]>
Commit | Line | Data |
---|---|---|
18080541 | 1 | /* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015 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 | |
a36de779 | 25 | #include "tnl-arp-cache.h" |
db7d4e46 | 26 | #include "bfd.h" |
9583bc14 EJ |
27 | #include "bitmap.h" |
28 | #include "bond.h" | |
29 | #include "bundle.h" | |
30 | #include "byte-order.h" | |
db7d4e46 | 31 | #include "cfm.h" |
9583bc14 EJ |
32 | #include "connmgr.h" |
33 | #include "coverage.h" | |
e14deea0 | 34 | #include "dp-packet.h" |
9583bc14 EJ |
35 | #include "dpif.h" |
36 | #include "dynamic-string.h" | |
f7f1ea29 | 37 | #include "in-band.h" |
db7d4e46 | 38 | #include "lacp.h" |
9583bc14 | 39 | #include "learn.h" |
46c88433 | 40 | #include "list.h" |
0477baa9 | 41 | #include "ovs-lldp.h" |
9583bc14 | 42 | #include "mac-learning.h" |
6d95c4e8 | 43 | #include "mcast-snooping.h" |
9583bc14 EJ |
44 | #include "meta-flow.h" |
45 | #include "multipath.h" | |
46 | #include "netdev-vport.h" | |
47 | #include "netlink.h" | |
48 | #include "nx-match.h" | |
49 | #include "odp-execute.h" | |
50 | #include "ofp-actions.h" | |
51 | #include "ofproto/ofproto-dpif-ipfix.h" | |
ec7ceaed | 52 | #include "ofproto/ofproto-dpif-mirror.h" |
60d02c72 | 53 | #include "ofproto/ofproto-dpif-monitor.h" |
9583bc14 EJ |
54 | #include "ofproto/ofproto-dpif-sflow.h" |
55 | #include "ofproto/ofproto-dpif.h" | |
6f00e29b | 56 | #include "ofproto/ofproto-provider.h" |
a36de779 PS |
57 | #include "ovs-router.h" |
58 | #include "tnl-ports.h" | |
9583bc14 | 59 | #include "tunnel.h" |
e6211adc | 60 | #include "openvswitch/vlog.h" |
9583bc14 | 61 | |
46c88433 | 62 | COVERAGE_DEFINE(xlate_actions); |
0f032e95 | 63 | COVERAGE_DEFINE(xlate_actions_oversize); |
7d031d7e | 64 | COVERAGE_DEFINE(xlate_actions_too_many_output); |
9583bc14 EJ |
65 | |
66 | VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate); | |
67 | ||
8a553e9a EJ |
68 | /* Maximum depth of flow table recursion (due to resubmit actions) in a |
69 | * flow translation. */ | |
70 | #define MAX_RESUBMIT_RECURSION 64 | |
adcf00ba AZ |
71 | #define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in |
72 | internal table. */ | |
8a553e9a | 73 | |
98b07853 BP |
74 | /* Maximum number of resubmit actions in a flow translation, whether they are |
75 | * recursive or not. */ | |
76 | #define MAX_RESUBMITS (MAX_RESUBMIT_RECURSION * MAX_RESUBMIT_RECURSION) | |
77 | ||
46c88433 EJ |
78 | struct xbridge { |
79 | struct hmap_node hmap_node; /* Node in global 'xbridges' map. */ | |
80 | struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */ | |
81 | ||
ca6ba700 | 82 | struct ovs_list xbundles; /* Owned xbundles. */ |
46c88433 EJ |
83 | struct hmap xports; /* Indexed by ofp_port. */ |
84 | ||
85 | char *name; /* Name used in log messages. */ | |
89a8a7f0 | 86 | struct dpif *dpif; /* Datapath interface. */ |
46c88433 | 87 | struct mac_learning *ml; /* Mac learning handle. */ |
6d95c4e8 | 88 | struct mcast_snooping *ms; /* Multicast Snooping handle. */ |
46c88433 EJ |
89 | struct mbridge *mbridge; /* Mirroring. */ |
90 | struct dpif_sflow *sflow; /* SFlow handle, or null. */ | |
91 | struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */ | |
ce3955be | 92 | struct netflow *netflow; /* Netflow handle, or null. */ |
9d189a50 | 93 | struct stp *stp; /* STP or null if disabled. */ |
9efd308e | 94 | struct rstp *rstp; /* RSTP or null if disabled. */ |
46c88433 | 95 | |
46c88433 EJ |
96 | bool has_in_band; /* Bridge has in band control? */ |
97 | bool forward_bpdu; /* Bridge forwards STP BPDUs? */ | |
4b97b70d | 98 | |
b440dd8c JS |
99 | /* Datapath feature support. */ |
100 | struct dpif_backer_support support; | |
46c88433 EJ |
101 | }; |
102 | ||
103 | struct xbundle { | |
104 | struct hmap_node hmap_node; /* In global 'xbundles' map. */ | |
105 | struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */ | |
106 | ||
ca6ba700 | 107 | struct ovs_list list_node; /* In parent 'xbridges' list. */ |
46c88433 EJ |
108 | struct xbridge *xbridge; /* Parent xbridge. */ |
109 | ||
ca6ba700 | 110 | struct ovs_list xports; /* Contains "struct xport"s. */ |
46c88433 EJ |
111 | |
112 | char *name; /* Name used in log messages. */ | |
113 | struct bond *bond; /* Nonnull iff more than one port. */ | |
114 | struct lacp *lacp; /* LACP handle or null. */ | |
115 | ||
116 | enum port_vlan_mode vlan_mode; /* VLAN mode. */ | |
117 | int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */ | |
118 | unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1. | |
119 | * NULL if all VLANs are trunked. */ | |
120 | bool use_priority_tags; /* Use 802.1p tag for frames in VLAN 0? */ | |
121 | bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */ | |
122 | }; | |
123 | ||
124 | struct xport { | |
125 | struct hmap_node hmap_node; /* Node in global 'xports' map. */ | |
126 | struct ofport_dpif *ofport; /* Key in global 'xports map. */ | |
127 | ||
128 | struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */ | |
129 | ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */ | |
130 | ||
131 | odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */ | |
132 | ||
ca6ba700 | 133 | struct ovs_list bundle_node; /* In parent xbundle (if it exists). */ |
46c88433 EJ |
134 | struct xbundle *xbundle; /* Parent xbundle or null. */ |
135 | ||
136 | struct netdev *netdev; /* 'ofport''s netdev. */ | |
137 | ||
138 | struct xbridge *xbridge; /* Parent bridge. */ | |
139 | struct xport *peer; /* Patch port peer or null. */ | |
140 | ||
141 | enum ofputil_port_config config; /* OpenFlow port configuration. */ | |
dd8cd4b4 | 142 | enum ofputil_port_state state; /* OpenFlow port state. */ |
92cf817b | 143 | int stp_port_no; /* STP port number or -1 if not in use. */ |
f025bcb7 | 144 | struct rstp_port *rstp_port; /* RSTP port or null. */ |
46c88433 | 145 | |
55954f6e EJ |
146 | struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */ |
147 | ||
46c88433 EJ |
148 | bool may_enable; /* May be enabled in bonds. */ |
149 | bool is_tunnel; /* Is a tunnel port. */ | |
150 | ||
151 | struct cfm *cfm; /* CFM handle or null. */ | |
152 | struct bfd *bfd; /* BFD handle or null. */ | |
0477baa9 | 153 | struct lldp *lldp; /* LLDP handle or null. */ |
46c88433 EJ |
154 | }; |
155 | ||
4d0acc70 EJ |
156 | struct xlate_ctx { |
157 | struct xlate_in *xin; | |
158 | struct xlate_out *xout; | |
159 | ||
46c88433 | 160 | const struct xbridge *xbridge; |
4d0acc70 | 161 | |
621b8064 | 162 | /* Flow tables version at the beginning of the translation. */ |
18721c4a | 163 | cls_version_t tables_version; |
621b8064 | 164 | |
4d0acc70 EJ |
165 | /* Flow at the last commit. */ |
166 | struct flow base_flow; | |
167 | ||
168 | /* Tunnel IP destination address as received. This is stored separately | |
169 | * as the base_flow.tunnel is cleared on init to reflect the datapath | |
170 | * behavior. Used to make sure not to send tunneled output to ourselves, | |
171 | * which might lead to an infinite loop. This could happen easily | |
172 | * if a tunnel is marked as 'ip_remote=flow', and the flow does not | |
173 | * actually set the tun_dst field. */ | |
174 | ovs_be32 orig_tunnel_ip_dst; | |
175 | ||
176 | /* Stack for the push and pop actions. Each stack element is of type | |
177 | * "union mf_subvalue". */ | |
4d0acc70 EJ |
178 | struct ofpbuf stack; |
179 | ||
180 | /* The rule that we are currently translating, or NULL. */ | |
181 | struct rule_dpif *rule; | |
182 | ||
49a73e0c BP |
183 | /* Flow translation populates this with wildcards relevant in translation. |
184 | * When 'xin->wc' is nonnull, this is the same pointer. When 'xin->wc' is | |
185 | * null, this is a pointer to uninitialized scratch memory. This allows | |
186 | * code to blindly write to 'ctx->wc' without worrying about whether the | |
187 | * caller really wants wildcards. */ | |
188 | struct flow_wildcards *wc; | |
189 | ||
1520ef4f BP |
190 | /* Output buffer for datapath actions. When 'xin->odp_actions' is nonnull, |
191 | * this is the same pointer. When 'xin->odp_actions' is null, this points | |
192 | * to a scratch ofpbuf. This allows code to add actions to | |
193 | * 'ctx->odp_actions' without worrying about whether the caller really | |
194 | * wants actions. */ | |
195 | struct ofpbuf *odp_actions; | |
196 | ||
98b07853 BP |
197 | /* Resubmit statistics, via xlate_table_action(). */ |
198 | int recurse; /* Current resubmit nesting depth. */ | |
199 | int resubmits; /* Total number of resubmits. */ | |
5a070238 | 200 | bool in_group; /* Currently translating ofgroup, if true. */ |
029ca940 | 201 | bool in_action_set; /* Currently translating action_set, if true. */ |
98b07853 | 202 | |
4d0acc70 | 203 | uint8_t table_id; /* OpenFlow table ID where flow was found. */ |
8b1e5560 JR |
204 | ovs_be64 rule_cookie; /* Cookie of the rule being translated. */ |
205 | uint32_t orig_skb_priority; /* Priority when packet arrived. */ | |
4d0acc70 | 206 | uint32_t sflow_n_outputs; /* Number of output ports. */ |
4e022ec0 | 207 | odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */ |
4d0acc70 | 208 | bool exit; /* No further actions should be processed. */ |
3d6151f3 | 209 | mirror_mask_t mirrors; /* Bitmap of associated mirrors. */ |
7fdb60a7 | 210 | |
e672ff9b JR |
211 | /* These are used for non-bond recirculation. The recirculation IDs are |
212 | * stored in xout and must be associated with a datapath flow (ukey), | |
213 | * otherwise they will be freed when the xout is uninitialized. | |
214 | * | |
215 | * | |
216 | * Steps in Recirculation Translation | |
217 | * ================================== | |
218 | * | |
219 | * At some point during translation, the code recognizes the need for | |
220 | * recirculation. For example, recirculation is necessary when, after | |
221 | * popping the last MPLS label, an action or a match tries to examine or | |
222 | * modify a field that has been newly revealed following the MPLS label. | |
223 | * | |
224 | * The simplest part of the work to be done is to commit existing changes to | |
225 | * the packet, which produces datapath actions corresponding to the changes, | |
226 | * and after this, add an OVS_ACTION_ATTR_RECIRC datapath action. | |
227 | * | |
228 | * The main problem here is preserving state. When the datapath executes | |
229 | * OVS_ACTION_ATTR_RECIRC, it will upcall to userspace to get a translation | |
230 | * for the post-recirculation actions. At this point userspace has to | |
231 | * resume the translation where it left off, which means that it has to | |
232 | * execute the following: | |
233 | * | |
234 | * - The action that prompted recirculation, and any actions following | |
235 | * it within the same flow. | |
236 | * | |
237 | * - If the action that prompted recirculation was invoked within a | |
238 | * NXAST_RESUBMIT, then any actions following the resubmit. These | |
239 | * "resubmit"s can be nested, so this has to go all the way up the | |
240 | * control stack. | |
241 | * | |
242 | * - The OpenFlow 1.1+ action set. | |
243 | * | |
244 | * State that actions and flow table lookups can depend on, such as the | |
245 | * following, must also be preserved: | |
246 | * | |
247 | * - Metadata fields (input port, registers, OF1.1+ metadata, ...). | |
248 | * | |
249 | * - Action set, stack | |
250 | * | |
251 | * - The table ID and cookie of the flow being translated at each level | |
252 | * of the control stack (since OFPAT_CONTROLLER actions send these to | |
253 | * the controller). | |
254 | * | |
255 | * Translation allows for the control of this state preservation via these | |
256 | * members. When a need for recirculation is identified, the translation | |
257 | * process: | |
258 | * | |
259 | * 1. Sets 'recirc_action_offset' to the current size of 'action_set'. The | |
260 | * action set is part of what needs to be preserved, so this allows the | |
261 | * action set and the additional state to share the 'action_set' buffer. | |
262 | * Later steps can tell that setup for recirculation is in progress from | |
263 | * the nonnegative value of 'recirc_action_offset'. | |
264 | * | |
265 | * 2. Sets 'exit' to true to tell later steps that we're exiting from the | |
266 | * translation process. | |
267 | * | |
268 | * 3. Adds an OFPACT_UNROLL_XLATE action to 'action_set'. This action | |
269 | * holds the current table ID and cookie so that they can be restored | |
270 | * during a post-recirculation upcall translation. | |
271 | * | |
272 | * 4. Adds the action that prompted recirculation and any actions following | |
273 | * it within the same flow to 'action_set', so that they can be executed | |
274 | * during a post-recirculation upcall translation. | |
275 | * | |
276 | * 5. Returns. | |
277 | * | |
278 | * 6. The action that prompted recirculation might be nested in a stack of | |
279 | * nested "resubmit"s that have actions remaining. Each of these notices | |
280 | * that we're exiting (from 'exit') and that recirculation setup is in | |
281 | * progress (from 'recirc_action_offset') and responds by adding more | |
282 | * OFPACT_UNROLL_XLATE actions to 'action_set', as necessary, and any | |
283 | * actions that were yet unprocessed. | |
284 | * | |
285 | * The caller stores all the state produced by this process associated with | |
286 | * the recirculation ID. For post-recirculation upcall translation, the | |
287 | * caller passes it back in for the new translation to execute. The | |
288 | * process yielded a set of ofpacts that can be translated directly, so it | |
289 | * is not much of a special case at that point. | |
290 | */ | |
291 | int recirc_action_offset; /* Offset in 'action_set' to actions to be | |
292 | * executed after recirculation, or -1. */ | |
293 | int last_unroll_offset; /* Offset in 'action_set' to the latest unroll | |
294 | * action, or -1. */ | |
295 | ||
7bbdd84f SH |
296 | /* True if a packet was but is no longer MPLS (due to an MPLS pop action). |
297 | * This is a trigger for recirculation in cases where translating an action | |
298 | * or looking up a flow requires access to the fields of the packet after | |
1d741d6d | 299 | * the MPLS label stack that was originally present. */ |
7bbdd84f SH |
300 | bool was_mpls; |
301 | ||
7fdb60a7 SH |
302 | /* OpenFlow 1.1+ action set. |
303 | * | |
304 | * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS. | |
305 | * When translation is otherwise complete, ofpacts_execute_action_set() | |
306 | * converts it to a set of "struct ofpact"s that can be translated into | |
ed9c9e3e | 307 | * datapath actions. */ |
c61f3870 | 308 | bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ |
7fdb60a7 | 309 | struct ofpbuf action_set; /* Action set. */ |
4d0acc70 EJ |
310 | }; |
311 | ||
ed9c9e3e JR |
312 | static void xlate_action_set(struct xlate_ctx *ctx); |
313 | ||
1d741d6d JR |
314 | static void |
315 | ctx_trigger_recirculation(struct xlate_ctx *ctx) | |
316 | { | |
317 | ctx->exit = true; | |
318 | ctx->recirc_action_offset = ctx->action_set.size; | |
319 | } | |
320 | ||
321 | static bool | |
322 | ctx_first_recirculation_action(const struct xlate_ctx *ctx) | |
323 | { | |
324 | return ctx->recirc_action_offset == ctx->action_set.size; | |
325 | } | |
326 | ||
e672ff9b JR |
327 | static inline bool |
328 | exit_recirculates(const struct xlate_ctx *ctx) | |
329 | { | |
330 | /* When recirculating the 'recirc_action_offset' has a non-negative value. | |
331 | */ | |
332 | return ctx->recirc_action_offset >= 0; | |
333 | } | |
334 | ||
335 | static void compose_recirculate_action(struct xlate_ctx *ctx); | |
336 | ||
9583bc14 EJ |
337 | /* A controller may use OFPP_NONE as the ingress port to indicate that |
338 | * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for | |
339 | * when an input bundle is needed for validation (e.g., mirroring or | |
340 | * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have | |
3548d242 BP |
341 | * any 'port' structs, so care must be taken when dealing with it. */ |
342 | static struct xbundle ofpp_none_bundle = { | |
343 | .name = "OFPP_NONE", | |
344 | .vlan_mode = PORT_VLAN_TRUNK | |
345 | }; | |
9583bc14 | 346 | |
55954f6e EJ |
347 | /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from |
348 | * 'priority' (the datapath's term for QoS queue) to the dscp bits which all | |
349 | * traffic egressing the 'ofport' with that priority should be marked with. */ | |
350 | struct skb_priority_to_dscp { | |
351 | struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */ | |
