct bfd *bfd, | |
0477baa9 DF |
1096 | const struct lldp *lldp, struct ofport_dpif *peer, |
1097 | int stp_port_no, const struct rstp_port *rstp_port, | |
55954f6e | 1098 | const struct ofproto_port_queue *qdscp_list, size_t n_qdscp, |
dd8cd4b4 SH |
1099 | enum ofputil_port_config config, |
1100 | enum ofputil_port_state state, bool is_tunnel, | |
9d189a50 | 1101 | bool may_enable) |
46c88433 | 1102 | { |
55954f6e | 1103 | size_t i; |
84f0f298 RW |
1104 | struct xport *xport; |
1105 | ||
1106 | ovs_assert(new_xcfg); | |
46c88433 | 1107 | |
84f0f298 | 1108 | xport = xport_lookup(new_xcfg, ofport); |
46c88433 EJ |
1109 | if (!xport) { |
1110 | xport = xzalloc(sizeof *xport); | |
1111 | xport->ofport = ofport; | |
84f0f298 | 1112 | xport->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 EJ |
1113 | xport->ofp_port = ofp_port; |
1114 | ||
84f0f298 | 1115 | xlate_xport_init(new_xcfg, xport); |
46c88433 EJ |
1116 | } |
1117 | ||
1118 | ovs_assert(xport->ofp_port == ofp_port); | |
1119 | ||
0477baa9 DF |
1120 | xlate_xport_set(xport, odp_port, netdev, cfm, bfd, lldp, |
1121 | stp_port_no, rstp_port, config, state, is_tunnel, | |
1122 | may_enable); | |
46c88433 EJ |
1123 | |
1124 | if (xport->peer) { | |
1125 | xport->peer->peer = NULL; | |
1126 | } | |
84f0f298 | 1127 | xport->peer = xport_lookup(new_xcfg, peer); |
46c88433 EJ |
1128 | if (xport->peer) { |
1129 | xport->peer->peer = xport; | |
1130 | } | |
1131 | ||
1132 | if (xport->xbundle) { | |
417e7e66 | 1133 | ovs_list_remove(&xport->bundle_node); |
46c88433 | 1134 | } |
84f0f298 | 1135 | xport->xbundle = xbundle_lookup(new_xcfg, ofbundle); |
46c88433 | 1136 | if (xport->xbundle) { |
417e7e66 | 1137 | ovs_list_insert(&xport->xbundle->xports, &xport->bundle_node); |
46c88433 | 1138 | } |
55954f6e EJ |
1139 | |
1140 | clear_skb_priorities(xport); | |
1141 | for (i = 0; i < n_qdscp; i++) { | |
1142 | struct skb_priority_to_dscp *pdscp; | |
1143 | uint32_t skb_priority; | |
1144 | ||
89a8a7f0 EJ |
1145 | if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue, |
1146 | &skb_priority)) { | |
55954f6e EJ |
1147 | continue; |
1148 | } | |
1149 | ||
1150 | pdscp = xmalloc(sizeof *pdscp); | |
1151 | pdscp->skb_priority = skb_priority; | |
1152 | pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK; | |
1153 | hmap_insert(&xport->skb_priorities, &pdscp->hmap_node, | |
1154 | hash_int(pdscp->skb_priority, 0)); | |
1155 | } | |
46c88433 EJ |
1156 | } |
1157 | ||
84f0f298 RW |
1158 | static void |
1159 | xlate_xport_remove(struct xlate_cfg *xcfg, struct xport *xport) | |
46c88433 | 1160 | { |
46c88433 EJ |
1161 | if (!xport) { |
1162 | return; | |
1163 | } | |
1164 | ||
1165 | if (xport->peer) { | |
1166 | xport->peer->peer = NULL; | |
1167 | xport->peer = NULL; | |
1168 | } | |
1169 | ||
e621a12d | 1170 | if (xport->xbundle) { |
417e7e66 | 1171 | ovs_list_remove(&xport->bundle_node); |
e621a12d EJ |
1172 | } |
1173 | ||
55954f6e EJ |
1174 | clear_skb_priorities(xport); |
1175 | hmap_destroy(&xport->skb_priorities); | |
1176 | ||
84f0f298 | 1177 | hmap_remove(&xcfg->xports, &xport->hmap_node); |
46c88433 EJ |
1178 | hmap_remove(&xport->xbridge->xports, &xport->ofp_node); |
1179 | ||
1180 | netdev_close(xport->netdev); | |
f025bcb7 | 1181 | rstp_port_unref(xport->rstp_port); |
46c88433 EJ |
1182 | cfm_unref(xport->cfm); |
1183 | bfd_unref(xport->bfd); | |
0477baa9 | 1184 | lldp_unref(xport->lldp); |
46c88433 EJ |
1185 | free(xport); |
1186 | } | |
1187 | ||
84f0f298 RW |
1188 | void |
1189 | xlate_ofport_remove(struct ofport_dpif *ofport) | |
1190 | { | |
1191 | struct xport *xport; | |
1192 | ||
1193 | ovs_assert(new_xcfg); | |
1194 | ||
1195 | xport = xport_lookup(new_xcfg, ofport); | |
1196 | xlate_xport_remove(new_xcfg, xport); | |
1197 | } | |
1198 | ||
ef377a58 JR |
1199 | static struct ofproto_dpif * |
1200 | xlate_lookup_ofproto_(const struct dpif_backer *backer, const struct flow *flow, | |
1201 | ofp_port_t *ofp_in_port, const struct xport **xportp) | |
1202 | { | |
e672ff9b | 1203 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
ef377a58 | 1204 | const struct xport *xport; |
f9038ef6 | 1205 | |
e672ff9b JR |
1206 | xport = xport_lookup(xcfg, tnl_port_should_receive(flow) |
1207 | ? tnl_port_receive(flow) | |
1208 | : odp_port_to_ofport(backer, flow->in_port.odp_port)); | |
1209 | if (OVS_UNLIKELY(!xport)) { | |
1210 | return NULL; | |
ef377a58 | 1211 | } |
e672ff9b | 1212 | *xportp = xport; |
f9038ef6 | 1213 | if (ofp_in_port) { |
e672ff9b | 1214 | *ofp_in_port = xport->ofp_port; |
f9038ef6 | 1215 | } |
e672ff9b | 1216 | return xport->xbridge->ofproto; |
ef377a58 JR |
1217 | } |
1218 | ||
1219 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively) | |
1220 | * returns the corresponding struct ofproto_dpif and OpenFlow port number. */ | |
1221 | struct ofproto_dpif * | |
1222 | xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow, | |
1223 | ofp_port_t *ofp_in_port) | |
1224 | { | |
1225 | const struct xport *xport; | |
1226 | ||
1227 | return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); | |
1228 | } | |
1229 | ||
cc377352 | 1230 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively), |
ef377a58 | 1231 | * optionally populates 'ofproto' with the ofproto_dpif, 'ofp_in_port' with the |
cc377352 | 1232 | * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate |
dcc2c6cd JR |
1233 | * handles for those protocols if they're enabled. Caller may use the returned |
1234 | * pointers until quiescing, for longer term use additional references must | |
1235 | * be taken. | |
8449c4d6 | 1236 | * |
f9038ef6 | 1237 | * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto. |
ef377a58 | 1238 | */ |
8449c4d6 | 1239 | int |
5c476ea3 JR |
1240 | xlate_lookup(const struct dpif_backer *backer, const struct flow *flow, |
1241 | struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix, | |
1242 | struct dpif_sflow **sflow, struct netflow **netflow, | |
1243 | ofp_port_t *ofp_in_port) | |
8449c4d6 | 1244 | { |
ef377a58 | 1245 | struct ofproto_dpif *ofproto; |
84f0f298 | 1246 | const struct xport *xport; |
8449c4d6 | 1247 | |
ef377a58 | 1248 | ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); |
8449c4d6 | 1249 | |
f9038ef6 | 1250 | if (!ofproto) { |
cc377352 | 1251 | return ENODEV; |
8449c4d6 | 1252 | } |
8449c4d6 | 1253 | |
ef377a58 JR |
1254 | if (ofprotop) { |
1255 | *ofprotop = ofproto; | |
8449c4d6 EJ |
1256 | } |
1257 | ||
1dfdb9b3 | 1258 | if (ipfix) { |
f9038ef6 | 1259 | *ipfix = xport ? xport->xbridge->ipfix : NULL; |
1dfdb9b3 EJ |
1260 | } |
1261 | ||
1262 | if (sflow) { | |
f9038ef6 | 1263 | *sflow = xport ? xport->xbridge->sflow : NULL; |
1dfdb9b3 EJ |
1264 | } |
1265 | ||
1266 | if (netflow) { | |
f9038ef6 | 1267 | *netflow = xport ? xport->xbridge->netflow : NULL; |
1dfdb9b3 | 1268 | } |
f9038ef6 | 1269 | |
cc377352 | 1270 | return 0; |
8449c4d6 EJ |
1271 | } |
1272 | ||
46c88433 | 1273 | static struct xbridge * |
84f0f298 | 1274 | xbridge_lookup(struct xlate_cfg *xcfg, const struct ofproto_dpif *ofproto) |
46c88433 | 1275 | { |
84f0f298 | 1276 | struct hmap *xbridges; |
46c88433 EJ |
1277 | struct xbridge *xbridge; |
1278 | ||
84f0f298 | 1279 | if (!ofproto || !xcfg) { |
5e6af486 EJ |
1280 | return NULL; |
1281 | } | |
1282 | ||
84f0f298 RW |
1283 | xbridges = &xcfg->xbridges; |
1284 | ||
46c88433 | 1285 | HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0), |
84f0f298 | 1286 | xbridges) { |
46c88433 EJ |
1287 | if (xbridge->ofproto == ofproto) { |
1288 | return xbridge; | |
1289 | } | |
1290 | } | |
1291 | return NULL; | |
1292 | } | |
1293 | ||
290835f9 BP |
1294 | static struct xbridge * |
1295 | xbridge_lookup_by_uuid(struct xlate_cfg *xcfg, const struct uuid *uuid) | |
1296 | { | |
1297 | struct xbridge *xbridge; | |
1298 | ||
1299 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
1300 | if (uuid_equals(ofproto_dpif_get_uuid(xbridge->ofproto), uuid)) { | |
1301 | return xbridge; | |
1302 | } | |
1303 | } | |
1304 | return NULL; | |
1305 | } | |
1306 | ||
46c88433 | 1307 | static struct xbundle * |
84f0f298 | 1308 | xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle) |
46c88433 | 1309 | { |
84f0f298 | 1310 | struct hmap *xbundles; |
46c88433 EJ |
1311 | struct xbundle *xbundle; |
1312 | ||
84f0f298 | 1313 | if (!ofbundle || !xcfg) { |
5e6af486 EJ |
1314 | return NULL; |
1315 | } | |
1316 | ||
84f0f298 RW |
1317 | xbundles = &xcfg->xbundles; |
1318 | ||
46c88433 | 1319 | HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), |
84f0f298 | 1320 | xbundles) { |
46c88433 EJ |
1321 | if (xbundle->ofbundle == ofbundle) { |
1322 | return xbundle; | |
1323 | } | |
1324 | } | |
1325 | return NULL; | |
1326 | } | |
1327 | ||
1328 | static struct xport * | |
84f0f298 | 1329 | xport_lookup(struct xlate_cfg *xcfg, const struct ofport_dpif *ofport) |
46c88433 | 1330 | { |
84f0f298 | 1331 | struct hmap *xports; |
46c88433 EJ |
1332 | struct xport *xport; |
1333 | ||
84f0f298 | 1334 | if (!ofport || !xcfg) { |
5e6af486 EJ |
1335 | return NULL; |
1336 | } | |
1337 | ||
84f0f298 RW |
1338 | xports = &xcfg->xports; |
1339 | ||
46c88433 | 1340 | HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0), |
84f0f298 | 1341 | xports) { |
46c88433 EJ |
1342 | if (xport->ofport == ofport) { |
1343 | return xport; | |
1344 | } | |
1345 | } | |
1346 | return NULL; | |
1347 | } | |
1348 | ||
40085e56 EJ |
1349 | static struct stp_port * |
1350 | xport_get_stp_port(const struct xport *xport) | |
1351 | { | |
92cf817b | 1352 | return xport->xbridge->stp && xport->stp_port_no != -1 |
40085e56 EJ |
1353 | ? stp_get_port(xport->xbridge->stp, xport->stp_port_no) |
1354 | : NULL; | |
1355 | } | |
9d189a50 | 1356 | |
0d1cee12 | 1357 | static bool |
9d189a50 EJ |
1358 | xport_stp_learn_state(const struct xport *xport) |
1359 | { | |
40085e56 | 1360 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1361 | return sp |
1362 | ? stp_learn_in_state(stp_port_get_state(sp)) | |
1363 | : true; | |
9d189a50 EJ |
1364 | } |
1365 | ||
1366 | static bool | |
1367 | xport_stp_forward_state(const struct xport *xport) | |
1368 | { | |
40085e56 | 1369 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1370 | return sp |
1371 | ? stp_forward_in_state(stp_port_get_state(sp)) | |
1372 | : true; | |
9d189a50 EJ |
1373 | } |
1374 | ||
0d1cee12 | 1375 | static bool |
bacdb85a | 1376 | xport_stp_should_forward_bpdu(const struct xport *xport) |
0d1cee12 K |
1377 | { |
1378 | struct stp_port *sp = xport_get_stp_port(xport); | |
bacdb85a | 1379 | return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED); |
0d1cee12 K |
1380 | } |
1381 | ||
9d189a50 EJ |
1382 | /* Returns true if STP should process 'flow'. Sets fields in 'wc' that |
1383 | * were used to make the determination.*/ | |
1384 | static bool | |
1385 | stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc) | |
1386 | { | |
bbbca389 | 1387 | /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */ |
9d189a50 | 1388 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
bbbca389 | 1389 | return is_stp(flow); |
9d189a50 EJ |
1390 | } |
1391 | ||
1392 | static void | |
cf62fa4c | 1393 | stp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9d189a50 | 1394 | { |
40085e56 | 1395 | struct stp_port *sp = xport_get_stp_port(xport); |
cf62fa4c PS |
1396 | struct dp_packet payload = *packet; |
1397 | struct eth_header *eth = dp_packet_data(&payload); | |
9d189a50 EJ |
1398 | |
1399 | /* Sink packets on ports that have STP disabled when the bridge has | |
1400 | * STP enabled. */ | |
1401 | if (!sp || stp_port_get_state(sp) == STP_DISABLED) { | |
1402 | return; | |
1403 | } | |
1404 | ||
1405 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1406 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1407 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9d189a50 EJ |
1408 | } |
1409 | ||
cf62fa4c PS |
1410 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1411 | stp_received_bpdu(sp, dp_packet_data(&payload), dp_packet_size(&payload)); | |
9d189a50 EJ |
1412 | } |
1413 | } | |
1414 | ||
f025bcb7 JR |
1415 | static enum rstp_state |
1416 | xport_get_rstp_port_state(const struct xport *xport) | |
9efd308e | 1417 | { |
f025bcb7 JR |
1418 | return xport->rstp_port |
1419 | ? rstp_port_get_state(xport->rstp_port) | |
1420 | : RSTP_DISABLED; | |
9efd308e DV |
1421 | } |
1422 | ||
1423 | static bool | |
1424 | xport_rstp_learn_state(const struct xport *xport) | |
1425 | { | |
4b5f1996 DV |
1426 | return xport->xbridge->rstp && xport->rstp_port |
1427 | ? rstp_learn_in_state(xport_get_rstp_port_state(xport)) | |
1428 | : true; | |
9efd308e DV |
1429 | } |
1430 | ||
1431 | static bool | |
1432 | xport_rstp_forward_state(const struct xport *xport) | |
1433 | { | |
4b5f1996 DV |
1434 | return xport->xbridge->rstp && xport->rstp_port |
1435 | ? rstp_forward_in_state(xport_get_rstp_port_state(xport)) | |
1436 | : true; | |
9efd308e DV |
1437 | } |
1438 | ||
1439 | static bool | |
1440 | xport_rstp_should_manage_bpdu(const struct xport *xport) | |
1441 | { | |
f025bcb7 | 1442 | return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport)); |
9efd308e DV |
1443 | } |
1444 | ||
1445 | static void | |
cf62fa4c | 1446 | rstp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9efd308e | 1447 | { |
cf62fa4c PS |
1448 | struct dp_packet payload = *packet; |
1449 | struct eth_header *eth = dp_packet_data(&payload); | |
9efd308e | 1450 | |
f025bcb7 JR |
1451 | /* Sink packets on ports that have no RSTP. */ |
1452 | if (!xport->rstp_port) { | |
9efd308e DV |
1453 | return; |
1454 | } | |
1455 | ||
1456 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1457 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1458 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9efd308e DV |
1459 | } |
1460 | ||
cf62fa4c PS |
1461 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1462 | rstp_port_received_bpdu(xport->rstp_port, dp_packet_data(&payload), | |
1463 | dp_packet_size(&payload)); | |
9efd308e DV |
1464 | } |
1465 | } | |
1466 | ||
46c88433 EJ |
1467 | static struct xport * |
1468 | get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1469 | { | |
1470 | struct xport *xport; | |
1471 | ||
1472 | HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port), | |
1473 | &xbridge->xports) { | |
1474 | if (xport->ofp_port == ofp_port) { | |
1475 | return xport; | |
1476 | } | |
1477 | } | |
1478 | return NULL; | |
1479 | } | |
1480 | ||
1481 | static odp_port_t | |
1482 | ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1483 | { | |
1484 | const struct xport *xport = get_ofp_port(xbridge, ofp_port); | |
1485 | return xport ? xport->odp_port : ODPP_NONE; | |
1486 | } | |
1487 | ||
dd8cd4b4 SH |
1488 | static bool |
1489 | odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port) | |
1490 | { | |
086fa873 BP |
1491 | struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1492 | return xport && xport->may_enable; | |
dd8cd4b4 SH |
1493 | } |
1494 | ||
1e684d7d | 1495 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1496 | group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *, |
1497 | int depth); | |
1498 | ||
1499 | static bool | |
1500 | group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) | |
1501 | { | |
1502 | struct group_dpif *group; | |
dd8cd4b4 | 1503 | |
dc25893e AZ |
1504 | if (group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group)) { |
1505 | struct ofputil_bucket *bucket; | |
dd8cd4b4 | 1506 | |
dc25893e AZ |
1507 | bucket = group_first_live_bucket(ctx, group, depth); |
1508 | group_dpif_unref(group); | |
62e2e746 | 1509 | return bucket != NULL; |
dc25893e | 1510 | } |
dd8cd4b4 | 1511 | |
dc25893e | 1512 | return false; |
dd8cd4b4 SH |
1513 | } |
1514 | ||
1515 | #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ | |
1516 | ||
1517 | static bool | |
1518 | bucket_is_alive(const struct xlate_ctx *ctx, | |
1e684d7d | 1519 | struct ofputil_bucket *bucket, int depth) |
dd8cd4b4 SH |
1520 | { |
1521 | if (depth >= MAX_LIVENESS_RECURSION) { | |
1522 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
1523 | ||
1524 | VLOG_WARN_RL(&rl, "bucket chaining exceeded %d links", | |
1525 | MAX_LIVENESS_RECURSION); | |
1526 | return false; | |
1527 | } | |
1528 | ||
fdb1999b AZ |
1529 | return (!ofputil_bucket_has_liveness(bucket) |
1530 | || (bucket->watch_port != OFPP_ANY | |
1531 | && odp_port_is_alive(ctx, bucket->watch_port)) | |
1532 | || (bucket->watch_group != OFPG_ANY | |
1533 | && group_is_alive(ctx, bucket->watch_group, depth + 1))); | |
dd8cd4b4 SH |
1534 | } |
1535 | ||
1e684d7d | 1536 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1537 | group_first_live_bucket(const struct xlate_ctx *ctx, |
1538 | const struct group_dpif *group, int depth) | |
1539 | { | |
1540 | struct ofputil_bucket *bucket; | |
ca6ba700 | 1541 | const struct ovs_list *buckets; |
dd8cd4b4 SH |
1542 | |
1543 | group_dpif_get_buckets(group, &buckets); | |
1544 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1545 | if (bucket_is_alive(ctx, bucket, depth)) { | |
1546 | return bucket; | |
1547 | } | |
1548 | } | |
1549 | ||
1550 | return NULL; | |
1551 | } | |
1552 | ||
1e684d7d | 1553 | static struct ofputil_bucket * |
fe7e5749 SH |
1554 | group_best_live_bucket(const struct xlate_ctx *ctx, |
1555 | const struct group_dpif *group, | |
1556 | uint32_t basis) | |
1557 | { | |
1e684d7d | 1558 | struct ofputil_bucket *best_bucket = NULL; |
fe7e5749 SH |
1559 | uint32_t best_score = 0; |
1560 | int i = 0; | |
1561 | ||
1e684d7d | 1562 | struct ofputil_bucket *bucket; |
ca6ba700 | 1563 | const struct ovs_list *buckets; |
fe7e5749 SH |
1564 | |
1565 | group_dpif_get_buckets(group, &buckets); | |
1566 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1567 | if (bucket_is_alive(ctx, bucket, 0)) { | |
7cb279c2 | 1568 | uint32_t score = (hash_int(i, basis) & 0xffff) * bucket->weight; |
fe7e5749 SH |
1569 | if (score >= best_score) { |
1570 | best_bucket = bucket; | |
1571 | best_score = score; | |
1572 | } | |
1573 | } | |
1574 | i++; | |
1575 | } | |
1576 | ||
1577 | return best_bucket; | |
1578 | } | |
1579 | ||
9583bc14 | 1580 | static bool |
46c88433 | 1581 | xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan) |
9583bc14 EJ |
1582 | { |
1583 | return (bundle->vlan_mode != PORT_VLAN_ACCESS | |
1584 | && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan))); | |
1585 | } | |
1586 | ||
1587 | static bool | |
46c88433 EJ |
1588 | xbundle_includes_vlan(const struct xbundle *xbundle, uint16_t vlan) |
1589 | { | |
1590 | return vlan == xbundle->vlan || xbundle_trunks_vlan(xbundle, vlan); | |
1591 | } | |
1592 | ||
1593 | static mirror_mask_t | |
1594 | xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle) | |
1595 | { | |
1596 | return xbundle != &ofpp_none_bundle | |
1597 | ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle) | |
1598 | : 0; | |
1599 | } | |
1600 | ||
1601 | static mirror_mask_t | |
1602 | xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1603 | { |
46c88433 EJ |
1604 | return xbundle != &ofpp_none_bundle |
1605 | ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle) | |
1606 | : 0; | |
9583bc14 EJ |
1607 | } |
1608 | ||
46c88433 EJ |
1609 | static mirror_mask_t |
1610 | xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1611 | { |
46c88433 EJ |
1612 | return xbundle != &ofpp_none_bundle |
1613 | ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle) | |
1614 | : 0; | |
1615 | } | |
1616 | ||
1617 | static struct xbundle * | |
1618 | lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port, | |
1619 | bool warn, struct xport **in_xportp) | |
1620 | { | |
1621 | struct xport *xport; | |
9583bc14 EJ |
1622 | |
1623 | /* Find the port and bundle for the received packet. */ | |
46c88433 EJ |
1624 | xport = get_ofp_port(xbridge, in_port); |
1625 | if (in_xportp) { | |
1626 | *in_xportp = xport; | |
9583bc14 | 1627 | } |
46c88433 EJ |
1628 | if (xport && xport->xbundle) { |
1629 | return xport->xbundle; | |
9583bc14 EJ |
1630 | } |
1631 | ||
6362203b YT |
1632 | /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+), |
1633 | * which a controller may use as the ingress port for traffic that | |
1634 | * it is sourcing. */ | |
1635 | if (in_port == OFPP_CONTROLLER || in_port == OFPP_NONE) { | |
9583bc14 EJ |
1636 | return &ofpp_none_bundle; |
1637 | } | |
1638 | ||
1639 | /* Odd. A few possible reasons here: | |
1640 | * | |
1641 | * - We deleted a port but there are still a few packets queued up | |
1642 | * from it. | |
1643 | * | |
1644 | * - Someone externally added a port (e.g. "ovs-dpctl add-if") that | |
1645 | * we don't know about. | |
1646 | * | |
1647 | * - The ofproto client didn't configure the port as part of a bundle. | |
1648 | * This is particularly likely to happen if a packet was received on the | |
1649 | * port after it was created, but before the client had a chance to | |
1650 | * configure its bundle. | |
1651 | */ | |
1652 | if (warn) { | |
1653 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1654 | ||
1655 | VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown " | |
46c88433 | 1656 | "port %"PRIu16, xbridge->name, in_port); |
9583bc14 EJ |
1657 | } |
1658 | return NULL; | |
1659 | } | |
1660 | ||
faa624b4 BP |
1661 | /* Mirrors the packet represented by 'ctx' to appropriate mirror destinations, |
1662 | * given the packet is ingressing or egressing on 'xbundle', which has ingress | |
1663 | * or egress (as appropriate) mirrors 'mirrors'. */ | |
9583bc14 | 1664 | static void |
7efbc3b7 BP |
1665 | mirror_packet(struct xlate_ctx *ctx, struct xbundle *xbundle, |
1666 | mirror_mask_t mirrors) | |
9583bc14 | 1667 | { |
faa624b4 BP |
1668 | /* Figure out what VLAN the packet is in (because mirrors can select |
1669 | * packets on basis of VLAN). */ | |
7efbc3b7 BP |
1670 | bool warn = ctx->xin->packet != NULL; |
1671 | uint16_t vid = vlan_tci_to_vid(ctx->xin->flow.vlan_tci); | |
1672 | if (!input_vid_is_valid(vid, xbundle, warn)) { | |
9583bc14 EJ |
1673 | return; |
1674 | } | |
7efbc3b7 | 1675 | uint16_t vlan = input_vid_to_vlan(xbundle, vid); |
9583bc14 | 1676 | |
7efbc3b7 | 1677 | const struct xbridge *xbridge = ctx->xbridge; |
9583bc14 | 1678 | |
7efbc3b7 BP |
1679 | /* Don't mirror to destinations that we've already mirrored to. */ |
1680 | mirrors &= ~ctx->mirrors; | |
9583bc14 EJ |
1681 | if (!mirrors) { |
1682 | return; | |
1683 | } | |
1684 | ||
7efbc3b7 BP |
1685 | if (ctx->xin->resubmit_stats) { |
1686 | mirror_update_stats(xbridge->mbridge, mirrors, | |
1687 | ctx->xin->resubmit_stats->n_packets, | |
1688 | ctx->xin->resubmit_stats->n_bytes); | |
1689 | } | |
1690 | if (ctx->xin->xcache) { | |
1691 | struct xc_entry *entry; | |
1692 | ||
1693 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_MIRROR); | |
1694 | entry->u.mirror.mbridge = mbridge_ref(xbridge->mbridge); | |
1695 | entry->u.mirror.mirrors = mirrors; | |
1696 | } | |
9583bc14 | 1697 | |
faa624b4 BP |
1698 | /* 'mirrors' is a bit-mask of candidates for mirroring. Iterate as long as |
1699 | * some candidates remain. */ | |
9583bc14 | 1700 | while (mirrors) { |
7efbc3b7 | 1701 | const unsigned long *vlans; |
ec7ceaed EJ |
1702 | mirror_mask_t dup_mirrors; |
1703 | struct ofbundle *out; | |
ec7ceaed EJ |
1704 | int out_vlan; |
1705 | ||
faa624b4 | 1706 | /* Get the details of the mirror represented by the rightmost 1-bit. */ |
7efbc3b7 BP |
1707 | bool has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors), |
1708 | &vlans, &dup_mirrors, &out, &out_vlan); | |
ec7ceaed EJ |
1709 | ovs_assert(has_mirror); |
1710 | ||
faa624b4 BP |
1711 | /* If this mirror selects on the basis of VLAN, and it does not select |
1712 | * 'vlan', then discard this mirror and go on to the next one. */ | |
ec7ceaed | 1713 | if (vlans) { |
49a73e0c | 1714 | ctx->wc->masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK); |
9583bc14 | 1715 | } |
7efbc3b7 | 1716 | if (vlans && !bitmap_is_set(vlans, vlan)) { |
9583bc14 EJ |
1717 | mirrors = zero_rightmost_1bit(mirrors); |
1718 | continue; | |
1719 | } | |
1720 | ||
faa624b4 BP |
1721 | /* Record the mirror, and the mirrors that output to the same |
1722 | * destination, so that we don't mirror to them again. This must be | |
1723 | * done now to ensure that output_normal(), below, doesn't recursively | |
1724 | * output to the same mirrors. */ | |
3d6151f3 | 1725 | ctx->mirrors |= dup_mirrors; |
faa624b4 BP |
1726 | |
1727 | /* Send the packet to the mirror. */ | |
ec7ceaed | 1728 | if (out) { |
84f0f298 RW |
1729 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
1730 | struct xbundle *out_xbundle = xbundle_lookup(xcfg, out); | |
46c88433 EJ |
1731 | if (out_xbundle) { |
1732 | output_normal(ctx, out_xbundle, vlan); | |
1733 | } | |
ec7ceaed | 1734 | } else if (vlan != out_vlan |
7efbc3b7 | 1735 | && !eth_addr_is_reserved(ctx->xin->flow.dl_dst)) { |
46c88433 | 1736 | struct xbundle *xbundle; |
9583bc14 | 1737 | |
46c88433 EJ |