352 | uint32_t skb_priority; /* Priority of this queue (see struct flow). */ | |
353 | ||
354 | uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */ | |
355 | }; | |
356 | ||
b256dc52 JS |
357 | enum xc_type { |
358 | XC_RULE, | |
359 | XC_BOND, | |
360 | XC_NETDEV, | |
361 | XC_NETFLOW, | |
362 | XC_MIRROR, | |
363 | XC_LEARN, | |
364 | XC_NORMAL, | |
365 | XC_FIN_TIMEOUT, | |
1e684d7d | 366 | XC_GROUP, |
a36de779 | 367 | XC_TNL_ARP, |
b256dc52 JS |
368 | }; |
369 | ||
370 | /* xlate_cache entries hold enough information to perform the side effects of | |
371 | * xlate_actions() for a rule, without needing to perform rule translation | |
372 | * from scratch. The primary usage of these is to submit statistics to objects | |
373 | * that a flow relates to, although they may be used for other effects as well | |
374 | * (for instance, refreshing hard timeouts for learned flows). */ | |
375 | struct xc_entry { | |
376 | enum xc_type type; | |
377 | union { | |
378 | struct rule_dpif *rule; | |
379 | struct { | |
380 | struct netdev *tx; | |
381 | struct netdev *rx; | |
382 | struct bfd *bfd; | |
383 | } dev; | |
384 | struct { | |
385 | struct netflow *netflow; | |
386 | struct flow *flow; | |
387 | ofp_port_t iface; | |
388 | } nf; | |
389 | struct { | |
390 | struct mbridge *mbridge; | |
391 | mirror_mask_t mirrors; | |
392 | } mirror; | |
393 | struct { | |
394 | struct bond *bond; | |
395 | struct flow *flow; | |
396 | uint16_t vid; | |
397 | } bond; | |
398 | struct { | |
4165b5e0 JS |
399 | struct ofproto_dpif *ofproto; |
400 | struct ofputil_flow_mod *fm; | |
401 | struct ofpbuf *ofpacts; | |
b256dc52 JS |
402 | } learn; |
403 | struct { | |
404 | struct ofproto_dpif *ofproto; | |
405 | struct flow *flow; | |
406 | int vlan; | |
407 | } normal; | |
408 | struct { | |
409 | struct rule_dpif *rule; | |
410 | uint16_t idle; | |
411 | uint16_t hard; | |
412 | } fin; | |
1e684d7d RW |
413 | struct { |
414 | struct group_dpif *group; | |
415 | struct ofputil_bucket *bucket; | |
416 | } group; | |
a36de779 PS |
417 | struct { |
418 | char br_name[IFNAMSIZ]; | |
419 | ovs_be32 d_ip; | |
420 | } tnl_arp_cache; | |
b256dc52 JS |
421 | } u; |
422 | }; | |
423 | ||
424 | #define XC_ENTRY_FOR_EACH(entry, entries, xcache) \ | |
425 | entries = xcache->entries; \ | |
426 | for (entry = ofpbuf_try_pull(&entries, sizeof *entry); \ | |
427 | entry; \ | |
428 | entry = ofpbuf_try_pull(&entries, sizeof *entry)) | |
429 | ||
430 | struct xlate_cache { | |
431 | struct ofpbuf entries; | |
432 | }; | |
433 | ||
84f0f298 RW |
434 | /* Xlate config contains hash maps of all bridges, bundles and ports. |
435 | * Xcfgp contains the pointer to the current xlate configuration. | |
436 | * When the main thread needs to change the configuration, it copies xcfgp to | |
437 | * new_xcfg and edits new_xcfg. This enables the use of RCU locking which | |
438 | * does not block handler and revalidator threads. */ | |
439 | struct xlate_cfg { | |
440 | struct hmap xbridges; | |
441 | struct hmap xbundles; | |
442 | struct hmap xports; | |
443 | }; | |
b1b72f2d | 444 | static OVSRCU_TYPE(struct xlate_cfg *) xcfgp = OVSRCU_INITIALIZER(NULL); |
f439f23b | 445 | static struct xlate_cfg *new_xcfg = NULL; |
46c88433 EJ |
446 | |
447 | static bool may_receive(const struct xport *, struct xlate_ctx *); | |
9583bc14 EJ |
448 | static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
449 | struct xlate_ctx *); | |
adcf00ba | 450 | static void xlate_normal(struct xlate_ctx *); |
c1b3756c BP |
451 | static inline void xlate_report(struct xlate_ctx *, const char *, ...) |
452 | OVS_PRINTF_FORMAT(2, 3); | |
6d328fa2 SH |
453 | static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port, |
454 | uint8_t table_id, bool may_packet_in, | |
455 | bool honor_table_miss); | |
46c88433 EJ |
456 | static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn); |
457 | static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid); | |
458 | static void output_normal(struct xlate_ctx *, const struct xbundle *, | |
9583bc14 | 459 | uint16_t vlan); |
e93ef1c7 JR |
460 | |
461 | /* Optional bond recirculation parameter to compose_output_action(). */ | |
462 | struct xlate_bond_recirc { | |
463 | uint32_t recirc_id; /* !0 Use recirculation instead of output. */ | |
464 | uint8_t hash_alg; /* !0 Compute hash for recirc before. */ | |
465 | uint32_t hash_basis; /* Compute hash for recirc before. */ | |
466 | }; | |
467 | ||
468 | static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port, | |
469 | const struct xlate_bond_recirc *xr); | |
9583bc14 | 470 | |
84f0f298 RW |
471 | static struct xbridge *xbridge_lookup(struct xlate_cfg *, |
472 | const struct ofproto_dpif *); | |
473 | static struct xbundle *xbundle_lookup(struct xlate_cfg *, | |
474 | const struct ofbundle *); | |
475 | static struct xport *xport_lookup(struct xlate_cfg *, | |
476 | const struct ofport_dpif *); | |
46c88433 | 477 | static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port); |
55954f6e EJ |
478 | static struct skb_priority_to_dscp *get_skb_priority(const struct xport *, |
479 | uint32_t skb_priority); | |
480 | static void clear_skb_priorities(struct xport *); | |
16194afd | 481 | static size_t count_skb_priorities(const struct xport *); |
55954f6e EJ |
482 | static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, |
483 | uint8_t *dscp); | |
46c88433 | 484 | |
b256dc52 JS |
485 | static struct xc_entry *xlate_cache_add_entry(struct xlate_cache *xc, |
486 | enum xc_type type); | |
84f0f298 RW |
487 | static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *); |
488 | static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *); | |
489 | static void xlate_xport_init(struct xlate_cfg *, struct xport *); | |
9efd308e | 490 | static void xlate_xbridge_set(struct xbridge *, struct dpif *, |
9efd308e DV |
491 | const struct mac_learning *, struct stp *, |
492 | struct rstp *, const struct mcast_snooping *, | |
493 | const struct mbridge *, | |
494 | const struct dpif_sflow *, | |
495 | const struct dpif_ipfix *, | |
2f47cdf4 | 496 | const struct netflow *, |
84f0f298 | 497 | bool forward_bpdu, bool has_in_band, |
b440dd8c | 498 | const struct dpif_backer_support *); |
84f0f298 RW |
499 | static void xlate_xbundle_set(struct xbundle *xbundle, |
500 | enum port_vlan_mode vlan_mode, int vlan, | |
501 | unsigned long *trunks, bool use_priority_tags, | |
502 | const struct bond *bond, const struct lacp *lacp, | |
503 | bool floodable); | |
504 | static void xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
505 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 DF |
506 | const struct bfd *bfd, const struct lldp *lldp, |
507 | int stp_port_no, const struct rstp_port *rstp_port, | |
84f0f298 RW |
508 | enum ofputil_port_config config, |
509 | enum ofputil_port_state state, bool is_tunnel, | |
510 | bool may_enable); | |
511 | static void xlate_xbridge_remove(struct xlate_cfg *, struct xbridge *); | |
512 | static void xlate_xbundle_remove(struct xlate_cfg *, struct xbundle *); | |
513 | static void xlate_xport_remove(struct xlate_cfg *, struct xport *); | |
514 | static void xlate_xbridge_copy(struct xbridge *); | |
515 | static void xlate_xbundle_copy(struct xbridge *, struct xbundle *); | |
516 | static void xlate_xport_copy(struct xbridge *, struct xbundle *, | |
517 | struct xport *); | |
518 | static void xlate_xcfg_free(struct xlate_cfg *); | |
b256dc52 | 519 | |
34dd0d78 | 520 | static inline void |
c1b3756c | 521 | xlate_report(struct xlate_ctx *ctx, const char *format, ...) |
34dd0d78 JR |
522 | { |
523 | if (OVS_UNLIKELY(ctx->xin->report_hook)) { | |
c1b3756c BP |
524 | va_list args; |
525 | ||
526 | va_start(args, format); | |
527 | ctx->xin->report_hook(ctx->xin, ctx->recurse, format, args); | |
528 | va_end(args); | |
34dd0d78 JR |
529 | } |
530 | } | |
84f0f298 RW |
531 | |
532 | static void | |
533 | xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
534 | { | |
535 | list_init(&xbridge->xbundles); | |
536 | hmap_init(&xbridge->xports); | |
537 | hmap_insert(&xcfg->xbridges, &xbridge->hmap_node, | |
538 | hash_pointer(xbridge->ofproto, 0)); | |
539 | } | |
540 | ||
541 | static void | |
542 | xlate_xbundle_init(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
543 | { | |
544 | list_init(&xbundle->xports); | |
545 | list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); | |
546 | hmap_insert(&xcfg->xbundles, &xbundle->hmap_node, | |
547 | hash_pointer(xbundle->ofbundle, 0)); | |
548 | } | |
549 | ||
550 | static void | |
551 | xlate_xport_init(struct xlate_cfg *xcfg, struct xport *xport) | |
552 | { | |
553 | hmap_init(&xport->skb_priorities); | |
554 | hmap_insert(&xcfg->xports, &xport->hmap_node, | |
555 | hash_pointer(xport->ofport, 0)); | |
556 | hmap_insert(&xport->xbridge->xports, &xport->ofp_node, | |
557 | hash_ofp_port(xport->ofp_port)); | |
558 | } | |
559 | ||
560 | static void | |
561 | xlate_xbridge_set(struct xbridge *xbridge, | |
562 | struct dpif *dpif, | |
ec89fc6f | 563 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 564 | struct rstp *rstp, const struct mcast_snooping *ms, |
ec89fc6f | 565 | const struct mbridge *mbridge, |
46c88433 | 566 | const struct dpif_sflow *sflow, |
ce3955be | 567 | const struct dpif_ipfix *ipfix, |
2f47cdf4 | 568 | const struct netflow *netflow, |
4b97b70d | 569 | bool forward_bpdu, bool has_in_band, |
b440dd8c | 570 | const struct dpif_backer_support *support) |
46c88433 | 571 | { |
46c88433 EJ |
572 | if (xbridge->ml != ml) { |
573 | mac_learning_unref(xbridge->ml); | |
574 | xbridge->ml = mac_learning_ref(ml); | |
575 | } | |
576 | ||
6d95c4e8 FL |
577 | if (xbridge->ms != ms) { |
578 | mcast_snooping_unref(xbridge->ms); | |
579 | xbridge->ms = mcast_snooping_ref(ms); | |
580 | } | |
581 | ||
46c88433 EJ |
582 | if (xbridge->mbridge != mbridge) { |
583 | mbridge_unref(xbridge->mbridge); | |
584 | xbridge->mbridge = mbridge_ref(mbridge); | |
585 | } | |
586 | ||
587 | if (xbridge->sflow != sflow) { | |
588 | dpif_sflow_unref(xbridge->sflow); | |
589 | xbridge->sflow = dpif_sflow_ref(sflow); | |
590 | } | |
591 | ||
592 | if (xbridge->ipfix != ipfix) { | |
593 | dpif_ipfix_unref(xbridge->ipfix); | |
594 | xbridge->ipfix = dpif_ipfix_ref(ipfix); | |
595 | } | |
596 | ||
9d189a50 EJ |
597 | if (xbridge->stp != stp) { |
598 | stp_unref(xbridge->stp); | |
599 | xbridge->stp = stp_ref(stp); | |
600 | } | |
601 | ||
9efd308e DV |
602 | if (xbridge->rstp != rstp) { |
603 | rstp_unref(xbridge->rstp); | |
604 | xbridge->rstp = rstp_ref(rstp); | |
605 | } | |
606 | ||
ce3955be EJ |
607 | if (xbridge->netflow != netflow) { |
608 | netflow_unref(xbridge->netflow); | |
609 | xbridge->netflow = netflow_ref(netflow); | |
610 | } | |
611 | ||
89a8a7f0 | 612 | xbridge->dpif = dpif; |
46c88433 EJ |
613 | xbridge->forward_bpdu = forward_bpdu; |
614 | xbridge->has_in_band = has_in_band; | |
b440dd8c | 615 | xbridge->support = *support; |
46c88433 EJ |
616 | } |
617 | ||
84f0f298 RW |
618 | static void |
619 | xlate_xbundle_set(struct xbundle *xbundle, | |
620 | enum port_vlan_mode vlan_mode, int vlan, | |
621 | unsigned long *trunks, bool use_priority_tags, | |
622 | const struct bond *bond, const struct lacp *lacp, | |
623 | bool floodable) | |
624 | { | |
625 | ovs_assert(xbundle->xbridge); | |
626 | ||
627 | xbundle->vlan_mode = vlan_mode; | |
628 | xbundle->vlan = vlan; | |
629 | xbundle->trunks = trunks; | |
630 | xbundle->use_priority_tags = use_priority_tags; | |
631 | xbundle->floodable = floodable; | |
632 | ||
633 | if (xbundle->bond != bond) { | |
634 | bond_unref(xbundle->bond); | |
635 | xbundle->bond = bond_ref(bond); | |
636 | } | |
637 | ||
638 | if (xbundle->lacp != lacp) { | |
639 | lacp_unref(xbundle->lacp); | |
640 | xbundle->lacp = lacp_ref(lacp); | |
641 | } | |
642 | } | |
643 | ||
644 | static void | |
645 | xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
646 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 | 647 | const struct bfd *bfd, const struct lldp *lldp, int stp_port_no, |
f025bcb7 | 648 | const struct rstp_port* rstp_port, |
84f0f298 RW |
649 | enum ofputil_port_config config, enum ofputil_port_state state, |
650 | bool is_tunnel, bool may_enable) | |
651 | { | |
652 | xport->config = config; | |
653 | xport->state = state; | |
654 | xport->stp_port_no = stp_port_no; | |
655 | xport->is_tunnel = is_tunnel; | |
656 | xport->may_enable = may_enable; | |
657 | xport->odp_port = odp_port; | |
658 | ||
f025bcb7 JR |
659 | if (xport->rstp_port != rstp_port) { |
660 | rstp_port_unref(xport->rstp_port); | |
661 | xport->rstp_port = rstp_port_ref(rstp_port); | |
662 | } | |
663 | ||
84f0f298 RW |
664 | if (xport->cfm != cfm) { |
665 | cfm_unref(xport->cfm); | |
666 | xport->cfm = cfm_ref(cfm); | |
667 | } | |
668 | ||
669 | if (xport->bfd != bfd) { | |
670 | bfd_unref(xport->bfd); | |
671 | xport->bfd = bfd_ref(bfd); | |
672 | } | |
673 | ||
0477baa9 DF |
674 | if (xport->lldp != lldp) { |
675 | lldp_unref(xport->lldp); | |
676 | xport->lldp = lldp_ref(lldp); | |
677 | } | |
678 | ||
84f0f298 RW |
679 | if (xport->netdev != netdev) { |
680 | netdev_close(xport->netdev); | |
681 | xport->netdev = netdev_ref(netdev); | |
682 | } | |
683 | } | |
684 | ||
685 | static void | |
686 | xlate_xbridge_copy(struct xbridge *xbridge) | |
687 | { | |
688 | struct xbundle *xbundle; | |
689 | struct xport *xport; | |
690 | struct xbridge *new_xbridge = xzalloc(sizeof *xbridge); | |
691 | new_xbridge->ofproto = xbridge->ofproto; | |
692 | new_xbridge->name = xstrdup(xbridge->name); | |
693 | xlate_xbridge_init(new_xcfg, new_xbridge); | |
694 | ||
695 | xlate_xbridge_set(new_xbridge, | |
34dd0d78 | 696 | xbridge->dpif, xbridge->ml, xbridge->stp, |
9efd308e DV |
697 | xbridge->rstp, xbridge->ms, xbridge->mbridge, |
698 | xbridge->sflow, xbridge->ipfix, xbridge->netflow, | |
b440dd8c JS |
699 | xbridge->forward_bpdu, xbridge->has_in_band, |
700 | &xbridge->support); | |
84f0f298 RW |
701 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
702 | xlate_xbundle_copy(new_xbridge, xbundle); | |
703 | } | |
704 | ||
705 | /* Copy xports which are not part of a xbundle */ | |
706 | HMAP_FOR_EACH (xport, ofp_node, &xbridge->xports) { | |
707 | if (!xport->xbundle) { | |
708 | xlate_xport_copy(new_xbridge, NULL, xport); | |
709 | } | |
710 | } | |
711 | } | |
712 | ||
713 | static void | |
714 | xlate_xbundle_copy(struct xbridge *xbridge, struct xbundle *xbundle) | |
715 | { | |
716 | struct xport *xport; | |
717 | struct xbundle *new_xbundle = xzalloc(sizeof *xbundle); | |
718 | new_xbundle->ofbundle = xbundle->ofbundle; | |
719 | new_xbundle->xbridge = xbridge; | |
720 | new_xbundle->name = xstrdup(xbundle->name); | |
721 | xlate_xbundle_init(new_xcfg, new_xbundle); | |
722 | ||
723 | xlate_xbundle_set(new_xbundle, xbundle->vlan_mode, | |
724 | xbundle->vlan, xbundle->trunks, | |
725 | xbundle->use_priority_tags, xbundle->bond, xbundle->lacp, | |
726 | xbundle->floodable); | |
727 | LIST_FOR_EACH (xport, bundle_node, &xbundle->xports) { | |
728 | xlate_xport_copy(xbridge, new_xbundle, xport); | |
729 | } | |
730 | } | |
731 | ||
732 | static void | |
733 | xlate_xport_copy(struct xbridge *xbridge, struct xbundle *xbundle, | |
734 | struct xport *xport) | |
735 | { | |
736 | struct skb_priority_to_dscp *pdscp, *new_pdscp; | |
737 | struct xport *new_xport = xzalloc(sizeof *xport); | |
738 | new_xport->ofport = xport->ofport; | |
739 | new_xport->ofp_port = xport->ofp_port; | |
740 | new_xport->xbridge = xbridge; | |
741 | xlate_xport_init(new_xcfg, new_xport); | |
742 | ||
743 | xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm, | |
0477baa9 DF |
744 | xport->bfd, xport->lldp, xport->stp_port_no, |
745 | xport->rstp_port, xport->config, xport->state, | |
746 | xport->is_tunnel, xport->may_enable); | |
84f0f298 RW |
747 | |
748 | if (xport->peer) { | |
749 | struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport); | |
750 | if (peer) { | |
751 | new_xport->peer = peer; | |
752 | new_xport->peer->peer = new_xport; | |
753 | } | |
754 | } | |
755 | ||
756 | if (xbundle) { | |
757 | new_xport->xbundle = xbundle; | |
758 | list_insert(&new_xport->xbundle->xports, &new_xport->bundle_node); | |
759 | } | |
760 | ||
761 | HMAP_FOR_EACH (pdscp, hmap_node, &xport->skb_priorities) { | |
762 | new_pdscp = xmalloc(sizeof *pdscp); | |
763 | new_pdscp->skb_priority = pdscp->skb_priority; | |
764 | new_pdscp->dscp = pdscp->dscp; | |
765 | hmap_insert(&new_xport->skb_priorities, &new_pdscp->hmap_node, | |
766 | hash_int(new_pdscp->skb_priority, 0)); | |
767 | } | |
768 | } | |
769 | ||
770 | /* Sets the current xlate configuration to new_xcfg and frees the old xlate | |
771 | * configuration in xcfgp. | |
772 | * | |
773 | * This needs to be called after editing the xlate configuration. | |
774 | * | |
775 | * Functions that edit the new xlate configuration are | |
776 | * xlate_<ofport/bundle/ofport>_set and xlate_<ofport/bundle/ofport>_remove. | |
777 | * | |
778 | * A sample workflow: | |
779 | * | |
780 | * xlate_txn_start(); | |
781 | * ... | |
782 | * edit_xlate_configuration(); | |
783 | * ... | |
784 | * xlate_txn_commit(); */ | |
46c88433 | 785 | void |
84f0f298 RW |
786 | xlate_txn_commit(void) |
787 | { | |
788 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
789 | ||
790 | ovsrcu_set(&xcfgp, new_xcfg); | |
40a9c4c2 AW |
791 | ovsrcu_synchronize(); |
792 | xlate_xcfg_free(xcfg); | |
84f0f298 RW |
793 | new_xcfg = NULL; |
794 | } | |
795 | ||
796 | /* Copies the current xlate configuration in xcfgp to new_xcfg. | |
797 | * | |
798 | * This needs to be called prior to editing the xlate configuration. */ | |
799 | void | |
800 | xlate_txn_start(void) | |
801 | { | |
802 | struct xbridge *xbridge; | |
803 | struct xlate_cfg *xcfg; | |
804 | ||
805 | ovs_assert(!new_xcfg); | |
806 | ||
807 | new_xcfg = xmalloc(sizeof *new_xcfg); | |
808 | hmap_init(&new_xcfg->xbridges); | |
809 | hmap_init(&new_xcfg->xbundles); | |
810 | hmap_init(&new_xcfg->xports); | |
811 | ||
812 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
813 | if (!xcfg) { | |
814 | return; | |
815 | } | |
816 | ||
817 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
818 | xlate_xbridge_copy(xbridge); | |
819 | } | |
820 | } | |
821 | ||
822 | ||
823 | static void | |
824 | xlate_xcfg_free(struct xlate_cfg *xcfg) | |
825 | { | |
826 | struct xbridge *xbridge, *next_xbridge; | |
827 | ||
828 | if (!xcfg) { | |
829 | return; | |
830 | } | |
831 | ||
832 | HMAP_FOR_EACH_SAFE (xbridge, next_xbridge, hmap_node, &xcfg->xbridges) { | |
833 | xlate_xbridge_remove(xcfg, xbridge); | |
834 | } | |
835 | ||
836 | hmap_destroy(&xcfg->xbridges); | |
837 | hmap_destroy(&xcfg->xbundles); | |
838 | hmap_destroy(&xcfg->xports); | |
839 | free(xcfg); | |
840 | } | |
841 | ||
842 | void | |
843 | xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, | |
34dd0d78 | 844 | struct dpif *dpif, |
84f0f298 | 845 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 846 | struct rstp *rstp, const struct mcast_snooping *ms, |
84f0f298 RW |
847 | const struct mbridge *mbridge, |
848 | const struct dpif_sflow *sflow, | |
849 | const struct dpif_ipfix *ipfix, | |
2f47cdf4 | 850 | const struct netflow *netflow, |
b440dd8c JS |
851 | bool forward_bpdu, bool has_in_band, |
852 | const struct dpif_backer_support *support) | |
84f0f298 RW |
853 | { |
854 | struct xbridge *xbridge; | |
855 | ||
856 | ovs_assert(new_xcfg); | |
857 | ||
858 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
859 | if (!xbridge) { | |
860 | xbridge = xzalloc(sizeof *xbridge); | |
861 | xbridge->ofproto = ofproto; | |
862 | ||
863 | xlate_xbridge_init(new_xcfg, xbridge); | |
864 | } | |
865 | ||
866 | free(xbridge->name); | |
867 | xbridge->name = xstrdup(name); | |
868 | ||
34dd0d78 | 869 | xlate_xbridge_set(xbridge, dpif, ml, stp, rstp, ms, mbridge, sflow, ipfix, |
b440dd8c | 870 | netflow, forward_bpdu, has_in_band, support); |
84f0f298 RW |
871 | } |
872 | ||
873 | static void | |
874 | xlate_xbridge_remove(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
46c88433 | 875 | { |
46c88433 EJ |
876 | struct xbundle *xbundle, *next_xbundle; |
877 | struct xport *xport, *next_xport; | |
878 | ||
879 | if (!xbridge) { | |
880 | return; | |
881 | } | |
882 | ||
883 | HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) { | |
84f0f298 | 884 | xlate_xport_remove(xcfg, xport); |
46c88433 EJ |
885 | } |
886 | ||
887 | LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) { | |
84f0f298 | 888 | xlate_xbundle_remove(xcfg, xbundle); |
46c88433 EJ |
889 | } |
890 | ||
84f0f298 | 891 | hmap_remove(&xcfg->xbridges, &xbridge->hmap_node); |
795cc5c1 | 892 | mac_learning_unref(xbridge->ml); |
6d95c4e8 | 893 | mcast_snooping_unref(xbridge->ms); |
795cc5c1 EJ |
894 | mbridge_unref(xbridge->mbridge); |
895 | dpif_sflow_unref(xbridge->sflow); | |
896 | dpif_ipfix_unref(xbridge->ipfix); | |
897 | stp_unref(xbridge->stp); | |
9efd308e | 898 | rstp_unref(xbridge->rstp); |
795cc5c1 | 899 | hmap_destroy(&xbridge->xports); |
46c88433 EJ |
900 | free(xbridge->name); |
901 | free(xbridge); | |
902 | } | |
903 | ||
84f0f298 RW |
904 | void |
905 | xlate_remove_ofproto(struct ofproto_dpif *ofproto) | |
906 | { | |
907 | struct xbridge *xbridge; | |
908 | ||
909 | ovs_assert(new_xcfg); | |
910 | ||
911 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
912 | xlate_xbridge_remove(new_xcfg, xbridge); | |
913 | } | |
914 | ||
46c88433 EJ |
915 | void |
916 | xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
917 | const char *name, enum port_vlan_mode vlan_mode, int vlan, | |
918 | unsigned long *trunks, bool use_priority_tags, | |
919 | const struct bond *bond, const struct lacp *lacp, | |
920 | bool floodable) | |
921 | { | |
84f0f298 | 922 | struct xbundle *xbundle; |
46c88433 | 923 | |
84f0f298 RW |
924 | ovs_assert(new_xcfg); |
925 | ||
926 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
46c88433 EJ |
927 | if (!xbundle) { |
928 | xbundle = xzalloc(sizeof *xbundle); | |
929 | xbundle->ofbundle = ofbundle; | |
84f0f298 | 930 | xbundle->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 931 | |
84f0f298 | 932 | xlate_xbundle_init(new_xcfg, xbundle); |
46c88433 EJ |
933 | } |
934 | ||
46c88433 EJ |
935 | free(xbundle->name); |
936 | xbundle->name = xstrdup(name); | |
937 | ||
84f0f298 RW |
938 | xlate_xbundle_set(xbundle, vlan_mode, vlan, trunks, |
939 | use_priority_tags, bond, lacp, floodable); | |
46c88433 EJ |
940 | } |
941 | ||
84f0f298 RW |
942 | static void |
943 | xlate_xbundle_remove(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
46c88433 | 944 | { |
5f03c983 | 945 | struct xport *xport; |
46c88433 EJ |
946 | |
947 | if (!xbundle) { | |
948 | return; | |
949 | } | |
950 | ||
5f03c983 | 951 | LIST_FOR_EACH_POP (xport, bundle_node, &xbundle->xports) { |
46c88433 EJ |
952 | xport->xbundle = NULL; |
953 | } | |
954 | ||
84f0f298 | 955 | hmap_remove(&xcfg->xbundles, &xbundle->hmap_node); |
46c88433 EJ |
956 | list_remove(&xbundle->list_node); |
957 | bond_unref(xbundle->bond); | |
958 | lacp_unref(xbundle->lacp); | |
959 | free(xbundle->name); | |
960 | free(xbundle); | |
961 | } | |
962 | ||
84f0f298 RW |
963 | void |
964 | xlate_bundle_remove(struct ofbundle *ofbundle) | |
965 | { | |
966 | struct xbundle *xbundle; | |
967 | ||
968 | ovs_assert(new_xcfg); | |
969 | ||
970 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
971 | xlate_xbundle_remove(new_xcfg, xbundle); | |
972 | } | |
973 | ||
46c88433 EJ |
974 | void |
975 | xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
976 | struct ofport_dpif *ofport, ofp_port_t ofp_port, | |
977 | odp_port_t odp_port, const struct netdev *netdev, | |
978 | const struct cfm *cfm, const struct bfd *bfd, | |
0477baa9 DF |
979 | const struct lldp *lldp, struct ofport_dpif *peer, |
980 | int stp_port_no, const struct rstp_port *rstp_port, | |
55954f6e | 981 | const struct ofproto_port_queue *qdscp_list, size_t n_qdscp, |
dd8cd4b4 SH |
982 | enum ofputil_port_config config, |
983 | enum ofputil_port_state state, bool is_tunnel, | |
9d189a50 | 984 | bool may_enable) |
46c88433 | 985 | { |
55954f6e | 986 | size_t i; |
84f0f298 RW |
987 | struct xport *xport; |
988 | ||
989 | ovs_assert(new_xcfg); | |
46c88433 | 990 | |
84f0f298 | 991 | xport = xport_lookup(new_xcfg, ofport); |
46c88433 EJ |
992 | if (!xport) { |
993 | xport = xzalloc(sizeof *xport); | |
994 | xport->ofport = ofport; | |
84f0f298 | 995 | xport->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 EJ |
996 | xport->ofp_port = ofp_port; |
997 | ||
84f0f298 | 998 | xlate_xport_init(new_xcfg, xport); |
46c88433 EJ |
999 | } |
1000 | ||
1001 | ovs_assert(xport->ofp_port == ofp_port); | |
1002 | ||
0477baa9 DF |
1003 | xlate_xport_set(xport, odp_port, netdev, cfm, bfd, lldp, |
1004 | stp_port_no, rstp_port, config, state, is_tunnel, | |
1005 | may_enable); | |
46c88433 EJ |
1006 | |
1007 | if (xport->peer) { | |
1008 | xport->peer->peer = NULL; | |
1009 | } | |
84f0f298 | 1010 | xport->peer = xport_lookup(new_xcfg, peer); |
46c88433 EJ |
1011 | if (xport->peer) { |
1012 | xport->peer->peer = xport; | |
1013 | } | |
1014 | ||
1015 | if (xport->xbundle) { | |
1016 | list_remove(&xport->bundle_node); | |
1017 | } | |
84f0f298 | 1018 | xport->xbundle = xbundle_lookup(new_xcfg, ofbundle); |
46c88433 EJ |
1019 | if (xport->xbundle) { |
1020 | list_insert(&xport->xbundle->xports, &xport->bundle_node); | |
1021 | } | |
55954f6e EJ |
1022 | |
1023 | clear_skb_priorities(xport); | |
1024 | for (i = 0; i < n_qdscp; i++) { | |
1025 | struct skb_priority_to_dscp *pdscp; | |
1026 | uint32_t skb_priority; | |
1027 | ||
89a8a7f0 EJ |
1028 | if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue, |
1029 | &skb_priority)) { | |
55954f6e EJ |
1030 | continue; |
1031 | } | |
1032 | ||
1033 | pdscp = xmalloc(sizeof *pdscp); | |
1034 | pdscp->skb_priority = skb_priority; | |
1035 | pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK; | |
1036 | hmap_insert(&xport->skb_priorities, &pdscp->hmap_node, | |
1037 | hash_int(pdscp->skb_priority, 0)); | |
1038 | } | |
46c88433 EJ |
1039 | } |
1040 | ||
84f0f298 RW |
1041 | static void |
1042 | xlate_xport_remove(struct xlate_cfg *xcfg, struct xport *xport) | |
46c88433 | 1043 | { |
46c88433 EJ |
1044 | if (!xport) { |
1045 | return; | |
1046 | } | |
1047 | ||
1048 | if (xport->peer) { | |
1049 | xport->peer->peer = NULL; | |
1050 | xport->peer = NULL; | |
1051 | } | |
1052 | ||
e621a12d EJ |
1053 | if (xport->xbundle) { |
1054 | list_remove(&xport->bundle_node); | |
1055 | } | |
1056 | ||
55954f6e EJ |
1057 | clear_skb_priorities(xport); |
1058 | hmap_destroy(&xport->skb_priorities); | |
1059 | ||
84f0f298 | 1060 | hmap_remove(&xcfg->xports, &xport->hmap_node); |
46c88433 EJ |
1061 | hmap_remove(&xport->xbridge->xports, &xport->ofp_node); |
1062 | ||
1063 | netdev_close(xport->netdev); | |
f025bcb7 | 1064 | rstp_port_unref(xport->rstp_port); |
46c88433 EJ |
1065 | cfm_unref(xport->cfm); |
1066 | bfd_unref(xport->bfd); | |
0477baa9 | 1067 | lldp_unref(xport->lldp); |
46c88433 EJ |
1068 | free(xport); |
1069 | } | |
1070 | ||
84f0f298 RW |
1071 | void |
1072 | xlate_ofport_remove(struct ofport_dpif *ofport) | |
1073 | { | |
1074 | struct xport *xport; | |
1075 | ||
1076 | ovs_assert(new_xcfg); | |
1077 | ||
1078 | xport = xport_lookup(new_xcfg, ofport); | |
1079 | xlate_xport_remove(new_xcfg, xport); | |
1080 | } | |
1081 | ||
ef377a58 JR |
1082 | static struct ofproto_dpif * |
1083 | xlate_lookup_ofproto_(const struct dpif_backer *backer, const struct flow *flow, | |
1084 | ofp_port_t *ofp_in_port, const struct xport **xportp) | |
1085 | { | |
e672ff9b | 1086 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
ef377a58 | 1087 | const struct xport *xport; |
f9038ef6 | 1088 | |
e672ff9b JR |
1089 | xport = xport_lookup(xcfg, tnl_port_should_receive(flow) |
1090 | ? tnl_port_receive(flow) | |
1091 | : odp_port_to_ofport(backer, flow->in_port.odp_port)); | |
1092 | if (OVS_UNLIKELY(!xport)) { | |
1093 | return NULL; | |
ef377a58 | 1094 | } |
e672ff9b | 1095 | *xportp = xport; |
f9038ef6 | 1096 | if (ofp_in_port) { |
e672ff9b | 1097 | *ofp_in_port = xport->ofp_port; |
f9038ef6 | 1098 | } |
e672ff9b | 1099 | return xport->xbridge->ofproto; |
ef377a58 JR |
1100 | } |
1101 | ||
1102 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively) | |
1103 | * returns the corresponding struct ofproto_dpif and OpenFlow port number. */ | |
1104 | struct ofproto_dpif * | |
1105 | xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow, | |
1106 | ofp_port_t *ofp_in_port) | |
1107 | { | |
1108 | const struct xport *xport; | |
1109 | ||
1110 | return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); | |
1111 | } | |
1112 | ||
cc377352 | 1113 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively), |
ef377a58 | 1114 | * optionally populates 'ofproto' with the ofproto_dpif, 'ofp_in_port' with the |
cc377352 | 1115 | * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate |
dcc2c6cd JR |
1116 | * handles for those protocols if they're enabled. Caller may use the returned |
1117 | * pointers until quiescing, for longer term use additional references must | |
1118 | * be taken. | |
8449c4d6 | 1119 | * |
f9038ef6 | 1120 | * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto. |
ef377a58 | 1121 | */ |
8449c4d6 | 1122 | int |
5c476ea3 JR |
1123 | xlate_lookup(const struct dpif_backer *backer, const struct flow *flow, |
1124 | struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix, | |
1125 | struct dpif_sflow **sflow, struct netflow **netflow, | |
1126 | ofp_port_t *ofp_in_port) | |
8449c4d6 | 1127 | { |
ef377a58 | 1128 | struct ofproto_dpif *ofproto; |
84f0f298 | 1129 | const struct xport *xport; |
8449c4d6 | 1130 | |
ef377a58 | 1131 | ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); |
8449c4d6 | 1132 | |
f9038ef6 | 1133 | if (!ofproto) { |
cc377352 | 1134 | return ENODEV; |
8449c4d6 | 1135 | } |
8449c4d6 | 1136 | |
ef377a58 JR |
1137 | if (ofprotop) { |
1138 | *ofprotop = ofproto; | |
8449c4d6 EJ |
1139 | } |
1140 | ||
1dfdb9b3 | 1141 | if (ipfix) { |
f9038ef6 | 1142 | *ipfix = xport ? xport->xbridge->ipfix : NULL; |
1dfdb9b3 EJ |
1143 | } |
1144 | ||
1145 | if (sflow) { | |
f9038ef6 | 1146 | *sflow = xport ? xport->xbridge->sflow : NULL; |
1dfdb9b3 EJ |
1147 | } |
1148 | ||
1149 | if (netflow) { | |
f9038ef6 | 1150 | *netflow = xport ? xport->xbridge->netflow : NULL; |
1dfdb9b3 | 1151 | } |
f9038ef6 | 1152 | |
cc377352 | 1153 | return 0; |
8449c4d6 EJ |
1154 | } |
1155 | ||
46c88433 | 1156 | static struct xbridge * |
84f0f298 | 1157 | xbridge_lookup(struct xlate_cfg *xcfg, const struct ofproto_dpif *ofproto) |
46c88433 | 1158 | { |
84f0f298 | 1159 | struct hmap *xbridges; |
46c88433 EJ |
1160 | struct xbridge *xbridge; |
1161 | ||
84f0f298 | 1162 | if (!ofproto || !xcfg) { |
5e6af486 EJ |
1163 | return NULL; |
1164 | } | |
1165 | ||
84f0f298 RW |
1166 | xbridges = &xcfg->xbridges; |
1167 | ||
46c88433 | 1168 | HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0), |
84f0f298 | 1169 | xbridges) { |
46c88433 EJ |
1170 | if (xbridge->ofproto == ofproto) { |
1171 | return xbridge; | |
1172 | } | |
1173 | } | |
1174 | return NULL; | |
1175 | } | |
1176 | ||
1177 | static struct xbundle * | |
84f0f298 | 1178 | xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle) |
46c88433 | 1179 | { |
84f0f298 | 1180 | struct hmap *xbundles; |
46c88433 EJ |
1181 | struct xbundle *xbundle; |
1182 | ||
84f0f298 | 1183 | if (!ofbundle || !xcfg) { |
5e6af486 EJ |
1184 | return NULL; |
1185 | } | |
1186 | ||
84f0f298 RW |
1187 | xbundles = &xcfg->xbundles; |
1188 | ||
46c88433 | 1189 | HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), |
84f0f298 | 1190 | xbundles) { |
46c88433 EJ |
1191 | if (xbundle->ofbundle == ofbundle) { |
1192 | return xbundle; | |
1193 | } | |
1194 | } | |
1195 | return NULL; | |
1196 | } | |
1197 | ||
1198 | static struct xport * | |
84f0f298 | 1199 | xport_lookup(struct xlate_cfg *xcfg, const struct ofport_dpif *ofport) |
46c88433 | 1200 | { |
84f0f298 | 1201 | struct hmap *xports; |
46c88433 EJ |
1202 | struct xport *xport; |
1203 | ||
84f0f298 | 1204 | if (!ofport || !xcfg) { |
5e6af486 EJ |
1205 | return NULL; |
1206 | } | |
1207 | ||
84f0f298 RW |
1208 | xports = &xcfg->xports; |
1209 | ||
46c88433 | 1210 | HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0), |
84f0f298 | 1211 | xports) { |
46c88433 EJ |
1212 | if (xport->ofport == ofport) { |
1213 | return xport; | |
1214 | } | |
1215 | } | |
1216 | return NULL; | |
1217 | } | |
1218 | ||
40085e56 EJ |
1219 | static struct stp_port * |
1220 | xport_get_stp_port(const struct xport *xport) | |
1221 | { | |
92cf817b | 1222 | return xport->xbridge->stp && xport->stp_port_no != -1 |
40085e56 EJ |
1223 | ? stp_get_port(xport->xbridge->stp, xport->stp_port_no) |
1224 | : NULL; | |
1225 | } | |
9d189a50 | 1226 | |
0d1cee12 | 1227 | static bool |
9d189a50 EJ |
1228 | xport_stp_learn_state(const struct xport *xport) |
1229 | { | |