1738 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
1739 | if (xbundle_includes_vlan(xbundle, out_vlan) | |
1740 | && !xbundle_mirror_out(xbridge, xbundle)) { | |
1741 | output_normal(ctx, xbundle, out_vlan); | |
9583bc14 EJ |
1742 | } |
1743 | } | |
1744 | } | |
faa624b4 BP |
1745 | |
1746 | /* output_normal() could have recursively output (to different | |
1747 | * mirrors), so make sure that we don't send duplicates. */ | |
1748 | mirrors &= ~ctx->mirrors; | |
9583bc14 EJ |
1749 | } |
1750 | } | |
1751 | ||
7efbc3b7 BP |
1752 | static void |
1753 | mirror_ingress_packet(struct xlate_ctx *ctx) | |
1754 | { | |
1755 | if (mbridge_has_mirrors(ctx->xbridge->mbridge)) { | |
1756 | bool warn = ctx->xin->packet != NULL; | |
1757 | struct xbundle *xbundle = lookup_input_bundle( | |
1758 | ctx->xbridge, ctx->xin->flow.in_port.ofp_port, warn, NULL); | |
1759 | if (xbundle) { | |
1760 | mirror_packet(ctx, xbundle, | |
1761 | xbundle_mirror_src(ctx->xbridge, xbundle)); | |
1762 | } | |
1763 | } | |
1764 | } | |
1765 | ||
9583bc14 | 1766 | /* Given 'vid', the VID obtained from the 802.1Q header that was received as |
46c88433 | 1767 | * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle', |
9583bc14 EJ |
1768 | * the bundle on which the packet was received, returns the VLAN to which the |
1769 | * packet belongs. | |
1770 | * | |
1771 | * Both 'vid' and the return value are in the range 0...4095. */ | |
1772 | static uint16_t | |
46c88433 | 1773 | input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid) |
9583bc14 | 1774 | { |
46c88433 | 1775 | switch (in_xbundle->vlan_mode) { |
9583bc14 | 1776 | case PORT_VLAN_ACCESS: |
46c88433 | 1777 | return in_xbundle->vlan; |
9583bc14 EJ |
1778 | break; |
1779 | ||
1780 | case PORT_VLAN_TRUNK: | |
1781 | return vid; | |
1782 | ||
1783 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1784 | case PORT_VLAN_NATIVE_TAGGED: | |
46c88433 | 1785 | return vid ? vid : in_xbundle->vlan; |
9583bc14 EJ |
1786 | |
1787 | default: | |
428b2edd | 1788 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1789 | } |
1790 | } | |
1791 | ||
46c88433 | 1792 | /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'. |
9583bc14 EJ |
1793 | * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs |
1794 | * a warning. | |
1795 | * | |
1796 | * 'vid' should be the VID obtained from the 802.1Q header that was received as | |
1797 | * part of a packet (specify 0 if there was no 802.1Q header), in the range | |
1798 | * 0...4095. */ | |
1799 | static bool | |
46c88433 | 1800 | input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn) |
9583bc14 EJ |
1801 | { |
1802 | /* Allow any VID on the OFPP_NONE port. */ | |
46c88433 | 1803 | if (in_xbundle == &ofpp_none_bundle) { |
9583bc14 EJ |
1804 | return true; |
1805 | } | |
1806 | ||
46c88433 | 1807 | switch (in_xbundle->vlan_mode) { |
9583bc14 EJ |
1808 | case PORT_VLAN_ACCESS: |
1809 | if (vid) { | |
1810 | if (warn) { | |
1811 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1812 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" tagged " |
9583bc14 | 1813 | "packet received on port %s configured as VLAN " |
46c88433 EJ |
1814 | "%"PRIu16" access port", vid, in_xbundle->name, |
1815 | in_xbundle->vlan); | |
9583bc14 EJ |
1816 | } |
1817 | return false; | |
1818 | } | |
1819 | return true; | |
1820 | ||
1821 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1822 | case PORT_VLAN_NATIVE_TAGGED: | |
1823 | if (!vid) { | |
1824 | /* Port must always carry its native VLAN. */ | |
1825 | return true; | |
1826 | } | |
1827 | /* Fall through. */ | |
1828 | case PORT_VLAN_TRUNK: | |
46c88433 | 1829 | if (!xbundle_includes_vlan(in_xbundle, vid)) { |
9583bc14 EJ |
1830 | if (warn) { |
1831 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1832 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" packet " |
9583bc14 | 1833 | "received on port %s not configured for trunking " |
46c88433 | 1834 | "VLAN %"PRIu16, vid, in_xbundle->name, vid); |
9583bc14 EJ |
1835 | } |
1836 | return false; | |
1837 | } | |
1838 | return true; | |
1839 | ||
1840 | default: | |
428b2edd | 1841 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1842 | } |
1843 | ||
1844 | } | |
1845 | ||
1846 | /* Given 'vlan', the VLAN that a packet belongs to, and | |
46c88433 | 1847 | * 'out_xbundle', a bundle on which the packet is to be output, returns the VID |
9583bc14 EJ |
1848 | * that should be included in the 802.1Q header. (If the return value is 0, |
1849 | * then the 802.1Q header should only be included in the packet if there is a | |
1850 | * nonzero PCP.) | |
1851 | * | |
1852 | * Both 'vlan' and the return value are in the range 0...4095. */ | |
1853 | static uint16_t | |
46c88433 | 1854 | output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan) |
9583bc14 | 1855 | { |
46c88433 | 1856 | switch (out_xbundle->vlan_mode) { |
9583bc14 EJ |
1857 | case PORT_VLAN_ACCESS: |
1858 | return 0; | |
1859 | ||
1860 | case PORT_VLAN_TRUNK: | |
1861 | case PORT_VLAN_NATIVE_TAGGED: | |
1862 | return vlan; | |
1863 | ||
1864 | case PORT_VLAN_NATIVE_UNTAGGED: | |
46c88433 | 1865 | return vlan == out_xbundle->vlan ? 0 : vlan; |
9583bc14 EJ |
1866 | |
1867 | default: | |
428b2edd | 1868 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1869 | } |
1870 | } | |
1871 | ||
1872 | static void | |
46c88433 | 1873 | output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, |
9583bc14 EJ |
1874 | uint16_t vlan) |
1875 | { | |
33bf9176 | 1876 | ovs_be16 *flow_tci = &ctx->xin->flow.vlan_tci; |
9583bc14 EJ |
1877 | uint16_t vid; |
1878 | ovs_be16 tci, old_tci; | |
46c88433 | 1879 | struct xport *xport; |
e93ef1c7 JR |
1880 | struct xlate_bond_recirc xr; |
1881 | bool use_recirc = false; | |
9583bc14 | 1882 | |
46c88433 | 1883 | vid = output_vlan_to_vid(out_xbundle, vlan); |
417e7e66 | 1884 | if (ovs_list_is_empty(&out_xbundle->xports)) { |
46c88433 EJ |
1885 | /* Partially configured bundle with no slaves. Drop the packet. */ |
1886 | return; | |
1887 | } else if (!out_xbundle->bond) { | |
417e7e66 | 1888 | xport = CONTAINER_OF(ovs_list_front(&out_xbundle->xports), struct xport, |
46c88433 | 1889 | bundle_node); |
9583bc14 | 1890 | } else { |
84f0f298 | 1891 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
49a73e0c | 1892 | struct flow_wildcards *wc = ctx->wc; |
84f0f298 | 1893 | struct ofport_dpif *ofport; |
adcf00ba | 1894 | |
2494ccd7 | 1895 | if (ctx->xbridge->support.odp.recirc) { |
e93ef1c7 JR |
1896 | use_recirc = bond_may_recirc( |
1897 | out_xbundle->bond, &xr.recirc_id, &xr.hash_basis); | |
adcf00ba | 1898 | |
e93ef1c7 | 1899 | if (use_recirc) { |
adcf00ba | 1900 | /* Only TCP mode uses recirculation. */ |
e93ef1c7 | 1901 | xr.hash_alg = OVS_HASH_ALG_L4; |
adcf00ba | 1902 | bond_update_post_recirc_rules(out_xbundle->bond, false); |
54ecb5a2 AZ |
1903 | |
1904 | /* Recirculation does not require unmasking hash fields. */ | |
1905 | wc = NULL; | |
adcf00ba AZ |
1906 | } |
1907 | } | |
46c88433 | 1908 | |
54ecb5a2 AZ |
1909 | ofport = bond_choose_output_slave(out_xbundle->bond, |
1910 | &ctx->xin->flow, wc, vid); | |
84f0f298 | 1911 | xport = xport_lookup(xcfg, ofport); |
46c88433 EJ |
1912 | |
1913 | if (!xport) { | |
9583bc14 EJ |
1914 | /* No slaves enabled, so drop packet. */ |
1915 | return; | |
1916 | } | |
d6fc5f57 | 1917 | |
e93ef1c7 | 1918 | /* If use_recirc is set, the main thread will handle stats |
b256dc52 | 1919 | * accounting for this bond. */ |
e93ef1c7 | 1920 | if (!use_recirc) { |
b256dc52 JS |
1921 | if (ctx->xin->resubmit_stats) { |
1922 | bond_account(out_xbundle->bond, &ctx->xin->flow, vid, | |
1923 | ctx->xin->resubmit_stats->n_bytes); | |
1924 | } | |
1925 | if (ctx->xin->xcache) { | |
1926 | struct xc_entry *entry; | |
1927 | struct flow *flow; | |
1928 | ||
1929 | flow = &ctx->xin->flow; | |
1930 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND); | |
1931 | entry->u.bond.bond = bond_ref(out_xbundle->bond); | |
1932 | entry->u.bond.flow = xmemdup(flow, sizeof *flow); | |
1933 | entry->u.bond.vid = vid; | |
1934 | } | |
d6fc5f57 | 1935 | } |
9583bc14 EJ |
1936 | } |
1937 | ||
33bf9176 | 1938 | old_tci = *flow_tci; |
9583bc14 | 1939 | tci = htons(vid); |
46c88433 | 1940 | if (tci || out_xbundle->use_priority_tags) { |
33bf9176 | 1941 | tci |= *flow_tci & htons(VLAN_PCP_MASK); |
9583bc14 EJ |
1942 | if (tci) { |
1943 | tci |= htons(VLAN_CFI); | |
1944 | } | |
1945 | } | |
33bf9176 | 1946 | *flow_tci = tci; |
9583bc14 | 1947 | |
e93ef1c7 | 1948 | compose_output_action(ctx, xport->ofp_port, use_recirc ? &xr : NULL); |
33bf9176 | 1949 | *flow_tci = old_tci; |
9583bc14 EJ |
1950 | } |
1951 | ||
1952 | /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after | |
1953 | * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to | |
1954 | * indicate this; newer upstream kernels use gratuitous ARP requests. */ | |
1955 | static bool | |
1956 | is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) | |
1957 | { | |
1958 | if (flow->dl_type != htons(ETH_TYPE_ARP)) { | |
1959 | return false; | |
1960 | } | |
1961 | ||
1962 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1963 | if (!eth_addr_is_broadcast(flow->dl_dst)) { | |
1964 | return false; | |
1965 | } | |
1966 | ||
1967 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
1968 | if (flow->nw_proto == ARP_OP_REPLY) { | |
1969 | return true; | |
1970 | } else if (flow->nw_proto == ARP_OP_REQUEST) { | |
1971 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
1972 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
1973 | ||
1974 | return flow->nw_src == flow->nw_dst; | |
1975 | } else { | |
1976 | return false; | |
1977 | } | |
1978 | } | |
1979 | ||
ff69c24a FL |
1980 | /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or |
1981 | * dropped. Returns true if they may be forwarded, false if they should be | |
1982 | * dropped. | |
1983 | * | |
1984 | * 'in_port' must be the xport that corresponds to flow->in_port. | |
1985 | * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). | |
1986 | * | |
1987 | * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as | |
1988 | * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as | |
1989 | * checked by input_vid_is_valid(). | |
1990 | * | |
1991 | * May also add tags to '*tags', although the current implementation only does | |
1992 | * so in one special case. | |
1993 | */ | |
1994 | static bool | |
1995 | is_admissible(struct xlate_ctx *ctx, struct xport *in_port, | |
1996 | uint16_t vlan) | |
1997 | { | |
1998 | struct xbundle *in_xbundle = in_port->xbundle; | |
1999 | const struct xbridge *xbridge = ctx->xbridge; | |
2000 | struct flow *flow = &ctx->xin->flow; | |
2001 | ||
2002 | /* Drop frames for reserved multicast addresses | |
2003 | * only if forward_bpdu option is absent. */ | |
2004 | if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { | |
2005 | xlate_report(ctx, "packet has reserved destination MAC, dropping"); | |
2006 | return false; | |
2007 | } | |
2008 | ||
2009 | if (in_xbundle->bond) { | |
2010 | struct mac_entry *mac; | |
2011 | ||
2012 | switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, | |
2013 | flow->dl_dst)) { | |
2014 | case BV_ACCEPT: | |
2015 | break; | |
2016 | ||
2017 | case BV_DROP: | |
2018 | xlate_report(ctx, "bonding refused admissibility, dropping"); | |
2019 | return false; | |
2020 | ||
2021 | case BV_DROP_IF_MOVED: | |
2022 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
2023 | mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan); | |
9d078ec2 BP |
2024 | if (mac |
2025 | && mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle | |
49a73e0c | 2026 | && (!is_gratuitous_arp(flow, ctx->wc) |
9d078ec2 | 2027 | || mac_entry_is_grat_arp_locked(mac))) { |
ff69c24a FL |
2028 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
2029 | xlate_report(ctx, "SLB bond thinks this packet looped back, " | |
2030 | "dropping"); | |
2031 | return false; | |
2032 | } | |
2033 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
2034 | break; | |
2035 | } | |
2036 | } | |
2037 | ||
2038 | return true; | |
2039 | } | |
2040 | ||
ee047520 BP |
2041 | /* Checks whether a MAC learning update is necessary for MAC learning table |
2042 | * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in | |
2043 | * 'vlan'. | |
2044 | * | |
2045 | * Most packets processed through the MAC learning table do not actually | |
2046 | * change it in any way. This function requires only a read lock on the MAC | |
2047 | * learning table, so it is much cheaper in this common case. | |
2048 | * | |
2049 | * Keep the code here synchronized with that in update_learning_table__() | |
2050 | * below. */ | |
2051 | static bool | |
2052 | is_mac_learning_update_needed(const struct mac_learning *ml, | |
2053 | const struct flow *flow, | |
2054 | struct flow_wildcards *wc, | |
2055 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 2056 | OVS_REQ_RDLOCK(ml->rwlock) |
9583bc14 EJ |
2057 | { |
2058 | struct mac_entry *mac; | |
2059 | ||
ee047520 BP |
2060 | if (!mac_learning_may_learn(ml, flow->dl_src, vlan)) { |
2061 | return false; | |
2062 | } | |
2063 | ||
2064 | mac = mac_learning_lookup(ml, flow->dl_src, vlan); | |
2065 | if (!mac || mac_entry_age(ml, mac)) { | |
2066 | return true; | |
9583bc14 EJ |
2067 | } |
2068 | ||
ee047520 BP |
2069 | if (is_gratuitous_arp(flow, wc)) { |
2070 | /* We don't want to learn from gratuitous ARP packets that are | |
2071 | * reflected back over bond slaves so we lock the learning table. */ | |
2072 | if (!in_xbundle->bond) { | |
2073 | return true; | |
2074 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
2075 | return false; | |
2076 | } | |
2077 | } | |
2078 | ||
9d078ec2 | 2079 | return mac_entry_get_port(ml, mac) != in_xbundle->ofbundle; |
ee047520 BP |
2080 | } |
2081 | ||
2082 | ||
2083 | /* Updates MAC learning table 'ml' given that a packet matching 'flow' was | |
2084 | * received on 'in_xbundle' in 'vlan'. | |
2085 | * | |
2086 | * This code repeats all the checks in is_mac_learning_update_needed() because | |
2087 | * the lock was released between there and here and thus the MAC learning state | |
2088 | * could have changed. | |
2089 | * | |
2090 | * Keep the code here synchronized with that in is_mac_learning_update_needed() | |
2091 | * above. */ | |
2092 | static void | |
2093 | update_learning_table__(const struct xbridge *xbridge, | |
2094 | const struct flow *flow, struct flow_wildcards *wc, | |
2095 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 2096 | OVS_REQ_WRLOCK(xbridge->ml->rwlock) |
ee047520 BP |
2097 | { |
2098 | struct mac_entry *mac; | |
2099 | ||
46c88433 | 2100 | if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) { |
ee047520 | 2101 | return; |
9583bc14 EJ |
2102 | } |
2103 | ||
46c88433 | 2104 | mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan); |
9583bc14 EJ |
2105 | if (is_gratuitous_arp(flow, wc)) { |
2106 | /* We don't want to learn from gratuitous ARP packets that are | |
2107 | * reflected back over bond slaves so we lock the learning table. */ | |
46c88433 | 2108 | if (!in_xbundle->bond) { |
9583bc14 EJ |
2109 | mac_entry_set_grat_arp_lock(mac); |
2110 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
ee047520 | 2111 | return; |
9583bc14 EJ |
2112 | } |
2113 | } | |
2114 | ||
9d078ec2 | 2115 | if (mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle) { |
9583bc14 EJ |
2116 | /* The log messages here could actually be useful in debugging, |
2117 | * so keep the rate limit relatively high. */ | |
2118 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
ee047520 | 2119 | |
9583bc14 EJ |
2120 | VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is " |
2121 | "on port %s in VLAN %d", | |
46c88433 EJ |
2122 | xbridge->name, ETH_ADDR_ARGS(flow->dl_src), |
2123 | in_xbundle->name, vlan); | |
9583bc14 | 2124 | |
9d078ec2 | 2125 | mac_entry_set_port(xbridge->ml, mac, in_xbundle->ofbundle); |
9583bc14 | 2126 | } |
ee047520 BP |
2127 | } |
2128 | ||
2129 | static void | |
2130 | update_learning_table(const struct xbridge *xbridge, | |
2131 | const struct flow *flow, struct flow_wildcards *wc, | |
2132 | int vlan, struct xbundle *in_xbundle) | |
2133 | { | |
2134 | bool need_update; | |
2135 | ||
2136 | /* Don't learn the OFPP_NONE port. */ | |
2137 | if (in_xbundle == &ofpp_none_bundle) { | |
2138 | return; | |
2139 | } | |
2140 | ||
2141 | /* First try the common case: no change to MAC learning table. */ | |
2142 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
2143 | need_update = is_mac_learning_update_needed(xbridge->ml, flow, wc, vlan, | |
2144 | in_xbundle); | |
509c0149 | 2145 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
ee047520 BP |
2146 | |
2147 | if (need_update) { | |
2148 | /* Slow path: MAC learning table might need an update. */ | |
2149 | ovs_rwlock_wrlock(&xbridge->ml->rwlock); | |
2150 | update_learning_table__(xbridge, flow, wc, vlan, in_xbundle); | |
2151 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
2152 | } | |
9583bc14 EJ |
2153 | } |
2154 | ||
86e2dcdd FL |
2155 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2156 | * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */ | |
2157 | static void | |
06994f87 TLSC |
2158 | update_mcast_snooping_table4__(const struct xbridge *xbridge, |
2159 | const struct flow *flow, | |
2160 | struct mcast_snooping *ms, int vlan, | |
2161 | struct xbundle *in_xbundle, | |
2162 | const struct dp_packet *packet) | |
86e2dcdd FL |
2163 | OVS_REQ_WRLOCK(ms->rwlock) |
2164 | { | |
2165 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
e3102e42 | 2166 | int count; |
06994f87 | 2167 | ovs_be32 ip4 = flow->igmp_group_ip4; |
86e2dcdd FL |
2168 | |
2169 | switch (ntohs(flow->tp_src)) { | |
2170 | case IGMP_HOST_MEMBERSHIP_REPORT: | |
2171 | case IGMPV2_HOST_MEMBERSHIP_REPORT: | |
964a4d5f | 2172 | if (mcast_snooping_add_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
86e2dcdd FL |
2173 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping learned that " |
2174 | IP_FMT" is on port %s in VLAN %d", | |
2175 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
2176 | } | |
2177 | break; | |
2178 | case IGMP_HOST_LEAVE_MESSAGE: | |
964a4d5f | 2179 | if (mcast_snooping_leave_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
86e2dcdd FL |
2180 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping leaving " |
2181 | IP_FMT" is on port %s in VLAN %d", | |
2182 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
2183 | } | |
2184 | break; | |
2185 | case IGMP_HOST_MEMBERSHIP_QUERY: | |
2186 | if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan, | |
2187 | in_xbundle->ofbundle)) { | |
2188 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query from " | |
2189 | IP_FMT" is on port %s in VLAN %d", | |
2190 | xbridge->name, IP_ARGS(flow->nw_src), | |
2191 | in_xbundle->name, vlan); | |
2192 | } | |
2193 | break; | |
e3102e42 TLSC |
2194 | case IGMPV3_HOST_MEMBERSHIP_REPORT: |
2195 | if ((count = mcast_snooping_add_report(ms, packet, vlan, | |
2196 | in_xbundle->ofbundle))) { | |
2197 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping processed %d " | |
2198 | "addresses on port %s in VLAN %d", | |
2199 | xbridge->name, count, in_xbundle->name, vlan); | |
2200 | } | |
2201 | break; | |
86e2dcdd FL |
2202 | } |
2203 | } | |
2204 | ||
06994f87 TLSC |
2205 | static void |
2206 | update_mcast_snooping_table6__(const struct xbridge *xbridge, | |
2207 | const struct flow *flow, | |
2208 | struct mcast_snooping *ms, int vlan, | |
2209 | struct xbundle *in_xbundle, | |
2210 | const struct dp_packet *packet) | |
2211 | OVS_REQ_WRLOCK(ms->rwlock) | |
2212 | { | |
2213 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
2214 | int count; | |
2215 | ||
2216 | switch (ntohs(flow->tp_src)) { | |
2217 | case MLD_QUERY: | |
2218 | if (!ipv6_addr_equals(&flow->ipv6_src, &in6addr_any) | |
2219 | && mcast_snooping_add_mrouter(ms, vlan, in_xbundle->ofbundle)) { | |
2220 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query on port %s" | |
2221 | "in VLAN %d", | |
2222 | xbridge->name, in_xbundle->name, vlan); | |
2223 | } | |
2224 | break; | |
2225 | case MLD_REPORT: | |
2226 | case MLD_DONE: | |
2227 | case MLD2_REPORT: | |
2228 | count = mcast_snooping_add_mld(ms, packet, vlan, in_xbundle->ofbundle); | |
2229 | if (count) { | |
2230 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping processed %d " | |
2231 | "addresses on port %s in VLAN %d", | |
2232 | xbridge->name, count, in_xbundle->name, vlan); | |
2233 | } | |
2234 | break; | |
2235 | } | |
2236 | } | |
2237 | ||
86e2dcdd FL |
2238 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2239 | * was received on 'in_xbundle' in 'vlan'. */ | |
2240 | static void | |
2241 | update_mcast_snooping_table(const struct xbridge *xbridge, | |
2242 | const struct flow *flow, int vlan, | |
e3102e42 TLSC |
2243 | struct xbundle *in_xbundle, |
2244 | const struct dp_packet *packet) | |
86e2dcdd FL |
2245 | { |
2246 | struct mcast_snooping *ms = xbridge->ms; | |
2247 | struct xlate_cfg *xcfg; | |
2248 | struct xbundle *mcast_xbundle; | |
f4ae6e23 | 2249 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2250 | |
2251 | /* Don't learn the OFPP_NONE port. */ | |
2252 | if (in_xbundle == &ofpp_none_bundle) { | |
2253 | return; | |
2254 | } | |
2255 | ||
2256 | /* Don't learn from flood ports */ | |
2257 | mcast_xbundle = NULL; | |
2258 | ovs_rwlock_wrlock(&ms->rwlock); | |
2259 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2260 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2261 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2262 | if (mcast_xbundle == in_xbundle) { | |
2263 | break; | |
2264 | } | |
2265 | } | |
2266 | ||
2267 | if (!mcast_xbundle || mcast_xbundle != in_xbundle) { | |
06994f87 TLSC |
2268 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2269 | update_mcast_snooping_table4__(xbridge, flow, ms, vlan, | |
2270 | in_xbundle, packet); | |
2271 | } else { | |
2272 | update_mcast_snooping_table6__(xbridge, flow, ms, vlan, | |
2273 | in_xbundle, packet); | |
2274 | } | |
86e2dcdd FL |
2275 | } |
2276 | ovs_rwlock_unlock(&ms->rwlock); | |
2277 | } | |
2278 | ||
2279 | /* send the packet to ports having the multicast group learned */ | |
2280 | static void | |
2281 | xlate_normal_mcast_send_group(struct xlate_ctx *ctx, | |
2282 | struct mcast_snooping *ms OVS_UNUSED, | |
2283 | struct mcast_group *grp, | |
2284 | struct xbundle *in_xbundle, uint16_t vlan) | |
2285 | OVS_REQ_RDLOCK(ms->rwlock) | |
2286 | { | |
2287 | struct xlate_cfg *xcfg; | |
2288 | struct mcast_group_bundle *b; | |
2289 | struct xbundle *mcast_xbundle; | |
2290 | ||
2291 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2292 | LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) { | |
2293 | mcast_xbundle = xbundle_lookup(xcfg, b->port); | |
2294 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2295 | xlate_report(ctx, "forwarding to mcast group port"); | |
2296 | output_normal(ctx, mcast_xbundle, vlan); | |
2297 | } else if (!mcast_xbundle) { | |
2298 | xlate_report(ctx, "mcast group port is unknown, dropping"); | |
2299 | } else { | |
2300 | xlate_report(ctx, "mcast group port is input port, dropping"); | |
2301 | } | |
2302 | } | |
2303 | } | |
2304 | ||
2305 | /* send the packet to ports connected to multicast routers */ | |
2306 | static void | |
2307 | xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx, | |
2308 | struct mcast_snooping *ms, | |
2309 | struct xbundle *in_xbundle, uint16_t vlan) | |
2310 | OVS_REQ_RDLOCK(ms->rwlock) | |
2311 | { | |
2312 | struct xlate_cfg *xcfg; | |
2313 | struct mcast_mrouter_bundle *mrouter; | |
2314 | struct xbundle *mcast_xbundle; | |
2315 | ||
2316 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2317 | LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) { | |
2318 | mcast_xbundle = xbundle_lookup(xcfg, mrouter->port); | |
2319 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2320 | xlate_report(ctx, "forwarding to mcast router port"); | |
2321 | output_normal(ctx, mcast_xbundle, vlan); | |
2322 | } else if (!mcast_xbundle) { | |
2323 | xlate_report(ctx, "mcast router port is unknown, dropping"); | |
2324 | } else { | |
2325 | xlate_report(ctx, "mcast router port is input port, dropping"); | |
2326 | } | |
2327 | } | |
2328 | } | |
2329 | ||
2330 | /* send the packet to ports flagged to be flooded */ | |
2331 | static void | |
2332 | xlate_normal_mcast_send_fports(struct xlate_ctx *ctx, | |
2333 | struct mcast_snooping *ms, | |
2334 | struct xbundle *in_xbundle, uint16_t vlan) | |
2335 | OVS_REQ_RDLOCK(ms->rwlock) | |
2336 | { | |
2337 | struct xlate_cfg *xcfg; | |
f4ae6e23 | 2338 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2339 | struct xbundle *mcast_xbundle; |
2340 | ||