40085e56 | 1230 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1231 | return sp |
1232 | ? stp_learn_in_state(stp_port_get_state(sp)) | |
1233 | : true; | |
9d189a50 EJ |
1234 | } |
1235 | ||
1236 | static bool | |
1237 | xport_stp_forward_state(const struct xport *xport) | |
1238 | { | |
40085e56 | 1239 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1240 | return sp |
1241 | ? stp_forward_in_state(stp_port_get_state(sp)) | |
1242 | : true; | |
9d189a50 EJ |
1243 | } |
1244 | ||
0d1cee12 | 1245 | static bool |
bacdb85a | 1246 | xport_stp_should_forward_bpdu(const struct xport *xport) |
0d1cee12 K |
1247 | { |
1248 | struct stp_port *sp = xport_get_stp_port(xport); | |
bacdb85a | 1249 | return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED); |
0d1cee12 K |
1250 | } |
1251 | ||
9d189a50 EJ |
1252 | /* Returns true if STP should process 'flow'. Sets fields in 'wc' that |
1253 | * were used to make the determination.*/ | |
1254 | static bool | |
1255 | stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc) | |
1256 | { | |
bbbca389 | 1257 | /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */ |
9d189a50 | 1258 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
bbbca389 | 1259 | return is_stp(flow); |
9d189a50 EJ |
1260 | } |
1261 | ||
1262 | static void | |
cf62fa4c | 1263 | stp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9d189a50 | 1264 | { |
40085e56 | 1265 | struct stp_port *sp = xport_get_stp_port(xport); |
cf62fa4c PS |
1266 | struct dp_packet payload = *packet; |
1267 | struct eth_header *eth = dp_packet_data(&payload); | |
9d189a50 EJ |
1268 | |
1269 | /* Sink packets on ports that have STP disabled when the bridge has | |
1270 | * STP enabled. */ | |
1271 | if (!sp || stp_port_get_state(sp) == STP_DISABLED) { | |
1272 | return; | |
1273 | } | |
1274 | ||
1275 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1276 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1277 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9d189a50 EJ |
1278 | } |
1279 | ||
cf62fa4c PS |
1280 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1281 | stp_received_bpdu(sp, dp_packet_data(&payload), dp_packet_size(&payload)); | |
9d189a50 EJ |
1282 | } |
1283 | } | |
1284 | ||
f025bcb7 JR |
1285 | static enum rstp_state |
1286 | xport_get_rstp_port_state(const struct xport *xport) | |
9efd308e | 1287 | { |
f025bcb7 JR |
1288 | return xport->rstp_port |
1289 | ? rstp_port_get_state(xport->rstp_port) | |
1290 | : RSTP_DISABLED; | |
9efd308e DV |
1291 | } |
1292 | ||
1293 | static bool | |
1294 | xport_rstp_learn_state(const struct xport *xport) | |
1295 | { | |
4b5f1996 DV |
1296 | return xport->xbridge->rstp && xport->rstp_port |
1297 | ? rstp_learn_in_state(xport_get_rstp_port_state(xport)) | |
1298 | : true; | |
9efd308e DV |
1299 | } |
1300 | ||
1301 | static bool | |
1302 | xport_rstp_forward_state(const struct xport *xport) | |
1303 | { | |
4b5f1996 DV |
1304 | return xport->xbridge->rstp && xport->rstp_port |
1305 | ? rstp_forward_in_state(xport_get_rstp_port_state(xport)) | |
1306 | : true; | |
9efd308e DV |
1307 | } |
1308 | ||
1309 | static bool | |
1310 | xport_rstp_should_manage_bpdu(const struct xport *xport) | |
1311 | { | |
f025bcb7 | 1312 | return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport)); |
9efd308e DV |
1313 | } |
1314 | ||
1315 | static void | |
cf62fa4c | 1316 | rstp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9efd308e | 1317 | { |
cf62fa4c PS |
1318 | struct dp_packet payload = *packet; |
1319 | struct eth_header *eth = dp_packet_data(&payload); | |
9efd308e | 1320 | |
f025bcb7 JR |
1321 | /* Sink packets on ports that have no RSTP. */ |
1322 | if (!xport->rstp_port) { | |
9efd308e DV |
1323 | return; |
1324 | } | |
1325 | ||
1326 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1327 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1328 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9efd308e DV |
1329 | } |
1330 | ||
cf62fa4c PS |
1331 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1332 | rstp_port_received_bpdu(xport->rstp_port, dp_packet_data(&payload), | |
1333 | dp_packet_size(&payload)); | |
9efd308e DV |
1334 | } |
1335 | } | |
1336 | ||
46c88433 EJ |
1337 | static struct xport * |
1338 | get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1339 | { | |
1340 | struct xport *xport; | |
1341 | ||
1342 | HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port), | |
1343 | &xbridge->xports) { | |
1344 | if (xport->ofp_port == ofp_port) { | |
1345 | return xport; | |
1346 | } | |
1347 | } | |
1348 | return NULL; | |
1349 | } | |
1350 | ||
1351 | static odp_port_t | |
1352 | ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1353 | { | |
1354 | const struct xport *xport = get_ofp_port(xbridge, ofp_port); | |
1355 | return xport ? xport->odp_port : ODPP_NONE; | |
1356 | } | |
1357 | ||
dd8cd4b4 SH |
1358 | static bool |
1359 | odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port) | |
1360 | { | |
086fa873 BP |
1361 | struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1362 | return xport && xport->may_enable; | |
dd8cd4b4 SH |
1363 | } |
1364 | ||
1e684d7d | 1365 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1366 | group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *, |
1367 | int depth); | |
1368 | ||
1369 | static bool | |
1370 | group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) | |
1371 | { | |
1372 | struct group_dpif *group; | |
dd8cd4b4 | 1373 | |
dc25893e AZ |
1374 | if (group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group)) { |
1375 | struct ofputil_bucket *bucket; | |
dd8cd4b4 | 1376 | |
dc25893e AZ |
1377 | bucket = group_first_live_bucket(ctx, group, depth); |
1378 | group_dpif_unref(group); | |
1379 | return bucket == NULL; | |
1380 | } | |
dd8cd4b4 | 1381 | |
dc25893e | 1382 | return false; |
dd8cd4b4 SH |
1383 | } |
1384 | ||
1385 | #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ | |
1386 | ||
1387 | static bool | |
1388 | bucket_is_alive(const struct xlate_ctx *ctx, | |
1e684d7d | 1389 | struct ofputil_bucket *bucket, int depth) |
dd8cd4b4 SH |
1390 | { |
1391 | if (depth >= MAX_LIVENESS_RECURSION) { | |
1392 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
1393 | ||
1394 | VLOG_WARN_RL(&rl, "bucket chaining exceeded %d links", | |
1395 | MAX_LIVENESS_RECURSION); | |
1396 | return false; | |
1397 | } | |
1398 | ||
fdb1999b AZ |
1399 | return (!ofputil_bucket_has_liveness(bucket) |
1400 | || (bucket->watch_port != OFPP_ANY | |
1401 | && odp_port_is_alive(ctx, bucket->watch_port)) | |
1402 | || (bucket->watch_group != OFPG_ANY | |
1403 | && group_is_alive(ctx, bucket->watch_group, depth + 1))); | |
dd8cd4b4 SH |
1404 | } |
1405 | ||
1e684d7d | 1406 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1407 | group_first_live_bucket(const struct xlate_ctx *ctx, |
1408 | const struct group_dpif *group, int depth) | |
1409 | { | |
1410 | struct ofputil_bucket *bucket; | |
ca6ba700 | 1411 | const struct ovs_list *buckets; |
dd8cd4b4 SH |
1412 | |
1413 | group_dpif_get_buckets(group, &buckets); | |
1414 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1415 | if (bucket_is_alive(ctx, bucket, depth)) { | |
1416 | return bucket; | |
1417 | } | |
1418 | } | |
1419 | ||
1420 | return NULL; | |
1421 | } | |
1422 | ||
1e684d7d | 1423 | static struct ofputil_bucket * |
fe7e5749 SH |
1424 | group_best_live_bucket(const struct xlate_ctx *ctx, |
1425 | const struct group_dpif *group, | |
1426 | uint32_t basis) | |
1427 | { | |
1e684d7d | 1428 | struct ofputil_bucket *best_bucket = NULL; |
fe7e5749 SH |
1429 | uint32_t best_score = 0; |
1430 | int i = 0; | |
1431 | ||
1e684d7d | 1432 | struct ofputil_bucket *bucket; |
ca6ba700 | 1433 | const struct ovs_list *buckets; |
fe7e5749 SH |
1434 | |
1435 | group_dpif_get_buckets(group, &buckets); | |
1436 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1437 | if (bucket_is_alive(ctx, bucket, 0)) { | |
7cb279c2 | 1438 | uint32_t score = (hash_int(i, basis) & 0xffff) * bucket->weight; |
fe7e5749 SH |
1439 | if (score >= best_score) { |
1440 | best_bucket = bucket; | |
1441 | best_score = score; | |
1442 | } | |
1443 | } | |
1444 | i++; | |
1445 | } | |
1446 | ||
1447 | return best_bucket; | |
1448 | } | |
1449 | ||
9583bc14 | 1450 | static bool |
46c88433 | 1451 | xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan) |
9583bc14 EJ |
1452 | { |
1453 | return (bundle->vlan_mode != PORT_VLAN_ACCESS | |
1454 | && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan))); | |
1455 | } | |
1456 | ||
1457 | static bool | |
46c88433 EJ |
1458 | xbundle_includes_vlan(const struct xbundle *xbundle, uint16_t vlan) |
1459 | { | |
1460 | return vlan == xbundle->vlan || xbundle_trunks_vlan(xbundle, vlan); | |
1461 | } | |
1462 | ||
1463 | static mirror_mask_t | |
1464 | xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle) | |
1465 | { | |
1466 | return xbundle != &ofpp_none_bundle | |
1467 | ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle) | |
1468 | : 0; | |
1469 | } | |
1470 | ||
1471 | static mirror_mask_t | |
1472 | xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1473 | { |
46c88433 EJ |
1474 | return xbundle != &ofpp_none_bundle |
1475 | ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle) | |
1476 | : 0; | |
9583bc14 EJ |
1477 | } |
1478 | ||
46c88433 EJ |
1479 | static mirror_mask_t |
1480 | xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1481 | { |
46c88433 EJ |
1482 | return xbundle != &ofpp_none_bundle |
1483 | ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle) | |
1484 | : 0; | |
1485 | } | |
1486 | ||
1487 | static struct xbundle * | |
1488 | lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port, | |
1489 | bool warn, struct xport **in_xportp) | |
1490 | { | |
1491 | struct xport *xport; | |
9583bc14 EJ |
1492 | |
1493 | /* Find the port and bundle for the received packet. */ | |
46c88433 EJ |
1494 | xport = get_ofp_port(xbridge, in_port); |
1495 | if (in_xportp) { | |
1496 | *in_xportp = xport; | |
9583bc14 | 1497 | } |
46c88433 EJ |
1498 | if (xport && xport->xbundle) { |
1499 | return xport->xbundle; | |
9583bc14 EJ |
1500 | } |
1501 | ||
6362203b YT |
1502 | /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+), |
1503 | * which a controller may use as the ingress port for traffic that | |
1504 | * it is sourcing. */ | |
1505 | if (in_port == OFPP_CONTROLLER || in_port == OFPP_NONE) { | |
9583bc14 EJ |
1506 | return &ofpp_none_bundle; |
1507 | } | |
1508 | ||
1509 | /* Odd. A few possible reasons here: | |
1510 | * | |
1511 | * - We deleted a port but there are still a few packets queued up | |
1512 | * from it. | |
1513 | * | |
1514 | * - Someone externally added a port (e.g. "ovs-dpctl add-if") that | |
1515 | * we don't know about. | |
1516 | * | |
1517 | * - The ofproto client didn't configure the port as part of a bundle. | |
1518 | * This is particularly likely to happen if a packet was received on the | |
1519 | * port after it was created, but before the client had a chance to | |
1520 | * configure its bundle. | |
1521 | */ | |
1522 | if (warn) { | |
1523 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1524 | ||
1525 | VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown " | |
46c88433 | 1526 | "port %"PRIu16, xbridge->name, in_port); |
9583bc14 EJ |
1527 | } |
1528 | return NULL; | |
1529 | } | |
1530 | ||
1531 | static void | |
1532 | add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) | |
1533 | { | |
46c88433 | 1534 | const struct xbridge *xbridge = ctx->xbridge; |
9583bc14 | 1535 | mirror_mask_t mirrors; |
46c88433 | 1536 | struct xbundle *in_xbundle; |
9583bc14 EJ |
1537 | uint16_t vlan; |
1538 | uint16_t vid; | |
cdf5d3a5 | 1539 | |
3d6151f3 BP |
1540 | mirrors = ctx->mirrors; |
1541 | ctx->mirrors = 0; | |
9583bc14 | 1542 | |
46c88433 EJ |
1543 | in_xbundle = lookup_input_bundle(xbridge, orig_flow->in_port.ofp_port, |
1544 | ctx->xin->packet != NULL, NULL); | |
1545 | if (!in_xbundle) { | |
9583bc14 EJ |
1546 | return; |
1547 | } | |
46c88433 | 1548 | mirrors |= xbundle_mirror_src(xbridge, in_xbundle); |
9583bc14 EJ |
1549 | |
1550 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 1551 | if (xbundle_mirror_out(xbridge, in_xbundle)) { |
9583bc14 EJ |
1552 | if (ctx->xin->packet != NULL) { |
1553 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1554 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
1555 | "%s, which is reserved exclusively for mirroring", | |
46c88433 | 1556 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 | 1557 | } |
1520ef4f | 1558 | ofpbuf_clear(ctx->odp_actions); |
9583bc14 EJ |
1559 | return; |
1560 | } | |
1561 | ||
1562 | /* Check VLAN. */ | |
1563 | vid = vlan_tci_to_vid(orig_flow->vlan_tci); | |
46c88433 | 1564 | if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { |
9583bc14 EJ |
1565 | return; |
1566 | } | |
46c88433 | 1567 | vlan = input_vid_to_vlan(in_xbundle, vid); |
9583bc14 | 1568 | |
9583bc14 EJ |
1569 | if (!mirrors) { |
1570 | return; | |
1571 | } | |
1572 | ||
1573 | /* Restore the original packet before adding the mirror actions. */ | |
1574 | ctx->xin->flow = *orig_flow; | |
1575 | ||
1576 | while (mirrors) { | |
ec7ceaed EJ |
1577 | mirror_mask_t dup_mirrors; |
1578 | struct ofbundle *out; | |
a4547540 | 1579 | const unsigned long *vlans; |
ec7ceaed EJ |
1580 | bool vlan_mirrored; |
1581 | bool has_mirror; | |
1582 | int out_vlan; | |
1583 | ||
4b6ab2b0 | 1584 | has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors), |
ec7ceaed EJ |
1585 | &vlans, &dup_mirrors, &out, &out_vlan); |
1586 | ovs_assert(has_mirror); | |
1587 | ||
1588 | if (vlans) { | |
49a73e0c | 1589 | ctx->wc->masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK); |
9583bc14 | 1590 | } |
ec7ceaed | 1591 | vlan_mirrored = !vlans || bitmap_is_set(vlans, vlan); |
9583bc14 | 1592 | |
ec7ceaed | 1593 | if (!vlan_mirrored) { |
9583bc14 EJ |
1594 | mirrors = zero_rightmost_1bit(mirrors); |
1595 | continue; | |
1596 | } | |
1597 | ||
ec7ceaed | 1598 | mirrors &= ~dup_mirrors; |
3d6151f3 | 1599 | ctx->mirrors |= dup_mirrors; |
ec7ceaed | 1600 | if (out) { |
84f0f298 RW |
1601 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
1602 | struct xbundle *out_xbundle = xbundle_lookup(xcfg, out); | |
46c88433 EJ |
1603 | if (out_xbundle) { |
1604 | output_normal(ctx, out_xbundle, vlan); | |
1605 | } | |
ec7ceaed | 1606 | } else if (vlan != out_vlan |
9583bc14 | 1607 | && !eth_addr_is_reserved(orig_flow->dl_dst)) { |
46c88433 | 1608 | struct xbundle *xbundle; |
9583bc14 | 1609 | |
46c88433 EJ |
1610 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
1611 | if (xbundle_includes_vlan(xbundle, out_vlan) | |
1612 | && !xbundle_mirror_out(xbridge, xbundle)) { | |
1613 | output_normal(ctx, xbundle, out_vlan); | |
9583bc14 EJ |
1614 | } |
1615 | } | |
1616 | } | |
1617 | } | |
1618 | } | |
1619 | ||
1620 | /* Given 'vid', the VID obtained from the 802.1Q header that was received as | |
46c88433 | 1621 | * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle', |
9583bc14 EJ |
1622 | * the bundle on which the packet was received, returns the VLAN to which the |
1623 | * packet belongs. | |
1624 | * | |
1625 | * Both 'vid' and the return value are in the range 0...4095. */ | |
1626 | static uint16_t | |
46c88433 | 1627 | input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid) |
9583bc14 | 1628 | { |
46c88433 | 1629 | switch (in_xbundle->vlan_mode) { |
9583bc14 | 1630 | case PORT_VLAN_ACCESS: |
46c88433 | 1631 | return in_xbundle->vlan; |
9583bc14 EJ |
1632 | break; |
1633 | ||
1634 | case PORT_VLAN_TRUNK: | |
1635 | return vid; | |
1636 | ||
1637 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1638 | case PORT_VLAN_NATIVE_TAGGED: | |
46c88433 | 1639 | return vid ? vid : in_xbundle->vlan; |
9583bc14 EJ |
1640 | |
1641 | default: | |
428b2edd | 1642 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1643 | } |
1644 | } | |
1645 | ||
46c88433 | 1646 | /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'. |
9583bc14 EJ |
1647 | * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs |
1648 | * a warning. | |
1649 | * | |
1650 | * 'vid' should be the VID obtained from the 802.1Q header that was received as | |
1651 | * part of a packet (specify 0 if there was no 802.1Q header), in the range | |
1652 | * 0...4095. */ | |
1653 | static bool | |
46c88433 | 1654 | input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn) |
9583bc14 EJ |
1655 | { |
1656 | /* Allow any VID on the OFPP_NONE port. */ | |
46c88433 | 1657 | if (in_xbundle == &ofpp_none_bundle) { |
9583bc14 EJ |
1658 | return true; |
1659 | } | |
1660 | ||
46c88433 | 1661 | switch (in_xbundle->vlan_mode) { |
9583bc14 EJ |
1662 | case PORT_VLAN_ACCESS: |
1663 | if (vid) { | |
1664 | if (warn) { | |
1665 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1666 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" tagged " |
9583bc14 | 1667 | "packet received on port %s configured as VLAN " |
46c88433 EJ |
1668 | "%"PRIu16" access port", vid, in_xbundle->name, |
1669 | in_xbundle->vlan); | |
9583bc14 EJ |
1670 | } |
1671 | return false; | |
1672 | } | |
1673 | return true; | |
1674 | ||
1675 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1676 | case PORT_VLAN_NATIVE_TAGGED: | |
1677 | if (!vid) { | |
1678 | /* Port must always carry its native VLAN. */ | |
1679 | return true; | |
1680 | } | |
1681 | /* Fall through. */ | |
1682 | case PORT_VLAN_TRUNK: | |
46c88433 | 1683 | if (!xbundle_includes_vlan(in_xbundle, vid)) { |
9583bc14 EJ |
1684 | if (warn) { |
1685 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1686 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" packet " |
9583bc14 | 1687 | "received on port %s not configured for trunking " |
46c88433 | 1688 | "VLAN %"PRIu16, vid, in_xbundle->name, vid); |
9583bc14 EJ |
1689 | } |
1690 | return false; | |
1691 | } | |
1692 | return true; | |
1693 | ||
1694 | default: | |
428b2edd | 1695 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1696 | } |
1697 | ||
1698 | } | |
1699 | ||
1700 | /* Given 'vlan', the VLAN that a packet belongs to, and | |
46c88433 | 1701 | * 'out_xbundle', a bundle on which the packet is to be output, returns the VID |
9583bc14 EJ |
1702 | * that should be included in the 802.1Q header. (If the return value is 0, |
1703 | * then the 802.1Q header should only be included in the packet if there is a | |
1704 | * nonzero PCP.) | |
1705 | * | |
1706 | * Both 'vlan' and the return value are in the range 0...4095. */ | |
1707 | static uint16_t | |
46c88433 | 1708 | output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan) |
9583bc14 | 1709 | { |
46c88433 | 1710 | switch (out_xbundle->vlan_mode) { |
9583bc14 EJ |
1711 | case PORT_VLAN_ACCESS: |
1712 | return 0; | |
1713 | ||
1714 | case PORT_VLAN_TRUNK: | |
1715 | case PORT_VLAN_NATIVE_TAGGED: | |
1716 | return vlan; | |
1717 | ||
1718 | case PORT_VLAN_NATIVE_UNTAGGED: | |
46c88433 | 1719 | return vlan == out_xbundle->vlan ? 0 : vlan; |
9583bc14 EJ |
1720 | |
1721 | default: | |
428b2edd | 1722 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1723 | } |
1724 | } | |
1725 | ||
1726 | static void | |
46c88433 | 1727 | output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, |
9583bc14 EJ |
1728 | uint16_t vlan) |
1729 | { | |
33bf9176 | 1730 | ovs_be16 *flow_tci = &ctx->xin->flow.vlan_tci; |
9583bc14 EJ |
1731 | uint16_t vid; |
1732 | ovs_be16 tci, old_tci; | |
46c88433 | 1733 | struct xport *xport; |
e93ef1c7 JR |
1734 | struct xlate_bond_recirc xr; |
1735 | bool use_recirc = false; | |
9583bc14 | 1736 | |
46c88433 EJ |
1737 | vid = output_vlan_to_vid(out_xbundle, vlan); |
1738 | if (list_is_empty(&out_xbundle->xports)) { | |
1739 | /* Partially configured bundle with no slaves. Drop the packet. */ | |
1740 | return; | |
1741 | } else if (!out_xbundle->bond) { | |
1742 | xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport, | |
1743 | bundle_node); | |
9583bc14 | 1744 | } else { |
84f0f298 | 1745 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
49a73e0c | 1746 | struct flow_wildcards *wc = ctx->wc; |
84f0f298 | 1747 | struct ofport_dpif *ofport; |
adcf00ba | 1748 | |
2494ccd7 | 1749 | if (ctx->xbridge->support.odp.recirc) { |
e93ef1c7 JR |
1750 | use_recirc = bond_may_recirc( |
1751 | out_xbundle->bond, &xr.recirc_id, &xr.hash_basis); | |
adcf00ba | 1752 | |
e93ef1c7 | 1753 | if (use_recirc) { |
adcf00ba | 1754 | /* Only TCP mode uses recirculation. */ |
e93ef1c7 | 1755 | xr.hash_alg = OVS_HASH_ALG_L4; |
adcf00ba | 1756 | bond_update_post_recirc_rules(out_xbundle->bond, false); |
54ecb5a2 AZ |
1757 | |
1758 | /* Recirculation does not require unmasking hash fields. */ | |
1759 | wc = NULL; | |
adcf00ba AZ |
1760 | } |
1761 | } | |
46c88433 | 1762 | |
54ecb5a2 AZ |
1763 | ofport = bond_choose_output_slave(out_xbundle->bond, |
1764 | &ctx->xin->flow, wc, vid); | |
84f0f298 | 1765 | xport = xport_lookup(xcfg, ofport); |
46c88433 EJ |
1766 | |
1767 | if (!xport) { | |
9583bc14 EJ |
1768 | /* No slaves enabled, so drop packet. */ |
1769 | return; | |
1770 | } | |
d6fc5f57 | 1771 | |
e93ef1c7 | 1772 | /* If use_recirc is set, the main thread will handle stats |
b256dc52 | 1773 | * accounting for this bond. */ |
e93ef1c7 | 1774 | if (!use_recirc) { |
b256dc52 JS |
1775 | if (ctx->xin->resubmit_stats) { |
1776 | bond_account(out_xbundle->bond, &ctx->xin->flow, vid, | |
1777 | ctx->xin->resubmit_stats->n_bytes); | |
1778 | } | |
1779 | if (ctx->xin->xcache) { | |
1780 | struct xc_entry *entry; | |
1781 | struct flow *flow; | |
1782 | ||
1783 | flow = &ctx->xin->flow; | |
1784 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND); | |
1785 | entry->u.bond.bond = bond_ref(out_xbundle->bond); | |
1786 | entry->u.bond.flow = xmemdup(flow, sizeof *flow); | |
1787 | entry->u.bond.vid = vid; | |
1788 | } | |
d6fc5f57 | 1789 | } |
9583bc14 EJ |
1790 | } |
1791 | ||
33bf9176 | 1792 | old_tci = *flow_tci; |
9583bc14 | 1793 | tci = htons(vid); |
46c88433 | 1794 | if (tci || out_xbundle->use_priority_tags) { |
33bf9176 | 1795 | tci |= *flow_tci & htons(VLAN_PCP_MASK); |
9583bc14 EJ |
1796 | if (tci) { |
1797 | tci |= htons(VLAN_CFI); | |
1798 | } | |
1799 | } | |
33bf9176 | 1800 | *flow_tci = tci; |
9583bc14 | 1801 | |
e93ef1c7 | 1802 | compose_output_action(ctx, xport->ofp_port, use_recirc ? &xr : NULL); |
33bf9176 | 1803 | *flow_tci = old_tci; |
9583bc14 EJ |
1804 | } |
1805 | ||
1806 | /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after | |
1807 | * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to | |
1808 | * indicate this; newer upstream kernels use gratuitous ARP requests. */ | |
1809 | static bool | |
1810 | is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) | |
1811 | { | |
1812 | if (flow->dl_type != htons(ETH_TYPE_ARP)) { | |
1813 | return false; | |
1814 | } | |
1815 | ||
1816 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1817 | if (!eth_addr_is_broadcast(flow->dl_dst)) { | |
1818 | return false; | |
1819 | } | |
1820 | ||
1821 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
1822 | if (flow->nw_proto == ARP_OP_REPLY) { | |
1823 | return true; | |
1824 | } else if (flow->nw_proto == ARP_OP_REQUEST) { | |
1825 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
1826 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
1827 | ||
1828 | return flow->nw_src == flow->nw_dst; | |
1829 | } else { | |
1830 | return false; | |
1831 | } | |
1832 | } | |
1833 | ||
ff69c24a FL |
1834 | /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or |
1835 | * dropped. Returns true if they may be forwarded, false if they should be | |
1836 | * dropped. | |
1837 | * | |
1838 | * 'in_port' must be the xport that corresponds to flow->in_port. | |
1839 | * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). | |
1840 | * | |
1841 | * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as | |
1842 | * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as | |
1843 | * checked by input_vid_is_valid(). | |
1844 | * | |
1845 | * May also add tags to '*tags', although the current implementation only does | |
1846 | * so in one special case. | |
1847 | */ | |
1848 | static bool | |
1849 | is_admissible(struct xlate_ctx *ctx, struct xport *in_port, | |
1850 | uint16_t vlan) | |
1851 | { | |
1852 | struct xbundle *in_xbundle = in_port->xbundle; | |
1853 | const struct xbridge *xbridge = ctx->xbridge; | |
1854 | struct flow *flow = &ctx->xin->flow; | |
1855 | ||
1856 | /* Drop frames for reserved multicast addresses | |
1857 | * only if forward_bpdu option is absent. */ | |
1858 | if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { | |
1859 | xlate_report(ctx, "packet has reserved destination MAC, dropping"); | |
1860 | return false; | |
1861 | } | |
1862 | ||
1863 | if (in_xbundle->bond) { | |
1864 | struct mac_entry *mac; | |
1865 | ||
1866 | switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, | |
1867 | flow->dl_dst)) { | |
1868 | case BV_ACCEPT: | |
1869 | break; | |
1870 | ||
1871 | case BV_DROP: | |
1872 | xlate_report(ctx, "bonding refused admissibility, dropping"); | |
1873 | return false; | |
1874 | ||
1875 | case BV_DROP_IF_MOVED: | |
1876 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
1877 | mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan); | |
9d078ec2 BP |
1878 | if (mac |
1879 | && mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle | |
49a73e0c | 1880 | && (!is_gratuitous_arp(flow, ctx->wc) |
9d078ec2 | 1881 | || mac_entry_is_grat_arp_locked(mac))) { |
ff69c24a FL |
1882 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
1883 | xlate_report(ctx, "SLB bond thinks this packet looped back, " | |
1884 | "dropping"); | |
1885 | return false; | |
1886 | } | |
1887 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
1888 | break; | |
1889 | } | |
1890 | } | |
1891 | ||
1892 | return true; | |
1893 | } | |
1894 | ||
ee047520 BP |
1895 | /* Checks whether a MAC learning update is necessary for MAC learning table |
1896 | * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in | |
1897 | * 'vlan'. | |
1898 | * | |
1899 | * Most packets processed through the MAC learning table do not actually | |
1900 | * change it in any way. This function requires only a read lock on the MAC | |
1901 | * learning table, so it is much cheaper in this common case. | |
1902 | * | |
1903 | * Keep the code here synchronized with that in update_learning_table__() | |
1904 | * below. */ | |
1905 | static bool | |
1906 | is_mac_learning_update_needed(const struct mac_learning *ml, | |
1907 | const struct flow *flow, | |
1908 | struct flow_wildcards *wc, | |
1909 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 1910 | OVS_REQ_RDLOCK(ml->rwlock) |
9583bc14 EJ |
1911 | { |
1912 | struct mac_entry *mac; | |
1913 | ||
ee047520 BP |
1914 | if (!mac_learning_may_learn(ml, flow->dl_src, vlan)) { |
1915 | return false; | |
1916 | } | |
1917 | ||
1918 | mac = mac_learning_lookup(ml, flow->dl_src, vlan); | |
1919 | if (!mac || mac_entry_age(ml, mac)) { | |
1920 | return true; | |
9583bc14 EJ |
1921 | } |
1922 | ||
ee047520 BP |
1923 | if (is_gratuitous_arp(flow, wc)) { |
1924 | /* We don't want to learn from gratuitous ARP packets that are | |
1925 | * reflected back over bond slaves so we lock the learning table. */ | |
1926 | if (!in_xbundle->bond) { | |
1927 | return true; | |
1928 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
1929 | return false; | |
1930 | } | |
1931 | } | |
1932 | ||
9d078ec2 | 1933 | return mac_entry_get_port(ml, mac) != in_xbundle->ofbundle; |
ee047520 BP |
1934 | } |
1935 | ||
1936 | ||
1937 | /* Updates MAC learning table 'ml' given that a packet matching 'flow' was | |
1938 | * received on 'in_xbundle' in 'vlan'. | |
1939 | * | |
1940 | * This code repeats all the checks in is_mac_learning_update_needed() because | |
1941 | * the lock was released between there and here and thus the MAC learning state | |
1942 | * could have changed. | |
1943 | * | |
1944 | * Keep the code here synchronized with that in is_mac_learning_update_needed() | |
1945 | * above. */ | |
1946 | static void | |
1947 | update_learning_table__(const struct xbridge *xbridge, | |
1948 | const struct flow *flow, struct flow_wildcards *wc, | |
1949 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 1950 | OVS_REQ_WRLOCK(xbridge->ml->rwlock) |
ee047520 BP |
1951 | { |
1952 | struct mac_entry *mac; | |
1953 | ||
46c88433 | 1954 | if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) { |
ee047520 | 1955 | return; |
9583bc14 EJ |
1956 | } |
1957 | ||
46c88433 | 1958 | mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan); |
9583bc14 EJ |
1959 | if (is_gratuitous_arp(flow, wc)) { |
1960 | /* We don't want to learn from gratuitous ARP packets that are | |
1961 | * reflected back over bond slaves so we lock the learning table. */ | |
46c88433 | 1962 | if (!in_xbundle->bond) { |
9583bc14 EJ |
1963 | mac_entry_set_grat_arp_lock(mac); |
1964 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
ee047520 | 1965 | return; |
9583bc14 EJ |
1966 | } |
1967 | } | |
1968 | ||
9d078ec2 | 1969 | if (mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle) { |
9583bc14 EJ |
1970 | /* The log messages here could actually be useful in debugging, |
1971 | * so keep the rate limit relatively high. */ | |
1972 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
ee047520 | 1973 | |
9583bc14 EJ |
1974 | VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is " |
1975 | "on port %s in VLAN %d", | |
46c88433 EJ |
1976 | xbridge->name, ETH_ADDR_ARGS(flow->dl_src), |
1977 | in_xbundle->name, vlan); | |
9583bc14 | 1978 | |
9d078ec2 | 1979 | mac_entry_set_port(xbridge->ml, mac, in_xbundle->ofbundle); |
9583bc14 | 1980 | } |
ee047520 BP |
1981 | } |
1982 | ||
1983 | static void | |
1984 | update_learning_table(const struct xbridge *xbridge, | |
1985 | const struct flow *flow, struct flow_wildcards *wc, | |
1986 | int vlan, struct xbundle *in_xbundle) | |
1987 | { | |
1988 | bool need_update; | |
1989 | ||
1990 | /* Don't learn the OFPP_NONE port. */ | |
1991 | if (in_xbundle == &ofpp_none_bundle) { | |
1992 | return; | |
1993 | } | |
1994 | ||
1995 | /* First try the common case: no change to MAC learning table. */ | |
1996 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
1997 | need_update = is_mac_learning_update_needed(xbridge->ml, flow, wc, vlan, | |
1998 | in_xbundle); | |
509c0149 | 1999 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
ee047520 BP |
2000 | |
2001 | if (need_update) { | |
2002 | /* Slow path: MAC learning table might need an update. */ | |
2003 | ovs_rwlock_wrlock(&xbridge->ml->rwlock); | |
2004 | update_learning_table__(xbridge, flow, wc, vlan, in_xbundle); | |
2005 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
2006 | } | |
9583bc14 EJ |
2007 | } |
2008 | ||
86e2dcdd FL |
2009 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2010 | * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */ | |
2011 | static void | |
06994f87 TLSC |
2012 | update_mcast_snooping_table4__(const struct xbridge *xbridge, |
2013 | const struct flow *flow, | |
2014 | struct mcast_snooping *ms, int vlan, | |
2015 | struct xbundle *in_xbundle, | |
2016 | const struct dp_packet *packet) | |
86e2dcdd FL |
2017 | OVS_REQ_WRLOCK(ms->rwlock) |
2018 | { | |
2019 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
e3102e42 | 2020 | int count; |
06994f87 | 2021 | ovs_be32 ip4 = flow->igmp_group_ip4; |
86e2dcdd FL |
2022 | |
2023 | switch (ntohs(flow->tp_src)) { | |
2024 | case IGMP_HOST_MEMBERSHIP_REPORT: | |
2025 | case IGMPV2_HOST_MEMBERSHIP_REPORT: | |
964a4d5f | 2026 | if (mcast_snooping_add_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
86e2dcdd FL |
2027 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping learned that " |
2028 | IP_FMT" is on port %s in VLAN %d", | |
2029 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
2030 | } | |
2031 | break; | |
2032 | case IGMP_HOST_LEAVE_MESSAGE: | |
964a4d5f | 2033 | if (mcast_snooping_leave_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
86e2dcdd FL |
2034 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping leaving " |
2035 | IP_FMT" is on port %s in VLAN %d", | |
2036 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
2037 | } | |
2038 | break; | |
2039 | case IGMP_HOST_MEMBERSHIP_QUERY: | |
2040 | if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan, | |
2041 | in_xbundle->ofbundle)) { | |
2042 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query from " | |
2043 | IP_FMT" is on port %s in VLAN %d", | |
2044 | xbridge->name, IP_ARGS(flow->nw_src), | |
2045 | in_xbundle->name, vlan); | |
2046 | } | |
2047 | break; | |
e3102e42 TLSC |
2048 | case IGMPV3_HOST_MEMBERSHIP_REPORT: |
2049 | if ((count = mcast_snooping_add_report(ms, packet, vlan, | |
2050 | in_xbundle->ofbundle))) { | |
2051 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping processed %d " | |
2052 | "addresses on port %s in VLAN %d", | |
2053 | xbridge->name, count, in_xbundle->name, vlan); | |
2054 | } | |
2055 | break; | |
86e2dcdd FL |
2056 | } |
2057 | } | |