2341 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2342 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2343 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2344 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2345 | xlate_report(ctx, "forwarding to mcast flood port"); | |
2346 | output_normal(ctx, mcast_xbundle, vlan); | |
2347 | } else if (!mcast_xbundle) { | |
2348 | xlate_report(ctx, "mcast flood port is unknown, dropping"); | |
2349 | } else { | |
2350 | xlate_report(ctx, "mcast flood port is input port, dropping"); | |
2351 | } | |
2352 | } | |
2353 | } | |
2354 | ||
8e04a33f FL |
2355 | /* forward the Reports to configured ports */ |
2356 | static void | |
2357 | xlate_normal_mcast_send_rports(struct xlate_ctx *ctx, | |
2358 | struct mcast_snooping *ms, | |
2359 | struct xbundle *in_xbundle, uint16_t vlan) | |
2360 | OVS_REQ_RDLOCK(ms->rwlock) | |
2361 | { | |
2362 | struct xlate_cfg *xcfg; | |
2363 | struct mcast_port_bundle *rport; | |
2364 | struct xbundle *mcast_xbundle; | |
2365 | ||
2366 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2367 | LIST_FOR_EACH(rport, node, &ms->rport_list) { | |
2368 | mcast_xbundle = xbundle_lookup(xcfg, rport->port); | |
2369 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2370 | xlate_report(ctx, "forwarding Report to mcast flagged port"); | |
2371 | output_normal(ctx, mcast_xbundle, vlan); | |
2372 | } else if (!mcast_xbundle) { | |
2373 | xlate_report(ctx, "mcast port is unknown, dropping the Report"); | |
2374 | } else { | |
2375 | xlate_report(ctx, "mcast port is input port, dropping the Report"); | |
2376 | } | |
2377 | } | |
2378 | } | |
2379 | ||
682800a4 FL |
2380 | static void |
2381 | xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, | |
2382 | uint16_t vlan) | |
2383 | { | |
2384 | struct xbundle *xbundle; | |
2385 | ||
2386 | LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { | |
2387 | if (xbundle != in_xbundle | |
2388 | && xbundle_includes_vlan(xbundle, vlan) | |
2389 | && xbundle->floodable | |
2390 | && !xbundle_mirror_out(ctx->xbridge, xbundle)) { | |
2391 | output_normal(ctx, xbundle, vlan); | |
2392 | } | |
2393 | } | |
2031ef97 | 2394 | ctx->nf_output_iface = NF_OUT_FLOOD; |
682800a4 FL |
2395 | } |
2396 | ||
a75636c8 BP |
2397 | static bool |
2398 | is_ip_local_multicast(const struct flow *flow, struct flow_wildcards *wc) | |
2399 | { | |
2400 | if (flow->dl_type == htons(ETH_TYPE_IP)) { | |
2401 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
2402 | return ip_is_local_multicast(flow->nw_dst); | |
2403 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
2404 | memset(&wc->masks.ipv6_dst, 0xff, sizeof wc->masks.ipv6_dst); | |
2405 | return ipv6_is_all_hosts(&flow->ipv6_dst); | |
2406 | } else { | |
2407 | return false; | |
2408 | } | |
2409 | } | |
2410 | ||
9583bc14 EJ |
2411 | static void |
2412 | xlate_normal(struct xlate_ctx *ctx) | |
2413 | { | |
49a73e0c | 2414 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 2415 | struct flow *flow = &ctx->xin->flow; |
46c88433 EJ |
2416 | struct xbundle *in_xbundle; |
2417 | struct xport *in_port; | |
9583bc14 | 2418 | struct mac_entry *mac; |
d6d5bbc9 | 2419 | void *mac_port; |
9583bc14 EJ |
2420 | uint16_t vlan; |
2421 | uint16_t vid; | |
2422 | ||
33bf9176 BP |
2423 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
2424 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1dd35f8a | 2425 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
9583bc14 | 2426 | |
46c88433 EJ |
2427 | in_xbundle = lookup_input_bundle(ctx->xbridge, flow->in_port.ofp_port, |
2428 | ctx->xin->packet != NULL, &in_port); | |
2429 | if (!in_xbundle) { | |
9583bc14 EJ |
2430 | xlate_report(ctx, "no input bundle, dropping"); |
2431 | return; | |
2432 | } | |
2433 | ||
2434 | /* Drop malformed frames. */ | |
33bf9176 BP |
2435 | if (flow->dl_type == htons(ETH_TYPE_VLAN) && |
2436 | !(flow->vlan_tci & htons(VLAN_CFI))) { | |
9583bc14 EJ |
2437 | if (ctx->xin->packet != NULL) { |
2438 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2439 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet with partial " | |
2440 | "VLAN tag received on port %s", | |
46c88433 | 2441 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2442 | } |
2443 | xlate_report(ctx, "partial VLAN tag, dropping"); | |
2444 | return; | |
2445 | } | |
2446 | ||
2447 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 2448 | if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) { |
9583bc14 EJ |
2449 | if (ctx->xin->packet != NULL) { |
2450 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2451 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
2452 | "%s, which is reserved exclusively for mirroring", | |
46c88433 | 2453 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2454 | } |
2455 | xlate_report(ctx, "input port is mirror output port, dropping"); | |
2456 | return; | |
2457 | } | |
2458 | ||
2459 | /* Check VLAN. */ | |
33bf9176 | 2460 | vid = vlan_tci_to_vid(flow->vlan_tci); |
46c88433 | 2461 | if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { |
9583bc14 EJ |
2462 | xlate_report(ctx, "disallowed VLAN VID for this input port, dropping"); |
2463 | return; | |
2464 | } | |
46c88433 | 2465 | vlan = input_vid_to_vlan(in_xbundle, vid); |
9583bc14 EJ |
2466 | |
2467 | /* Check other admissibility requirements. */ | |
2468 | if (in_port && !is_admissible(ctx, in_port, vlan)) { | |
2469 | return; | |
2470 | } | |
2471 | ||
2472 | /* Learn source MAC. */ | |
2473 | if (ctx->xin->may_learn) { | |
46c88433 | 2474 | update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle); |
9583bc14 | 2475 | } |
b256dc52 JS |
2476 | if (ctx->xin->xcache) { |
2477 | struct xc_entry *entry; | |
2478 | ||
2479 | /* Save enough info to update mac learning table later. */ | |
2480 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL); | |
9edf6b48 | 2481 | entry->u.normal.ofproto = ctx->xbridge->ofproto; |
b256dc52 JS |
2482 | entry->u.normal.flow = xmemdup(flow, sizeof *flow); |
2483 | entry->u.normal.vlan = vlan; | |
2484 | } | |
9583bc14 EJ |
2485 | |
2486 | /* Determine output bundle. */ | |
86e2dcdd FL |
2487 | if (mcast_snooping_enabled(ctx->xbridge->ms) |
2488 | && !eth_addr_is_broadcast(flow->dl_dst) | |
2489 | && eth_addr_is_multicast(flow->dl_dst) | |
06994f87 | 2490 | && is_ip_any(flow)) { |
86e2dcdd | 2491 | struct mcast_snooping *ms = ctx->xbridge->ms; |
06994f87 | 2492 | struct mcast_group *grp = NULL; |
86e2dcdd | 2493 | |
a75636c8 BP |
2494 | if (is_igmp(flow, wc)) { |
2495 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
d29f137b TLSC |
2496 | if (mcast_snooping_is_membership(flow->tp_src) || |
2497 | mcast_snooping_is_query(flow->tp_src)) { | |
f7c5f6aa | 2498 | if (ctx->xin->may_learn && ctx->xin->packet) { |
86e2dcdd | 2499 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, |
e3102e42 | 2500 | in_xbundle, ctx->xin->packet); |
d29f137b TLSC |
2501 | } |
2502 | /* | |
2503 | * IGMP packets need to take the slow path, in order to be | |
2504 | * processed for mdb updates. That will prevent expires | |
2505 | * firing off even after hosts have sent reports. | |
2506 | */ | |
2507 | ctx->xout->slow |= SLOW_ACTION; | |
86e2dcdd | 2508 | } |
d6d5bbc9 | 2509 | |
86e2dcdd FL |
2510 | if (mcast_snooping_is_membership(flow->tp_src)) { |
2511 | ovs_rwlock_rdlock(&ms->rwlock); | |
2512 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
8e04a33f FL |
2513 | /* RFC4541: section 2.1.1, item 1: A snooping switch should |
2514 | * forward IGMP Membership Reports only to those ports where | |
2515 | * multicast routers are attached. Alternatively stated: a | |
2516 | * snooping switch should not forward IGMP Membership Reports | |
2517 | * to ports on which only hosts are attached. | |
2518 | * An administrative control may be provided to override this | |
2519 | * restriction, allowing the report messages to be flooded to | |
2520 | * other ports. */ | |
2521 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
86e2dcdd FL |
2522 | ovs_rwlock_unlock(&ms->rwlock); |
2523 | } else { | |
2524 | xlate_report(ctx, "multicast traffic, flooding"); | |
2525 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2526 | } | |
2527 | return; | |
a75636c8 | 2528 | } else if (is_mld(flow, wc)) { |
06994f87 | 2529 | ctx->xout->slow |= SLOW_ACTION; |
f7c5f6aa | 2530 | if (ctx->xin->may_learn && ctx->xin->packet) { |
06994f87 TLSC |
2531 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, |
2532 | in_xbundle, ctx->xin->packet); | |
2533 | } | |
a75636c8 | 2534 | if (is_mld_report(flow, wc)) { |
06994f87 TLSC |
2535 | ovs_rwlock_rdlock(&ms->rwlock); |
2536 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2537 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
2538 | ovs_rwlock_unlock(&ms->rwlock); | |
2539 | } else { | |
2540 | xlate_report(ctx, "MLD query, flooding"); | |
2541 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2542 | } | |
86e2dcdd | 2543 | } else { |
a75636c8 | 2544 | if (is_ip_local_multicast(flow, wc)) { |
86e2dcdd FL |
2545 | /* RFC4541: section 2.1.2, item 2: Packets with a dst IP |
2546 | * address in the 224.0.0.x range which are not IGMP must | |
2547 | * be forwarded on all ports */ | |
2548 | xlate_report(ctx, "RFC4541: section 2.1.2, item 2, flooding"); | |
2549 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2550 | return; | |
2551 | } | |
2552 | } | |
2553 | ||
2554 | /* forwarding to group base ports */ | |
2555 | ovs_rwlock_rdlock(&ms->rwlock); | |
06994f87 TLSC |
2556 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2557 | grp = mcast_snooping_lookup4(ms, flow->nw_dst, vlan); | |
2558 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
2559 | grp = mcast_snooping_lookup(ms, &flow->ipv6_dst, vlan); | |
2560 | } | |
86e2dcdd FL |
2561 | if (grp) { |
2562 | xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, vlan); | |
2563 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2564 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
9583bc14 | 2565 | } else { |
86e2dcdd FL |
2566 | if (mcast_snooping_flood_unreg(ms)) { |
2567 | xlate_report(ctx, "unregistered multicast, flooding"); | |
2568 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2569 | } else { | |
2570 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2571 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2572 | } | |
9583bc14 | 2573 | } |
86e2dcdd | 2574 | ovs_rwlock_unlock(&ms->rwlock); |
9583bc14 | 2575 | } else { |
86e2dcdd FL |
2576 | ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); |
2577 | mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); | |
9d078ec2 | 2578 | mac_port = mac ? mac_entry_get_port(ctx->xbridge->ml, mac) : NULL; |
86e2dcdd FL |
2579 | ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); |
2580 | ||
2581 | if (mac_port) { | |
2582 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2583 | struct xbundle *mac_xbundle = xbundle_lookup(xcfg, mac_port); | |
2584 | if (mac_xbundle && mac_xbundle != in_xbundle) { | |
2585 | xlate_report(ctx, "forwarding to learned port"); | |
2586 | output_normal(ctx, mac_xbundle, vlan); | |
2587 | } else if (!mac_xbundle) { | |
2588 | xlate_report(ctx, "learned port is unknown, dropping"); | |
2589 | } else { | |
2590 | xlate_report(ctx, "learned port is input port, dropping"); | |
2591 | } | |
2592 | } else { | |
2593 | xlate_report(ctx, "no learned MAC for destination, flooding"); | |
2594 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2595 | } | |
9583bc14 EJ |
2596 | } |
2597 | } | |
2598 | ||
a6092018 BP |
2599 | /* Appends a "sample" action for sFlow or IPFIX to 'ctx->odp_actions'. The |
2600 | * 'probability' is the number of packets out of UINT32_MAX to sample. The | |
2601 | * 'cookie' (of length 'cookie_size' bytes) is passed back in the callback for | |
2602 | * each sampled packet. 'tunnel_out_port', if not ODPP_NONE, is added as the | |
2603 | * OVS_USERSPACE_ATTR_EGRESS_TUN_PORT attribute. If 'include_actions', an | |
f69f713b BY |
2604 | * OVS_USERSPACE_ATTR_ACTIONS attribute is added. If 'emit_set_tunnel', |
2605 | * sample(sampling_port=1) would translate into datapath sample action | |
2606 | * set(tunnel(...)), sample(...) and it is used for sampling egress tunnel | |
2607 | * information. | |
9583bc14 EJ |
2608 | */ |
2609 | static size_t | |
a6092018 | 2610 | compose_sample_action(struct xlate_ctx *ctx, |
9583bc14 EJ |
2611 | const uint32_t probability, |
2612 | const union user_action_cookie *cookie, | |
8b7ea2d4 | 2613 | const size_t cookie_size, |
7321bda3 NM |
2614 | const odp_port_t tunnel_out_port, |
2615 | bool include_actions) | |
9583bc14 | 2616 | { |
a6092018 BP |
2617 | size_t sample_offset = nl_msg_start_nested(ctx->odp_actions, |
2618 | OVS_ACTION_ATTR_SAMPLE); | |
9583bc14 | 2619 | |
a6092018 | 2620 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability); |
9583bc14 | 2621 | |
a6092018 BP |
2622 | size_t actions_offset = nl_msg_start_nested(ctx->odp_actions, |
2623 | OVS_SAMPLE_ATTR_ACTIONS); | |
9583bc14 | 2624 | |
a6092018 BP |
2625 | odp_port_t odp_port = ofp_port_to_odp_port( |
2626 | ctx->xbridge, ctx->xin->flow.in_port.ofp_port); | |
2627 | uint32_t pid = dpif_port_get_pid(ctx->xbridge->dpif, odp_port, | |
2628 | flow_hash_5tuple(&ctx->xin->flow, 0)); | |
2629 | int cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, | |
2630 | tunnel_out_port, | |
2631 | include_actions, | |
2632 | ctx->odp_actions); | |
89a8a7f0 | 2633 | |
a6092018 BP |
2634 | nl_msg_end_nested(ctx->odp_actions, actions_offset); |
2635 | nl_msg_end_nested(ctx->odp_actions, sample_offset); | |
9583bc14 | 2636 | |
9583bc14 EJ |
2637 | return cookie_offset; |
2638 | } | |
2639 | ||
a6092018 BP |
2640 | /* If sFLow is not enabled, returns 0 without doing anything. |
2641 | * | |
2642 | * If sFlow is enabled, appends a template "sample" action to the ODP actions | |
2643 | * in 'ctx'. This action is a template because some of the information needed | |
2644 | * to fill it out is not available until flow translation is complete. In this | |
2645 | * case, this functions returns an offset, which is always nonzero, to pass | |
2646 | * later to fix_sflow_action() to fill in the rest of the template. */ | |
9583bc14 | 2647 | static size_t |
a6092018 | 2648 | compose_sflow_action(struct xlate_ctx *ctx) |
9583bc14 | 2649 | { |
a6092018 BP |
2650 | struct dpif_sflow *sflow = ctx->xbridge->sflow; |
2651 | if (!sflow || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { | |
9583bc14 EJ |
2652 | return 0; |
2653 | } | |
2654 | ||
a6092018 BP |
2655 | union user_action_cookie cookie = { .type = USER_ACTION_COOKIE_SFLOW }; |
2656 | return compose_sample_action(ctx, dpif_sflow_get_probability(sflow), | |
7321bda3 NM |
2657 | &cookie, sizeof cookie.sflow, ODPP_NONE, |
2658 | true); | |
9583bc14 EJ |
2659 | } |
2660 | ||
f69f713b BY |
2661 | /* If flow IPFIX is enabled, make sure IPFIX flow sample action |
2662 | * at egress point of tunnel port is just in front of corresponding | |
2663 | * output action. If bridge IPFIX is enabled, this appends an IPFIX | |
2664 | * sample action to 'ctx->odp_actions'. */ | |
9583bc14 | 2665 | static void |
a6092018 | 2666 | compose_ipfix_action(struct xlate_ctx *ctx, odp_port_t output_odp_port) |
9583bc14 | 2667 | { |
a6092018 | 2668 | struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; |
8b7ea2d4 | 2669 | odp_port_t tunnel_out_port = ODPP_NONE; |
9583bc14 | 2670 | |
a6092018 | 2671 | if (!ipfix || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
2672 | return; |
2673 | } | |
2674 | ||
8b7ea2d4 WZ |
2675 | /* For input case, output_odp_port is ODPP_NONE, which is an invalid port |
2676 | * number. */ | |
2677 | if (output_odp_port == ODPP_NONE && | |
a6092018 | 2678 | !dpif_ipfix_get_bridge_exporter_input_sampling(ipfix)) { |
8b7ea2d4 WZ |
2679 | return; |
2680 | } | |
2681 | ||
f69f713b | 2682 | /* For output case, output_odp_port is valid. */ |
8b7ea2d4 | 2683 | if (output_odp_port != ODPP_NONE) { |
a6092018 | 2684 | if (!dpif_ipfix_get_bridge_exporter_output_sampling(ipfix)) { |
8b7ea2d4 WZ |
2685 | return; |
2686 | } | |
2687 | /* If tunnel sampling is enabled, put an additional option attribute: | |
2688 | * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT | |
2689 | */ | |
a6092018 BP |
2690 | if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(ipfix) && |
2691 | dpif_ipfix_get_tunnel_port(ipfix, output_odp_port) ) { | |
8b7ea2d4 WZ |
2692 | tunnel_out_port = output_odp_port; |
2693 | } | |
2694 | } | |
2695 | ||
a6092018 BP |
2696 | union user_action_cookie cookie = { |
2697 | .ipfix = { | |
2698 | .type = USER_ACTION_COOKIE_IPFIX, | |
2699 | .output_odp_port = output_odp_port, | |
2700 | } | |
2701 | }; | |
2702 | compose_sample_action(ctx, | |
2703 | dpif_ipfix_get_bridge_exporter_probability(ipfix), | |
7321bda3 NM |
2704 | &cookie, sizeof cookie.ipfix, tunnel_out_port, |
2705 | false); | |
9583bc14 EJ |
2706 | } |
2707 | ||
a6092018 BP |
2708 | /* Fix "sample" action according to data collected while composing ODP actions, |
2709 | * as described in compose_sflow_action(). | |
2710 | * | |
2711 | * 'user_cookie_offset' must be the offset returned by add_sflow_action(). */ | |
9583bc14 | 2712 | static void |
a6092018 | 2713 | fix_sflow_action(struct xlate_ctx *ctx, unsigned int user_cookie_offset) |
9583bc14 EJ |
2714 | { |
2715 | const struct flow *base = &ctx->base_flow; | |
2716 | union user_action_cookie *cookie; | |
2717 | ||
a6092018 | 2718 | cookie = ofpbuf_at(ctx->odp_actions, user_cookie_offset, |
9583bc14 EJ |
2719 | sizeof cookie->sflow); |
2720 | ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); | |
2721 | ||
a6092018 BP |
2722 | cookie->type = USER_ACTION_COOKIE_SFLOW; |
2723 | cookie->sflow.vlan_tci = base->vlan_tci; | |
2724 | ||
2725 | /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output | |
2726 | * port information") for the interpretation of cookie->output. */ | |
2727 | switch (ctx->sflow_n_outputs) { | |
2728 | case 0: | |
2729 | /* 0x40000000 | 256 means "packet dropped for unknown reason". */ | |
2730 | cookie->sflow.output = 0x40000000 | 256; | |
2731 | break; | |
2732 | ||
2733 | case 1: | |
2734 | cookie->sflow.output = dpif_sflow_odp_port_to_ifindex( | |
2735 | ctx->xbridge->sflow, ctx->sflow_odp_port); | |
2736 | if (cookie->sflow.output) { | |
2737 | break; | |
2738 | } | |
2739 | /* Fall through. */ | |
2740 | default: | |
2741 | /* 0x80000000 means "multiple output ports. */ | |
2742 | cookie->sflow.output = 0x80000000 | ctx->sflow_n_outputs; | |
2743 | break; | |
2744 | } | |
9583bc14 EJ |
2745 | } |
2746 | ||
515793d5 BP |
2747 | static bool |
2748 | process_special(struct xlate_ctx *ctx, const struct xport *xport) | |
db7d4e46 | 2749 | { |
515793d5 | 2750 | const struct flow *flow = &ctx->xin->flow; |
49a73e0c | 2751 | struct flow_wildcards *wc = ctx->wc; |
46c88433 | 2752 | const struct xbridge *xbridge = ctx->xbridge; |
515793d5 BP |
2753 | const struct dp_packet *packet = ctx->xin->packet; |
2754 | enum slow_path_reason slow; | |
642dc74d | 2755 | |
46c88433 | 2756 | if (!xport) { |
515793d5 | 2757 | slow = 0; |
46c88433 | 2758 | } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) { |
db7d4e46 | 2759 | if (packet) { |
46c88433 | 2760 | cfm_process_heartbeat(xport->cfm, packet); |
db7d4e46 | 2761 | } |
515793d5 | 2762 | slow = SLOW_CFM; |
fab52e16 | 2763 | } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) { |
db7d4e46 | 2764 | if (packet) { |
46c88433 | 2765 | bfd_process_packet(xport->bfd, flow, packet); |
60d02c72 AW |
2766 | /* If POLL received, immediately sends FINAL back. */ |
2767 | if (bfd_should_send_packet(xport->bfd)) { | |
6d308b28 | 2768 | ofproto_dpif_monitor_port_send_soon(xport->ofport); |
60d02c72 | 2769 | } |
db7d4e46 | 2770 | } |
515793d5 | 2771 | slow = SLOW_BFD; |
46c88433 | 2772 | } else if (xport->xbundle && xport->xbundle->lacp |
db7d4e46 JP |
2773 | && flow->dl_type == htons(ETH_TYPE_LACP)) { |
2774 | if (packet) { | |
46c88433 | 2775 | lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet); |
db7d4e46 | 2776 | } |
515793d5 | 2777 | slow = SLOW_LACP; |
9efd308e DV |
2778 | } else if ((xbridge->stp || xbridge->rstp) && |
2779 | stp_should_process_flow(flow, wc)) { | |
db7d4e46 | 2780 | if (packet) { |
f025bcb7 JR |
2781 | xbridge->stp |
2782 | ? stp_process_packet(xport, packet) | |
2783 | : rstp_process_packet(xport, packet); | |
db7d4e46 | 2784 | } |
515793d5 | 2785 | slow = SLOW_STP; |
19aef6ef | 2786 | } else if (xport->lldp && lldp_should_process_flow(xport->lldp, flow)) { |
0477baa9 DF |
2787 | if (packet) { |
2788 | lldp_process_packet(xport->lldp, packet); | |
2789 | } | |
515793d5 | 2790 | slow = SLOW_LLDP; |
db7d4e46 | 2791 | } else { |
515793d5 BP |
2792 | slow = 0; |
2793 | } | |
2794 | ||
2795 | if (slow) { | |
2796 | ctx->xout->slow |= slow; | |
2797 | return true; | |
2798 | } else { | |
2799 | return false; | |
db7d4e46 JP |
2800 | } |
2801 | } | |
2802 | ||
a36de779 PS |
2803 | static int |
2804 | tnl_route_lookup_flow(const struct flow *oflow, | |
a8704b50 PS |
2805 | struct in6_addr *ip, struct in6_addr *src, |
2806 | struct xport **out_port) | |
a36de779 PS |
2807 | { |
2808 | char out_dev[IFNAMSIZ]; | |
2809 | struct xbridge *xbridge; | |
2810 | struct xlate_cfg *xcfg; | |
c2b878e0 TLSC |
2811 | struct in6_addr gw; |
2812 | struct in6_addr dst; | |
a36de779 | 2813 | |
c2b878e0 | 2814 | dst = flow_tnl_dst(&oflow->tunnel); |
a8704b50 | 2815 | if (!ovs_router_lookup(&dst, out_dev, src, &gw)) { |
a36de779 PS |
2816 | return -ENOENT; |
2817 | } | |
2818 | ||
c2b878e0 TLSC |
2819 | if (ipv6_addr_is_set(&gw) && |
2820 | (!IN6_IS_ADDR_V4MAPPED(&gw) || in6_addr_get_mapped_ipv4(&gw))) { | |
a36de779 PS |
2821 | *ip = gw; |
2822 | } else { | |
c2b878e0 | 2823 | *ip = dst; |
a36de779 PS |
2824 | } |
2825 | ||
2826 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2827 | ovs_assert(xcfg); | |
2828 | ||
2829 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
2830 | if (!strncmp(xbridge->name, out_dev, IFNAMSIZ)) { | |
2831 | struct xport *port; | |
2832 | ||
2833 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
2834 | if (!strncmp(netdev_get_name(port->netdev), out_dev, IFNAMSIZ)) { | |
2835 | *out_port = port; | |
2836 | return 0; | |
2837 | } | |
2838 | } | |
2839 | } | |
2840 | } | |
2841 | return -ENOENT; | |
2842 | } | |
2843 | ||
2844 | static int | |
cdd42eda JG |
2845 | compose_table_xlate(struct xlate_ctx *ctx, const struct xport *out_dev, |
2846 | struct dp_packet *packet) | |
a36de779 | 2847 | { |
cdd42eda | 2848 | struct xbridge *xbridge = out_dev->xbridge; |
a36de779 PS |
2849 | struct ofpact_output output; |
2850 | struct flow flow; | |
2851 | ||
2852 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); | |
cf62fa4c | 2853 | flow_extract(packet, &flow); |
cdd42eda JG |
2854 | flow.in_port.ofp_port = out_dev->ofp_port; |
2855 | output.port = OFPP_TABLE; | |
a36de779 PS |
2856 | output.max_len = 0; |
2857 | ||
cdd42eda JG |
2858 | return ofproto_dpif_execute_actions__(xbridge->ofproto, &flow, NULL, |
2859 | &output.ofpact, sizeof output, | |
790c5d26 BP |
2860 | ctx->indentation, ctx->depth, |
2861 | ctx->resubmits, packet); | |
a36de779 PS |
2862 | } |
2863 | ||
c2b878e0 TLSC |
2864 | static void |
2865 | tnl_send_nd_request(struct xlate_ctx *ctx, const struct xport *out_dev, | |
2866 | const struct eth_addr eth_src, | |
2867 | struct in6_addr * ipv6_src, struct in6_addr * ipv6_dst) | |
2868 | { | |
2869 | struct dp_packet packet; | |
2870 | ||
2871 | dp_packet_init(&packet, 0); | |
2872 | compose_nd(&packet, eth_src, ipv6_src, ipv6_dst); | |
2873 | compose_table_xlate(ctx, out_dev, &packet); | |
2874 | dp_packet_uninit(&packet); | |
2875 | } | |
2876 | ||
a36de779 | 2877 | static void |
cdd42eda | 2878 | tnl_send_arp_request(struct xlate_ctx *ctx, const struct xport *out_dev, |
74ff3298 | 2879 | const struct eth_addr eth_src, |
a36de779 PS |
2880 | ovs_be32 ip_src, ovs_be32 ip_dst) |
2881 | { | |
cf62fa4c | 2882 | struct dp_packet packet; |
a36de779 | 2883 | |
cf62fa4c | 2884 | dp_packet_init(&packet, 0); |
eb0b295e BP |
2885 | compose_arp(&packet, ARP_OP_REQUEST, |
2886 | eth_src, eth_addr_zero, true, ip_src, ip_dst); | |
a36de779 | 2887 | |
cdd42eda | 2888 | compose_table_xlate(ctx, out_dev, &packet); |
cf62fa4c | 2889 | dp_packet_uninit(&packet); |
a36de779 PS |
2890 | } |
2891 | ||
2892 | static int | |
81de18ec | 2893 | build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport, |
a36de779 PS |
2894 | const struct flow *flow, odp_port_t tunnel_odp_port) |
2895 | { | |
4975aa3e | 2896 | struct netdev_tnl_build_header_params tnl_params; |
a36de779 PS |
2897 | struct ovs_action_push_tnl tnl_push_data; |
2898 | struct xport *out_dev = NULL; | |
c2b878e0 TLSC |
2899 | ovs_be32 s_ip = 0, d_ip = 0; |
2900 | struct in6_addr s_ip6 = in6addr_any; | |
2901 | struct in6_addr d_ip6 = in6addr_any; | |
74ff3298 JR |
2902 | struct eth_addr smac; |
2903 | struct eth_addr dmac; | |
a36de779 | 2904 | int err; |
c2b878e0 TLSC |
2905 | char buf_sip6[INET6_ADDRSTRLEN]; |