2058 | ||
06994f87 TLSC |
2059 | static void |
2060 | update_mcast_snooping_table6__(const struct xbridge *xbridge, | |
2061 | const struct flow *flow, | |
2062 | struct mcast_snooping *ms, int vlan, | |
2063 | struct xbundle *in_xbundle, | |
2064 | const struct dp_packet *packet) | |
2065 | OVS_REQ_WRLOCK(ms->rwlock) | |
2066 | { | |
2067 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
2068 | int count; | |
2069 | ||
2070 | switch (ntohs(flow->tp_src)) { | |
2071 | case MLD_QUERY: | |
2072 | if (!ipv6_addr_equals(&flow->ipv6_src, &in6addr_any) | |
2073 | && mcast_snooping_add_mrouter(ms, vlan, in_xbundle->ofbundle)) { | |
2074 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query on port %s" | |
2075 | "in VLAN %d", | |
2076 | xbridge->name, in_xbundle->name, vlan); | |
2077 | } | |
2078 | break; | |
2079 | case MLD_REPORT: | |
2080 | case MLD_DONE: | |
2081 | case MLD2_REPORT: | |
2082 | count = mcast_snooping_add_mld(ms, packet, vlan, in_xbundle->ofbundle); | |
2083 | if (count) { | |
2084 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping processed %d " | |
2085 | "addresses on port %s in VLAN %d", | |
2086 | xbridge->name, count, in_xbundle->name, vlan); | |
2087 | } | |
2088 | break; | |
2089 | } | |
2090 | } | |
2091 | ||
86e2dcdd FL |
2092 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2093 | * was received on 'in_xbundle' in 'vlan'. */ | |
2094 | static void | |
2095 | update_mcast_snooping_table(const struct xbridge *xbridge, | |
2096 | const struct flow *flow, int vlan, | |
e3102e42 TLSC |
2097 | struct xbundle *in_xbundle, |
2098 | const struct dp_packet *packet) | |
86e2dcdd FL |
2099 | { |
2100 | struct mcast_snooping *ms = xbridge->ms; | |
2101 | struct xlate_cfg *xcfg; | |
2102 | struct xbundle *mcast_xbundle; | |
f4ae6e23 | 2103 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2104 | |
2105 | /* Don't learn the OFPP_NONE port. */ | |
2106 | if (in_xbundle == &ofpp_none_bundle) { | |
2107 | return; | |
2108 | } | |
2109 | ||
2110 | /* Don't learn from flood ports */ | |
2111 | mcast_xbundle = NULL; | |
2112 | ovs_rwlock_wrlock(&ms->rwlock); | |
2113 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2114 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2115 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2116 | if (mcast_xbundle == in_xbundle) { | |
2117 | break; | |
2118 | } | |
2119 | } | |
2120 | ||
2121 | if (!mcast_xbundle || mcast_xbundle != in_xbundle) { | |
06994f87 TLSC |
2122 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2123 | update_mcast_snooping_table4__(xbridge, flow, ms, vlan, | |
2124 | in_xbundle, packet); | |
2125 | } else { | |
2126 | update_mcast_snooping_table6__(xbridge, flow, ms, vlan, | |
2127 | in_xbundle, packet); | |
2128 | } | |
86e2dcdd FL |
2129 | } |
2130 | ovs_rwlock_unlock(&ms->rwlock); | |
2131 | } | |
2132 | ||
2133 | /* send the packet to ports having the multicast group learned */ | |
2134 | static void | |
2135 | xlate_normal_mcast_send_group(struct xlate_ctx *ctx, | |
2136 | struct mcast_snooping *ms OVS_UNUSED, | |
2137 | struct mcast_group *grp, | |
2138 | struct xbundle *in_xbundle, uint16_t vlan) | |
2139 | OVS_REQ_RDLOCK(ms->rwlock) | |
2140 | { | |
2141 | struct xlate_cfg *xcfg; | |
2142 | struct mcast_group_bundle *b; | |
2143 | struct xbundle *mcast_xbundle; | |
2144 | ||
2145 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2146 | LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) { | |
2147 | mcast_xbundle = xbundle_lookup(xcfg, b->port); | |
2148 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2149 | xlate_report(ctx, "forwarding to mcast group port"); | |
2150 | output_normal(ctx, mcast_xbundle, vlan); | |
2151 | } else if (!mcast_xbundle) { | |
2152 | xlate_report(ctx, "mcast group port is unknown, dropping"); | |
2153 | } else { | |
2154 | xlate_report(ctx, "mcast group port is input port, dropping"); | |
2155 | } | |
2156 | } | |
2157 | } | |
2158 | ||
2159 | /* send the packet to ports connected to multicast routers */ | |
2160 | static void | |
2161 | xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx, | |
2162 | struct mcast_snooping *ms, | |
2163 | struct xbundle *in_xbundle, uint16_t vlan) | |
2164 | OVS_REQ_RDLOCK(ms->rwlock) | |
2165 | { | |
2166 | struct xlate_cfg *xcfg; | |
2167 | struct mcast_mrouter_bundle *mrouter; | |
2168 | struct xbundle *mcast_xbundle; | |
2169 | ||
2170 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2171 | LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) { | |
2172 | mcast_xbundle = xbundle_lookup(xcfg, mrouter->port); | |
2173 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2174 | xlate_report(ctx, "forwarding to mcast router port"); | |
2175 | output_normal(ctx, mcast_xbundle, vlan); | |
2176 | } else if (!mcast_xbundle) { | |
2177 | xlate_report(ctx, "mcast router port is unknown, dropping"); | |
2178 | } else { | |
2179 | xlate_report(ctx, "mcast router port is input port, dropping"); | |
2180 | } | |
2181 | } | |
2182 | } | |
2183 | ||
2184 | /* send the packet to ports flagged to be flooded */ | |
2185 | static void | |
2186 | xlate_normal_mcast_send_fports(struct xlate_ctx *ctx, | |
2187 | struct mcast_snooping *ms, | |
2188 | struct xbundle *in_xbundle, uint16_t vlan) | |
2189 | OVS_REQ_RDLOCK(ms->rwlock) | |
2190 | { | |
2191 | struct xlate_cfg *xcfg; | |
f4ae6e23 | 2192 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2193 | struct xbundle *mcast_xbundle; |
2194 | ||
2195 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2196 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2197 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2198 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2199 | xlate_report(ctx, "forwarding to mcast flood port"); | |
2200 | output_normal(ctx, mcast_xbundle, vlan); | |
2201 | } else if (!mcast_xbundle) { | |
2202 | xlate_report(ctx, "mcast flood port is unknown, dropping"); | |
2203 | } else { | |
2204 | xlate_report(ctx, "mcast flood port is input port, dropping"); | |
2205 | } | |
2206 | } | |
2207 | } | |
2208 | ||
8e04a33f FL |
2209 | /* forward the Reports to configured ports */ |
2210 | static void | |
2211 | xlate_normal_mcast_send_rports(struct xlate_ctx *ctx, | |
2212 | struct mcast_snooping *ms, | |
2213 | struct xbundle *in_xbundle, uint16_t vlan) | |
2214 | OVS_REQ_RDLOCK(ms->rwlock) | |
2215 | { | |
2216 | struct xlate_cfg *xcfg; | |
2217 | struct mcast_port_bundle *rport; | |
2218 | struct xbundle *mcast_xbundle; | |
2219 | ||
2220 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2221 | LIST_FOR_EACH(rport, node, &ms->rport_list) { | |
2222 | mcast_xbundle = xbundle_lookup(xcfg, rport->port); | |
2223 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2224 | xlate_report(ctx, "forwarding Report to mcast flagged port"); | |
2225 | output_normal(ctx, mcast_xbundle, vlan); | |
2226 | } else if (!mcast_xbundle) { | |
2227 | xlate_report(ctx, "mcast port is unknown, dropping the Report"); | |
2228 | } else { | |
2229 | xlate_report(ctx, "mcast port is input port, dropping the Report"); | |
2230 | } | |
2231 | } | |
2232 | } | |
2233 | ||
682800a4 FL |
2234 | static void |
2235 | xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, | |
2236 | uint16_t vlan) | |
2237 | { | |
2238 | struct xbundle *xbundle; | |
2239 | ||
2240 | LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { | |
2241 | if (xbundle != in_xbundle | |
2242 | && xbundle_includes_vlan(xbundle, vlan) | |
2243 | && xbundle->floodable | |
2244 | && !xbundle_mirror_out(ctx->xbridge, xbundle)) { | |
2245 | output_normal(ctx, xbundle, vlan); | |
2246 | } | |
2247 | } | |
2248 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
2249 | } | |
2250 | ||
9583bc14 EJ |
2251 | static void |
2252 | xlate_normal(struct xlate_ctx *ctx) | |
2253 | { | |
49a73e0c | 2254 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 2255 | struct flow *flow = &ctx->xin->flow; |
46c88433 EJ |
2256 | struct xbundle *in_xbundle; |
2257 | struct xport *in_port; | |
9583bc14 | 2258 | struct mac_entry *mac; |
d6d5bbc9 | 2259 | void *mac_port; |
9583bc14 EJ |
2260 | uint16_t vlan; |
2261 | uint16_t vid; | |
2262 | ||
33bf9176 BP |
2263 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
2264 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1dd35f8a | 2265 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
9583bc14 | 2266 | |
46c88433 EJ |
2267 | in_xbundle = lookup_input_bundle(ctx->xbridge, flow->in_port.ofp_port, |
2268 | ctx->xin->packet != NULL, &in_port); | |
2269 | if (!in_xbundle) { | |
9583bc14 EJ |
2270 | xlate_report(ctx, "no input bundle, dropping"); |
2271 | return; | |
2272 | } | |
2273 | ||
2274 | /* Drop malformed frames. */ | |
33bf9176 BP |
2275 | if (flow->dl_type == htons(ETH_TYPE_VLAN) && |
2276 | !(flow->vlan_tci & htons(VLAN_CFI))) { | |
9583bc14 EJ |
2277 | if (ctx->xin->packet != NULL) { |
2278 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2279 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet with partial " | |
2280 | "VLAN tag received on port %s", | |
46c88433 | 2281 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2282 | } |
2283 | xlate_report(ctx, "partial VLAN tag, dropping"); | |
2284 | return; | |
2285 | } | |
2286 | ||
2287 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 2288 | if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) { |
9583bc14 EJ |
2289 | if (ctx->xin->packet != NULL) { |
2290 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2291 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
2292 | "%s, which is reserved exclusively for mirroring", | |
46c88433 | 2293 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2294 | } |
2295 | xlate_report(ctx, "input port is mirror output port, dropping"); | |
2296 | return; | |
2297 | } | |
2298 | ||
2299 | /* Check VLAN. */ | |
33bf9176 | 2300 | vid = vlan_tci_to_vid(flow->vlan_tci); |
46c88433 | 2301 | if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { |
9583bc14 EJ |
2302 | xlate_report(ctx, "disallowed VLAN VID for this input port, dropping"); |
2303 | return; | |
2304 | } | |
46c88433 | 2305 | vlan = input_vid_to_vlan(in_xbundle, vid); |
9583bc14 EJ |
2306 | |
2307 | /* Check other admissibility requirements. */ | |
2308 | if (in_port && !is_admissible(ctx, in_port, vlan)) { | |
2309 | return; | |
2310 | } | |
2311 | ||
2312 | /* Learn source MAC. */ | |
2313 | if (ctx->xin->may_learn) { | |
46c88433 | 2314 | update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle); |
9583bc14 | 2315 | } |
b256dc52 JS |
2316 | if (ctx->xin->xcache) { |
2317 | struct xc_entry *entry; | |
2318 | ||
2319 | /* Save enough info to update mac learning table later. */ | |
2320 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL); | |
9edf6b48 | 2321 | entry->u.normal.ofproto = ctx->xbridge->ofproto; |
b256dc52 JS |
2322 | entry->u.normal.flow = xmemdup(flow, sizeof *flow); |
2323 | entry->u.normal.vlan = vlan; | |
2324 | } | |
9583bc14 EJ |
2325 | |
2326 | /* Determine output bundle. */ | |
86e2dcdd FL |
2327 | if (mcast_snooping_enabled(ctx->xbridge->ms) |
2328 | && !eth_addr_is_broadcast(flow->dl_dst) | |
2329 | && eth_addr_is_multicast(flow->dl_dst) | |
06994f87 | 2330 | && is_ip_any(flow)) { |
86e2dcdd | 2331 | struct mcast_snooping *ms = ctx->xbridge->ms; |
06994f87 | 2332 | struct mcast_group *grp = NULL; |
86e2dcdd | 2333 | |
06994f87 | 2334 | if (is_igmp(flow)) { |
d29f137b TLSC |
2335 | if (mcast_snooping_is_membership(flow->tp_src) || |
2336 | mcast_snooping_is_query(flow->tp_src)) { | |
2337 | if (ctx->xin->may_learn) { | |
86e2dcdd | 2338 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, |
e3102e42 | 2339 | in_xbundle, ctx->xin->packet); |
d29f137b TLSC |
2340 | } |
2341 | /* | |
2342 | * IGMP packets need to take the slow path, in order to be | |
2343 | * processed for mdb updates. That will prevent expires | |
2344 | * firing off even after hosts have sent reports. | |
2345 | */ | |
2346 | ctx->xout->slow |= SLOW_ACTION; | |
86e2dcdd | 2347 | } |
d6d5bbc9 | 2348 | |
86e2dcdd FL |
2349 | if (mcast_snooping_is_membership(flow->tp_src)) { |
2350 | ovs_rwlock_rdlock(&ms->rwlock); | |
2351 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
8e04a33f FL |
2352 | /* RFC4541: section 2.1.1, item 1: A snooping switch should |
2353 | * forward IGMP Membership Reports only to those ports where | |
2354 | * multicast routers are attached. Alternatively stated: a | |
2355 | * snooping switch should not forward IGMP Membership Reports | |
2356 | * to ports on which only hosts are attached. | |
2357 | * An administrative control may be provided to override this | |
2358 | * restriction, allowing the report messages to be flooded to | |
2359 | * other ports. */ | |
2360 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
86e2dcdd FL |
2361 | ovs_rwlock_unlock(&ms->rwlock); |
2362 | } else { | |
2363 | xlate_report(ctx, "multicast traffic, flooding"); | |
2364 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2365 | } | |
2366 | return; | |
06994f87 TLSC |
2367 | } else if (is_mld(flow)) { |
2368 | ctx->xout->slow |= SLOW_ACTION; | |
2369 | if (ctx->xin->may_learn) { | |
2370 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, | |
2371 | in_xbundle, ctx->xin->packet); | |
2372 | } | |
2373 | if (is_mld_report(flow)) { | |
2374 | ovs_rwlock_rdlock(&ms->rwlock); | |
2375 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2376 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
2377 | ovs_rwlock_unlock(&ms->rwlock); | |
2378 | } else { | |
2379 | xlate_report(ctx, "MLD query, flooding"); | |
2380 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2381 | } | |
86e2dcdd | 2382 | } else { |
06994f87 TLSC |
2383 | if ((flow->dl_type == htons(ETH_TYPE_IP) |
2384 | && ip_is_local_multicast(flow->nw_dst)) | |
2385 | || (flow->dl_type == htons(ETH_TYPE_IPV6) | |
2386 | && ipv6_is_all_hosts(&flow->ipv6_dst))) { | |
86e2dcdd FL |
2387 | /* RFC4541: section 2.1.2, item 2: Packets with a dst IP |
2388 | * address in the 224.0.0.x range which are not IGMP must | |
2389 | * be forwarded on all ports */ | |
2390 | xlate_report(ctx, "RFC4541: section 2.1.2, item 2, flooding"); | |
2391 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2392 | return; | |
2393 | } | |
2394 | } | |
2395 | ||
2396 | /* forwarding to group base ports */ | |
2397 | ovs_rwlock_rdlock(&ms->rwlock); | |
06994f87 TLSC |
2398 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2399 | grp = mcast_snooping_lookup4(ms, flow->nw_dst, vlan); | |
2400 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
2401 | grp = mcast_snooping_lookup(ms, &flow->ipv6_dst, vlan); | |
2402 | } | |
86e2dcdd FL |
2403 | if (grp) { |
2404 | xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, vlan); | |
2405 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2406 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
9583bc14 | 2407 | } else { |
86e2dcdd FL |
2408 | if (mcast_snooping_flood_unreg(ms)) { |
2409 | xlate_report(ctx, "unregistered multicast, flooding"); | |
2410 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2411 | } else { | |
2412 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2413 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2414 | } | |
9583bc14 | 2415 | } |
86e2dcdd | 2416 | ovs_rwlock_unlock(&ms->rwlock); |
9583bc14 | 2417 | } else { |
86e2dcdd FL |
2418 | ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); |
2419 | mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); | |
9d078ec2 | 2420 | mac_port = mac ? mac_entry_get_port(ctx->xbridge->ml, mac) : NULL; |
86e2dcdd FL |
2421 | ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); |
2422 | ||
2423 | if (mac_port) { | |
2424 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2425 | struct xbundle *mac_xbundle = xbundle_lookup(xcfg, mac_port); | |
2426 | if (mac_xbundle && mac_xbundle != in_xbundle) { | |
2427 | xlate_report(ctx, "forwarding to learned port"); | |
2428 | output_normal(ctx, mac_xbundle, vlan); | |
2429 | } else if (!mac_xbundle) { | |
2430 | xlate_report(ctx, "learned port is unknown, dropping"); | |
2431 | } else { | |
2432 | xlate_report(ctx, "learned port is input port, dropping"); | |
2433 | } | |
2434 | } else { | |
2435 | xlate_report(ctx, "no learned MAC for destination, flooding"); | |
2436 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2437 | } | |
9583bc14 EJ |
2438 | } |
2439 | } | |
2440 | ||
a6092018 BP |
2441 | /* Appends a "sample" action for sFlow or IPFIX to 'ctx->odp_actions'. The |