2906 | char buf_dip6[INET6_ADDRSTRLEN]; | |
a36de779 | 2907 | |
a8704b50 | 2908 | err = tnl_route_lookup_flow(flow, &d_ip6, &s_ip6, &out_dev); |
a36de779 | 2909 | if (err) { |
81de18ec | 2910 | xlate_report(ctx, "native tunnel routing failed"); |
a36de779 PS |
2911 | return err; |
2912 | } | |
c2b878e0 TLSC |
2913 | |
2914 | xlate_report(ctx, "tunneling to %s via %s", | |
2915 | ipv6_string_mapped(buf_dip6, &d_ip6), | |
2916 | netdev_get_name(out_dev->netdev)); | |
a36de779 PS |
2917 | |
2918 | /* Use mac addr of bridge port of the peer. */ | |
74ff3298 | 2919 | err = netdev_get_etheraddr(out_dev->netdev, &smac); |
a36de779 | 2920 | if (err) { |
81de18ec | 2921 | xlate_report(ctx, "tunnel output device lacks Ethernet address"); |
a36de779 PS |
2922 | return err; |
2923 | } | |
2924 | ||
c2b878e0 TLSC |
2925 | d_ip = in6_addr_get_mapped_ipv4(&d_ip6); |
2926 | if (d_ip) { | |
a8704b50 | 2927 | s_ip = in6_addr_get_mapped_ipv4(&s_ip6); |
a36de779 PS |
2928 | } |
2929 | ||
c2b878e0 | 2930 | err = tnl_neigh_lookup(out_dev->xbridge->name, &d_ip6, &dmac); |
a36de779 | 2931 | if (err) { |
c2b878e0 TLSC |
2932 | xlate_report(ctx, "neighbor cache miss for %s on bridge %s, " |
2933 | "sending %s request", | |
2934 | buf_dip6, out_dev->xbridge->name, d_ip ? "ARP" : "ND"); | |
2935 | if (d_ip) { | |
2936 | tnl_send_arp_request(ctx, out_dev, smac, s_ip, d_ip); | |
2937 | } else { | |
2938 | tnl_send_nd_request(ctx, out_dev, smac, &s_ip6, &d_ip6); | |
2939 | } | |
a36de779 PS |
2940 | return err; |
2941 | } | |
c2b878e0 | 2942 | |
a36de779 PS |
2943 | if (ctx->xin->xcache) { |
2944 | struct xc_entry *entry; | |
2945 | ||
53902038 TLSC |
2946 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_NEIGH); |
2947 | ovs_strlcpy(entry->u.tnl_neigh_cache.br_name, out_dev->xbridge->name, | |
2948 | sizeof entry->u.tnl_neigh_cache.br_name); | |
c2b878e0 | 2949 | entry->u.tnl_neigh_cache.d_ipv6 = d_ip6; |
a36de779 | 2950 | } |
81de18ec | 2951 | |
c2b878e0 TLSC |
2952 | xlate_report(ctx, "tunneling from "ETH_ADDR_FMT" %s" |
2953 | " to "ETH_ADDR_FMT" %s", | |
2954 | ETH_ADDR_ARGS(smac), ipv6_string_mapped(buf_sip6, &s_ip6), | |
2955 | ETH_ADDR_ARGS(dmac), buf_dip6); | |
2956 | ||
4975aa3e PS |
2957 | netdev_init_tnl_build_header_params(&tnl_params, flow, &s_ip6, dmac, smac); |
2958 | err = tnl_port_build_header(xport->ofport, &tnl_push_data, &tnl_params); | |
a36de779 PS |
2959 | if (err) { |
2960 | return err; | |
2961 | } | |
2962 | tnl_push_data.tnl_port = odp_to_u32(tunnel_odp_port); | |
2963 | tnl_push_data.out_port = odp_to_u32(out_dev->odp_port); | |
1520ef4f | 2964 | odp_put_tnl_push_action(ctx->odp_actions, &tnl_push_data); |
a36de779 PS |
2965 | return 0; |
2966 | } | |
2967 | ||
704bb0bf JS |
2968 | static void |
2969 | xlate_commit_actions(struct xlate_ctx *ctx) | |
2970 | { | |
2971 | bool use_masked = ctx->xbridge->support.masked_set_action; | |
2972 | ||
2973 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, | |
2974 | ctx->odp_actions, ctx->wc, | |
2975 | use_masked); | |
2976 | } | |
2977 | ||
07659514 JS |
2978 | static void |
2979 | clear_conntrack(struct flow *flow) | |
2980 | { | |
2981 | flow->ct_state = 0; | |
2982 | flow->ct_zone = 0; | |
8e53fe8c | 2983 | flow->ct_mark = 0; |
9daf2348 | 2984 | memset(&flow->ct_label, 0, sizeof flow->ct_label); |
07659514 JS |
2985 | } |
2986 | ||
9583bc14 | 2987 | static void |
4e022ec0 | 2988 | compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
e93ef1c7 | 2989 | const struct xlate_bond_recirc *xr, bool check_stp) |
9583bc14 | 2990 | { |
46c88433 | 2991 | const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
49a73e0c | 2992 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 2993 | struct flow *flow = &ctx->xin->flow; |
a36de779 | 2994 | struct flow_tnl flow_tnl; |
9583bc14 | 2995 | ovs_be16 flow_vlan_tci; |
1362e248 | 2996 | uint32_t flow_pkt_mark; |
9583bc14 | 2997 | uint8_t flow_nw_tos; |
4e022ec0 | 2998 | odp_port_t out_port, odp_port; |
a36de779 | 2999 | bool tnl_push_pop_send = false; |
ca077186 | 3000 | uint8_t dscp; |
9583bc14 EJ |
3001 | |
3002 | /* If 'struct flow' gets additional metadata, we'll need to zero it out | |
3003 | * before traversing a patch port. */ | |
ffe4c74f | 3004 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 35); |
a36de779 | 3005 | memset(&flow_tnl, 0, sizeof flow_tnl); |
9583bc14 | 3006 | |
46c88433 | 3007 | if (!xport) { |
9583bc14 EJ |
3008 | xlate_report(ctx, "Nonexistent output port"); |
3009 | return; | |
46c88433 | 3010 | } else if (xport->config & OFPUTIL_PC_NO_FWD) { |
9583bc14 EJ |
3011 | xlate_report(ctx, "OFPPC_NO_FWD set, skipping output"); |
3012 | return; | |
0d1cee12 | 3013 | } else if (check_stp) { |
bbbca389 | 3014 | if (is_stp(&ctx->base_flow)) { |
9efd308e DV |
3015 | if (!xport_stp_should_forward_bpdu(xport) && |
3016 | !xport_rstp_should_manage_bpdu(xport)) { | |
3017 | if (ctx->xbridge->stp != NULL) { | |
3018 | xlate_report(ctx, "STP not in listening state, " | |
3019 | "skipping bpdu output"); | |
3020 | } else if (ctx->xbridge->rstp != NULL) { | |
3021 | xlate_report(ctx, "RSTP not managing BPDU in this state, " | |
3022 | "skipping bpdu output"); | |
3023 | } | |
0d1cee12 K |
3024 | return; |
3025 | } | |
9efd308e DV |
3026 | } else if (!xport_stp_forward_state(xport) || |
3027 | !xport_rstp_forward_state(xport)) { | |
3028 | if (ctx->xbridge->stp != NULL) { | |
3029 | xlate_report(ctx, "STP not in forwarding state, " | |
3030 | "skipping output"); | |
3031 | } else if (ctx->xbridge->rstp != NULL) { | |
3032 | xlate_report(ctx, "RSTP not in forwarding state, " | |
3033 | "skipping output"); | |
3034 | } | |
0d1cee12 K |
3035 | return; |
3036 | } | |
9583bc14 EJ |
3037 | } |
3038 | ||
46c88433 EJ |
3039 | if (xport->peer) { |
3040 | const struct xport *peer = xport->peer; | |
9583bc14 | 3041 | struct flow old_flow = ctx->xin->flow; |
07659514 | 3042 | bool old_conntrack = ctx->conntracked; |
e12ec36b | 3043 | bool old_was_mpls = ctx->was_mpls; |
18721c4a | 3044 | cls_version_t old_version = ctx->tables_version; |
1774d762 JR |
3045 | struct ofpbuf old_stack = ctx->stack; |
3046 | union mf_subvalue new_stack[1024 / sizeof(union mf_subvalue)]; | |
ed9c9e3e JR |
3047 | struct ofpbuf old_action_set = ctx->action_set; |
3048 | uint64_t actset_stub[1024 / 8]; | |
9583bc14 | 3049 | |
1774d762 | 3050 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); |
ed9c9e3e | 3051 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); |
46c88433 EJ |
3052 | ctx->xbridge = peer->xbridge; |
3053 | flow->in_port.ofp_port = peer->ofp_port; | |
33bf9176 BP |
3054 | flow->metadata = htonll(0); |
3055 | memset(&flow->tunnel, 0, sizeof flow->tunnel); | |
3056 | memset(flow->regs, 0, sizeof flow->regs); | |
c61f3870 | 3057 | flow->actset_output = OFPP_UNSET; |
07659514 JS |
3058 | ctx->conntracked = false; |
3059 | clear_conntrack(flow); | |
9583bc14 | 3060 | |
621b8064 JR |
3061 | /* The bridge is now known so obtain its table version. */ |
3062 | ctx->tables_version | |
3063 | = ofproto_dpif_get_tables_version(ctx->xbridge->ofproto); | |
3064 | ||
515793d5 | 3065 | if (!process_special(ctx, peer) && may_receive(peer, ctx)) { |
9efd308e | 3066 | if (xport_stp_forward_state(peer) && xport_rstp_forward_state(peer)) { |
e672ff9b | 3067 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); |
1d361a81 | 3068 | if (!ctx->freezing) { |
8a5fb3b4 | 3069 | xlate_action_set(ctx); |
e672ff9b | 3070 | } |
1d361a81 | 3071 | if (ctx->freezing) { |
77ab5fd2 | 3072 | finish_freezing(ctx); |
ed9c9e3e | 3073 | } |
9583bc14 | 3074 | } else { |
9efd308e DV |
3075 | /* Forwarding is disabled by STP and RSTP. Let OFPP_NORMAL and |
3076 | * the learning action look at the packet, then drop it. */ | |
9583bc14 | 3077 | struct flow old_base_flow = ctx->base_flow; |
1520ef4f | 3078 | size_t old_size = ctx->odp_actions->size; |
3d6151f3 | 3079 | mirror_mask_t old_mirrors = ctx->mirrors; |
f3d5b473 | 3080 | |
e672ff9b | 3081 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); |
3d6151f3 | 3082 | ctx->mirrors = old_mirrors; |
9583bc14 | 3083 | ctx->base_flow = old_base_flow; |
1520ef4f | 3084 | ctx->odp_actions->size = old_size; |
e672ff9b | 3085 | |
1d361a81 BP |
3086 | /* Undo changes that may have been done for freezing. */ |
3087 | ctx_cancel_freeze(ctx); | |
9583bc14 EJ |
3088 | } |
3089 | } | |
3090 | ||
3091 | ctx->xin->flow = old_flow; | |
832554e3 | 3092 | ctx->xbridge = xport->xbridge; |
ed9c9e3e JR |
3093 | ofpbuf_uninit(&ctx->action_set); |
3094 | ctx->action_set = old_action_set; | |
1774d762 JR |
3095 | ofpbuf_uninit(&ctx->stack); |
3096 | ctx->stack = old_stack; | |
9583bc14 | 3097 | |
621b8064 JR |
3098 | /* Restore calling bridge's lookup version. */ |
3099 | ctx->tables_version = old_version; | |
3100 | ||
e12ec36b SH |
3101 | /* The peer bridge popping MPLS should have no effect on the original |
3102 | * bridge. */ | |
3103 | ctx->was_mpls = old_was_mpls; | |
3104 | ||
07659514 JS |
3105 | /* The peer bridge's conntrack execution should have no effect on the |
3106 | * original bridge. */ | |
3107 | ctx->conntracked = old_conntrack; | |
3108 | ||
f3d5b473 JR |
3109 | /* The fact that the peer bridge exits (for any reason) does not mean |
3110 | * that the original bridge should exit. Specifically, if the peer | |
1d361a81 BP |
3111 | * bridge freezes translation, the original bridge must continue |
3112 | * processing with the original, not the frozen packet! */ | |
f3d5b473 JR |
3113 | ctx->exit = false; |
3114 | ||
fff1b9c0 JR |
3115 | /* Peer bridge errors do not propagate back. */ |
3116 | ctx->error = XLATE_OK; | |
3117 | ||
9583bc14 | 3118 | if (ctx->xin->resubmit_stats) { |
46c88433 EJ |
3119 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
3120 | netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats); | |
a1aeea86 AW |
3121 | if (peer->bfd) { |
3122 | bfd_account_rx(peer->bfd, ctx->xin->resubmit_stats); | |
3123 | } | |
9583bc14 | 3124 | } |
b256dc52 JS |
3125 | if (ctx->xin->xcache) { |
3126 | struct xc_entry *entry; | |
3127 | ||
3128 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
3129 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
3130 | entry->u.dev.rx = netdev_ref(peer->netdev); | |
3131 | entry->u.dev.bfd = bfd_ref(peer->bfd); | |
3132 | } | |
9583bc14 EJ |
3133 | return; |
3134 | } | |
3135 | ||
33bf9176 | 3136 | flow_vlan_tci = flow->vlan_tci; |
1362e248 | 3137 | flow_pkt_mark = flow->pkt_mark; |
33bf9176 | 3138 | flow_nw_tos = flow->nw_tos; |
9583bc14 | 3139 | |
16194afd DDP |
3140 | if (count_skb_priorities(xport)) { |
3141 | memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority); | |
3142 | if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { | |
3143 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
3144 | flow->nw_tos &= ~IP_DSCP_MASK; | |
3145 | flow->nw_tos |= dscp; | |
3146 | } | |
9583bc14 EJ |
3147 | } |
3148 | ||
46c88433 | 3149 | if (xport->is_tunnel) { |
c2b878e0 | 3150 | struct in6_addr dst; |
9583bc14 EJ |
3151 | /* Save tunnel metadata so that changes made due to |
3152 | * the Logical (tunnel) Port are not visible for any further | |
3153 | * matches, while explicit set actions on tunnel metadata are. | |
3154 | */ | |
a36de779 | 3155 | flow_tnl = flow->tunnel; |
49a73e0c | 3156 | odp_port = tnl_port_send(xport->ofport, flow, ctx->wc); |
4e022ec0 | 3157 | if (odp_port == ODPP_NONE) { |
9583bc14 EJ |
3158 | xlate_report(ctx, "Tunneling decided against output"); |
3159 | goto out; /* restore flow_nw_tos */ | |
3160 | } | |
c2b878e0 TLSC |
3161 | dst = flow_tnl_dst(&flow->tunnel); |
3162 | if (ipv6_addr_equals(&dst, &ctx->orig_tunnel_ipv6_dst)) { | |
9583bc14 EJ |
3163 | xlate_report(ctx, "Not tunneling to our own address"); |
3164 | goto out; /* restore flow_nw_tos */ | |
3165 | } | |
3166 | if (ctx->xin->resubmit_stats) { | |
46c88433 | 3167 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
9583bc14 | 3168 | } |
b256dc52 JS |
3169 | if (ctx->xin->xcache) { |
3170 | struct xc_entry *entry; | |
3171 | ||
3172 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
3173 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
3174 | } | |
9583bc14 | 3175 | out_port = odp_port; |
a36de779 | 3176 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
81de18ec | 3177 | xlate_report(ctx, "output to native tunnel"); |
a36de779 PS |
3178 | tnl_push_pop_send = true; |
3179 | } else { | |
81de18ec | 3180 | xlate_report(ctx, "output to kernel tunnel"); |
1520ef4f | 3181 | commit_odp_tunnel_action(flow, &ctx->base_flow, ctx->odp_actions); |
a36de779 PS |
3182 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ |
3183 | } | |
9583bc14 | 3184 | } else { |
46c88433 | 3185 | odp_port = xport->odp_port; |
7614e5d0 | 3186 | out_port = odp_port; |
9583bc14 | 3187 | } |
9583bc14 | 3188 | |
4e022ec0 | 3189 | if (out_port != ODPP_NONE) { |
704bb0bf | 3190 | xlate_commit_actions(ctx); |
adcf00ba | 3191 | |
e93ef1c7 | 3192 | if (xr) { |
347bf289 | 3193 | struct ovs_action_hash *act_hash; |
adcf00ba | 3194 | |
347bf289 | 3195 | /* Hash action. */ |
1520ef4f | 3196 | act_hash = nl_msg_put_unspec_uninit(ctx->odp_actions, |
347bf289 AZ |
3197 | OVS_ACTION_ATTR_HASH, |
3198 | sizeof *act_hash); | |
3199 | act_hash->hash_alg = xr->hash_alg; | |
62ac1f20 | 3200 | act_hash->hash_basis = xr->hash_basis; |
347bf289 AZ |
3201 | |
3202 | /* Recirc action. */ | |
1520ef4f | 3203 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, |
347bf289 | 3204 | xr->recirc_id); |
adcf00ba | 3205 | } else { |
a36de779 PS |
3206 | |
3207 | if (tnl_push_pop_send) { | |
3208 | build_tunnel_send(ctx, xport, flow, odp_port); | |
3209 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ | |
3210 | } else { | |
3211 | odp_port_t odp_tnl_port = ODPP_NONE; | |
3212 | ||
3213 | /* XXX: Write better Filter for tunnel port. We can use inport | |
3214 | * int tunnel-port flow to avoid these checks completely. */ | |
3215 | if (ofp_port == OFPP_LOCAL && | |
3216 | ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { | |
3217 | ||
3218 | odp_tnl_port = tnl_port_map_lookup(flow, wc); | |
3219 | } | |
3220 | ||
3221 | if (odp_tnl_port != ODPP_NONE) { | |
1520ef4f | 3222 | nl_msg_put_odp_port(ctx->odp_actions, |
a36de779 PS |
3223 | OVS_ACTION_ATTR_TUNNEL_POP, |
3224 | odp_tnl_port); | |
3225 | } else { | |
3226 | /* Tunnel push-pop action is not compatible with | |
3227 | * IPFIX action. */ | |
a6092018 | 3228 | compose_ipfix_action(ctx, out_port); |
1520ef4f | 3229 | nl_msg_put_odp_port(ctx->odp_actions, |
a36de779 PS |
3230 | OVS_ACTION_ATTR_OUTPUT, |
3231 | out_port); | |
3232 | } | |
3233 | } | |
adcf00ba | 3234 | } |
9583bc14 | 3235 | |
6cbbf4fa EJ |
3236 | ctx->sflow_odp_port = odp_port; |
3237 | ctx->sflow_n_outputs++; | |
2031ef97 | 3238 | ctx->nf_output_iface = ofp_port; |
6cbbf4fa EJ |
3239 | } |
3240 | ||
7efbc3b7 BP |
3241 | if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { |
3242 | mirror_packet(ctx, xport->xbundle, | |
3243 | xbundle_mirror_dst(xport->xbundle->xbridge, | |
3244 | xport->xbundle)); | |
3245 | } | |
3246 | ||
6cbbf4fa | 3247 | out: |
9583bc14 | 3248 | /* Restore flow */ |
33bf9176 | 3249 | flow->vlan_tci = flow_vlan_tci; |
1362e248 | 3250 | flow->pkt_mark = flow_pkt_mark; |
33bf9176 | 3251 | flow->nw_tos = flow_nw_tos; |
9583bc14 EJ |
3252 | } |
3253 | ||
3254 | static void | |
e93ef1c7 JR |
3255 | compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
3256 | const struct xlate_bond_recirc *xr) | |
9583bc14 | 3257 | { |
e93ef1c7 | 3258 | compose_output_action__(ctx, ofp_port, xr, true); |
9583bc14 EJ |
3259 | } |
3260 | ||
bb61b33d | 3261 | static void |
790c5d26 | 3262 | xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule, bool deepens) |
bb61b33d BP |
3263 | { |
3264 | struct rule_dpif *old_rule = ctx->rule; | |
8b1e5560 | 3265 | ovs_be64 old_cookie = ctx->rule_cookie; |
dc723c44 | 3266 | const struct rule_actions *actions; |
bb61b33d BP |
3267 | |
3268 | if (ctx->xin->resubmit_stats) { | |
70742c7f | 3269 | rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats); |
bb61b33d BP |
3270 | } |
3271 | ||
98b07853 | 3272 | ctx->resubmits++; |
790c5d26 | 3273 | |
9a9b001a | 3274 | ctx->indentation++; |
790c5d26 | 3275 | ctx->depth += deepens; |
bb61b33d | 3276 | ctx->rule = rule; |
8b1e5560 | 3277 | ctx->rule_cookie = rule_dpif_get_flow_cookie(rule); |
6f00e29b BP |
3278 | actions = rule_dpif_get_actions(rule); |
3279 | do_xlate_actions(actions->ofpacts, actions->ofpacts_len, ctx); | |
8b1e5560 | 3280 | ctx->rule_cookie = old_cookie; |
bb61b33d | 3281 | ctx->rule = old_rule; |
790c5d26 | 3282 | ctx->depth -= deepens; |
9a9b001a | 3283 | ctx->indentation--; |
bb61b33d BP |
3284 | } |
3285 | ||
bd3240ba SH |
3286 | static bool |
3287 | xlate_resubmit_resource_check(struct xlate_ctx *ctx) | |
9583bc14 | 3288 | { |
790c5d26 BP |
3289 | if (ctx->depth >= MAX_DEPTH) { |
3290 | XLATE_REPORT_ERROR(ctx, "over max translation depth %d", MAX_DEPTH); | |
fff1b9c0 | 3291 | ctx->error = XLATE_RECURSION_TOO_DEEP; |
790c5d26 | 3292 | } else if (ctx->resubmits >= MAX_RESUBMITS) { |
fff1b9c0 JR |
3293 | XLATE_REPORT_ERROR(ctx, "over %d resubmit actions", MAX_RESUBMITS); |
3294 | ctx->error = XLATE_TOO_MANY_RESUBMITS; | |
1520ef4f | 3295 | } else if (ctx->odp_actions->size > UINT16_MAX) { |
fff1b9c0 JR |
3296 | XLATE_REPORT_ERROR(ctx, "resubmits yielded over 64 kB of actions"); |
3297 | /* NOT an error, as we'll be slow-pathing the flow in this case? */ | |
3298 | ctx->exit = true; /* XXX: translation still terminated! */ | |
6fd6ed71 | 3299 | } else if (ctx->stack.size >= 65536) { |
fff1b9c0 JR |
3300 | XLATE_REPORT_ERROR(ctx, "resubmits yielded over 64 kB of stack"); |
3301 | ctx->error = XLATE_STACK_TOO_DEEP; | |
98b07853 | 3302 | } else { |
bd3240ba SH |
3303 | return true; |
3304 | } | |
3305 | ||
3306 | return false; | |
3307 | } | |
3308 | ||
3309 | static void | |
6d328fa2 SH |
3310 | xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, |
3311 | bool may_packet_in, bool honor_table_miss) | |
bd3240ba | 3312 | { |
e12ec36b SH |
3313 | /* Check if we need to recirculate before matching in a table. */ |
3314 | if (ctx->was_mpls) { | |
3315 | ctx_trigger_freeze(ctx); | |
3316 | return; | |
3317 | } | |
bd3240ba | 3318 | if (xlate_resubmit_resource_check(ctx)) { |
9583bc14 | 3319 | uint8_t old_table_id = ctx->table_id; |
3f207910 | 3320 | struct rule_dpif *rule; |
9583bc14 EJ |
3321 | |
3322 | ctx->table_id = table_id; | |
3323 | ||
34dd0d78 | 3324 | rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, |
621b8064 | 3325 | ctx->tables_version, |
c0e638aa | 3326 | &ctx->xin->flow, ctx->wc, |
34dd0d78 JR |
3327 | ctx->xin->resubmit_stats, |
3328 | &ctx->table_id, in_port, | |
3329 | may_packet_in, honor_table_miss); | |
9583bc14 | 3330 | |
a8c31348 | 3331 | if (OVS_UNLIKELY(ctx->xin->resubmit_hook)) { |
9a9b001a | 3332 | ctx->xin->resubmit_hook(ctx->xin, rule, ctx->indentation + 1); |
ad3efdcb EJ |
3333 | } |
3334 | ||
a2143702 | 3335 | if (rule) { |
83709dfa JR |
3336 | /* Fill in the cache entry here instead of xlate_recursively |
3337 | * to make the reference counting more explicit. We take a | |
3338 | * reference in the lookups above if we are going to cache the | |
3339 | * rule. */ | |
3340 | if (ctx->xin->xcache) { | |
3341 | struct xc_entry *entry; | |
3342 | ||
3343 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE); | |
3344 | entry->u.rule = rule; | |
1e1e1d19 | 3345 | rule_dpif_ref(rule); |
83709dfa | 3346 | } |
790c5d26 | 3347 | xlate_recursively(ctx, rule, table_id <= old_table_id); |
ad3efdcb EJ |
3348 | } |
3349 | ||
9583bc14 | 3350 | ctx->table_id = old_table_id; |
98b07853 | 3351 | return; |
9583bc14 EJ |
3352 | } |
3353 | } | |
3354 | ||
f4fb341b | 3355 | static void |
1e684d7d RW |
3356 | xlate_group_stats(struct xlate_ctx *ctx, struct group_dpif *group, |
3357 | struct ofputil_bucket *bucket) | |
3358 | { | |
3359 | if (ctx->xin->resubmit_stats) { | |
3360 | group_dpif_credit_stats(group, bucket, ctx->xin->resubmit_stats); | |
3361 | } | |
3362 | if (ctx->xin->xcache) { | |
3363 | struct xc_entry *entry; | |
3364 | ||
3365 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_GROUP); | |
3366 | entry->u.group.group = group_dpif_ref(group); | |
3367 | entry->u.group.bucket = bucket; | |
3368 | } | |
3369 | } | |
3370 | ||
3371 | static void | |
3372 | xlate_group_bucket(struct xlate_ctx *ctx, struct ofputil_bucket *bucket) | |
f4fb341b SH |
3373 | { |
3374 | uint64_t action_list_stub[1024 / 8]; | |
0a2869d5 BP |
3375 | struct ofpbuf action_list = OFPBUF_STUB_INITIALIZER(action_list_stub); |
3376 | struct ofpbuf action_set = ofpbuf_const_initializer(bucket->ofpacts, | |
3377 | bucket->ofpacts_len); | |
5b09e569 | 3378 | struct flow old_flow = ctx->xin->flow; |
e12ec36b | 3379 | bool old_was_mpls = ctx->was_mpls; |
f4fb341b | 3380 | |
f4fb341b | 3381 | ofpacts_execute_action_set(&action_list, &action_set); |
9a9b001a | 3382 | ctx->indentation++; |
790c5d26 | 3383 | ctx->depth++; |
6fd6ed71 | 3384 | do_xlate_actions(action_list.data, action_list.size, ctx); |
790c5d26 | 3385 | ctx->depth--; |
9a9b001a | 3386 | ctx->indentation--; |
f4fb341b | 3387 | |
f4fb341b | 3388 | ofpbuf_uninit(&action_list); |
5b09e569 | 3389 | |
77ab5fd2 | 3390 | /* Check if need to freeze. */ |
1d361a81 | 3391 | if (ctx->freezing) { |
77ab5fd2 | 3392 | finish_freezing(ctx); |
e672ff9b JR |
3393 | } |
3394 | ||
5b09e569 JR |
3395 | /* Roll back flow to previous state. |
3396 | * This is equivalent to cloning the packet for each bucket. | |
3397 | * | |
3398 | * As a side effect any subsequently applied actions will | |
3399 | * also effectively be applied to a clone of the packet taken | |
3400 | * just before applying the all or indirect group. | |
3401 | * | |
3402 | * Note that group buckets are action sets, hence they cannot modify the | |
3403 | * main action set. Also any stack actions are ignored when executing an | |
3404 | * action set, so group buckets cannot change the stack either. | |
3405 | * However, we do allow resubmit actions in group buckets, which could | |
3406 | * break the above assumptions. It is up to the controller to not mess up | |
3407 | * with the action_set and stack in the tables resubmitted to from | |
3408 | * group buckets. */ | |
3409 | ctx->xin->flow = old_flow; | |
3410 | ||
e12ec36b SH |
3411 | /* The group bucket popping MPLS should have no effect after bucket |
3412 | * execution. */ | |
3413 | ctx->was_mpls = old_was_mpls; | |
3414 | ||
5b09e569 JR |
3415 | /* The fact that the group bucket exits (for any reason) does not mean that |
3416 | * the translation after the group action should exit. Specifically, if | |
1d361a81 BP |
3417 | * the group bucket freezes translation, the actions after the group action |
3418 | * must continue processing with the original, not the frozen packet! */ | |
5b09e569 | 3419 | ctx->exit = false; |
f4fb341b SH |
3420 | } |
3421 | ||
3422 | static void | |
3423 | xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3424 | { | |
1e684d7d | 3425 | struct ofputil_bucket *bucket; |
ca6ba700 | 3426 | const struct ovs_list *buckets; |
f4fb341b SH |
3427 | |
3428 | group_dpif_get_buckets(group, &buckets); | |
3429 | ||
3430 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
3431 | xlate_group_bucket(ctx, bucket); | |
f4fb341b | 3432 | } |
1e684d7d | 3433 | xlate_group_stats(ctx, group, NULL); |
f4fb341b SH |
3434 | } |
3435 | ||
dd8cd4b4 SH |
3436 | static void |
3437 | xlate_ff_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3438 | { | |
1e684d7d | 3439 | struct ofputil_bucket *bucket; |
dd8cd4b4 SH |
3440 | |
3441 | bucket = group_first_live_bucket(ctx, group, 0); | |
3442 | if (bucket) { | |
3443 | xlate_group_bucket(ctx, bucket); | |
1e684d7d | 3444 | xlate_group_stats(ctx, group, bucket); |
dd8cd4b4 SH |
3445 | } |
3446 | } | |
3447 | ||
fe7e5749 | 3448 | static void |
7565c3e4 | 3449 | xlate_default_select_group(struct xlate_ctx *ctx, struct group_dpif *group) |
fe7e5749 | 3450 | { |
49a73e0c | 3451 | struct flow_wildcards *wc = ctx->wc; |
1e684d7d | 3452 | struct ofputil_bucket *bucket; |
fe7e5749 SH |
3453 | uint32_t basis; |
3454 | ||
1d1aae0b | 3455 | basis = flow_hash_symmetric_l4(&ctx->xin->flow, 0); |
80e3509d | 3456 | flow_mask_hash_fields(&ctx->xin->flow, wc, NX_HASH_FIELDS_SYMMETRIC_L4); |
fe7e5749 SH |
3457 | bucket = group_best_live_bucket(ctx, group, basis); |
3458 | if (bucket) { | |
fe7e5749 | 3459 | xlate_group_bucket(ctx, bucket); |
1e684d7d | 3460 | xlate_group_stats(ctx, group, bucket); |
fe7e5749 SH |
3461 | } |
3462 | } | |
3463 | ||
0c4b9393 SH |
3464 | static void |
3465 | xlate_hash_fields_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3466 | { | |
3467 | struct mf_bitmap hash_fields = MF_BITMAP_INITIALIZER; | |
0c4b9393 SH |
3468 | const struct field_array *fields; |
3469 | struct ofputil_bucket *bucket; | |
3470 | uint32_t basis; | |
3471 | int i; | |
3472 | ||
3473 | fields = group_dpif_get_fields(group); | |
3474 | basis = hash_uint64(group_dpif_get_selection_method_param(group)); | |
3475 | ||
3476 | /* Determine which fields to hash */ | |
3477 | for (i = 0; i < MFF_N_IDS; i++) { | |
3478 | if (bitmap_is_set(fields->used.bm, i)) { | |
3479 | const struct mf_field *mf; | |
3480 | ||
3481 | /* If the field is already present in 'hash_fields' then | |
3482 | * this loop has already checked that it and its pre-requisites | |
3483 | * are present in the flow and its pre-requisites have | |
3484 | * already been added to 'hash_fields'. There is nothing more | |
3485 | * to do here and as an optimisation the loop can continue. */ | |
3486 | if (bitmap_is_set(hash_fields.bm, i)) { | |
3487 | continue; | |
3488 | } | |
3489 | ||
3490 | mf = mf_from_id(i); | |
3491 | ||
3492 | /* Only hash a field if it and its pre-requisites are present | |
3493 | * in the flow. */ | |
3494 | if (!mf_are_prereqs_ok(mf, &ctx->xin->flow)) { | |
3495 | continue; | |
3496 | } | |
3497 | ||
3498 | /* Hash both the field and its pre-requisites */ | |
3499 | mf_bitmap_set_field_and_prereqs(mf, &hash_fields); | |
3500 | } | |
3501 | } | |
3502 | ||
3503 | /* Hash the fields */ | |
3504 | for (i = 0; i < MFF_N_IDS; i++) { | |
3505 | if (bitmap_is_set(hash_fields.bm, i)) { | |
3506 | const struct mf_field *mf = mf_from_id(i); | |
3507 | union mf_value value; | |
3508 | int j; | |
3509 | ||
3510 | mf_get_value(mf, &ctx->xin->flow, &value); | |
3511 | /* This seems inefficient but so does apply_mask() */ | |
3512 | for (j = 0; j < mf->n_bytes; j++) { | |
3513 | ((uint8_t *) &value)[j] &= ((uint8_t *) &fields->value[i])[j]; | |
3514 | } | |
3515 | basis = hash_bytes(&value, mf->n_bytes, basis); | |
3516 | ||
1cb20095 JG |
3517 | /* For tunnels, hash in whether the field is present. */ |
3518 | if (mf_is_tun_metadata(mf)) { | |
3519 | basis = hash_boolean(mf_is_set(mf, &ctx->xin->flow), basis); | |
3520 | } | |
3521 | ||
49a73e0c | 3522 | mf_mask_field(mf, &ctx->wc->masks); |
0c4b9393 SH |
3523 | } |
3524 | } | |
3525 | ||
3526 | bucket = group_best_live_bucket(ctx, group, basis); | |
3527 | if (bucket) { | |
3528 | xlate_group_bucket(ctx, bucket); | |
3529 | xlate_group_stats(ctx, group, bucket); | |
3530 | } | |
3531 | } | |
3532 | ||
7565c3e4 SH |
3533 | static void |
3534 | xlate_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3535 | { | |
3536 | const char *selection_method = group_dpif_get_selection_method(group); | |
3537 | ||
e12ec36b SH |
3538 | /* Select groups may access flow keys beyond L2 in order to |
3539 | * select a bucket. Recirculate as appropriate to make this possible. | |
3540 | */ | |
3541 | if (ctx->was_mpls) { | |
3542 | ctx_trigger_freeze(ctx); | |
3543 | } | |
3544 | ||
7565c3e4 SH |
3545 | if (selection_method[0] == '\0') { |
3546 | xlate_default_select_group(ctx, group); | |
0c4b9393 SH |
3547 | } else if (!strcasecmp("hash", selection_method)) { |
3548 | xlate_hash_fields_select_group(ctx, group); | |
7565c3e4 SH |
3549 | } else { |
3550 | /* Parsing of groups should ensure this never happens */ | |
3551 | OVS_NOT_REACHED(); | |
3552 | } | |
3553 | } | |
3554 | ||
f4fb341b SH |
3555 | static void |
3556 | xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group) | |
3557 | { | |
0eb48fe1 | 3558 | bool was_in_group = ctx->in_group; |
5a070238 BP |
3559 | ctx->in_group = true; |
3560 | ||
f4fb341b SH |
3561 | switch (group_dpif_get_type(group)) { |
3562 | case OFPGT11_ALL: | |
3563 | case OFPGT11_INDIRECT: | |
3564 | xlate_all_group(ctx, group); | |
3565 | break; | |
3566 | case OFPGT11_SELECT: | |
fe7e5749 | 3567 | xlate_select_group(ctx, group); |
f4fb341b | 3568 | break; |
dd8cd4b4 SH |
3569 | case OFPGT11_FF: |
3570 | xlate_ff_group(ctx, group); | |
3571 | break; | |
f4fb341b | 3572 | default: |
428b2edd | 3573 | OVS_NOT_REACHED(); |
f4fb341b | 3574 | } |
809c7548 | 3575 | group_dpif_unref(group); |
5a070238 | 3576 | |
0eb48fe1 | 3577 | ctx->in_group = was_in_group; |
f4fb341b SH |
3578 | } |
3579 | ||
3580 | static bool | |
3581 | xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id) | |
3582 | { | |
0eb48fe1 | 3583 | if (xlate_resubmit_resource_check(ctx)) { |
f4fb341b SH |
3584 | struct group_dpif *group; |
3585 | bool got_group; | |
3586 | ||
3587 | got_group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group); | |
3588 | if (got_group) { | |
3589 | xlate_group_action__(ctx, group); | |
3590 | } else { | |
3591 | return true; | |
3592 | } | |
3593 | } | |
3594 | ||
3595 | return false; | |
3596 | } | |
3597 | ||
9583bc14 EJ |
3598 | static void |
3599 | xlate_ofpact_resubmit(struct xlate_ctx *ctx, | |
3600 | const struct ofpact_resubmit *resubmit) | |
3601 | { | |
4e022ec0 | 3602 | ofp_port_t in_port; |
9583bc14 | 3603 | uint8_t table_id; |
adcf00ba AZ |
3604 | bool may_packet_in = false; |
3605 | bool honor_table_miss = false; | |
3606 | ||
3607 | if (ctx->rule && rule_dpif_is_internal(ctx->rule)) { | |
3608 | /* Still allow missed packets to be sent to the controller | |
3609 | * if resubmitting from an internal table. */ | |
3610 | may_packet_in = true; | |
3611 | honor_table_miss = true; | |
3612 | } | |
9583bc14 EJ |
3613 | |
3614 | in_port = resubmit->in_port; | |
3615 | if (in_port == OFPP_IN_PORT) { | |
4e022ec0 | 3616 | in_port = ctx->xin->flow.in_port.ofp_port; |
9583bc14 EJ |
3617 | } |
3618 | ||
3619 | table_id = resubmit->table_id; | |
3620 | if (table_id == 255) { | |
3621 | table_id = ctx->table_id; | |
3622 | } | |
3623 | ||
adcf00ba AZ |
3624 | xlate_table_action(ctx, in_port, table_id, may_packet_in, |
3625 | honor_table_miss); | |
9583bc14 EJ |
3626 | } |
3627 | ||
3628 | static void | |
3629 | flood_packets(struct xlate_ctx *ctx, bool all) | |
3630 | { | |
46c88433 | 3631 | const struct xport *xport; |
9583bc14 | 3632 | |
46c88433 EJ |
3633 | HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) { |
3634 | if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3635 | continue; |
3636 | } | |
3637 | ||
3638 | if (all) { | |
e93ef1c7 | 3639 | compose_output_action__(ctx, xport->ofp_port, NULL, false); |
46c88433 | 3640 | } else if (!(xport->config & OFPUTIL_PC_NO_FLOOD)) { |
e93ef1c7 | 3641 | compose_output_action(ctx, xport->ofp_port, NULL); |
9583bc14 EJ |
3642 | } |
3643 | } | |
3644 | ||
2031ef97 | 3645 | ctx->nf_output_iface = NF_OUT_FLOOD; |
9583bc14 EJ |
3646 | } |
3647 | ||
3648 | static void | |
3649 | execute_controller_action(struct xlate_ctx *ctx, int len, | |
3650 | enum ofp_packet_in_reason reason, | |
bdcad671 BP |
3651 | uint16_t controller_id, |
3652 | const uint8_t *userdata, size_t userdata_len) | |
9583bc14 | 3653 | { |
1895cc8d | 3654 | struct dp_packet_batch batch; |
e14deea0 | 3655 | struct dp_packet *packet; |
9583bc14 | 3656 | |
04594cd5 | 3657 | ctx->xout->slow |= SLOW_CONTROLLER; |
b476e2f2 | 3658 | xlate_commit_actions(ctx); |
9583bc14 EJ |
3659 | if (!ctx->xin->packet) { |
3660 | return; | |
3661 | } | |
3662 | ||
cf62fa4c | 3663 | packet = dp_packet_clone(ctx->xin->packet); |
1895cc8d PS |
3664 | packet_batch_init_packet(&batch, packet); |
3665 | odp_execute_actions(NULL, &batch, false, | |
1520ef4f | 3666 | ctx->odp_actions->data, ctx->odp_actions->size, NULL); |
9583bc14 | 3667 | |
9bfe9334 BP |
3668 | /* A packet sent by an action in a table-miss rule is considered an |
3669 | * explicit table miss. OpenFlow before 1.3 doesn't have that concept so | |
3670 | * it will get translated back to OFPR_ACTION for those versions. */ | |
3671 | if (reason == OFPR_ACTION | |
3672 | && ctx->rule && rule_dpif_is_table_miss(ctx->rule)) { | |
3673 | reason = OFPR_EXPLICIT_MISS; | |
3674 | } | |
3675 | ||
3676 | size_t packet_len = dp_packet_size(packet); | |
0fb7792a | 3677 | |
a2b53dec BP |
3678 | struct ofproto_async_msg *am = xmalloc(sizeof *am); |
3679 | *am = (struct ofproto_async_msg) { | |
9bfe9334 | 3680 | .controller_id = controller_id, |
a2b53dec BP |
3681 | .oam = OAM_PACKET_IN, |
3682 | .pin = { | |
3683 | .up = { | |
77ab5fd2 BP |
3684 | .public = { |
3685 | .packet = dp_packet_steal_data(packet), | |
3686 | .packet_len = packet_len, | |
3687 | .reason = reason, | |
3688 | .table_id = ctx->table_id, | |
3689 | .cookie = ctx->rule_cookie, | |
3690 | .userdata = (userdata_len | |
3691 | ? xmemdup(userdata, userdata_len) | |
3692 | : NULL), | |
3693 | .userdata_len = userdata_len, | |
3694 | } | |
a2b53dec BP |
3695 | }, |
3696 | .max_len = len, | |
9bfe9334 | 3697 | }, |
9bfe9334 | 3698 | }; |
77ab5fd2 | 3699 | flow_get_metadata(&ctx->xin->flow, &am->pin.up.public.flow_metadata); |
9583bc14 | 3700 | |
a2b53dec | 3701 | ofproto_dpif_send_async_msg(ctx->xbridge->ofproto, am); |
e14deea0 | 3702 | dp_packet_delete(packet); |
9583bc14 EJ |
3703 | } |
3704 | ||
7bbdd84f | 3705 | static void |
77ab5fd2 | 3706 | emit_continuation(struct xlate_ctx *ctx, const struct frozen_state *state) |
7bbdd84f | 3707 | { |
77ab5fd2 BP |
3708 | struct ofproto_async_msg *am = xmalloc(sizeof *am); |
3709 | *am = (struct ofproto_async_msg) { | |
3710 | .controller_id = ctx->pause->controller_id, | |
3711 | .oam = OAM_PACKET_IN, | |
3712 | .pin = { | |
3713 | .up = { | |
3714 | .public = { | |
3715 | .userdata = xmemdup(ctx->pause->userdata, | |
3716 | ctx->pause->userdata_len), | |
3717 | .userdata_len = ctx->pause->userdata_len, | |
3718 | .packet = xmemdup(dp_packet_data(ctx->xin->packet), | |
3719 | dp_packet_size(ctx->xin->packet)), | |
3720 | .packet_len = dp_packet_size(ctx->xin->packet), | |
0b024e49 | 3721 | .reason = ctx->pause->reason, |
77ab5fd2 BP |
3722 | }, |
3723 | .bridge = *ofproto_dpif_get_uuid(ctx->xbridge->ofproto), | |
3724 | .stack = xmemdup(state->stack, | |
3725 | state->n_stack * sizeof *state->stack), | |
3726 | .n_stack = state->n_stack, | |
3727 | .mirrors = state->mirrors, | |
3728 | .conntracked = state->conntracked, | |
3729 | .actions = xmemdup(state->ofpacts, state->ofpacts_len), | |
3730 | .actions_len = state->ofpacts_len, | |
3731 | .action_set = xmemdup(state->action_set, | |
3732 | state->action_set_len), | |
3733 | .action_set_len = state->action_set_len, | |
3734 | }, | |
3735 | .max_len = UINT16_MAX, | |
3736 | }, | |
3737 | }; | |
3738 | flow_get_metadata(&ctx->xin->flow, &am->pin.up.public.flow_metadata); | |
3739 | ofproto_dpif_send_async_msg(ctx->xbridge->ofproto, am); | |
3740 | } | |
7bbdd84f | 3741 | |
77ab5fd2 BP |
3742 | static void |
3743 | finish_freezing__(struct xlate_ctx *ctx, uint8_t table) | |
3744 | { | |
1d361a81 | 3745 | ovs_assert(ctx->freezing); |
7bbdd84f | 3746 | |
1d361a81 | 3747 | struct frozen_state state = { |
07659514 | 3748 | .table_id = table, |
290835f9 | 3749 | .ofproto_uuid = *ofproto_dpif_get_uuid(ctx->xbridge->ofproto), |
5c1b2314 BP |
3750 | .stack = ctx->stack.data, |
3751 | .n_stack = ctx->stack.size / sizeof(union mf_subvalue), | |
29bae541 | 3752 | .mirrors = ctx->mirrors, |
07659514 | 3753 | .conntracked = ctx->conntracked, |
1d361a81 BP |
3754 | .ofpacts = ctx->frozen_actions.data, |
3755 | .ofpacts_len = ctx->frozen_actions.size, | |
417509fa | 3756 | .action_set = ctx->action_set.data, |
8a5fb3b4 | 3757 | .action_set_len = ctx->action_set.size, |
2082425c | 3758 | }; |
77ab5fd2 | 3759 | frozen_metadata_from_flow(&state.metadata, &ctx->xin->flow); |
2082425c | 3760 | |
77ab5fd2 BP |
3761 | if (ctx->pause) { |
3762 | if (ctx->xin->packet) { | |
3763 | emit_continuation(ctx, &state); | |
3764 | } | |
3765 | } else { | |
3766 | /* Allocate a unique recirc id for the given metadata state in the | |
3767 | * flow. An existing id, with a new reference to the corresponding | |
3768 | * recirculation context, will be returned if possible. | |
3769 | * The life-cycle of this recirc id is managed by associating it | |
3770 | * with the udpif key ('ukey') created for each new datapath flow. */ | |
3771 | uint32_t id = recirc_alloc_id_ctx(&state); | |
3772 | if (!id) { | |
3773 | XLATE_REPORT_ERROR(ctx, "Failed to allocate recirculation id"); | |
3774 | ctx->error = XLATE_NO_RECIRCULATION_CONTEXT; | |
3775 | return; | |
3776 | } | |
3777 | recirc_refs_add(&ctx->xout->recircs, id); | |
7bbdd84f | 3778 | |
77ab5fd2 BP |
3779 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, id); |
3780 | } | |
e672ff9b | 3781 | |
1d361a81 BP |
3782 | /* Undo changes done by freezing. */ |
3783 | ctx_cancel_freeze(ctx); | |
7bbdd84f SH |
3784 | } |
3785 | ||
1d361a81 | 3786 | /* Called only when we're freezing. */ |
07659514 | 3787 | static void |
77ab5fd2 | 3788 | finish_freezing(struct xlate_ctx *ctx) |
07659514 JS |
3789 | { |
3790 | xlate_commit_actions(ctx); | |
77ab5fd2 | 3791 | finish_freezing__(ctx, 0); |
07659514 JS |
3792 | } |
3793 | ||
e37b8437 JS |
3794 | /* Fork the pipeline here. The current packet will continue processing the |
3795 | * current action list. A clone of the current packet will recirculate, skip | |
3796 | * the remainder of the current action list and asynchronously resume pipeline | |
3797 | * processing in 'table' with the current metadata and action set. */ | |
3798 | static void | |
3799 | compose_recirculate_and_fork(struct xlate_ctx *ctx, uint8_t table) | |
3800 | { | |
1d361a81 | 3801 | ctx->freezing = true; |
77ab5fd2 | 3802 | finish_freezing__(ctx, table); |
e37b8437 JS |
3803 | } |
3804 | ||
8bfd0fda BP |
3805 | static void |
3806 | compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) | |
9583bc14 | 3807 | { |
33bf9176 | 3808 | struct flow *flow = &ctx->xin->flow; |
8bfd0fda | 3809 | int n; |
33bf9176 | 3810 | |
8bfd0fda | 3811 | ovs_assert(eth_type_mpls(mpls->ethertype)); |
b0a17866 | 3812 | |
49a73e0c | 3813 | n = flow_count_mpls_labels(flow, ctx->wc); |
8bfd0fda | 3814 | if (!n) { |
704bb0bf | 3815 | xlate_commit_actions(ctx); |
8bfd0fda BP |
3816 | } else if (n >= FLOW_MAX_MPLS_LABELS) { |
3817 | if (ctx->xin->packet != NULL) { | |
fff1b9c0 | 3818 | XLATE_REPORT_ERROR(ctx, "bridge %s: dropping packet on which an " |
8bfd0fda BP |
3819 | "MPLS push action can't be performed as it would " |
3820 | "have more MPLS LSEs than the %d supported.", | |
3821 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
9583bc14 | 3822 | } |
fff1b9c0 | 3823 | ctx->error = XLATE_TOO_MANY_MPLS_LABELS; |
8bfd0fda | 3824 | return; |
9583bc14 | 3825 | } |
b0a17866 | 3826 | |
49a73e0c | 3827 | flow_push_mpls(flow, n, mpls->ethertype, ctx->wc); |
9583bc14 EJ |
3828 | } |
3829 | ||
8bfd0fda | 3830 | static void |
9cfef3d0 | 3831 | compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) |
9583bc14 | 3832 | { |
8bfd0fda | 3833 | struct flow *flow = &ctx->xin->flow; |
49a73e0c | 3834 | int n = flow_count_mpls_labels(flow, ctx->wc); |
33bf9176 | 3835 | |
49a73e0c | 3836 | if (flow_pop_mpls(flow, n, eth_type, ctx->wc)) { |
8bf009bf | 3837 | if (!eth_type_mpls(eth_type) && ctx->xbridge->support.odp.recirc) { |
e12ec36b | 3838 | ctx->was_mpls = true; |
7bbdd84f SH |
3839 | } |
3840 | } else if (n >= FLOW_MAX_MPLS_LABELS) { | |
8bfd0fda | 3841 | if (ctx->xin->packet != NULL) { |
fff1b9c0 | 3842 | XLATE_REPORT_ERROR(ctx, "bridge %s: dropping packet on which an " |
8bfd0fda BP |
3843 | "MPLS pop action can't be performed as it has " |
3844 | "more MPLS LSEs than the %d supported.", | |
3845 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
3846 | } | |
fff1b9c0 | 3847 | ctx->error = XLATE_TOO_MANY_MPLS_LABELS; |
1520ef4f | 3848 | ofpbuf_clear(ctx->odp_actions); |
9583bc14 EJ |
3849 | } |
3850 | } | |
3851 | ||
3852 | static bool | |
3853 | compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids) | |
3854 | { | |
33bf9176 BP |
3855 | struct flow *flow = &ctx->xin->flow; |
3856 | ||
3857 | if (!is_ip_any(flow)) { | |
9583bc14 EJ |
3858 | return false; |
3859 | } | |
3860 | ||
49a73e0c | 3861 | ctx->wc->masks.nw_ttl = 0xff; |
33bf9176 BP |
3862 | if (flow->nw_ttl > 1) { |
3863 | flow->nw_ttl--; | |
9583bc14 EJ |
3864 | return false; |
3865 | } else { | |
3866 | size_t i; | |
3867 | ||
3868 | for (i = 0; i < ids->n_controllers; i++) { | |
3869 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, | |
bdcad671 | 3870 | ids->cnt_ids[i], NULL, 0); |
9583bc14 EJ |
3871 | } |
3872 | ||
3873 | /* Stop processing for current table. */ | |
3874 | return true; | |
3875 | } | |
3876 | } | |
3877 | ||
8bfd0fda | 3878 | static void |
097d4939 JR |
3879 | compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label) |
3880 | { | |
8bfd0fda | 3881 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3882 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK); |
8bfd0fda | 3883 | set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label); |
097d4939 | 3884 | } |
097d4939 JR |
3885 | } |
3886 | ||
8bfd0fda | 3887 | static void |
097d4939 JR |
3888 | compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc) |
3889 | { | |
8bfd0fda | 3890 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3891 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TC_MASK); |
8bfd0fda | 3892 | set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc); |
097d4939 | 3893 | } |
097d4939 JR |
3894 | } |
3895 | ||
8bfd0fda | 3896 | static void |
9cfef3d0 | 3897 | compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl) |
9583bc14 | 3898 | { |
8bfd0fda | 3899 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3900 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda | 3901 | set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl); |
b0a17866 | 3902 | } |
9583bc14 EJ |
3903 | } |
3904 | ||
3905 | static bool | |
9cfef3d0 | 3906 | compose_dec_mpls_ttl_action(struct xlate_ctx *ctx) |
9583bc14 | 3907 | { |
33bf9176 | 3908 | struct flow *flow = &ctx->xin->flow; |
1dd35f8a | 3909 | |
8bfd0fda | 3910 | if (eth_type_mpls(flow->dl_type)) { |
22d38fca JR |
3911 | uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]); |
3912 | ||
49a73e0c | 3913 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda BP |
3914 | if (ttl > 1) { |
3915 | ttl--; | |
3916 | set_mpls_lse_ttl(&flow->mpls_lse[0], ttl); | |
3917 | return false; | |
3918 | } else { | |
bdcad671 BP |
3919 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0, |
3920 | NULL, 0); | |
8bfd0fda | 3921 | } |
9583bc14 | 3922 | } |
22d38fca JR |
3923 | |
3924 | /* Stop processing for current table. */ | |
3925 | return true; | |
9583bc14 EJ |
3926 | } |
3927 | ||
3928 | static void | |
3929 | xlate_output_action(struct xlate_ctx *ctx, | |
4e022ec0 | 3930 | ofp_port_t port, uint16_t max_len, bool may_packet_in) |
9583bc14 | 3931 | { |
2031ef97 | 3932 | ofp_port_t prev_nf_output_iface = ctx->nf_output_iface; |
9583bc14 | 3933 | |
2031ef97 | 3934 | ctx->nf_output_iface = NF_OUT_DROP; |
9583bc14 EJ |
3935 | |
3936 | switch (port) { | |
3937 | case OFPP_IN_PORT: | |
e93ef1c7 | 3938 | compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port, NULL); |
9583bc14 EJ |
3939 | break; |
3940 | case OFPP_TABLE: | |
4e022ec0 | 3941 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 3942 | 0, may_packet_in, true); |
9583bc14 EJ |
3943 | break; |
3944 | case OFPP_NORMAL: | |
3945 | xlate_normal(ctx); | |
3946 | break; | |
3947 | case OFPP_FLOOD: | |
3948 | flood_packets(ctx, false); | |
3949 | break; | |
3950 | case OFPP_ALL: | |
3951 | flood_packets(ctx, true); | |
3952 | break; | |
3953 | case OFPP_CONTROLLER: | |
3a11fd5b | 3954 | execute_controller_action(ctx, max_len, |
029ca940 SS |
3955 | (ctx->in_group ? OFPR_GROUP |
3956 | : ctx->in_action_set ? OFPR_ACTION_SET | |
3957 | : OFPR_ACTION), | |
bdcad671 | 3958 | 0, NULL, 0); |
9583bc14 EJ |
3959 | break; |
3960 | case OFPP_NONE: | |
3961 | break; | |
3962 | case OFPP_LOCAL: | |
3963 | default: | |
4e022ec0 | 3964 | if (port != ctx->xin->flow.in_port.ofp_port) { |
e93ef1c7 | 3965 | compose_output_action(ctx, port, NULL); |
9583bc14 EJ |
3966 | } else { |
3967 | xlate_report(ctx, "skipping output to input port"); | |
3968 | } | |
3969 | break; | |
3970 | } | |
3971 | ||
3972 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
2031ef97 BP |
3973 | ctx->nf_output_iface = NF_OUT_FLOOD; |
3974 | } else if (ctx->nf_output_iface == NF_OUT_DROP) { | |
3975 | ctx->nf_output_iface = prev_nf_output_iface; | |
9583bc14 | 3976 | } else if (prev_nf_output_iface != NF_OUT_DROP && |
2031ef97 BP |
3977 | ctx->nf_output_iface != NF_OUT_FLOOD) { |
3978 | ctx->nf_output_iface = NF_OUT_MULTI; | |
9583bc14 EJ |
3979 | } |
3980 | } | |
3981 | ||
3982 | static void | |
3983 | xlate_output_reg_action(struct xlate_ctx *ctx, | |
3984 | const struct ofpact_output_reg *or) | |
3985 | { | |
3986 | uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow); | |
3987 | if (port <= UINT16_MAX) { | |
3988 | union mf_subvalue value; | |
3989 | ||
3990 | memset(&value, 0xff, sizeof value); | |
49a73e0c | 3991 | mf_write_subfield_flow(&or->src, &value, &ctx->wc->masks); |
4e022ec0 AW |
3992 | xlate_output_action(ctx, u16_to_ofp(port), |
3993 | or->max_len, false); | |
9583bc14 EJ |
3994 | } |
3995 | } | |
3996 | ||
aaca4fe0 WT |
3997 | static void |
3998 | xlate_output_trunc_action(struct xlate_ctx *ctx, | |
3999 | ofp_port_t port, uint32_t max_len) | |
4000 | { | |
4001 | bool support_trunc = ctx->xbridge->support.trunc; | |
4002 | struct ovs_action_trunc *trunc; | |
4003 | char name[OFP_MAX_PORT_NAME_LEN]; | |
4004 | ||
4005 | switch (port) { | |
4006 | case OFPP_TABLE: | |
4007 | case OFPP_NORMAL: | |
4008 | case OFPP_FLOOD: | |
4009 | case OFPP_ALL: | |
4010 | case OFPP_CONTROLLER: | |
4011 | case OFPP_NONE: | |
4012 | ofputil_port_to_string(port, name, sizeof name); | |
4013 | xlate_report(ctx, "output_trunc does not support port: %s", name); | |
4014 | break; | |
4015 | case OFPP_LOCAL: | |
4016 | case OFPP_IN_PORT: | |
4017 | default: | |
4018 | if (port != ctx->xin->flow.in_port.ofp_port) { | |
4019 | const struct xport *xport = get_ofp_port(ctx->xbridge, port); | |
4020 | ||
4021 | if (xport == NULL || xport->odp_port == ODPP_NONE) { | |
4022 | /* Since truncate happens at its following output action, if | |
4023 | * the output port is a patch port, the behavior is somehow | |
4024 | * unpredicable. For simpilicity, disallow this case. */ | |
4025 | ofputil_port_to_string(port, name, sizeof name); | |
4026 | XLATE_REPORT_ERROR(ctx, "bridge %s: " | |
4027 | "output_trunc does not support port: %s", | |
4028 | ctx->xbridge->name, name); | |
4029 | break; | |
4030 | } | |
4031 | ||
4032 | trunc = nl_msg_put_unspec_uninit(ctx->odp_actions, | |
4033 | OVS_ACTION_ATTR_TRUNC, | |
4034 | sizeof *trunc); | |
4035 | trunc->max_len = max_len; | |
4036 | xlate_output_action(ctx, port, max_len, false); | |
4037 | if (!support_trunc) { | |
4038 | ctx->xout->slow |= SLOW_ACTION; | |
4039 | } | |
4040 | } else { | |
4041 | xlate_report(ctx, "skipping output to input port"); | |
4042 | } | |
4043 | break; | |
4044 | } | |
4045 | } | |
4046 | ||
9583bc14 EJ |
4047 | static void |
4048 | xlate_enqueue_action(struct xlate_ctx *ctx, | |
4049 | const struct ofpact_enqueue *enqueue) | |
4050 | { | |
4e022ec0 | 4051 | ofp_port_t ofp_port = enqueue->port; |
9583bc14 EJ |
4052 | uint32_t queue_id = enqueue->queue; |
4053 | uint32_t flow_priority, priority; | |
4054 | int error; | |
4055 | ||
4056 | /* Translate queue to priority. */ | |
89a8a7f0 | 4057 | error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority); |
9583bc14 EJ |
4058 | if (error) { |
4059 | /* Fall back to ordinary output action. */ | |
4060 | xlate_output_action(ctx, enqueue->port, 0, false); | |
4061 | return; | |
4062 | } | |
4063 | ||
4064 | /* Check output port. */ | |
4065 | if (ofp_port == OFPP_IN_PORT) { | |
4e022ec0 AW |
4066 | ofp_port = ctx->xin->flow.in_port.ofp_port; |
4067 | } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
4068 | return; |
4069 | } | |
4070 | ||
4071 | /* Add datapath actions. */ | |
4072 | flow_priority = ctx->xin->flow.skb_priority; | |
4073 | ctx->xin->flow.skb_priority = priority; | |
e93ef1c7 | 4074 | compose_output_action(ctx, ofp_port, NULL); |
9583bc14 EJ |
4075 | ctx->xin->flow.skb_priority = flow_priority; |
4076 | ||
4077 | /* Update NetFlow output port. */ | |
2031ef97 BP |
4078 | if (ctx->nf_output_iface == NF_OUT_DROP) { |
4079 | ctx->nf_output_iface = ofp_port; | |
4080 | } else if (ctx->nf_output_iface != NF_OUT_FLOOD) { | |
4081 | ctx->nf_output_iface = NF_OUT_MULTI; | |
9583bc14 EJ |
4082 | } |
4083 | } | |
4084 | ||
4085 | static void | |
4086 | xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id) | |
4087 | { | |
4088 | uint32_t skb_priority; | |
4089 | ||
89a8a7f0 | 4090 | if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) { |
9583bc14 EJ |
4091 | ctx->xin->flow.skb_priority = skb_priority; |
4092 | } else { | |
4093 | /* Couldn't translate queue to a priority. Nothing to do. A warning | |
4094 | * has already been logged. */ | |
4095 | } | |
4096 | } | |
4097 | ||
4098 | static bool | |
46c88433 | 4099 | slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_) |
9583bc14 | 4100 | { |
46c88433 EJ |
4101 | const struct xbridge *xbridge = xbridge_; |
4102 | struct xport *port; | |
9583bc14 EJ |
4103 | |