2442 | * 'probability' is the number of packets out of UINT32_MAX to sample. The | |
2443 | * 'cookie' (of length 'cookie_size' bytes) is passed back in the callback for | |
2444 | * each sampled packet. 'tunnel_out_port', if not ODPP_NONE, is added as the | |
2445 | * OVS_USERSPACE_ATTR_EGRESS_TUN_PORT attribute. If 'include_actions', an | |
2446 | * OVS_USERSPACE_ATTR_ACTIONS attribute is added. | |
9583bc14 EJ |
2447 | */ |
2448 | static size_t | |
a6092018 | 2449 | compose_sample_action(struct xlate_ctx *ctx, |
9583bc14 EJ |
2450 | const uint32_t probability, |
2451 | const union user_action_cookie *cookie, | |
8b7ea2d4 | 2452 | const size_t cookie_size, |
7321bda3 NM |
2453 | const odp_port_t tunnel_out_port, |
2454 | bool include_actions) | |
9583bc14 | 2455 | { |
a6092018 BP |
2456 | size_t sample_offset = nl_msg_start_nested(ctx->odp_actions, |
2457 | OVS_ACTION_ATTR_SAMPLE); | |
9583bc14 | 2458 | |
a6092018 | 2459 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability); |
9583bc14 | 2460 | |
a6092018 BP |
2461 | size_t actions_offset = nl_msg_start_nested(ctx->odp_actions, |
2462 | OVS_SAMPLE_ATTR_ACTIONS); | |
9583bc14 | 2463 | |
a6092018 BP |
2464 | odp_port_t odp_port = ofp_port_to_odp_port( |
2465 | ctx->xbridge, ctx->xin->flow.in_port.ofp_port); | |
2466 | uint32_t pid = dpif_port_get_pid(ctx->xbridge->dpif, odp_port, | |
2467 | flow_hash_5tuple(&ctx->xin->flow, 0)); | |
2468 | int cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, | |
2469 | tunnel_out_port, | |
2470 | include_actions, | |
2471 | ctx->odp_actions); | |
89a8a7f0 | 2472 | |
a6092018 BP |
2473 | nl_msg_end_nested(ctx->odp_actions, actions_offset); |
2474 | nl_msg_end_nested(ctx->odp_actions, sample_offset); | |
9583bc14 | 2475 | |
9583bc14 EJ |
2476 | return cookie_offset; |
2477 | } | |
2478 | ||
a6092018 BP |
2479 | /* If sFLow is not enabled, returns 0 without doing anything. |
2480 | * | |
2481 | * If sFlow is enabled, appends a template "sample" action to the ODP actions | |
2482 | * in 'ctx'. This action is a template because some of the information needed | |
2483 | * to fill it out is not available until flow translation is complete. In this | |
2484 | * case, this functions returns an offset, which is always nonzero, to pass | |
2485 | * later to fix_sflow_action() to fill in the rest of the template. */ | |
9583bc14 | 2486 | static size_t |
a6092018 | 2487 | compose_sflow_action(struct xlate_ctx *ctx) |
9583bc14 | 2488 | { |
a6092018 BP |
2489 | struct dpif_sflow *sflow = ctx->xbridge->sflow; |
2490 | if (!sflow || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { | |
9583bc14 EJ |
2491 | return 0; |
2492 | } | |
2493 | ||
a6092018 BP |
2494 | union user_action_cookie cookie = { .type = USER_ACTION_COOKIE_SFLOW }; |
2495 | return compose_sample_action(ctx, dpif_sflow_get_probability(sflow), | |
7321bda3 NM |
2496 | &cookie, sizeof cookie.sflow, ODPP_NONE, |
2497 | true); | |
9583bc14 EJ |
2498 | } |
2499 | ||
a6092018 BP |
2500 | /* If IPFIX is enabled, this appends a "sample" action to implement IPFIX to |
2501 | * 'ctx->odp_actions'. */ | |
9583bc14 | 2502 | static void |
a6092018 | 2503 | compose_ipfix_action(struct xlate_ctx *ctx, odp_port_t output_odp_port) |
9583bc14 | 2504 | { |
a6092018 | 2505 | struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; |
8b7ea2d4 | 2506 | odp_port_t tunnel_out_port = ODPP_NONE; |
9583bc14 | 2507 | |
a6092018 | 2508 | if (!ipfix || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
2509 | return; |
2510 | } | |
2511 | ||
8b7ea2d4 WZ |
2512 | /* For input case, output_odp_port is ODPP_NONE, which is an invalid port |
2513 | * number. */ | |
2514 | if (output_odp_port == ODPP_NONE && | |
a6092018 | 2515 | !dpif_ipfix_get_bridge_exporter_input_sampling(ipfix)) { |
8b7ea2d4 WZ |
2516 | return; |
2517 | } | |
2518 | ||
2519 | /* For output case, output_odp_port is valid*/ | |
2520 | if (output_odp_port != ODPP_NONE) { | |
a6092018 | 2521 | if (!dpif_ipfix_get_bridge_exporter_output_sampling(ipfix)) { |
8b7ea2d4 WZ |
2522 | return; |
2523 | } | |
2524 | /* If tunnel sampling is enabled, put an additional option attribute: | |
2525 | * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT | |
2526 | */ | |
a6092018 BP |
2527 | if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(ipfix) && |
2528 | dpif_ipfix_get_tunnel_port(ipfix, output_odp_port) ) { | |
8b7ea2d4 WZ |
2529 | tunnel_out_port = output_odp_port; |
2530 | } | |
2531 | } | |
2532 | ||
a6092018 BP |
2533 | union user_action_cookie cookie = { |
2534 | .ipfix = { | |
2535 | .type = USER_ACTION_COOKIE_IPFIX, | |
2536 | .output_odp_port = output_odp_port, | |
2537 | } | |
2538 | }; | |
2539 | compose_sample_action(ctx, | |
2540 | dpif_ipfix_get_bridge_exporter_probability(ipfix), | |
7321bda3 NM |
2541 | &cookie, sizeof cookie.ipfix, tunnel_out_port, |
2542 | false); | |
9583bc14 EJ |
2543 | } |
2544 | ||
a6092018 BP |
2545 | /* Fix "sample" action according to data collected while composing ODP actions, |
2546 | * as described in compose_sflow_action(). | |
2547 | * | |
2548 | * 'user_cookie_offset' must be the offset returned by add_sflow_action(). */ | |
9583bc14 | 2549 | static void |
a6092018 | 2550 | fix_sflow_action(struct xlate_ctx *ctx, unsigned int user_cookie_offset) |
9583bc14 EJ |
2551 | { |
2552 | const struct flow *base = &ctx->base_flow; | |
2553 | union user_action_cookie *cookie; | |
2554 | ||
a6092018 | 2555 | cookie = ofpbuf_at(ctx->odp_actions, user_cookie_offset, |
9583bc14 EJ |
2556 | sizeof cookie->sflow); |
2557 | ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); | |
2558 | ||
a6092018 BP |
2559 | cookie->type = USER_ACTION_COOKIE_SFLOW; |
2560 | cookie->sflow.vlan_tci = base->vlan_tci; | |
2561 | ||
2562 | /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output | |
2563 | * port information") for the interpretation of cookie->output. */ | |
2564 | switch (ctx->sflow_n_outputs) { | |
2565 | case 0: | |
2566 | /* 0x40000000 | 256 means "packet dropped for unknown reason". */ | |
2567 | cookie->sflow.output = 0x40000000 | 256; | |
2568 | break; | |
2569 | ||
2570 | case 1: | |
2571 | cookie->sflow.output = dpif_sflow_odp_port_to_ifindex( | |
2572 | ctx->xbridge->sflow, ctx->sflow_odp_port); | |
2573 | if (cookie->sflow.output) { | |
2574 | break; | |
2575 | } | |
2576 | /* Fall through. */ | |
2577 | default: | |
2578 | /* 0x80000000 means "multiple output ports. */ | |
2579 | cookie->sflow.output = 0x80000000 | ctx->sflow_n_outputs; | |
2580 | break; | |
2581 | } | |
9583bc14 EJ |
2582 | } |
2583 | ||
515793d5 BP |
2584 | static bool |
2585 | process_special(struct xlate_ctx *ctx, const struct xport *xport) | |
db7d4e46 | 2586 | { |
515793d5 | 2587 | const struct flow *flow = &ctx->xin->flow; |
49a73e0c | 2588 | struct flow_wildcards *wc = ctx->wc; |
46c88433 | 2589 | const struct xbridge *xbridge = ctx->xbridge; |
515793d5 BP |
2590 | const struct dp_packet *packet = ctx->xin->packet; |
2591 | enum slow_path_reason slow; | |
642dc74d | 2592 | |
46c88433 | 2593 | if (!xport) { |
515793d5 | 2594 | slow = 0; |
46c88433 | 2595 | } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) { |
db7d4e46 | 2596 | if (packet) { |
46c88433 | 2597 | cfm_process_heartbeat(xport->cfm, packet); |
db7d4e46 | 2598 | } |
515793d5 | 2599 | slow = SLOW_CFM; |
fab52e16 | 2600 | } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) { |
db7d4e46 | 2601 | if (packet) { |
46c88433 | 2602 | bfd_process_packet(xport->bfd, flow, packet); |
60d02c72 AW |
2603 | /* If POLL received, immediately sends FINAL back. */ |
2604 | if (bfd_should_send_packet(xport->bfd)) { | |
6d308b28 | 2605 | ofproto_dpif_monitor_port_send_soon(xport->ofport); |
60d02c72 | 2606 | } |
db7d4e46 | 2607 | } |
515793d5 | 2608 | slow = SLOW_BFD; |
46c88433 | 2609 | } else if (xport->xbundle && xport->xbundle->lacp |
db7d4e46 JP |
2610 | && flow->dl_type == htons(ETH_TYPE_LACP)) { |
2611 | if (packet) { | |
46c88433 | 2612 | lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet); |
db7d4e46 | 2613 | } |
515793d5 | 2614 | slow = SLOW_LACP; |
9efd308e DV |
2615 | } else if ((xbridge->stp || xbridge->rstp) && |
2616 | stp_should_process_flow(flow, wc)) { | |
db7d4e46 | 2617 | if (packet) { |
f025bcb7 JR |
2618 | xbridge->stp |
2619 | ? stp_process_packet(xport, packet) | |
2620 | : rstp_process_packet(xport, packet); | |
db7d4e46 | 2621 | } |
515793d5 | 2622 | slow = SLOW_STP; |
19aef6ef | 2623 | } else if (xport->lldp && lldp_should_process_flow(xport->lldp, flow)) { |
0477baa9 DF |
2624 | if (packet) { |
2625 | lldp_process_packet(xport->lldp, packet); | |
2626 | } | |
515793d5 | 2627 | slow = SLOW_LLDP; |
db7d4e46 | 2628 | } else { |
515793d5 BP |
2629 | slow = 0; |
2630 | } | |
2631 | ||
2632 | if (slow) { | |
2633 | ctx->xout->slow |= slow; | |
2634 | return true; | |
2635 | } else { | |
2636 | return false; | |
db7d4e46 JP |
2637 | } |
2638 | } | |
2639 | ||
a36de779 PS |
2640 | static int |
2641 | tnl_route_lookup_flow(const struct flow *oflow, | |
2642 | ovs_be32 *ip, struct xport **out_port) | |
2643 | { | |
2644 | char out_dev[IFNAMSIZ]; | |
2645 | struct xbridge *xbridge; | |
2646 | struct xlate_cfg *xcfg; | |
2647 | ovs_be32 gw; | |
2648 | ||
2649 | if (!ovs_router_lookup(oflow->tunnel.ip_dst, out_dev, &gw)) { | |
2650 | return -ENOENT; | |
2651 | } | |
2652 | ||
2653 | if (gw) { | |
2654 | *ip = gw; | |
2655 | } else { | |
2656 | *ip = oflow->tunnel.ip_dst; | |
2657 | } | |
2658 | ||
2659 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2660 | ovs_assert(xcfg); | |
2661 | ||
2662 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
2663 | if (!strncmp(xbridge->name, out_dev, IFNAMSIZ)) { | |
2664 | struct xport *port; | |
2665 | ||
2666 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
2667 | if (!strncmp(netdev_get_name(port->netdev), out_dev, IFNAMSIZ)) { | |
2668 | *out_port = port; | |
2669 | return 0; | |
2670 | } | |
2671 | } | |
2672 | } | |
2673 | } | |
2674 | return -ENOENT; | |
2675 | } | |
2676 | ||
2677 | static int | |
cf62fa4c | 2678 | xlate_flood_packet(struct xbridge *xbridge, struct dp_packet *packet) |
a36de779 PS |
2679 | { |
2680 | struct ofpact_output output; | |
2681 | struct flow flow; | |
2682 | ||
2683 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); | |
2684 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ | |
cf62fa4c | 2685 | flow_extract(packet, &flow); |
a36de779 PS |
2686 | flow.in_port.ofp_port = OFPP_NONE; |
2687 | output.port = OFPP_FLOOD; | |
2688 | output.max_len = 0; | |
2689 | ||
2690 | return ofproto_dpif_execute_actions(xbridge->ofproto, &flow, NULL, | |
2691 | &output.ofpact, sizeof output, | |
2692 | packet); | |
2693 | } | |
2694 | ||
2695 | static void | |
2696 | tnl_send_arp_request(const struct xport *out_dev, const uint8_t eth_src[ETH_ADDR_LEN], | |
2697 | ovs_be32 ip_src, ovs_be32 ip_dst) | |
2698 | { | |
2699 | struct xbridge *xbridge = out_dev->xbridge; | |
cf62fa4c | 2700 | struct dp_packet packet; |
a36de779 | 2701 | |
cf62fa4c | 2702 | dp_packet_init(&packet, 0); |
eb0b295e BP |
2703 | compose_arp(&packet, ARP_OP_REQUEST, |
2704 | eth_src, eth_addr_zero, true, ip_src, ip_dst); | |
a36de779 PS |
2705 | |
2706 | xlate_flood_packet(xbridge, &packet); | |
cf62fa4c | 2707 | dp_packet_uninit(&packet); |
a36de779 PS |
2708 | } |
2709 | ||
2710 | static int | |
81de18ec | 2711 | build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport, |
a36de779 PS |
2712 | const struct flow *flow, odp_port_t tunnel_odp_port) |
2713 | { | |
2714 | struct ovs_action_push_tnl tnl_push_data; | |
2715 | struct xport *out_dev = NULL; | |
2716 | ovs_be32 s_ip, d_ip = 0; | |
2717 | uint8_t smac[ETH_ADDR_LEN]; | |
2718 | uint8_t dmac[ETH_ADDR_LEN]; | |
2719 | int err; | |
2720 | ||
2721 | err = tnl_route_lookup_flow(flow, &d_ip, &out_dev); | |
2722 | if (err) { | |
81de18ec | 2723 | xlate_report(ctx, "native tunnel routing failed"); |
a36de779 PS |
2724 | return err; |
2725 | } | |
81de18ec BP |
2726 | xlate_report(ctx, "tunneling to "IP_FMT" via %s", |
2727 | IP_ARGS(d_ip), netdev_get_name(out_dev->netdev)); | |
a36de779 PS |
2728 | |
2729 | /* Use mac addr of bridge port of the peer. */ | |
2730 | err = netdev_get_etheraddr(out_dev->netdev, smac); | |
2731 | if (err) { | |
81de18ec | 2732 | xlate_report(ctx, "tunnel output device lacks Ethernet address"); |
a36de779 PS |
2733 | return err; |
2734 | } | |
2735 | ||
2736 | err = netdev_get_in4(out_dev->netdev, (struct in_addr *) &s_ip, NULL); | |
2737 | if (err) { | |
81de18ec | 2738 | xlate_report(ctx, "tunnel output device lacks IPv4 address"); |
a36de779 PS |
2739 | return err; |
2740 | } | |
2741 | ||
2742 | err = tnl_arp_lookup(out_dev->xbridge->name, d_ip, dmac); | |
2743 | if (err) { | |
81de18ec BP |
2744 | xlate_report(ctx, "ARP cache miss for "IP_FMT" on bridge %s, " |
2745 | "sending ARP request", | |
2746 | IP_ARGS(d_ip), out_dev->xbridge->name); | |
a36de779 PS |
2747 | tnl_send_arp_request(out_dev, smac, s_ip, d_ip); |
2748 | return err; | |
2749 | } | |
2750 | if (ctx->xin->xcache) { | |
2751 | struct xc_entry *entry; | |
2752 | ||
2753 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_ARP); | |
8742957c BP |
2754 | ovs_strlcpy(entry->u.tnl_arp_cache.br_name, out_dev->xbridge->name, |
2755 | sizeof entry->u.tnl_arp_cache.br_name); | |
a36de779 PS |
2756 | entry->u.tnl_arp_cache.d_ip = d_ip; |
2757 | } | |
81de18ec BP |
2758 | |
2759 | xlate_report(ctx, "tunneling from "ETH_ADDR_FMT" "IP_FMT | |
2760 | " to "ETH_ADDR_FMT" "IP_FMT, | |
2761 | ETH_ADDR_ARGS(smac), IP_ARGS(s_ip), | |
2762 | ETH_ADDR_ARGS(dmac), IP_ARGS(d_ip)); | |
a36de779 PS |
2763 | err = tnl_port_build_header(xport->ofport, flow, |
2764 | dmac, smac, s_ip, &tnl_push_data); | |
2765 | if (err) { | |
2766 | return err; | |
2767 | } | |
2768 | tnl_push_data.tnl_port = odp_to_u32(tunnel_odp_port); | |
2769 | tnl_push_data.out_port = odp_to_u32(out_dev->odp_port); | |
1520ef4f | 2770 | odp_put_tnl_push_action(ctx->odp_actions, &tnl_push_data); |
a36de779 PS |
2771 | return 0; |
2772 | } | |
2773 | ||
9583bc14 | 2774 | static void |
4e022ec0 | 2775 | compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
e93ef1c7 | 2776 | const struct xlate_bond_recirc *xr, bool check_stp) |
9583bc14 | 2777 | { |
46c88433 | 2778 | const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
49a73e0c | 2779 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 2780 | struct flow *flow = &ctx->xin->flow; |
a36de779 | 2781 | struct flow_tnl flow_tnl; |
9583bc14 | 2782 | ovs_be16 flow_vlan_tci; |
1362e248 | 2783 | uint32_t flow_pkt_mark; |
9583bc14 | 2784 | uint8_t flow_nw_tos; |
4e022ec0 | 2785 | odp_port_t out_port, odp_port; |
a36de779 | 2786 | bool tnl_push_pop_send = false; |
ca077186 | 2787 | uint8_t dscp; |
9583bc14 EJ |
2788 | |
2789 | /* If 'struct flow' gets additional metadata, we'll need to zero it out | |
2790 | * before traversing a patch port. */ | |
b666962b | 2791 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 33); |
a36de779 | 2792 | memset(&flow_tnl, 0, sizeof flow_tnl); |
9583bc14 | 2793 | |
46c88433 | 2794 | if (!xport) { |
9583bc14 EJ |
2795 | xlate_report(ctx, "Nonexistent output port"); |
2796 | return; | |
46c88433 | 2797 | } else if (xport->config & OFPUTIL_PC_NO_FWD) { |
9583bc14 EJ |
2798 | xlate_report(ctx, "OFPPC_NO_FWD set, skipping output"); |
2799 | return; | |
0d1cee12 | 2800 | } else if (check_stp) { |
bbbca389 | 2801 | if (is_stp(&ctx->base_flow)) { |
9efd308e DV |
2802 | if (!xport_stp_should_forward_bpdu(xport) && |
2803 | !xport_rstp_should_manage_bpdu(xport)) { | |
2804 | if (ctx->xbridge->stp != NULL) { | |
2805 | xlate_report(ctx, "STP not in listening state, " | |
2806 | "skipping bpdu output"); | |
2807 | } else if (ctx->xbridge->rstp != NULL) { | |
2808 | xlate_report(ctx, "RSTP not managing BPDU in this state, " | |
2809 | "skipping bpdu output"); | |
2810 | } | |
0d1cee12 K |
2811 | return; |
2812 | } | |
9efd308e DV |
2813 | } else if (!xport_stp_forward_state(xport) || |
2814 | !xport_rstp_forward_state(xport)) { | |
2815 | if (ctx->xbridge->stp != NULL) { | |
2816 | xlate_report(ctx, "STP not in forwarding state, " | |
2817 | "skipping output"); | |
2818 | } else if (ctx->xbridge->rstp != NULL) { | |
2819 | xlate_report(ctx, "RSTP not in forwarding state, " | |
2820 | "skipping output"); | |
2821 | } | |
0d1cee12 K |
2822 | return; |
2823 | } | |
9583bc14 EJ |
2824 | } |
2825 | ||
46c88433 | 2826 | if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { |
3d6151f3 BP |
2827 | ctx->mirrors |= xbundle_mirror_dst(xport->xbundle->xbridge, |