4104 | switch (ofp_port) { | |
4105 | case OFPP_IN_PORT: | |
4106 | case OFPP_TABLE: | |
4107 | case OFPP_NORMAL: | |
4108 | case OFPP_FLOOD: | |
4109 | case OFPP_ALL: | |
4110 | case OFPP_NONE: | |
4111 | return true; | |
4112 | case OFPP_CONTROLLER: /* Not supported by the bundle action. */ | |
4113 | return false; | |
4114 | default: | |
46c88433 | 4115 | port = get_ofp_port(xbridge, ofp_port); |
9583bc14 EJ |
4116 | return port ? port->may_enable : false; |
4117 | } | |
4118 | } | |
4119 | ||
4120 | static void | |
4121 | xlate_bundle_action(struct xlate_ctx *ctx, | |
4122 | const struct ofpact_bundle *bundle) | |
4123 | { | |
4e022ec0 | 4124 | ofp_port_t port; |
9583bc14 | 4125 | |
49a73e0c | 4126 | port = bundle_execute(bundle, &ctx->xin->flow, ctx->wc, slave_enabled_cb, |
46c88433 | 4127 | CONST_CAST(struct xbridge *, ctx->xbridge)); |
9583bc14 | 4128 | if (bundle->dst.field) { |
49a73e0c | 4129 | nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow, ctx->wc); |
9583bc14 EJ |
4130 | } else { |
4131 | xlate_output_action(ctx, port, 0, false); | |
4132 | } | |
4133 | } | |
4134 | ||
4135 | static void | |
4165b5e0 JS |
4136 | xlate_learn_action__(struct xlate_ctx *ctx, const struct ofpact_learn *learn, |
4137 | struct ofputil_flow_mod *fm, struct ofpbuf *ofpacts) | |
9583bc14 | 4138 | { |
4165b5e0 JS |
4139 | learn_execute(learn, &ctx->xin->flow, fm, ofpacts); |
4140 | if (ctx->xin->may_learn) { | |
4141 | ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm); | |
4142 | } | |
4143 | } | |
9583bc14 | 4144 | |
4165b5e0 JS |
4145 | static void |
4146 | xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) | |
4147 | { | |
49a73e0c | 4148 | learn_mask(learn, ctx->wc); |
9583bc14 | 4149 | |
b256dc52 JS |
4150 | if (ctx->xin->xcache) { |
4151 | struct xc_entry *entry; | |
4152 | ||
4153 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN); | |
4165b5e0 JS |
4154 | entry->u.learn.ofproto = ctx->xbridge->ofproto; |
4155 | entry->u.learn.fm = xmalloc(sizeof *entry->u.learn.fm); | |
4156 | entry->u.learn.ofpacts = ofpbuf_new(64); | |
4157 | xlate_learn_action__(ctx, learn, entry->u.learn.fm, | |
4158 | entry->u.learn.ofpacts); | |
4159 | } else if (ctx->xin->may_learn) { | |
4160 | uint64_t ofpacts_stub[1024 / 8]; | |
4161 | struct ofputil_flow_mod fm; | |
4162 | struct ofpbuf ofpacts; | |
4163 | ||
4164 | ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); | |
4165 | xlate_learn_action__(ctx, learn, &fm, &ofpacts); | |
4166 | ofpbuf_uninit(&ofpacts); | |
b256dc52 JS |
4167 | } |
4168 | } | |
4169 | ||
4170 | static void | |
4171 | xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags, | |
4172 | uint16_t idle_timeout, uint16_t hard_timeout) | |
4173 | { | |
4174 | if (tcp_flags & (TCP_FIN | TCP_RST)) { | |
4175 | rule_dpif_reduce_timeouts(rule, idle_timeout, hard_timeout); | |
4176 | } | |
9583bc14 EJ |
4177 | } |
4178 | ||
9583bc14 EJ |
4179 | static void |
4180 | xlate_fin_timeout(struct xlate_ctx *ctx, | |
4181 | const struct ofpact_fin_timeout *oft) | |
4182 | { | |
b256dc52 JS |
4183 | if (ctx->rule) { |
4184 | xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags, | |
4185 | oft->fin_idle_timeout, oft->fin_hard_timeout); | |
4186 | if (ctx->xin->xcache) { | |
4187 | struct xc_entry *entry; | |
4188 | ||
4189 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT); | |
83709dfa JR |
4190 | /* XC_RULE already holds a reference on the rule, none is taken |
4191 | * here. */ | |
b256dc52 JS |
4192 | entry->u.fin.rule = ctx->rule; |
4193 | entry->u.fin.idle = oft->fin_idle_timeout; | |
4194 | entry->u.fin.hard = oft->fin_hard_timeout; | |
b256dc52 | 4195 | } |
9583bc14 EJ |
4196 | } |
4197 | } | |
4198 | ||
4199 | static void | |
4200 | xlate_sample_action(struct xlate_ctx *ctx, | |
4201 | const struct ofpact_sample *os) | |
4202 | { | |
f69f713b BY |
4203 | odp_port_t output_odp_port = ODPP_NONE; |
4204 | odp_port_t tunnel_out_port = ODPP_NONE; | |
4205 | struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; | |
4206 | bool emit_set_tunnel = false; | |
4207 | ||
4208 | if (!ipfix || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { | |
4209 | return; | |
4210 | } | |
4211 | ||
e824d78d JR |
4212 | /* Scale the probability from 16-bit to 32-bit while representing |
4213 | * the same percentage. */ | |
4214 | uint32_t probability = (os->probability << 16) | os->probability; | |
4215 | ||
b440dd8c | 4216 | if (!ctx->xbridge->support.variable_length_userdata) { |
e824d78d JR |
4217 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); |
4218 | ||
4219 | VLOG_ERR_RL(&rl, "ignoring NXAST_SAMPLE action because datapath " | |
4220 | "lacks support (needs Linux 3.10+ or kernel module from " | |
4221 | "OVS 1.11+)"); | |
4222 | return; | |
4223 | } | |
4224 | ||
f69f713b BY |
4225 | /* If ofp_port in flow sample action is equel to ofp_port, |
4226 | * this sample action is a input port action. */ | |
4227 | if (os->sampling_port != OFPP_NONE && | |
4228 | os->sampling_port != ctx->xin->flow.in_port.ofp_port) { | |
4229 | output_odp_port = ofp_port_to_odp_port(ctx->xbridge, | |
4230 | os->sampling_port); | |
4231 | if (output_odp_port == ODPP_NONE) { | |
4232 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4233 | VLOG_WARN_RL(&rl, "can't use unknown port %d in flow sample " | |
4234 | "action", os->sampling_port); | |
4235 | return; | |
4236 | } | |
4237 | ||
4238 | if (dpif_ipfix_get_flow_exporter_tunnel_sampling(ipfix, | |
4239 | os->collector_set_id) | |
4240 | && dpif_ipfix_get_tunnel_port(ipfix, output_odp_port)) { | |
4241 | tunnel_out_port = output_odp_port; | |
4242 | emit_set_tunnel = true; | |
4243 | } | |
4244 | } | |
4245 | ||
4246 | xlate_commit_actions(ctx); | |
4247 | /* If 'emit_set_tunnel', sample(sampling_port=1) would translate | |
4248 | * into datapath sample action set(tunnel(...)), sample(...) and | |
4249 | * it is used for sampling egress tunnel information. */ | |
4250 | if (emit_set_tunnel) { | |
4251 | const struct xport *xport = get_ofp_port(ctx->xbridge, | |
4252 | os->sampling_port); | |
4253 | ||
4254 | if (xport && xport->is_tunnel) { | |
4255 | struct flow *flow = &ctx->xin->flow; | |
4256 | tnl_port_send(xport->ofport, flow, ctx->wc); | |
4257 | if (!ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { | |
4258 | struct flow_tnl flow_tnl = flow->tunnel; | |
4259 | ||
4260 | commit_odp_tunnel_action(flow, &ctx->base_flow, | |
4261 | ctx->odp_actions); | |
4262 | flow->tunnel = flow_tnl; | |
4263 | } | |
4264 | } else { | |
4265 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4266 | VLOG_WARN_RL(&rl, "sampling_port:%d should be a tunnel port.", | |
4267 | os->sampling_port); | |
4268 | } | |
4269 | } | |
e824d78d | 4270 | |
a6092018 BP |
4271 | union user_action_cookie cookie = { |
4272 | .flow_sample = { | |
4273 | .type = USER_ACTION_COOKIE_FLOW_SAMPLE, | |
4274 | .probability = os->probability, | |
4275 | .collector_set_id = os->collector_set_id, | |
4276 | .obs_domain_id = os->obs_domain_id, | |
4277 | .obs_point_id = os->obs_point_id, | |
f69f713b | 4278 | .output_odp_port = output_odp_port, |
a6092018 BP |
4279 | } |
4280 | }; | |
4281 | compose_sample_action(ctx, probability, &cookie, sizeof cookie.flow_sample, | |
f69f713b | 4282 | tunnel_out_port, false); |
9583bc14 EJ |
4283 | } |
4284 | ||
4285 | static bool | |
46c88433 | 4286 | may_receive(const struct xport *xport, struct xlate_ctx *ctx) |
9583bc14 | 4287 | { |
bbbca389 | 4288 | if (xport->config & (is_stp(&ctx->xin->flow) |
46c88433 EJ |
4289 | ? OFPUTIL_PC_NO_RECV_STP |
4290 | : OFPUTIL_PC_NO_RECV)) { | |
9583bc14 EJ |
4291 | return false; |
4292 | } | |
4293 | ||
4294 | /* Only drop packets here if both forwarding and learning are | |
4295 | * disabled. If just learning is enabled, we need to have | |
4296 | * OFPP_NORMAL and the learning action have a look at the packet | |
4297 | * before we can drop it. */ | |
9efd308e DV |
4298 | if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) || |
4299 | (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) { | |
9583bc14 EJ |
4300 | return false; |
4301 | } | |
4302 | ||
4303 | return true; | |
4304 | } | |
4305 | ||
7fdb60a7 | 4306 | static void |
7e7e8dbb BP |
4307 | xlate_write_actions__(struct xlate_ctx *ctx, |
4308 | const struct ofpact *ofpacts, size_t ofpacts_len) | |
7fdb60a7 | 4309 | { |
c61f3870 BP |
4310 | /* Maintain actset_output depending on the contents of the action set: |
4311 | * | |
4312 | * - OFPP_UNSET, if there is no "output" action. | |
4313 | * | |
4314 | * - The output port, if there is an "output" action and no "group" | |
4315 | * action. | |
4316 | * | |
4317 | * - OFPP_UNSET, if there is a "group" action. | |
4318 | */ | |
4319 | if (!ctx->action_set_has_group) { | |
7e7e8dbb BP |
4320 | const struct ofpact *a; |
4321 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { | |
4322 | if (a->type == OFPACT_OUTPUT) { | |
4323 | ctx->xin->flow.actset_output = ofpact_get_OUTPUT(a)->port; | |
4324 | } else if (a->type == OFPACT_GROUP) { | |
c61f3870 BP |
4325 | ctx->xin->flow.actset_output = OFPP_UNSET; |
4326 | ctx->action_set_has_group = true; | |
9055ca9a | 4327 | break; |
c61f3870 BP |
4328 | } |
4329 | } | |
4330 | } | |
4331 | ||
7e7e8dbb BP |
4332 | ofpbuf_put(&ctx->action_set, ofpacts, ofpacts_len); |
4333 | } | |
4334 | ||
4335 | static void | |
4336 | xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact_nest *a) | |
4337 | { | |
4338 | xlate_write_actions__(ctx, a->actions, ofpact_nest_get_action_len(a)); | |
7fdb60a7 SH |
4339 | } |
4340 | ||
4341 | static void | |
4342 | xlate_action_set(struct xlate_ctx *ctx) | |
4343 | { | |
4344 | uint64_t action_list_stub[1024 / 64]; | |
4345 | struct ofpbuf action_list; | |
4346 | ||
029ca940 | 4347 | ctx->in_action_set = true; |
7fdb60a7 SH |
4348 | ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub); |
4349 | ofpacts_execute_action_set(&action_list, &ctx->action_set); | |
ed9c9e3e JR |
4350 | /* Clear the action set, as it is not needed any more. */ |
4351 | ofpbuf_clear(&ctx->action_set); | |
6fd6ed71 | 4352 | do_xlate_actions(action_list.data, action_list.size, ctx); |
029ca940 | 4353 | ctx->in_action_set = false; |
7fdb60a7 SH |
4354 | ofpbuf_uninit(&action_list); |
4355 | } | |
4356 | ||
e672ff9b | 4357 | static void |
1d361a81 | 4358 | freeze_put_unroll_xlate(struct xlate_ctx *ctx) |
e672ff9b | 4359 | { |
1d361a81 | 4360 | struct ofpact_unroll_xlate *unroll = ctx->frozen_actions.header; |
e672ff9b JR |
4361 | |
4362 | /* Restore the table_id and rule cookie for a potential PACKET | |
4363 | * IN if needed. */ | |
4364 | if (!unroll || | |
4365 | (ctx->table_id != unroll->rule_table_id | |
4366 | || ctx->rule_cookie != unroll->rule_cookie)) { | |
1d361a81 | 4367 | unroll = ofpact_put_UNROLL_XLATE(&ctx->frozen_actions); |
e672ff9b JR |
4368 | unroll->rule_table_id = ctx->table_id; |
4369 | unroll->rule_cookie = ctx->rule_cookie; | |
1d361a81 | 4370 | ctx->frozen_actions.header = unroll; |
e672ff9b JR |
4371 | } |
4372 | } | |
4373 | ||
4374 | ||
1d361a81 BP |
4375 | /* Copy actions 'a' through 'end' to ctx->frozen_actions, which will be |
4376 | * executed after thawing. Inserts an UNROLL_XLATE action, if none is already | |
4377 | * present, before any action that may depend on the current table ID or flow | |
4378 | * cookie. */ | |
e672ff9b | 4379 | static void |
1d361a81 | 4380 | freeze_unroll_actions(const struct ofpact *a, const struct ofpact *end, |
e672ff9b JR |
4381 | struct xlate_ctx *ctx) |
4382 | { | |
c2b283b7 | 4383 | for (; a < end; a = ofpact_next(a)) { |
e672ff9b | 4384 | switch (a->type) { |
e672ff9b | 4385 | case OFPACT_OUTPUT_REG: |
aaca4fe0 | 4386 | case OFPACT_OUTPUT_TRUNC: |
e672ff9b JR |
4387 | case OFPACT_GROUP: |
4388 | case OFPACT_OUTPUT: | |
4389 | case OFPACT_CONTROLLER: | |
4390 | case OFPACT_DEC_MPLS_TTL: | |
4391 | case OFPACT_DEC_TTL: | |
83a31283 BP |
4392 | /* These actions may generate asynchronous messages, which include |
4393 | * table ID and flow cookie information. */ | |
1d361a81 | 4394 | freeze_put_unroll_xlate(ctx); |
e672ff9b JR |
4395 | break; |
4396 | ||
83a31283 BP |
4397 | case OFPACT_RESUBMIT: |
4398 | if (ofpact_get_RESUBMIT(a)->table_id == 0xff) { | |
4399 | /* This resubmit action is relative to the current table, so we | |
4400 | * need to track what table that is.*/ | |
1d361a81 | 4401 | freeze_put_unroll_xlate(ctx); |
83a31283 BP |
4402 | } |
4403 | break; | |
4404 | ||
e672ff9b JR |
4405 | case OFPACT_SET_TUNNEL: |
4406 | case OFPACT_REG_MOVE: | |
4407 | case OFPACT_SET_FIELD: | |
4408 | case OFPACT_STACK_PUSH: | |
4409 | case OFPACT_STACK_POP: | |
4410 | case OFPACT_LEARN: | |
4411 | case OFPACT_WRITE_METADATA: | |
83a31283 | 4412 | case OFPACT_GOTO_TABLE: |
e672ff9b JR |
4413 | case OFPACT_ENQUEUE: |
4414 | case OFPACT_SET_VLAN_VID: | |
4415 | case OFPACT_SET_VLAN_PCP: | |
4416 | case OFPACT_STRIP_VLAN: | |
4417 | case OFPACT_PUSH_VLAN: | |
4418 | case OFPACT_SET_ETH_SRC: | |
4419 | case OFPACT_SET_ETH_DST: | |
4420 | case OFPACT_SET_IPV4_SRC: | |
4421 | case OFPACT_SET_IPV4_DST: | |
4422 | case OFPACT_SET_IP_DSCP: | |
4423 | case OFPACT_SET_IP_ECN: | |
4424 | case OFPACT_SET_IP_TTL: | |
4425 | case OFPACT_SET_L4_SRC_PORT: | |
4426 | case OFPACT_SET_L4_DST_PORT: | |
4427 | case OFPACT_SET_QUEUE: | |
4428 | case OFPACT_POP_QUEUE: | |
4429 | case OFPACT_PUSH_MPLS: | |
4430 | case OFPACT_POP_MPLS: | |
4431 | case OFPACT_SET_MPLS_LABEL: | |
4432 | case OFPACT_SET_MPLS_TC: | |
4433 | case OFPACT_SET_MPLS_TTL: | |
4434 | case OFPACT_MULTIPATH: | |
4435 | case OFPACT_BUNDLE: | |
4436 | case OFPACT_EXIT: | |
4437 | case OFPACT_UNROLL_XLATE: | |
4438 | case OFPACT_FIN_TIMEOUT: | |
4439 | case OFPACT_CLEAR_ACTIONS: | |
4440 | case OFPACT_WRITE_ACTIONS: | |
4441 | case OFPACT_METER: | |
4442 | case OFPACT_SAMPLE: | |
d4abaff5 | 4443 | case OFPACT_DEBUG_RECIRC: |
07659514 | 4444 | case OFPACT_CT: |
9ac0aada | 4445 | case OFPACT_NAT: |
83a31283 | 4446 | /* These may not generate PACKET INs. */ |
e672ff9b JR |
4447 | break; |
4448 | ||
e672ff9b JR |
4449 | case OFPACT_NOTE: |
4450 | case OFPACT_CONJUNCTION: | |
83a31283 | 4451 | /* These need not be copied for restoration. */ |
e672ff9b JR |
4452 | continue; |
4453 | } | |
4454 | /* Copy the action over. */ | |
1d361a81 | 4455 | ofpbuf_put(&ctx->frozen_actions, a, OFPACT_ALIGN(a->len)); |
e672ff9b JR |
4456 | } |
4457 | } | |
4458 | ||
8e53fe8c | 4459 | static void |
f2d105b5 JS |
4460 | put_ct_mark(const struct flow *flow, struct ofpbuf *odp_actions, |
4461 | struct flow_wildcards *wc) | |
8e53fe8c | 4462 | { |
2a754f4a JS |
4463 | if (wc->masks.ct_mark) { |
4464 | struct { | |
4465 | uint32_t key; | |
4466 | uint32_t mask; | |
4467 | } *odp_ct_mark; | |
4468 | ||
4469 | odp_ct_mark = nl_msg_put_unspec_uninit(odp_actions, OVS_CT_ATTR_MARK, | |
4470 | sizeof(*odp_ct_mark)); | |
4471 | odp_ct_mark->key = flow->ct_mark & wc->masks.ct_mark; | |
4472 | odp_ct_mark->mask = wc->masks.ct_mark; | |
8e53fe8c JS |
4473 | } |
4474 | } | |
4475 | ||
9daf2348 | 4476 | static void |
f2d105b5 JS |
4477 | put_ct_label(const struct flow *flow, struct ofpbuf *odp_actions, |
4478 | struct flow_wildcards *wc) | |
9daf2348 | 4479 | { |
2ff8484b | 4480 | if (!ovs_u128_is_zero(wc->masks.ct_label)) { |
9daf2348 JS |
4481 | struct { |
4482 | ovs_u128 key; | |
4483 | ovs_u128 mask; | |
4484 | } *odp_ct_label; | |
4485 | ||
4486 | odp_ct_label = nl_msg_put_unspec_uninit(odp_actions, | |
4487 | OVS_CT_ATTR_LABELS, | |
4488 | sizeof(*odp_ct_label)); | |
f2d105b5 | 4489 | odp_ct_label->key = ovs_u128_and(flow->ct_label, wc->masks.ct_label); |
9daf2348 JS |
4490 | odp_ct_label->mask = wc->masks.ct_label; |
4491 | } | |
4492 | } | |
4493 | ||
d787ad39 JS |
4494 | static void |
4495 | put_ct_helper(struct ofpbuf *odp_actions, struct ofpact_conntrack *ofc) | |
4496 | { | |
4497 | if (ofc->alg) { | |
4498 | if (ofc->alg == IPPORT_FTP) { | |
4499 | nl_msg_put_string(odp_actions, OVS_CT_ATTR_HELPER, "ftp"); | |
4500 | } else { | |
4501 | VLOG_WARN("Cannot serialize ct_helper %d\n", ofc->alg); | |
4502 | } | |
4503 | } | |
4504 | } | |
4505 | ||
9ac0aada JR |
4506 | static void |
4507 | put_ct_nat(struct xlate_ctx *ctx) | |
4508 | { | |
4509 | struct ofpact_nat *ofn = ctx->ct_nat_action; | |
4510 | size_t nat_offset; | |
4511 | ||
4512 | if (!ofn) { | |
4513 | return; | |
4514 | } | |
4515 | ||
4516 | nat_offset = nl_msg_start_nested(ctx->odp_actions, OVS_CT_ATTR_NAT); | |
4517 | if (ofn->flags & NX_NAT_F_SRC || ofn->flags & NX_NAT_F_DST) { | |
4518 | nl_msg_put_flag(ctx->odp_actions, ofn->flags & NX_NAT_F_SRC | |
4519 | ? OVS_NAT_ATTR_SRC : OVS_NAT_ATTR_DST); | |
4520 | if (ofn->flags & NX_NAT_F_PERSISTENT) { | |
4521 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PERSISTENT); | |
4522 | } | |
4523 | if (ofn->flags & NX_NAT_F_PROTO_HASH) { | |
4524 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_HASH); | |
4525 | } else if (ofn->flags & NX_NAT_F_PROTO_RANDOM) { | |
4526 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_RANDOM); | |
4527 | } | |
4528 | if (ofn->range_af == AF_INET) { | |
73e8bc23 | 4529 | nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, |
9ac0aada JR |
4530 | ofn->range.addr.ipv4.min); |
4531 | if (ofn->range.addr.ipv4.max && | |
73e8bc23 BP |
4532 | (ntohl(ofn->range.addr.ipv4.max) |
4533 | > ntohl(ofn->range.addr.ipv4.min))) { | |
4534 | nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, | |
4535 | ofn->range.addr.ipv4.max); | |
9ac0aada JR |
4536 | } |
4537 | } else if (ofn->range_af == AF_INET6) { | |
4538 | nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, | |
4539 | &ofn->range.addr.ipv6.min, | |
4540 | sizeof ofn->range.addr.ipv6.min); | |
4541 | if (!ipv6_mask_is_any(&ofn->range.addr.ipv6.max) && | |
4542 | memcmp(&ofn->range.addr.ipv6.max, &ofn->range.addr.ipv6.min, | |
4543 | sizeof ofn->range.addr.ipv6.max) > 0) { | |
4544 | nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, | |
4545 | &ofn->range.addr.ipv6.max, | |
4546 | sizeof ofn->range.addr.ipv6.max); | |
4547 | } | |
4548 | } | |
4549 | if (ofn->range_af != AF_UNSPEC && ofn->range.proto.min) { | |
4550 | nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MIN, | |
4551 | ofn->range.proto.min); | |
4552 | if (ofn->range.proto.max && | |
4553 | ofn->range.proto.max > ofn->range.proto.min) { | |
4554 | nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MAX, | |
4555 | ofn->range.proto.max); | |
4556 | } | |
4557 | } | |
4558 | } | |
4559 | nl_msg_end_nested(ctx->odp_actions, nat_offset); | |
4560 | } | |
4561 | ||
07659514 JS |
4562 | static void |
4563 | compose_conntrack_action(struct xlate_ctx *ctx, struct ofpact_conntrack *ofc) | |
4564 | { | |
9daf2348 | 4565 | ovs_u128 old_ct_label = ctx->base_flow.ct_label; |
f2d105b5 | 4566 | ovs_u128 old_ct_label_mask = ctx->wc->masks.ct_label; |
8e53fe8c | 4567 | uint32_t old_ct_mark = ctx->base_flow.ct_mark; |
f2d105b5 | 4568 | uint32_t old_ct_mark_mask = ctx->wc->masks.ct_mark; |
07659514 JS |
4569 | size_t ct_offset; |
4570 | uint16_t zone; | |
4571 | ||
4572 | /* Ensure that any prior actions are applied before composing the new | |
4573 | * conntrack action. */ | |
4574 | xlate_commit_actions(ctx); | |
4575 | ||
8e53fe8c | 4576 | /* Process nested actions first, to populate the key. */ |
9ac0aada | 4577 | ctx->ct_nat_action = NULL; |
f2d105b5 JS |
4578 | ctx->wc->masks.ct_mark = 0; |
4579 | ctx->wc->masks.ct_label.u64.hi = ctx->wc->masks.ct_label.u64.lo = 0; | |
8e53fe8c JS |
4580 | do_xlate_actions(ofc->actions, ofpact_ct_get_action_len(ofc), ctx); |
4581 | ||
07659514 JS |
4582 | if (ofc->zone_src.field) { |
4583 | zone = mf_get_subfield(&ofc->zone_src, &ctx->xin->flow); | |
4584 | } else { | |
4585 | zone = ofc->zone_imm; | |
4586 | } | |
4587 | ||
4588 | ct_offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CT); | |
4589 | if (ofc->flags & NX_CT_F_COMMIT) { | |
4590 | nl_msg_put_flag(ctx->odp_actions, OVS_CT_ATTR_COMMIT); | |
4591 | } | |
4592 | nl_msg_put_u16(ctx->odp_actions, OVS_CT_ATTR_ZONE, zone); | |
f2d105b5 JS |
4593 | put_ct_mark(&ctx->xin->flow, ctx->odp_actions, ctx->wc); |
4594 | put_ct_label(&ctx->xin->flow, ctx->odp_actions, ctx->wc); | |
d787ad39 | 4595 | put_ct_helper(ctx->odp_actions, ofc); |
9ac0aada JR |
4596 | put_ct_nat(ctx); |
4597 | ctx->ct_nat_action = NULL; | |
07659514 JS |
4598 | nl_msg_end_nested(ctx->odp_actions, ct_offset); |
4599 | ||
8e53fe8c JS |
4600 | /* Restore the original ct fields in the key. These should only be exposed |
4601 | * after recirculation to another table. */ | |
4602 | ctx->base_flow.ct_mark = old_ct_mark; | |
f2d105b5 | 4603 | ctx->wc->masks.ct_mark = old_ct_mark_mask; |
9daf2348 | 4604 | ctx->base_flow.ct_label = old_ct_label; |
f2d105b5 | 4605 | ctx->wc->masks.ct_label = old_ct_label_mask; |
8e53fe8c | 4606 | |
07659514 JS |
4607 | if (ofc->recirc_table == NX_CT_RECIRC_NONE) { |
4608 | /* If we do not recirculate as part of this action, hide the results of | |
4609 | * connection tracking from subsequent recirculations. */ | |
4610 | ctx->conntracked = false; | |
4611 | } else { | |
4612 | /* Use ct_* fields from datapath during recirculation upcall. */ | |
4613 | ctx->conntracked = true; | |
e37b8437 | 4614 | compose_recirculate_and_fork(ctx, ofc->recirc_table); |
07659514 JS |
4615 | } |
4616 | } | |
4617 | ||
e12ec36b SH |
4618 | static void |
4619 | recirc_for_mpls(const struct ofpact *a, struct xlate_ctx *ctx) | |
4620 | { | |
4621 | /* No need to recirculate if already exiting. */ | |
4622 | if (ctx->exit) { | |
4623 | return; | |
4624 | } | |
4625 | ||
4626 | /* Do not consider recirculating unless the packet was previously MPLS. */ | |
4627 | if (!ctx->was_mpls) { | |
4628 | return; | |
4629 | } | |
4630 | ||
4631 | /* Special case these actions, only recirculating if necessary. | |
4632 | * This avoids the overhead of recirculation in common use-cases. | |
4633 | */ | |
4634 | switch (a->type) { | |
4635 | ||
4636 | /* Output actions do not require recirculation. */ | |
4637 | case OFPACT_OUTPUT: | |
aaca4fe0 | 4638 | case OFPACT_OUTPUT_TRUNC: |
e12ec36b SH |
4639 | case OFPACT_ENQUEUE: |
4640 | case OFPACT_OUTPUT_REG: | |
4641 | /* Set actions that don't touch L3+ fields do not require recirculation. */ | |
4642 | case OFPACT_SET_VLAN_VID: | |
4643 | case OFPACT_SET_VLAN_PCP: | |
4644 | case OFPACT_SET_ETH_SRC: | |
4645 | case OFPACT_SET_ETH_DST: | |
4646 | case OFPACT_SET_TUNNEL: | |
4647 | case OFPACT_SET_QUEUE: | |
4648 | /* If actions of a group require recirculation that can be detected | |
4649 | * when translating them. */ | |
4650 | case OFPACT_GROUP: | |
4651 | return; | |
4652 | ||
4653 | /* Set field that don't touch L3+ fields don't require recirculation. */ | |
4654 | case OFPACT_SET_FIELD: | |
4655 | if (mf_is_l3_or_higher(ofpact_get_SET_FIELD(a)->field)) { | |
4656 | break; | |
4657 | } | |
4658 | return; | |
4659 | ||
4660 | /* For simplicity, recirculate in all other cases. */ | |
4661 | case OFPACT_CONTROLLER: | |
4662 | case OFPACT_BUNDLE: | |
4663 | case OFPACT_STRIP_VLAN: | |
4664 | case OFPACT_PUSH_VLAN: | |
4665 | case OFPACT_SET_IPV4_SRC: | |
4666 | case OFPACT_SET_IPV4_DST: | |
4667 | case OFPACT_SET_IP_DSCP: | |
4668 | case OFPACT_SET_IP_ECN: | |
4669 | case OFPACT_SET_IP_TTL: | |
4670 | case OFPACT_SET_L4_SRC_PORT: | |
4671 | case OFPACT_SET_L4_DST_PORT: | |
4672 | case OFPACT_REG_MOVE: | |
4673 | case OFPACT_STACK_PUSH: | |
4674 | case OFPACT_STACK_POP: | |
4675 | case OFPACT_DEC_TTL: | |
4676 | case OFPACT_SET_MPLS_LABEL: | |
4677 | case OFPACT_SET_MPLS_TC: | |
4678 | case OFPACT_SET_MPLS_TTL: | |
4679 | case OFPACT_DEC_MPLS_TTL: | |
4680 | case OFPACT_PUSH_MPLS: | |
4681 | case OFPACT_POP_MPLS: | |
4682 | case OFPACT_POP_QUEUE: | |
4683 | case OFPACT_FIN_TIMEOUT: | |
4684 | case OFPACT_RESUBMIT: | |
4685 | case OFPACT_LEARN: | |
4686 | case OFPACT_CONJUNCTION: | |
4687 | case OFPACT_MULTIPATH: | |
4688 | case OFPACT_NOTE: | |
4689 | case OFPACT_EXIT: | |
4690 | case OFPACT_SAMPLE: | |
4691 | case OFPACT_UNROLL_XLATE: | |
4692 | case OFPACT_CT: | |
4693 | case OFPACT_NAT: | |
4694 | case OFPACT_DEBUG_RECIRC: | |
4695 | case OFPACT_METER: | |
4696 | case OFPACT_CLEAR_ACTIONS: | |
4697 | case OFPACT_WRITE_ACTIONS: | |
4698 | case OFPACT_WRITE_METADATA: | |
4699 | case OFPACT_GOTO_TABLE: | |
4700 | default: | |
4701 | break; | |
4702 | } | |
4703 | ||
4704 | /* Recirculate */ | |
4705 | ctx_trigger_freeze(ctx); | |
4706 | } | |
4707 | ||
9583bc14 EJ |
4708 | static void |
4709 | do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
4710 | struct xlate_ctx *ctx) | |
4711 | { | |
49a73e0c | 4712 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 4713 | struct flow *flow = &ctx->xin->flow; |
9583bc14 EJ |
4714 | const struct ofpact *a; |
4715 | ||
a36de779 | 4716 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
53902038 | 4717 | tnl_neigh_snoop(flow, wc, ctx->xbridge->name); |
a36de779 | 4718 | } |
f47ea021 JR |
4719 | /* dl_type already in the mask, not set below. */ |
4720 | ||
9583bc14 EJ |
4721 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
4722 | struct ofpact_controller *controller; | |
4723 | const struct ofpact_metadata *metadata; | |
b2dd70be JR |
4724 | const struct ofpact_set_field *set_field; |
4725 | const struct mf_field *mf; | |
9583bc14 | 4726 | |
fff1b9c0 JR |
4727 | if (ctx->error) { |
4728 | break; | |
4729 | } | |
4730 | ||
e12ec36b SH |
4731 | recirc_for_mpls(a, ctx); |
4732 | ||
e672ff9b JR |
4733 | if (ctx->exit) { |
4734 | /* Check if need to store the remaining actions for later | |
4735 | * execution. */ | |
1d361a81 BP |
4736 | if (ctx->freezing) { |
4737 | freeze_unroll_actions(a, ofpact_end(ofpacts, ofpacts_len), | |
e672ff9b JR |
4738 | ctx); |
4739 | } | |
4740 | break; | |
7bbdd84f SH |
4741 | } |
4742 | ||
9583bc14 EJ |
4743 | switch (a->type) { |
4744 | case OFPACT_OUTPUT: | |
4745 | xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port, | |
4746 | ofpact_get_OUTPUT(a)->max_len, true); | |
4747 | break; | |
4748 | ||
7395c052 | 4749 | case OFPACT_GROUP: |
f4fb341b | 4750 | if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id)) { |
1d741d6d | 4751 | /* Group could not be found. */ |
f4fb341b SH |
4752 | return; |
4753 | } | |
7395c052 NZ |
4754 | break; |
4755 | ||
9583bc14 EJ |
4756 | case OFPACT_CONTROLLER: |
4757 | controller = ofpact_get_CONTROLLER(a); | |
77ab5fd2 BP |
4758 | if (controller->pause) { |
4759 | ctx->pause = controller; | |
4760 | ctx->xout->slow |= SLOW_CONTROLLER; | |
4761 | ctx_trigger_freeze(ctx); | |
4762 | a = ofpact_next(a); | |
4763 | } else { | |
4764 | execute_controller_action(ctx, controller->max_len, | |
4765 | controller->reason, | |
4766 | controller->controller_id, | |
4767 | controller->userdata, | |
4768 | controller->userdata_len); | |
4769 | } | |
9583bc14 EJ |
4770 | break; |
4771 | ||
4772 | case OFPACT_ENQUEUE: | |
16194afd DDP |
4773 | memset(&wc->masks.skb_priority, 0xff, |
4774 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4775 | xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a)); |
4776 | break; | |
4777 | ||
4778 | case OFPACT_SET_VLAN_VID: | |
f74e7df7 | 4779 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
ca287d20 JR |
4780 | if (flow->vlan_tci & htons(VLAN_CFI) || |
4781 | ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) { | |
4782 | flow->vlan_tci &= ~htons(VLAN_VID_MASK); | |
4783 | flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid) | |
4784 | | htons(VLAN_CFI)); | |
4785 | } | |
9583bc14 EJ |
4786 | break; |
4787 | ||
4788 | case OFPACT_SET_VLAN_PCP: | |
f74e7df7 | 4789 | wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI); |
ca287d20 JR |
4790 | if (flow->vlan_tci & htons(VLAN_CFI) || |
4791 | ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) { | |
4792 | flow->vlan_tci &= ~htons(VLAN_PCP_MASK); | |
4793 | flow->vlan_tci |= htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp | |
4794 | << VLAN_PCP_SHIFT) | VLAN_CFI); | |
4795 | } | |
9583bc14 EJ |
4796 | break; |
4797 | ||
4798 | case OFPACT_STRIP_VLAN: | |
f74e7df7 | 4799 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 4800 | flow->vlan_tci = htons(0); |
9583bc14 EJ |
4801 | break; |
4802 | ||
4803 | case OFPACT_PUSH_VLAN: | |
4804 | /* XXX 802.1AD(QinQ) */ | |
f74e7df7 | 4805 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 4806 | flow->vlan_tci = htons(VLAN_CFI); |
9583bc14 EJ |
4807 | break; |
4808 | ||
4809 | case OFPACT_SET_ETH_SRC: | |
74ff3298 JR |
4810 | WC_MASK_FIELD(wc, dl_src); |
4811 | flow->dl_src = ofpact_get_SET_ETH_SRC(a)->mac; | |
9583bc14 EJ |
4812 | break; |
4813 | ||
4814 | case OFPACT_SET_ETH_DST: | |
74ff3298 JR |
4815 | WC_MASK_FIELD(wc, dl_dst); |
4816 | flow->dl_dst = ofpact_get_SET_ETH_DST(a)->mac; | |
9583bc14 EJ |
4817 | break; |
4818 | ||
4819 | case OFPACT_SET_IPV4_SRC: | |
33bf9176 | 4820 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 4821 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); |
33bf9176 | 4822 | flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4; |
9583bc14 EJ |
4823 | } |
4824 | break; | |
4825 | ||
4826 | case OFPACT_SET_IPV4_DST: | |
33bf9176 | 4827 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 4828 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); |
33bf9176 | 4829 | flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4; |
9583bc14 EJ |
4830 | } |
4831 | break; | |
4832 | ||
04f01c24 BP |
4833 | case OFPACT_SET_IP_DSCP: |
4834 | if (is_ip_any(flow)) { | |
f47ea021 | 4835 | wc->masks.nw_tos |= IP_DSCP_MASK; |
33bf9176 | 4836 | flow->nw_tos &= ~IP_DSCP_MASK; |
04f01c24 | 4837 | flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp; |
9583bc14 EJ |
4838 | } |
4839 | break; | |
4840 | ||
ff14eb7a JR |
4841 | case OFPACT_SET_IP_ECN: |
4842 | if (is_ip_any(flow)) { | |
4843 | wc->masks.nw_tos |= IP_ECN_MASK; | |
4844 | flow->nw_tos &= ~IP_ECN_MASK; | |
4845 | flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn; | |
4846 | } | |
4847 | break; | |
4848 | ||
0c20dbe4 JR |
4849 | case OFPACT_SET_IP_TTL: |
4850 | if (is_ip_any(flow)) { | |
4851 | wc->masks.nw_ttl = 0xff; | |
4852 | flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl; | |
4853 | } | |
4854 | break; | |
4855 | ||
9583bc14 | 4856 | case OFPACT_SET_L4_SRC_PORT: |
b8778a0d | 4857 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
4858 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
4859 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
33bf9176 | 4860 | flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port); |
9583bc14 EJ |
4861 | } |
4862 | break; | |
4863 | ||
4864 | case OFPACT_SET_L4_DST_PORT: | |
b8778a0d | 4865 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
4866 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
4867 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
33bf9176 | 4868 | flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port); |
9583bc14 EJ |
4869 | } |
4870 | break; | |
4871 | ||
4872 | case OFPACT_RESUBMIT: | |
8bf009bf JR |
4873 | /* Freezing complicates resubmit. Some action in the flow |
4874 | * entry found by resubmit might trigger freezing. If that | |
4875 | * happens, then we do not want to execute the resubmit again after | |
4876 | * during thawing, so we want to skip back to the head of the loop | |
4877 | * to avoid that, only adding any actions that follow the resubmit | |
4878 | * to the frozen actions. | |
6b1c5734 | 4879 | */ |
9583bc14 | 4880 | xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a)); |
6b1c5734 | 4881 | continue; |
9583bc14 EJ |
4882 | |
4883 | case OFPACT_SET_TUNNEL: | |
33bf9176 | 4884 | flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id); |
9583bc14 EJ |
4885 | break; |
4886 | ||
4887 | case OFPACT_SET_QUEUE: | |
16194afd DDP |
4888 | memset(&wc->masks.skb_priority, 0xff, |
4889 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4890 | xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id); |
4891 | break; | |
4892 | ||
4893 | case OFPACT_POP_QUEUE: | |
16194afd DDP |
4894 | memset(&wc->masks.skb_priority, 0xff, |
4895 | sizeof wc->masks.skb_priority); | |
33bf9176 | 4896 | flow->skb_priority = ctx->orig_skb_priority; |
9583bc14 EJ |
4897 | break; |
4898 | ||
4899 | case OFPACT_REG_MOVE: | |
33bf9176 | 4900 | nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc); |
9583bc14 EJ |
4901 | break; |
4902 | ||
b2dd70be JR |
4903 | case OFPACT_SET_FIELD: |
4904 | set_field = ofpact_get_SET_FIELD(a); | |
4905 | mf = set_field->field; | |
b2dd70be JR |
4906 | |
4907 | /* Set field action only ever overwrites packet's outermost | |
4908 | * applicable header fields. Do nothing if no header exists. */ | |
4479846e JR |
4909 | if (mf->id == MFF_VLAN_VID) { |
4910 | wc->masks.vlan_tci |= htons(VLAN_CFI); | |
4911 | if (!(flow->vlan_tci & htons(VLAN_CFI))) { | |
4912 | break; | |
4913 | } | |
4914 | } else if ((mf->id == MFF_MPLS_LABEL || mf->id == MFF_MPLS_TC) | |
4915 | /* 'dl_type' is already unwildcarded. */ | |
4916 | && !eth_type_mpls(flow->dl_type)) { | |
4917 | break; | |
b2dd70be | 4918 | } |
67794d56 | 4919 | /* A flow may wildcard nw_frag. Do nothing if setting a transport |
b8778a0d | 4920 | * header field on a packet that does not have them. */ |
4d182934 | 4921 | mf_mask_field_and_prereqs__(mf, &set_field->mask, wc); |
b8778a0d JR |
4922 | if (mf_are_prereqs_ok(mf, flow)) { |
4923 | mf_set_flow_value_masked(mf, &set_field->value, | |
4924 | &set_field->mask, flow); | |
4925 | } | |
b2dd70be JR |
4926 | break; |
4927 | ||
9583bc14 | 4928 | case OFPACT_STACK_PUSH: |
33bf9176 BP |
4929 | nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc, |
4930 | &ctx->stack); | |
9583bc14 EJ |
4931 | break; |
4932 | ||
4933 | case OFPACT_STACK_POP: | |
f74e7df7 JP |
4934 | nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc, |
4935 | &ctx->stack); | |
9583bc14 EJ |
4936 | break; |
4937 | ||
4938 | case OFPACT_PUSH_MPLS: | |
8bfd0fda | 4939 | compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)); |
9583bc14 EJ |
4940 | break; |
4941 | ||
4942 | case OFPACT_POP_MPLS: | |
8bfd0fda | 4943 | compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype); |
9583bc14 EJ |
4944 | break; |
4945 | ||
097d4939 | 4946 | case OFPACT_SET_MPLS_LABEL: |
8bfd0fda BP |
4947 | compose_set_mpls_label_action( |
4948 | ctx, ofpact_get_SET_MPLS_LABEL(a)->label); | |
1d741d6d | 4949 | break; |
097d4939 JR |
4950 | |
4951 | case OFPACT_SET_MPLS_TC: | |
8bfd0fda | 4952 | compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc); |
097d4939 JR |
4953 | break; |
4954 | ||
9583bc14 | 4955 | case OFPACT_SET_MPLS_TTL: |
8bfd0fda | 4956 | compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl); |
9583bc14 EJ |
4957 | break; |
4958 | ||
4959 | case OFPACT_DEC_MPLS_TTL: | |
9cfef3d0 | 4960 | if (compose_dec_mpls_ttl_action(ctx)) { |
ad3efdcb | 4961 | return; |
9583bc14 EJ |
4962 | } |
4963 | break; | |
4964 | ||
4965 | case OFPACT_DEC_TTL: | |
f74e7df7 | 4966 | wc->masks.nw_ttl = 0xff; |
9583bc14 | 4967 | if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) { |
ad3efdcb | 4968 | return; |
9583bc14 EJ |
4969 | } |
4970 | break; | |
4971 | ||
4972 | case OFPACT_NOTE: | |
4973 | /* Nothing to do. */ | |
4974 | break; | |
4975 | ||
4976 | case OFPACT_MULTIPATH: | |
33bf9176 | 4977 | multipath_execute(ofpact_get_MULTIPATH(a), flow, wc); |
9583bc14 EJ |
4978 | break; |
4979 | ||
4980 | case OFPACT_BUNDLE: | |
9583bc14 EJ |
4981 | xlate_bundle_action(ctx, ofpact_get_BUNDLE(a)); |
4982 | break; | |
4983 | ||
4984 | case OFPACT_OUTPUT_REG: | |
4985 | xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a)); | |
4986 | break; | |
4987 | ||
aaca4fe0 WT |
4988 | case OFPACT_OUTPUT_TRUNC: |
4989 | xlate_output_trunc_action(ctx, ofpact_get_OUTPUT_TRUNC(a)->port, | |
4990 | ofpact_get_OUTPUT_TRUNC(a)->max_len); | |
4991 | break; | |
4992 | ||
9583bc14 EJ |
4993 | case OFPACT_LEARN: |
4994 | xlate_learn_action(ctx, ofpact_get_LEARN(a)); | |
4995 | break; | |
4996 | ||
afc3987b BP |
4997 | case OFPACT_CONJUNCTION: { |
4998 | /* A flow with a "conjunction" action represents part of a special | |
4999 | * kind of "set membership match". Such a flow should not actually | |
5000 | * get executed, but it could via, say, a "packet-out", even though | |
5001 | * that wouldn't be useful. Log it to help debugging. */ | |
5002 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
5003 | VLOG_INFO_RL(&rl, "executing no-op conjunction action"); | |
18080541 | 5004 | break; |
afc3987b | 5005 | } |
18080541 | 5006 | |
9583bc14 EJ |
5007 | case OFPACT_EXIT: |
5008 | ctx->exit = true; | |
5009 | break; | |
5010 | ||
e672ff9b JR |
5011 | case OFPACT_UNROLL_XLATE: { |
5012 | struct ofpact_unroll_xlate *unroll = ofpact_get_UNROLL_XLATE(a); | |
5013 | ||
5014 | /* Restore translation context data that was stored earlier. */ | |
5015 | ctx->table_id = unroll->rule_table_id; | |
5016 | ctx->rule_cookie = unroll->rule_cookie; | |
5017 | break; | |
5018 | } | |
9583bc14 | 5019 | case OFPACT_FIN_TIMEOUT: |
33bf9176 | 5020 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
9583bc14 EJ |
5021 | xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a)); |
5022 | break; | |
5023 | ||
5024 | case OFPACT_CLEAR_ACTIONS: | |
7fdb60a7 | 5025 | ofpbuf_clear(&ctx->action_set); |
c61f3870 BP |
5026 | ctx->xin->flow.actset_output = OFPP_UNSET; |
5027 | ctx->action_set_has_group = false; | |
7fdb60a7 SH |
5028 | break; |
5029 | ||
5030 | case OFPACT_WRITE_ACTIONS: | |
7e7e8dbb | 5031 | xlate_write_actions(ctx, ofpact_get_WRITE_ACTIONS(a)); |
9583bc14 EJ |
5032 | break; |
5033 | ||
5034 | case OFPACT_WRITE_METADATA: | |
5035 | metadata = ofpact_get_WRITE_METADATA(a); | |
33bf9176 BP |
5036 | flow->metadata &= ~metadata->mask; |
5037 | flow->metadata |= metadata->metadata & metadata->mask; | |
9583bc14 EJ |
5038 | break; |
5039 | ||
638a19b0 JR |
5040 | case OFPACT_METER: |
5041 | /* Not implemented yet. */ | |
5042 | break; | |
5043 | ||
9583bc14 | 5044 | case OFPACT_GOTO_TABLE: { |
9583bc14 | 5045 | struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a); |
9583bc14 | 5046 | |
9167fc1a JR |
5047 | ovs_assert(ctx->table_id < ogt->table_id); |
5048 | ||
4468099e | 5049 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 5050 | ogt->table_id, true, true); |
9583bc14 EJ |
5051 | break; |
5052 | } | |
5053 | ||
5054 | case OFPACT_SAMPLE: | |
5055 | xlate_sample_action(ctx, ofpact_get_SAMPLE(a)); | |
5056 | break; | |
d4abaff5 | 5057 | |
07659514 | 5058 | case OFPACT_CT: |
07659514 JS |
5059 | compose_conntrack_action(ctx, ofpact_get_CT(a)); |
5060 | break; | |
5061 | ||
9ac0aada JR |
5062 | case OFPACT_NAT: |
5063 | /* This will be processed by compose_conntrack_action(). */ | |
5064 | ctx->ct_nat_action = ofpact_get_NAT(a); | |
5065 | break; | |
5066 | ||
d4abaff5 | 5067 | case OFPACT_DEBUG_RECIRC: |
1d361a81 | 5068 | ctx_trigger_freeze(ctx); |
d4abaff5 BP |
5069 | a = ofpact_next(a); |
5070 | break; | |
9583bc14 | 5071 | } |
1d741d6d JR |
5072 | |
5073 | /* Check if need to store this and the remaining actions for later | |
5074 | * execution. */ | |
1d361a81 BP |
5075 | if (!ctx->error && ctx->exit && ctx_first_frozen_action(ctx)) { |
5076 | freeze_unroll_actions(a, ofpact_end(ofpacts, ofpacts_len), ctx); | |
1d741d6d JR |
5077 | break; |
5078 | } | |
9583bc14 | 5079 | } |
9583bc14 EJ |
5080 | } |
5081 | ||
5082 | void | |
5083 | xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, | |
cc377352 EJ |
5084 | const struct flow *flow, ofp_port_t in_port, |
5085 | struct rule_dpif *rule, uint16_t tcp_flags, | |
1520ef4f BP |
5086 | const struct dp_packet *packet, struct flow_wildcards *wc, |
5087 | struct ofpbuf *odp_actions) | |
9583bc14 EJ |
5088 | { |
5089 | xin->ofproto = ofproto; | |
5090 | xin->flow = *flow; | |
cc377352 | 5091 | xin->flow.in_port.ofp_port = in_port; |
c61f3870 | 5092 | xin->flow.actset_output = OFPP_UNSET; |
9583bc14 EJ |
5093 | xin->packet = packet; |
5094 | xin->may_learn = packet != NULL; | |
5095 | xin->rule = rule; | |
b256dc52 | 5096 | xin->xcache = NULL; |
9583bc14 EJ |
5097 | xin->ofpacts = NULL; |
5098 | xin->ofpacts_len = 0; | |
5099 | xin->tcp_flags = tcp_flags; | |
5100 | xin->resubmit_hook = NULL; | |
5101 | xin->report_hook = NULL; | |
5102 | xin->resubmit_stats = NULL; | |
9a9b001a | 5103 | xin->indentation = 0; |
790c5d26 | 5104 | xin->depth = 0; |
cdd42eda | 5105 | xin->resubmits = 0; |
49a73e0c | 5106 | xin->wc = wc; |
1520ef4f | 5107 | xin->odp_actions = odp_actions; |
e672ff9b JR |
5108 | |
5109 | /* Do recirc lookup. */ | |
1d361a81 | 5110 | xin->frozen_state = NULL; |
29b1ea3f BP |
5111 | if (flow->recirc_id) { |
5112 | const struct recirc_id_node *node | |
5113 | = recirc_id_node_find(flow->recirc_id); | |
5114 | if (node) { | |
1d361a81 | 5115 | xin->frozen_state = &node->state; |
29b1ea3f BP |
5116 | } |
5117 | } | |
9583bc14 EJ |
5118 | } |
5119 | ||
5120 | void | |
5121 | xlate_out_uninit(struct xlate_out *xout) | |
5122 | { | |
e672ff9b | 5123 | if (xout) { |
fbf5d6ec | 5124 | recirc_refs_unref(&xout->recircs); |
9583bc14 EJ |
5125 | } |
5126 | } | |
5127 | ||
5128 | /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts' | |
5129 | * into datapath actions, using 'ctx', and discards the datapath actions. */ | |
5130 | void | |
5131 | xlate_actions_for_side_effects(struct xlate_in *xin) | |
5132 | { | |
5133 | struct xlate_out xout; | |
fff1b9c0 JR |
5134 | enum xlate_error error; |
5135 | ||
5136 | error = xlate_actions(xin, &xout); | |
5137 | if (error) { | |
5138 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
5139 | ||
5140 | VLOG_WARN_RL(&rl, "xlate_actions failed (%s)!", xlate_strerror(error)); | |
5141 | } | |
9583bc14 | 5142 | |
9583bc14 EJ |
5143 | xlate_out_uninit(&xout); |
5144 | } | |
9583bc14 | 5145 | \f |
55954f6e EJ |
5146 | static struct skb_priority_to_dscp * |
5147 | get_skb_priority(const struct xport *xport, uint32_t skb_priority) | |
5148 | { | |
5149 | struct skb_priority_to_dscp *pdscp; | |
5150 | uint32_t hash; | |
5151 | ||
5152 | hash = hash_int(skb_priority, 0); | |
5153 | HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) { | |
5154 | if (pdscp->skb_priority == skb_priority) { | |
5155 | return pdscp; | |
5156 | } | |
5157 | } | |
5158 | return NULL; | |
5159 | } | |
5160 | ||
5161 | static bool | |
5162 | dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority, | |
5163 | uint8_t *dscp) | |
5164 | { | |
5165 | struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority); | |
5166 | *dscp = pdscp ? pdscp->dscp : 0; | |
5167 | return pdscp != NULL; | |
5168 | } | |
5169 | ||
16194afd DDP |
5170 | static size_t |
5171 | count_skb_priorities(const struct xport *xport) | |
5172 | { | |
5173 | return hmap_count(&xport->skb_priorities); | |
5174 | } | |
5175 | ||
55954f6e EJ |
5176 | static void |
5177 | clear_skb_priorities(struct xport *xport) | |
5178 | { | |
4ec3d7c7 | 5179 | struct skb_priority_to_dscp *pdscp; |
55954f6e | 5180 | |
4ec3d7c7 | 5181 | HMAP_FOR_EACH_POP (pdscp, hmap_node, &xport->skb_priorities) { |
55954f6e EJ |
5182 | free(pdscp); |
5183 | } | |
5184 | } | |
5185 | ||
ce4a6b76 BP |
5186 | static bool |
5187 | actions_output_to_local_port(const struct xlate_ctx *ctx) | |
5188 | { | |
46c88433 | 5189 | odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL); |
ce4a6b76 BP |
5190 | const struct nlattr *a; |
5191 | unsigned int left; | |
5192 | ||
1520ef4f BP |
5193 | NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->odp_actions->data, |
5194 | ctx->odp_actions->size) { | |
ce4a6b76 BP |
5195 | if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT |
5196 | && nl_attr_get_odp_port(a) == local_odp_port) { | |
5197 | return true; | |
5198 | } | |
5199 | } | |
5200 | return false; | |
5201 | } | |
9583bc14 | 5202 | |
5e2a6702 | 5203 | #if defined(__linux__) |
7d031d7e BP |
5204 | /* Returns the maximum number of packets that the Linux kernel is willing to |
5205 | * queue up internally to certain kinds of software-implemented ports, or the | |
5206 | * default (and rarely modified) value if it cannot be determined. */ | |
5207 | static int | |
5208 | netdev_max_backlog(void) | |
5209 | { | |
5210 | static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; | |
5211 | static int max_backlog = 1000; /* The normal default value. */ | |
5212 | ||
5213 | if (ovsthread_once_start(&once)) { | |
5214 | static const char filename[] = "/proc/sys/net/core/netdev_max_backlog"; | |
5215 | FILE *stream; | |
5216 | int n; | |
5217 | ||
5218 | stream = fopen(filename, "r"); | |
5219 | if (!stream) { | |
120c348f | 5220 | VLOG_INFO("%s: open failed (%s)", filename, ovs_strerror(errno)); |
7d031d7e BP |
5221 | } else { |
5222 | if (fscanf(stream, "%d", &n) != 1) { | |
5223 | VLOG_WARN("%s: read error", filename); | |
5224 | } else if (n <= 100) { | |
5225 | VLOG_WARN("%s: unexpectedly small value %d", filename, n); | |
5226 | } else { | |
5227 | max_backlog = n; | |
5228 | } | |
5229 | fclose(stream); | |
5230 | } | |
5231 | ovsthread_once_done(&once); | |
5232 | ||
5233 | VLOG_DBG("%s: using %d max_backlog", filename, max_backlog); | |
5234 | } | |
5235 | ||
5236 | return max_backlog; | |
5237 | } | |
5238 | ||
5239 | /* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in | |
5240 | * 'odp_actions'. */ | |
5241 | static int | |
5242 | count_output_actions(const struct ofpbuf *odp_actions) | |
5243 | { | |
5244 | const struct nlattr *a; | |
5245 | size_t left; | |
5246 | int n = 0; | |
5247 | ||
6fd6ed71 | 5248 | NL_ATTR_FOR_EACH_UNSAFE (a, left, odp_actions->data, odp_actions->size) { |
7d031d7e BP |
5249 | if (a->nla_type == OVS_ACTION_ATTR_OUTPUT) { |
5250 | n++; | |
5251 | } | |
5252 | } | |
5253 | return n; | |
5254 | } | |
5e2a6702 | 5255 | #endif /* defined(__linux__) */ |
7d031d7e BP |
5256 | |
5257 | /* Returns true if 'odp_actions' contains more output actions than the datapath | |
5258 | * can reliably handle in one go. On Linux, this is the value of the | |
5259 | * net.core.netdev_max_backlog sysctl, which limits the maximum number of | |
5260 | * packets that the kernel is willing to queue up for processing while the | |
5261 | * datapath is processing a set of actions. */ | |
5262 | static bool | |
5e2a6702 | 5263 | too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) |
7d031d7e BP |
5264 | { |
5265 | #ifdef __linux__ | |
6fd6ed71 | 5266 | return (odp_actions->size / NL_A_U32_SIZE > netdev_max_backlog() |
7d031d7e BP |
5267 | && count_output_actions(odp_actions) > netdev_max_backlog()); |
5268 | #else | |
5269 | /* OSes other than Linux might have similar limits, but we don't know how | |
5270 | * to determine them.*/ | |
5271 | return false; | |
5272 | #endif | |
5273 | } | |
5274 | ||
234c3da9 BP |
5275 | static void |
5276 | xlate_wc_init(struct xlate_ctx *ctx) | |
5277 | { | |
5278 | flow_wildcards_init_catchall(ctx->wc); | |
5279 | ||
5280 | /* Some fields we consider to always be examined. */ | |
5e2e998a JR |
5281 | WC_MASK_FIELD(ctx->wc, in_port); |
5282 | WC_MASK_FIELD(ctx->wc, dl_type); | |
234c3da9 | 5283 | if (is_ip_any(&ctx->xin->flow)) { |
5e2e998a | 5284 | WC_MASK_FIELD_MASK(ctx->wc, nw_frag, FLOW_NW_FRAG_MASK); |
234c3da9 BP |
5285 | } |
5286 | ||
5287 | if (ctx->xbridge->support.odp.recirc) { | |
5288 | /* Always exactly match recirc_id when datapath supports | |
5289 | * recirculation. */ | |
5e2e998a | 5290 | WC_MASK_FIELD(ctx->wc, recirc_id); |
234c3da9 BP |
5291 | } |
5292 | ||
5293 | if (ctx->xbridge->netflow) { | |
5294 | netflow_mask_wc(&ctx->xin->flow, ctx->wc); | |
5295 | } | |
5296 | ||
5297 | tnl_wc_init(&ctx->xin->flow, ctx->wc); | |
5298 | } | |
5299 | ||
5300 | static void | |
5301 | xlate_wc_finish(struct xlate_ctx *ctx) | |
5302 | { | |
5303 | /* Clear the metadata and register wildcard masks, because we won't | |
5304 | * use non-header fields as part of the cache. */ | |
5305 | flow_wildcards_clear_non_packet_fields(ctx->wc); | |
5306 | ||
5307 | /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow | |
5308 | * uses the low 8 bits of the 16-bit tp_src and tp_dst members to | |
5309 | * represent these fields. The datapath interface, on the other hand, | |
5310 | * represents them with just 8 bits each. This means that if the high | |
5311 | * 8 bits of the masks for these fields somehow become set, then they | |
5312 | * will get chopped off by a round trip through the datapath, and | |
5313 | * revalidation will spot that as an inconsistency and delete the flow. | |
5314 | * Avoid the problem here by making sure that only the low 8 bits of | |
5315 | * either field can be unwildcarded for ICMP. | |
5316 | */ | |
a75636c8 | 5317 | if (is_icmpv4(&ctx->xin->flow, NULL) || is_icmpv6(&ctx->xin->flow, NULL)) { |
234c3da9 BP |
5318 | ctx->wc->masks.tp_src &= htons(UINT8_MAX); |
5319 | ctx->wc->masks.tp_dst &= htons(UINT8_MAX); | |
5320 | } | |
5321 | /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */ | |
5322 | if (ctx->wc->masks.vlan_tci) { | |
5323 | ctx->wc->masks.vlan_tci |= htons(VLAN_CFI); | |
5324 | } | |
5325 | } | |
5326 | ||
e672ff9b JR |
5327 | /* Translates the flow, actions, or rule in 'xin' into datapath actions in |
5328 | * 'xout'. | |
56450a41 | 5329 | * The caller must take responsibility for eventually freeing 'xout', with |
fff1b9c0 JR |
5330 | * xlate_out_uninit(). |
5331 | * Returns 'XLATE_OK' if translation was successful. In case of an error an | |
5332 | * empty set of actions will be returned in 'xin->odp_actions' (if non-NULL), | |
5333 | * so that most callers may ignore the return value and transparently install a | |
5334 | * drop flow when the translation fails. */ | |
5335 | enum xlate_error | |
84f0f298 | 5336 | xlate_actions(struct xlate_in *xin, struct xlate_out *xout) |
9583bc14 | 5337 | { |
e467ea42 BP |
5338 | *xout = (struct xlate_out) { |
5339 | .slow = 0, | |
fbf5d6ec | 5340 | .recircs = RECIRC_REFS_EMPTY_INITIALIZER, |
e467ea42 BP |
5341 | }; |
5342 | ||
84f0f298 | 5343 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
bb00fdef BP |
5344 | struct xbridge *xbridge = xbridge_lookup(xcfg, xin->ofproto); |
5345 | if (!xbridge) { | |
fff1b9c0 | 5346 | return XLATE_BRIDGE_NOT_FOUND; |
bb00fdef BP |
5347 | } |
5348 | ||
33bf9176 BP |
5349 | struct flow *flow = &xin->flow; |
5350 | ||
bb00fdef BP |
5351 | union mf_subvalue stack_stub[1024 / sizeof(union mf_subvalue)]; |
5352 | uint64_t action_set_stub[1024 / 8]; | |
1d361a81 | 5353 | uint64_t frozen_actions_stub[1024 / 8]; |
1520ef4f BP |
5354 | uint64_t actions_stub[256 / 8]; |
5355 | struct ofpbuf scratch_actions = OFPBUF_STUB_INITIALIZER(actions_stub); | |
bb00fdef BP |
5356 | struct xlate_ctx ctx = { |
5357 | .xin = xin, | |
5358 | .xout = xout, | |
5359 | .base_flow = *flow, | |
c2b878e0 | 5360 | .orig_tunnel_ipv6_dst = flow_tnl_dst(&flow->tunnel), |
bb00fdef BP |
5361 | .xbridge = xbridge, |
5362 | .stack = OFPBUF_STUB_INITIALIZER(stack_stub), | |
5363 | .rule = xin->rule, | |
c0e638aa BP |
5364 | .wc = (xin->wc |
5365 | ? xin->wc | |
f36efd90 | 5366 | : &(struct flow_wildcards) { .masks = { .dl_type = 0 } }), |
1520ef4f | 5367 | .odp_actions = xin->odp_actions ? xin->odp_actions : &scratch_actions, |
bb00fdef | 5368 | |
9a9b001a | 5369 | .indentation = xin->indentation, |
790c5d26 | 5370 | .depth = xin->depth, |
cdd42eda | 5371 | .resubmits = xin->resubmits, |
bb00fdef BP |
5372 | .in_group = false, |
5373 | .in_action_set = false, | |
5374 | ||
5375 | .table_id = 0, | |
5376 | .rule_cookie = OVS_BE64_MAX, | |
5377 | .orig_skb_priority = flow->skb_priority, | |
5378 | .sflow_n_outputs = 0, | |
5379 | .sflow_odp_port = 0, | |
2031ef97 | 5380 | .nf_output_iface = NF_OUT_DROP, |
bb00fdef | 5381 | .exit = false, |
fff1b9c0 | 5382 | .error = XLATE_OK, |
3d6151f3 | 5383 | .mirrors = 0, |
bb00fdef | 5384 | |
1d361a81 BP |
5385 | .freezing = false, |
5386 | .frozen_actions = OFPBUF_STUB_INITIALIZER(frozen_actions_stub), | |
77ab5fd2 | 5387 | .pause = NULL, |
bb00fdef | 5388 | |
e12ec36b | 5389 | .was_mpls = false, |
07659514 | 5390 | .conntracked = false, |
bb00fdef | 5391 | |
9ac0aada JR |
5392 | .ct_nat_action = NULL, |
5393 | ||
bb00fdef BP |
5394 | .action_set_has_group = false, |
5395 | .action_set = OFPBUF_STUB_INITIALIZER(action_set_stub), | |
5396 | }; | |
865ca6cf BP |
5397 | |
5398 | /* 'base_flow' reflects the packet as it came in, but we need it to reflect | |
42deb67d PS |
5399 | * the packet as the datapath will treat it for output actions. Our |
5400 | * datapath doesn't retain tunneling information without us re-setting | |
5401 | * it, so clear the tunnel data. | |
865ca6cf | 5402 | */ |
42deb67d | 5403 | |
bb00fdef | 5404 | memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); |
865ca6cf | 5405 | |
1520ef4f | 5406 | ofpbuf_reserve(ctx.odp_actions, NL_A_U32_SIZE); |
c0e638aa | 5407 | xlate_wc_init(&ctx); |
bb00fdef | 5408 | |
46c88433 | 5409 | COVERAGE_INC(xlate_actions); |
9583bc14 | 5410 | |
1d361a81 BP |
5411 | if (xin->frozen_state) { |
5412 | const struct frozen_state *state = xin->frozen_state; | |
e672ff9b | 5413 | |
1d361a81 | 5414 | xlate_report(&ctx, "Thawing frozen state:"); |
d6bef3cc | 5415 | |
e672ff9b JR |
5416 | if (xin->ofpacts_len > 0 || ctx.rule) { |
5417 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
d6bef3cc | 5418 | const char *conflict = xin->ofpacts_len ? "actions" : "rule"; |
e672ff9b | 5419 | |
d6bef3cc BP |
5420 | VLOG_WARN_RL(&rl, "Recirculation conflict (%s)!", conflict); |
5421 | xlate_report(&ctx, "- Recirculation conflict (%s)!", conflict); | |
fff1b9c0 | 5422 | ctx.error = XLATE_RECIRCULATION_CONFLICT; |
1520ef4f | 5423 | goto exit; |
e672ff9b JR |
5424 | } |
5425 | ||
5426 | /* Set the bridge for post-recirculation processing if needed. */ | |
290835f9 BP |
5427 | if (!uuid_equals(ofproto_dpif_get_uuid(ctx.xbridge->ofproto), |
5428 | &state->ofproto_uuid)) { | |
e672ff9b | 5429 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
2082425c | 5430 | const struct xbridge *new_bridge |
290835f9 | 5431 | = xbridge_lookup_by_uuid(xcfg, &state->ofproto_uuid); |
e672ff9b JR |
5432 | |
5433 | if (OVS_UNLIKELY(!new_bridge)) { | |
5434 | /* Drop the packet if the bridge cannot be found. */ | |
5435 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
1d361a81 BP |
5436 | VLOG_WARN_RL(&rl, "Frozen bridge no longer exists."); |
5437 | xlate_report(&ctx, "- Frozen bridge no longer exists."); | |
fff1b9c0 | 5438 | ctx.error = XLATE_BRIDGE_NOT_FOUND; |
1520ef4f | 5439 | goto exit; |
e672ff9b JR |
5440 | } |
5441 | ctx.xbridge = new_bridge; | |
5442 | } | |
5443 | ||
1d361a81 BP |
5444 | /* Set the thawed table id. Note: A table lookup is done only if there |
5445 | * are no frozen actions. */ | |
2082425c | 5446 | ctx.table_id = state->table_id; |
d6bef3cc | 5447 | xlate_report(&ctx, "- Resuming from table %"PRIu8, ctx.table_id); |
e672ff9b | 5448 | |
07659514 JS |
5449 | if (!state->conntracked) { |
5450 | clear_conntrack(flow); | |
5451 | } | |
5452 | ||
e672ff9b | 5453 | /* Restore pipeline metadata. May change flow's in_port and other |
1d361a81 BP |
5454 | * metadata to the values that existed when freezing was triggered. */ |
5455 | frozen_metadata_to_flow(&state->metadata, flow); | |
e672ff9b JR |
5456 | |
5457 | /* Restore stack, if any. */ | |
2082425c | 5458 | if (state->stack) { |
5c1b2314 BP |
5459 | ofpbuf_put(&ctx.stack, state->stack, |
5460 | state->n_stack * sizeof *state->stack); | |
e672ff9b JR |
5461 | } |
5462 | ||
29bae541 BP |
5463 | /* Restore mirror state. */ |
5464 | ctx.mirrors = state->mirrors; | |
5465 | ||
e672ff9b | 5466 | /* Restore action set, if any. */ |
2082425c | 5467 | if (state->action_set_len) { |
d6bef3cc | 5468 | xlate_report_actions(&ctx, "- Restoring action set", |
417509fa | 5469 | state->action_set, state->action_set_len); |
d6bef3cc | 5470 | |
7e7e8dbb BP |
5471 | flow->actset_output = OFPP_UNSET; |
5472 | xlate_write_actions__(&ctx, state->action_set, | |
5473 | state->action_set_len); | |
e672ff9b JR |
5474 | } |
5475 | ||
1d361a81 BP |
5476 | /* Restore frozen actions. If there are no actions, processing will |
5477 | * start with a lookup in the table set above. */ | |
417509fa BP |
5478 | xin->ofpacts = state->ofpacts; |
5479 | xin->ofpacts_len = state->ofpacts_len; | |
5480 | if (state->ofpacts_len) { | |
d6bef3cc BP |
5481 | xlate_report_actions(&ctx, "- Restoring actions", |
5482 | xin->ofpacts, xin->ofpacts_len); | |
e672ff9b JR |
5483 | } |
5484 | } else if (OVS_UNLIKELY(flow->recirc_id)) { | |
5485 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
5486 | ||
5487 | VLOG_WARN_RL(&rl, "Recirculation context not found for ID %"PRIx32, | |
5488 | flow->recirc_id); | |
fff1b9c0 | 5489 | ctx.error = XLATE_NO_RECIRCULATION_CONTEXT; |
1520ef4f | 5490 | goto exit; |
e672ff9b | 5491 | } |
621b8064 JR |
5492 | /* The bridge is now known so obtain its table version. */ |
5493 | ctx.tables_version = ofproto_dpif_get_tables_version(ctx.xbridge->ofproto); | |
9583bc14 | 5494 | |
10c44245 | 5495 | if (!xin->ofpacts && !ctx.rule) { |
b2e89cc9 | 5496 | ctx.rule = rule_dpif_lookup_from_table( |
c0e638aa | 5497 | ctx.xbridge->ofproto, ctx.tables_version, flow, ctx.wc, |
1e1e1d19 | 5498 | ctx.xin->resubmit_stats, &ctx.table_id, |
b2e89cc9 | 5499 | flow->in_port.ofp_port, true, true); |
10c44245 | 5500 | if (ctx.xin->resubmit_stats) { |
b2e89cc9 | 5501 | rule_dpif_credit_stats(ctx.rule, ctx.xin->resubmit_stats); |
10c44245 | 5502 | } |
b256dc52 JS |
5503 | if (ctx.xin->xcache) { |
5504 | struct xc_entry *entry; | |
5505 | ||
5506 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE); | |
b2e89cc9 | 5507 | entry->u.rule = ctx.rule; |
1e1e1d19 | 5508 | rule_dpif_ref(ctx.rule); |
b256dc52 | 5509 | } |
a8c31348 BP |
5510 | |
5511 | if (OVS_UNLIKELY(ctx.xin->resubmit_hook)) { | |
b2e89cc9 | 5512 | ctx.xin->resubmit_hook(ctx.xin, ctx.rule, 0); |
a8c31348 | 5513 | } |
10c44245 | 5514 | } |
10c44245 | 5515 | |
1d361a81 | 5516 | /* Get the proximate input port of the packet. (If xin->frozen_state, |
14d2b8b2 BP |
5517 | * flow->in_port is the ultimate input port of the packet.) */ |
5518 | struct xport *in_port = get_ofp_port(xbridge, | |
5519 | ctx.base_flow.in_port.ofp_port); | |
5520 | ||
1d361a81 BP |
5521 | /* Tunnel stats only for not-thawed packets. */ |
5522 | if (!xin->frozen_state && in_port && in_port->is_tunnel) { | |
b256dc52 JS |
5523 | if (ctx.xin->resubmit_stats) { |
5524 | netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats); | |
5525 | if (in_port->bfd) { | |
5526 | bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats); | |
5527 | } | |
5528 | } | |
5529 | if (ctx.xin->xcache) { | |
5530 | struct xc_entry *entry; | |
5531 | ||
5532 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV); | |
5533 | entry->u.dev.rx = netdev_ref(in_port->netdev); | |
5534 | entry->u.dev.bfd = bfd_ref(in_port->bfd); | |
d6fc5f57 EJ |
5535 | } |
5536 | } | |
5537 | ||
1d361a81 | 5538 | if (!xin->frozen_state && process_special(&ctx, in_port)) { |
bef1403e BP |
5539 | /* process_special() did all the processing for this packet. |
5540 | * | |
1d361a81 BP |
5541 | * We do not perform special processing on thawed packets, since that |
5542 | * was done before they were frozen and should not be redone. */ | |
bef1403e BP |
5543 | } else if (in_port && in_port->xbundle |
5544 | && xbundle_mirror_out(xbridge, in_port->xbundle)) { | |
5545 | if (ctx.xin->packet != NULL) { | |
5546 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
5547 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
5548 | "%s, which is reserved exclusively for mirroring", | |
5549 | ctx.xbridge->name, in_port->xbundle->name); | |
5550 | } | |
5551 | } else { | |
1d361a81 | 5552 | /* Sampling is done on initial reception; don't redo after thawing. */ |
a6092018 | 5553 | unsigned int user_cookie_offset = 0; |
1d361a81 | 5554 | if (!xin->frozen_state) { |
a6092018 BP |
5555 | user_cookie_offset = compose_sflow_action(&ctx); |
5556 | compose_ipfix_action(&ctx, ODPP_NONE); | |
e672ff9b | 5557 | } |
0731abc5 | 5558 | size_t sample_actions_len = ctx.odp_actions->size; |
9583bc14 | 5559 | |
234c3da9 BP |
5560 | if (tnl_process_ecn(flow) |
5561 | && (!in_port || may_receive(in_port, &ctx))) { | |
1806291d BP |
5562 | const struct ofpact *ofpacts; |
5563 | size_t ofpacts_len; | |
5564 | ||
5565 | if (xin->ofpacts) { | |
5566 | ofpacts = xin->ofpacts; | |
5567 | ofpacts_len = xin->ofpacts_len; | |
5568 | } else if (ctx.rule) { | |
5569 | const struct rule_actions *actions | |
5570 | = rule_dpif_get_actions(ctx.rule); | |
5571 | ofpacts = actions->ofpacts; | |
5572 | ofpacts_len = actions->ofpacts_len; | |
5573 | ctx.rule_cookie = rule_dpif_get_flow_cookie(ctx.rule); | |
5574 | } else { | |
5575 | OVS_NOT_REACHED(); | |
5576 | } | |
5577 | ||
7efbc3b7 | 5578 | mirror_ingress_packet(&ctx); |
9583bc14 | 5579 | do_xlate_actions(ofpacts, ofpacts_len, &ctx); |
fff1b9c0 JR |
5580 | if (ctx.error) { |
5581 | goto exit; | |
5582 | } | |
9583bc14 EJ |
5583 | |
5584 | /* We've let OFPP_NORMAL and the learning action look at the | |
1d361a81 | 5585 | * packet, so cancel all actions and freezing if forwarding is |
8a5fb3b4 | 5586 | * disabled. */ |
9efd308e DV |
5587 | if (in_port && (!xport_stp_forward_state(in_port) || |
5588 | !xport_rstp_forward_state(in_port))) { | |
1520ef4f | 5589 | ctx.odp_actions->size = sample_actions_len; |
1d361a81 | 5590 | ctx_cancel_freeze(&ctx); |
8a5fb3b4 BP |
5591 | ofpbuf_clear(&ctx.action_set); |
5592 | } | |
5593 | ||
1d361a81 | 5594 | if (!ctx.freezing) { |
8a5fb3b4 | 5595 | xlate_action_set(&ctx); |
e672ff9b | 5596 | } |
1d361a81 | 5597 | if (ctx.freezing) { |
77ab5fd2 | 5598 | finish_freezing(&ctx); |
9583bc14 EJ |
5599 | } |
5600 | } | |
5601 | ||
e672ff9b | 5602 | /* Output only fully processed packets. */ |
1d361a81 | 5603 | if (!ctx.freezing |
e672ff9b | 5604 | && xbridge->has_in_band |
ce4a6b76 BP |
5605 | && in_band_must_output_to_local_port(flow) |
5606 | && !actions_output_to_local_port(&ctx)) { | |
e93ef1c7 | 5607 | compose_output_action(&ctx, OFPP_LOCAL, NULL); |
9583bc14 | 5608 | } |
aaa0fbae | 5609 | |
a6092018 BP |
5610 | if (user_cookie_offset) { |
5611 | fix_sflow_action(&ctx, user_cookie_offset); | |
e672ff9b | 5612 | } |
9583bc14 EJ |
5613 | } |
5614 | ||
1520ef4f | 5615 | if (nl_attr_oversized(ctx.odp_actions->size)) { |
542024c4 | 5616 | /* These datapath actions are too big for a Netlink attribute, so we |
0f032e95 BP |
5617 | * can't hand them to the kernel directly. dpif_execute() can execute |
5618 | * them one by one with help, so just mark the result as SLOW_ACTION to | |
5619 | * prevent the flow from being installed. */ | |
5620 | COVERAGE_INC(xlate_actions_oversize); | |
5621 | ctx.xout->slow |= SLOW_ACTION; | |
1520ef4f | 5622 | } else if (too_many_output_actions(ctx.odp_actions)) { |
7d031d7e BP |
5623 | COVERAGE_INC(xlate_actions_too_many_output); |
5624 | ctx.xout->slow |= SLOW_ACTION; | |
542024c4 BP |
5625 | } |
5626 | ||
1d361a81 BP |
5627 | /* Do netflow only for packets on initial reception, that are not sent to |
5628 | * the controller. We consider packets sent to the controller to be part | |
5629 | * of the control plane rather than the data plane. */ | |
5630 | if (!xin->frozen_state | |
5631 | && xbridge->netflow | |
5632 | && !(xout->slow & SLOW_CONTROLLER)) { | |
1806291d BP |
5633 | if (ctx.xin->resubmit_stats) { |
5634 | netflow_flow_update(xbridge->netflow, flow, | |
2031ef97 | 5635 | ctx.nf_output_iface, |
1806291d BP |
5636 | ctx.xin->resubmit_stats); |
5637 | } | |
5638 | if (ctx.xin->xcache) { | |
5639 | struct xc_entry *entry; | |
b256dc52 | 5640 | |
1806291d BP |
5641 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW); |
5642 | entry->u.nf.netflow = netflow_ref(xbridge->netflow); | |
5643 | entry->u.nf.flow = xmemdup(flow, sizeof *flow); | |
2031ef97 | 5644 | entry->u.nf.iface = ctx.nf_output_iface; |
d6fc5f57 EJ |
5645 | } |
5646 | } | |
5647 | ||
c0e638aa | 5648 | xlate_wc_finish(&ctx); |
1520ef4f BP |
5649 | |
5650 | exit: | |
5651 | ofpbuf_uninit(&ctx.stack); | |
5652 | ofpbuf_uninit(&ctx.action_set); | |
1d361a81 | 5653 | ofpbuf_uninit(&ctx.frozen_actions); |
1520ef4f | 5654 | ofpbuf_uninit(&scratch_actions); |
fff1b9c0 JR |
5655 | |
5656 | /* Make sure we return a "drop flow" in case of an error. */ | |
5657 | if (ctx.error) { | |
5658 | xout->slow = 0; | |
5659 | if (xin->odp_actions) { | |
5660 | ofpbuf_clear(xin->odp_actions); | |
5661 | } | |
5662 | } | |
5663 | return ctx.error; | |
91d6cd12 AW |
5664 | } |
5665 | ||
77ab5fd2 BP |
5666 | enum ofperr |
5667 | xlate_resume(struct ofproto_dpif *ofproto, | |
5668 | const struct ofputil_packet_in_private *pin, | |
5669 | struct ofpbuf *odp_actions, | |
5670 | enum slow_path_reason *slow) | |
5671 | { | |
5672 | struct dp_packet packet; | |
5673 | dp_packet_use_const(&packet, pin->public.packet, | |
5674 | pin->public.packet_len); | |
5675 | ||
5676 | struct flow flow; | |
5677 | flow_extract(&packet, &flow); | |
5678 | ||
5679 | struct xlate_in xin; | |
5680 | xlate_in_init(&xin, ofproto, &flow, 0, NULL, ntohs(flow.tcp_flags), | |
5681 | &packet, NULL, odp_actions); | |
5682 | ||
5683 | struct ofpact_note noop; | |
5684 | ofpact_init_NOTE(&noop); | |
5685 | noop.length = 0; | |
5686 | ||
5687 | bool any_actions = pin->actions_len > 0; | |
5688 | struct frozen_state state = { | |
5689 | .table_id = 0, /* Not the table where NXAST_PAUSE was executed. */ | |
5690 | .ofproto_uuid = pin->bridge, | |
5691 | .stack = pin->stack, | |
5692 | .n_stack = pin->n_stack, | |
5693 | .mirrors = pin->mirrors, | |
5694 | .conntracked = pin->conntracked, | |
5695 | ||
5696 | /* When there are no actions, xlate_actions() will search the flow | |
5697 | * table. We don't want it to do that (we want it to resume), so | |
5698 | * supply a no-op action if there aren't any. | |
5699 | * | |
5700 | * (We can't necessarily avoid translating actions entirely if there | |
5701 | * aren't any actions, because there might be some finishing-up to do | |
5702 | * at the end of the pipeline, and we don't check for those | |
5703 | * conditions.) */ | |
5704 | .ofpacts = any_actions ? pin->actions : &noop.ofpact, | |
5705 | .ofpacts_len = any_actions ? pin->actions_len : sizeof noop, | |
5706 | ||
5707 | .action_set = pin->action_set, | |
5708 | .action_set_len = pin->action_set_len, | |
5709 | }; | |
5710 | frozen_metadata_from_flow(&state.metadata, | |
5711 | &pin->public.flow_metadata.flow); | |
5712 | xin.frozen_state = &state; | |
5713 | ||
5714 | struct xlate_out xout; | |
5715 | enum xlate_error error = xlate_actions(&xin, &xout); | |
5716 | *slow = xout.slow; | |
5717 | xlate_out_uninit(&xout); | |
5718 | ||
5719 | /* xlate_actions() can generate a number of errors, but only | |
5720 | * XLATE_BRIDGE_NOT_FOUND really stands out to me as one that we should be | |
5721 | * sure to report over OpenFlow. The others could come up in packet-outs | |
5722 | * or regular flow translation and I don't think that it's going to be too | |
5723 | * useful to report them to the controller. */ | |
5724 | return error == XLATE_BRIDGE_NOT_FOUND ? OFPERR_NXR_STALE : 0; | |
5725 | } | |
5726 | ||
91d6cd12 AW |
5727 | /* Sends 'packet' out 'ofport'. |
5728 | * May modify 'packet'. | |
5729 | * Returns 0 if successful, otherwise a positive errno value. */ | |
5730 | int | |
cf62fa4c | 5731 | xlate_send_packet(const struct ofport_dpif *ofport, struct dp_packet *packet) |
91d6cd12 | 5732 | { |
84f0f298 | 5733 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
91d6cd12 | 5734 | struct xport *xport; |
91d6cd12 | 5735 | struct ofpact_output output; |
91d6cd12 | 5736 | struct flow flow; |
91d6cd12 | 5737 | |
91d6cd12 AW |
5738 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); |
5739 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ | |
cf62fa4c | 5740 | flow_extract(packet, &flow); |
b5e7e61a | 5741 | flow.in_port.ofp_port = OFPP_NONE; |
91d6cd12 | 5742 | |
84f0f298 | 5743 | xport = xport_lookup(xcfg, ofport); |
91d6cd12 | 5744 | if (!xport) { |
02ea2703 | 5745 | return EINVAL; |
91d6cd12 | 5746 | } |
91d6cd12 AW |
5747 | output.port = xport->ofp_port; |
5748 | output.max_len = 0; | |
e491a67a YT |
5749 | |
5750 | return ofproto_dpif_execute_actions(xport->xbridge->ofproto, &flow, NULL, | |
5751 | &output.ofpact, sizeof output, | |
5752 | packet); | |
9583bc14 | 5753 | } |
b256dc52 JS |
5754 | |
5755 | struct xlate_cache * | |
5756 | xlate_cache_new(void) | |
5757 | { | |
5758 | struct xlate_cache *xcache = xmalloc(sizeof *xcache); | |
5759 | ||
5760 | ofpbuf_init(&xcache->entries, 512); | |
5761 | return xcache; | |
5762 | } | |
5763 | ||
5764 | static struct xc_entry * | |
5765 | xlate_cache_add_entry(struct xlate_cache *xcache, enum xc_type type) | |
5766 | { | |
5767 | struct xc_entry *entry; | |
5768 | ||
5769 | entry = ofpbuf_put_zeros(&xcache->entries, sizeof *entry); | |
5770 | entry->type = type; | |
5771 | ||
5772 | return entry; | |
5773 | } | |
5774 | ||
5775 | static void | |
5776 | xlate_cache_netdev(struct xc_entry *entry, const struct dpif_flow_stats *stats) | |
5777 | { | |
5778 | if (entry->u.dev.tx) { | |
5779 | netdev_vport_inc_tx(entry->u.dev.tx, stats); | |
5780 | } | |
5781 | if (entry->u.dev.rx) { | |
5782 | netdev_vport_inc_rx(entry->u.dev.rx, stats); | |
5783 | } | |
5784 | if (entry->u.dev.bfd) { | |
5785 | bfd_account_rx(entry->u.dev.bfd, stats); | |
5786 | } | |
5787 | } | |
5788 | ||
5789 | static void | |
5790 | xlate_cache_normal(struct ofproto_dpif *ofproto, struct flow *flow, int vlan) | |
5791 | { | |
84f0f298 | 5792 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
b256dc52 JS |
5793 | struct xbridge *xbridge; |
5794 | struct xbundle *xbundle; | |
5795 | struct flow_wildcards wc; | |
5796 | ||
84f0f298 | 5797 | xbridge = xbridge_lookup(xcfg, ofproto); |
b256dc52 JS |
5798 | if (!xbridge) { |
5799 | return; | |
5800 | } | |
5801 | ||
5802 | xbundle = lookup_input_bundle(xbridge, flow->in_port.ofp_port, false, | |
5803 | NULL); | |
5804 | if (!xbundle) { | |
5805 | return; | |
5806 | } | |
5807 | ||
5808 | update_learning_table(xbridge, flow, &wc, vlan, xbundle); | |
5809 | } | |
5810 | ||
5811 | /* Push stats and perform side effects of flow translation. */ | |
5812 | void | |
0725e747 | 5813 | xlate_push_stats(struct xlate_cache *xcache, |
b256dc52 JS |
5814 | const struct dpif_flow_stats *stats) |
5815 | { | |
5816 | struct xc_entry *entry; | |
5817 | struct ofpbuf entries = xcache->entries; | |
74ff3298 | 5818 | struct eth_addr dmac; |
b256dc52 | 5819 | |
bc388583 BP |
5820 | if (!stats->n_packets) { |
5821 | return; | |
5822 | } | |
5823 | ||
b256dc52 JS |
5824 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { |
5825 | switch (entry->type) { | |
5826 | case XC_RULE: | |
5827 | rule_dpif_credit_stats(entry->u.rule, stats); | |
5828 | break; | |
5829 | case XC_BOND: | |
5830 | bond_account(entry->u.bond.bond, entry->u.bond.flow, | |
5831 | entry->u.bond.vid, stats->n_bytes); | |
5832 | break; | |
5833 | case XC_NETDEV: | |
5834 | xlate_cache_netdev(entry, stats); | |
5835 | break; | |
5836 | case XC_NETFLOW: | |
5837 | netflow_flow_update(entry->u.nf.netflow, entry->u.nf.flow, | |
5838 | entry->u.nf.iface, stats); | |
5839 | break; | |
5840 | case XC_MIRROR: | |
5841 | mirror_update_stats(entry->u.mirror.mbridge, | |
5842 | entry->u.mirror.mirrors, | |
5843 | stats->n_packets, stats->n_bytes); | |
5844 | break; | |
5845 | case XC_LEARN: | |
0725e747 | 5846 | ofproto_dpif_flow_mod(entry->u.learn.ofproto, entry->u.learn.fm); |
b256dc52 JS |
5847 | break; |
5848 | case XC_NORMAL: | |
0725e747 BP |
5849 | xlate_cache_normal(entry->u.normal.ofproto, entry->u.normal.flow, |
5850 | entry->u.normal.vlan); | |
b256dc52 JS |
5851 | break; |
5852 | case XC_FIN_TIMEOUT: | |
5853 | xlate_fin_timeout__(entry->u.fin.rule, stats->tcp_flags, | |
5854 | entry->u.fin.idle, entry->u.fin.hard); | |
5855 | break; | |
1e684d7d RW |
5856 | case XC_GROUP: |
5857 | group_dpif_credit_stats(entry->u.group.group, entry->u.group.bucket, | |
5858 | stats); | |
5859 | break; | |
53902038 TLSC |
5860 | case XC_TNL_NEIGH: |
5861 | /* Lookup neighbor to avoid timeout. */ | |
0b411137 TLSC |
5862 | tnl_neigh_lookup(entry->u.tnl_neigh_cache.br_name, |
5863 | &entry->u.tnl_neigh_cache.d_ipv6, &dmac); | |
a36de779 | 5864 | break; |
b256dc52 JS |
5865 | default: |
5866 | OVS_NOT_REACHED(); | |
5867 | } | |
5868 | } | |
5869 | } | |
5870 | ||
5871 | static void | |
5872 | xlate_dev_unref(struct xc_entry *entry) | |
5873 | { | |
5874 | if (entry->u.dev.tx) { | |
5875 | netdev_close(entry->u.dev.tx); | |
5876 | } | |
5877 | if (entry->u.dev.rx) { | |
5878 | netdev_close(entry->u.dev.rx); | |
5879 | } | |
5880 | if (entry->u.dev.bfd) { | |
5881 | bfd_unref(entry->u.dev.bfd); | |
5882 | } | |
5883 | } | |
5884 | ||
5885 | static void | |
5886 | xlate_cache_clear_netflow(struct netflow *netflow, struct flow *flow) | |
5887 | { | |
b256dc52 JS |
5888 | netflow_flow_clear(netflow, flow); |
5889 | netflow_unref(netflow); | |
5890 | free(flow); | |
5891 | } | |
5892 | ||
5893 | void | |
5894 | xlate_cache_clear(struct xlate_cache *xcache) | |
5895 | { | |
5896 | struct xc_entry *entry; | |
5897 | struct ofpbuf entries; | |
5898 | ||
5899 | if (!xcache) { | |
5900 | return; | |
5901 | } | |
5902 | ||
5903 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { | |
5904 | switch (entry->type) { | |
5905 | case XC_RULE: | |
5906 | rule_dpif_unref(entry->u.rule); | |
5907 | break; | |
5908 | case XC_BOND: | |
5909 | free(entry->u.bond.flow); | |
5910 | bond_unref(entry->u.bond.bond); | |
5911 | break; | |
5912 | case XC_NETDEV: | |
5913 | xlate_dev_unref(entry); | |
5914 | break; | |
5915 | case XC_NETFLOW: | |
5916 | xlate_cache_clear_netflow(entry->u.nf.netflow, entry->u.nf.flow); | |
5917 | break; | |
5918 | case XC_MIRROR: | |
5919 | mbridge_unref(entry->u.mirror.mbridge); | |
5920 | break; | |
5921 | case XC_LEARN: | |
4165b5e0 JS |
5922 | free(entry->u.learn.fm); |
5923 | ofpbuf_delete(entry->u.learn.ofpacts); | |
b256dc52 JS |
5924 | break; |
5925 | case XC_NORMAL: | |
5926 | free(entry->u.normal.flow); | |
5927 | break; | |
5928 | case XC_FIN_TIMEOUT: | |
83709dfa JR |
5929 | /* 'u.fin.rule' is always already held as a XC_RULE, which |
5930 | * has already released it's reference above. */ | |
b256dc52 | 5931 | break; |
1e684d7d RW |
5932 | case XC_GROUP: |
5933 | group_dpif_unref(entry->u.group.group); | |
5934 | break; | |
53902038 | 5935 | case XC_TNL_NEIGH: |
a36de779 | 5936 | break; |
b256dc52 JS |
5937 | default: |
5938 | OVS_NOT_REACHED(); | |
5939 | } | |
5940 | } | |
5941 | ||
5942 | ofpbuf_clear(&xcache->entries); | |
5943 | } | |
5944 | ||
5945 | void | |
5946 | xlate_cache_delete(struct xlate_cache *xcache) | |
5947 | { | |
5948 | xlate_cache_clear(xcache); | |
5949 | ofpbuf_uninit(&xcache->entries); | |
5950 | free(xcache); | |
5951 | } |