2828 | xport->xbundle); | |
cdf5d3a5 EJ |
2829 | } |
2830 | ||
46c88433 EJ |
2831 | if (xport->peer) { |
2832 | const struct xport *peer = xport->peer; | |
9583bc14 | 2833 | struct flow old_flow = ctx->xin->flow; |
7f5fe537 | 2834 | bool old_was_mpls = ctx->was_mpls; |
18721c4a | 2835 | cls_version_t old_version = ctx->tables_version; |
1774d762 JR |
2836 | struct ofpbuf old_stack = ctx->stack; |
2837 | union mf_subvalue new_stack[1024 / sizeof(union mf_subvalue)]; | |
ed9c9e3e JR |
2838 | struct ofpbuf old_action_set = ctx->action_set; |
2839 | uint64_t actset_stub[1024 / 8]; | |
9583bc14 | 2840 | |
1774d762 | 2841 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); |
ed9c9e3e | 2842 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); |
46c88433 EJ |
2843 | ctx->xbridge = peer->xbridge; |
2844 | flow->in_port.ofp_port = peer->ofp_port; | |
33bf9176 BP |
2845 | flow->metadata = htonll(0); |
2846 | memset(&flow->tunnel, 0, sizeof flow->tunnel); | |
2847 | memset(flow->regs, 0, sizeof flow->regs); | |
c61f3870 | 2848 | flow->actset_output = OFPP_UNSET; |
9583bc14 | 2849 | |
621b8064 JR |
2850 | /* The bridge is now known so obtain its table version. */ |
2851 | ctx->tables_version | |
2852 | = ofproto_dpif_get_tables_version(ctx->xbridge->ofproto); | |
2853 | ||
515793d5 | 2854 | if (!process_special(ctx, peer) && may_receive(peer, ctx)) { |
9efd308e | 2855 | if (xport_stp_forward_state(peer) && xport_rstp_forward_state(peer)) { |
e672ff9b | 2856 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); |
ed9c9e3e | 2857 | if (ctx->action_set.size) { |
e672ff9b JR |
2858 | /* Translate action set only if not dropping the packet and |
2859 | * not recirculating. */ | |
2860 | if (!exit_recirculates(ctx)) { | |
2861 | xlate_action_set(ctx); | |
2862 | } | |
2863 | } | |
2864 | /* Check if need to recirculate. */ | |
2865 | if (exit_recirculates(ctx)) { | |
2866 | compose_recirculate_action(ctx); | |
ed9c9e3e | 2867 | } |
9583bc14 | 2868 | } else { |
9efd308e DV |
2869 | /* Forwarding is disabled by STP and RSTP. Let OFPP_NORMAL and |
2870 | * the learning action look at the packet, then drop it. */ | |
9583bc14 | 2871 | struct flow old_base_flow = ctx->base_flow; |
1520ef4f | 2872 | size_t old_size = ctx->odp_actions->size; |
3d6151f3 | 2873 | mirror_mask_t old_mirrors = ctx->mirrors; |
f3d5b473 | 2874 | |
e672ff9b | 2875 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); |
3d6151f3 | 2876 | ctx->mirrors = old_mirrors; |
9583bc14 | 2877 | ctx->base_flow = old_base_flow; |
1520ef4f | 2878 | ctx->odp_actions->size = old_size; |
e672ff9b JR |
2879 | |
2880 | /* Undo changes that may have been done for recirculation. */ | |
2881 | if (exit_recirculates(ctx)) { | |
2882 | ctx->action_set.size = ctx->recirc_action_offset; | |
2883 | ctx->recirc_action_offset = -1; | |
2884 | ctx->last_unroll_offset = -1; | |
2885 | } | |
9583bc14 EJ |
2886 | } |
2887 | } | |
2888 | ||
2889 | ctx->xin->flow = old_flow; | |
832554e3 | 2890 | ctx->xbridge = xport->xbridge; |
ed9c9e3e JR |
2891 | ofpbuf_uninit(&ctx->action_set); |
2892 | ctx->action_set = old_action_set; | |
1774d762 JR |
2893 | ofpbuf_uninit(&ctx->stack); |
2894 | ctx->stack = old_stack; | |
9583bc14 | 2895 | |
621b8064 JR |
2896 | /* Restore calling bridge's lookup version. */ |
2897 | ctx->tables_version = old_version; | |
2898 | ||
7f5fe537 JR |
2899 | /* The peer bridge popping MPLS should have no effect on the original |
2900 | * bridge. */ | |
2901 | ctx->was_mpls = old_was_mpls; | |
2902 | ||
f3d5b473 JR |
2903 | /* The fact that the peer bridge exits (for any reason) does not mean |
2904 | * that the original bridge should exit. Specifically, if the peer | |
2905 | * bridge recirculates (which typically modifies the packet), the | |
2906 | * original bridge must continue processing with the original, not the | |
2907 | * recirculated packet! */ | |
2908 | ctx->exit = false; | |
2909 | ||
9583bc14 | 2910 | if (ctx->xin->resubmit_stats) { |
46c88433 EJ |
2911 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
2912 | netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats); | |
a1aeea86 AW |
2913 | if (peer->bfd) { |
2914 | bfd_account_rx(peer->bfd, ctx->xin->resubmit_stats); | |
2915 | } | |
9583bc14 | 2916 | } |
b256dc52 JS |
2917 | if (ctx->xin->xcache) { |
2918 | struct xc_entry *entry; | |
2919 | ||
2920 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
2921 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
2922 | entry->u.dev.rx = netdev_ref(peer->netdev); | |
2923 | entry->u.dev.bfd = bfd_ref(peer->bfd); | |
2924 | } | |
9583bc14 EJ |
2925 | return; |
2926 | } | |
2927 | ||
33bf9176 | 2928 | flow_vlan_tci = flow->vlan_tci; |
1362e248 | 2929 | flow_pkt_mark = flow->pkt_mark; |
33bf9176 | 2930 | flow_nw_tos = flow->nw_tos; |
9583bc14 | 2931 | |
16194afd DDP |
2932 | if (count_skb_priorities(xport)) { |
2933 | memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority); | |
2934 | if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { | |
2935 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
2936 | flow->nw_tos &= ~IP_DSCP_MASK; | |
2937 | flow->nw_tos |= dscp; | |
2938 | } | |
9583bc14 EJ |
2939 | } |
2940 | ||
46c88433 | 2941 | if (xport->is_tunnel) { |
9583bc14 EJ |
2942 | /* Save tunnel metadata so that changes made due to |
2943 | * the Logical (tunnel) Port are not visible for any further | |
2944 | * matches, while explicit set actions on tunnel metadata are. | |
2945 | */ | |
a36de779 | 2946 | flow_tnl = flow->tunnel; |
49a73e0c | 2947 | odp_port = tnl_port_send(xport->ofport, flow, ctx->wc); |
4e022ec0 | 2948 | if (odp_port == ODPP_NONE) { |
9583bc14 EJ |
2949 | xlate_report(ctx, "Tunneling decided against output"); |
2950 | goto out; /* restore flow_nw_tos */ | |
2951 | } | |
33bf9176 | 2952 | if (flow->tunnel.ip_dst == ctx->orig_tunnel_ip_dst) { |
9583bc14 EJ |
2953 | xlate_report(ctx, "Not tunneling to our own address"); |
2954 | goto out; /* restore flow_nw_tos */ | |
2955 | } | |
2956 | if (ctx->xin->resubmit_stats) { | |
46c88433 | 2957 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
9583bc14 | 2958 | } |
b256dc52 JS |
2959 | if (ctx->xin->xcache) { |
2960 | struct xc_entry *entry; | |
2961 | ||
2962 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
2963 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
2964 | } | |
9583bc14 | 2965 | out_port = odp_port; |
a36de779 | 2966 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
81de18ec | 2967 | xlate_report(ctx, "output to native tunnel"); |
a36de779 PS |
2968 | tnl_push_pop_send = true; |
2969 | } else { | |
81de18ec | 2970 | xlate_report(ctx, "output to kernel tunnel"); |
1520ef4f | 2971 | commit_odp_tunnel_action(flow, &ctx->base_flow, ctx->odp_actions); |
a36de779 PS |
2972 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ |
2973 | } | |
9583bc14 | 2974 | } else { |
46c88433 | 2975 | odp_port = xport->odp_port; |
7614e5d0 | 2976 | out_port = odp_port; |
46c88433 | 2977 | if (ofproto_has_vlan_splinters(ctx->xbridge->ofproto)) { |
7614e5d0 JR |
2978 | ofp_port_t vlandev_port; |
2979 | ||
1dd35f8a | 2980 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
7614e5d0 JR |
2981 | vlandev_port = vsp_realdev_to_vlandev(ctx->xbridge->ofproto, |
2982 | ofp_port, flow->vlan_tci); | |
2983 | if (vlandev_port != ofp_port) { | |
2984 | out_port = ofp_port_to_odp_port(ctx->xbridge, vlandev_port); | |
2985 | flow->vlan_tci = htons(0); | |
2986 | } | |
9583bc14 | 2987 | } |
9583bc14 | 2988 | } |
9583bc14 | 2989 | |
4e022ec0 | 2990 | if (out_port != ODPP_NONE) { |
b440dd8c JS |
2991 | bool use_masked = ctx->xbridge->support.masked_set_action; |
2992 | ||
7fd91025 | 2993 | ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow, |
1520ef4f | 2994 | ctx->odp_actions, |
b440dd8c | 2995 | wc, use_masked); |
adcf00ba | 2996 | |
e93ef1c7 | 2997 | if (xr) { |
347bf289 | 2998 | struct ovs_action_hash *act_hash; |
adcf00ba | 2999 | |
347bf289 | 3000 | /* Hash action. */ |
1520ef4f | 3001 | act_hash = nl_msg_put_unspec_uninit(ctx->odp_actions, |
347bf289 AZ |
3002 | OVS_ACTION_ATTR_HASH, |
3003 | sizeof *act_hash); | |
3004 | act_hash->hash_alg = xr->hash_alg; | |
62ac1f20 | 3005 | act_hash->hash_basis = xr->hash_basis; |
347bf289 AZ |
3006 | |
3007 | /* Recirc action. */ | |
1520ef4f | 3008 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, |
347bf289 | 3009 | xr->recirc_id); |
adcf00ba | 3010 | } else { |
a36de779 PS |
3011 | |
3012 | if (tnl_push_pop_send) { | |
3013 | build_tunnel_send(ctx, xport, flow, odp_port); | |
3014 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ | |
3015 | } else { | |
3016 | odp_port_t odp_tnl_port = ODPP_NONE; | |
3017 | ||
3018 | /* XXX: Write better Filter for tunnel port. We can use inport | |
3019 | * int tunnel-port flow to avoid these checks completely. */ | |
3020 | if (ofp_port == OFPP_LOCAL && | |
3021 | ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { | |
3022 | ||
3023 | odp_tnl_port = tnl_port_map_lookup(flow, wc); | |
3024 | } | |
3025 | ||
3026 | if (odp_tnl_port != ODPP_NONE) { | |
1520ef4f | 3027 | nl_msg_put_odp_port(ctx->odp_actions, |
a36de779 PS |
3028 | OVS_ACTION_ATTR_TUNNEL_POP, |
3029 | odp_tnl_port); | |
3030 | } else { | |
3031 | /* Tunnel push-pop action is not compatible with | |
3032 | * IPFIX action. */ | |
a6092018 | 3033 | compose_ipfix_action(ctx, out_port); |
1520ef4f | 3034 | nl_msg_put_odp_port(ctx->odp_actions, |
a36de779 PS |
3035 | OVS_ACTION_ATTR_OUTPUT, |
3036 | out_port); | |
3037 | } | |
3038 | } | |
adcf00ba | 3039 | } |
9583bc14 | 3040 | |
6cbbf4fa EJ |
3041 | ctx->sflow_odp_port = odp_port; |
3042 | ctx->sflow_n_outputs++; | |
3043 | ctx->xout->nf_output_iface = ofp_port; | |
3044 | } | |
3045 | ||
3046 | out: | |
9583bc14 | 3047 | /* Restore flow */ |
33bf9176 | 3048 | flow->vlan_tci = flow_vlan_tci; |
1362e248 | 3049 | flow->pkt_mark = flow_pkt_mark; |
33bf9176 | 3050 | flow->nw_tos = flow_nw_tos; |
9583bc14 EJ |
3051 | } |
3052 | ||
3053 | static void | |
e93ef1c7 JR |
3054 | compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
3055 | const struct xlate_bond_recirc *xr) | |
9583bc14 | 3056 | { |
e93ef1c7 | 3057 | compose_output_action__(ctx, ofp_port, xr, true); |
9583bc14 EJ |
3058 | } |
3059 | ||
bb61b33d BP |
3060 | static void |
3061 | xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule) | |
bb61b33d BP |
3062 | { |
3063 | struct rule_dpif *old_rule = ctx->rule; | |
8b1e5560 | 3064 | ovs_be64 old_cookie = ctx->rule_cookie; |
dc723c44 | 3065 | const struct rule_actions *actions; |
bb61b33d BP |
3066 | |
3067 | if (ctx->xin->resubmit_stats) { | |
70742c7f | 3068 | rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats); |
bb61b33d BP |
3069 | } |
3070 | ||
98b07853 | 3071 | ctx->resubmits++; |
bb61b33d BP |
3072 | ctx->recurse++; |
3073 | ctx->rule = rule; | |
8b1e5560 | 3074 | ctx->rule_cookie = rule_dpif_get_flow_cookie(rule); |
6f00e29b BP |
3075 | actions = rule_dpif_get_actions(rule); |
3076 | do_xlate_actions(actions->ofpacts, actions->ofpacts_len, ctx); | |
8b1e5560 | 3077 | ctx->rule_cookie = old_cookie; |
bb61b33d BP |
3078 | ctx->rule = old_rule; |
3079 | ctx->recurse--; | |
bb61b33d BP |
3080 | } |
3081 | ||
bd3240ba SH |
3082 | static bool |
3083 | xlate_resubmit_resource_check(struct xlate_ctx *ctx) | |
9583bc14 | 3084 | { |
98b07853 BP |
3085 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); |
3086 | ||
adcf00ba | 3087 | if (ctx->recurse >= MAX_RESUBMIT_RECURSION + MAX_INTERNAL_RESUBMITS) { |
98b07853 BP |
3088 | VLOG_ERR_RL(&rl, "resubmit actions recursed over %d times", |
3089 | MAX_RESUBMIT_RECURSION); | |
adcf00ba | 3090 | } else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) { |
98b07853 | 3091 | VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS); |
1520ef4f | 3092 | } else if (ctx->odp_actions->size > UINT16_MAX) { |
98b07853 | 3093 | VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of actions"); |
6fd6ed71 | 3094 | } else if (ctx->stack.size >= 65536) { |
98b07853 BP |
3095 | VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of stack"); |
3096 | } else { | |
bd3240ba SH |
3097 | return true; |
3098 | } | |
3099 | ||
3100 | return false; | |
3101 | } | |
3102 | ||
3103 | static void | |
6d328fa2 SH |
3104 | xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, |
3105 | bool may_packet_in, bool honor_table_miss) | |
bd3240ba | 3106 | { |
1d741d6d JR |
3107 | /* Check if we need to recirculate before matching in a table. */ |
3108 | if (ctx->was_mpls) { | |
3109 | ctx_trigger_recirculation(ctx); | |
3110 | return; | |
3111 | } | |
bd3240ba | 3112 | if (xlate_resubmit_resource_check(ctx)) { |
9583bc14 | 3113 | uint8_t old_table_id = ctx->table_id; |
3f207910 | 3114 | struct rule_dpif *rule; |
9583bc14 EJ |
3115 | |
3116 | ctx->table_id = table_id; | |
3117 | ||
34dd0d78 | 3118 | rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, |
621b8064 | 3119 | ctx->tables_version, |
49a73e0c | 3120 | &ctx->xin->flow, ctx->xin->wc, |
34dd0d78 JR |
3121 | ctx->xin->xcache != NULL, |
3122 | ctx->xin->resubmit_stats, | |
3123 | &ctx->table_id, in_port, | |
3124 | may_packet_in, honor_table_miss); | |
9583bc14 | 3125 | |
a8c31348 BP |
3126 | if (OVS_UNLIKELY(ctx->xin->resubmit_hook)) { |
3127 | ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse + 1); | |
ad3efdcb EJ |
3128 | } |
3129 | ||
a2143702 | 3130 | if (rule) { |
83709dfa JR |
3131 | /* Fill in the cache entry here instead of xlate_recursively |
3132 | * to make the reference counting more explicit. We take a | |
3133 | * reference in the lookups above if we are going to cache the | |
3134 | * rule. */ | |
3135 | if (ctx->xin->xcache) { | |
3136 | struct xc_entry *entry; | |
3137 | ||
3138 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE); | |
3139 | entry->u.rule = rule; | |
3140 | } | |
bb61b33d | 3141 | xlate_recursively(ctx, rule); |
ad3efdcb EJ |
3142 | } |
3143 | ||
9583bc14 | 3144 | ctx->table_id = old_table_id; |
98b07853 | 3145 | return; |
9583bc14 | 3146 | } |
98b07853 BP |
3147 | |
3148 | ctx->exit = true; | |
9583bc14 EJ |
3149 | } |
3150 | ||
f4fb341b | 3151 | static void |
1e684d7d RW |
3152 | xlate_group_stats(struct xlate_ctx *ctx, struct group_dpif *group, |
3153 | struct ofputil_bucket *bucket) | |
3154 | { | |
3155 | if (ctx->xin->resubmit_stats) { | |
3156 | group_dpif_credit_stats(group, bucket, ctx->xin->resubmit_stats); | |
3157 | } | |
3158 | if (ctx->xin->xcache) { | |
3159 | struct xc_entry *entry; | |
3160 | ||
3161 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_GROUP); | |
3162 | entry->u.group.group = group_dpif_ref(group); | |
3163 | entry->u.group.bucket = bucket; | |
3164 | } | |
3165 | } | |
3166 | ||
3167 | static void | |
3168 | xlate_group_bucket(struct xlate_ctx *ctx, struct ofputil_bucket *bucket) | |
f4fb341b SH |
3169 | { |
3170 | uint64_t action_list_stub[1024 / 8]; | |
3171 | struct ofpbuf action_list, action_set; | |
5b09e569 JR |
3172 | struct flow old_flow = ctx->xin->flow; |
3173 | bool old_was_mpls = ctx->was_mpls; | |
f4fb341b SH |
3174 | |
3175 | ofpbuf_use_const(&action_set, bucket->ofpacts, bucket->ofpacts_len); | |
3176 | ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub); | |
3177 | ||
3178 | ofpacts_execute_action_set(&action_list, &action_set); | |
3179 | ctx->recurse++; | |
6fd6ed71 | 3180 | do_xlate_actions(action_list.data, action_list.size, ctx); |
f4fb341b SH |
3181 | ctx->recurse--; |
3182 | ||
3183 | ofpbuf_uninit(&action_set); | |
3184 | ofpbuf_uninit(&action_list); | |
5b09e569 | 3185 | |
e672ff9b JR |
3186 | /* Check if need to recirculate. */ |
3187 | if (exit_recirculates(ctx)) { | |
3188 | compose_recirculate_action(ctx); | |
3189 | } | |
3190 | ||
5b09e569 JR |
3191 | /* Roll back flow to previous state. |
3192 | * This is equivalent to cloning the packet for each bucket. | |
3193 | * | |
3194 | * As a side effect any subsequently applied actions will | |
3195 | * also effectively be applied to a clone of the packet taken | |
3196 | * just before applying the all or indirect group. | |
3197 | * | |
3198 | * Note that group buckets are action sets, hence they cannot modify the | |
3199 | * main action set. Also any stack actions are ignored when executing an | |
3200 | * action set, so group buckets cannot change the stack either. | |
3201 | * However, we do allow resubmit actions in group buckets, which could | |
3202 | * break the above assumptions. It is up to the controller to not mess up | |
3203 | * with the action_set and stack in the tables resubmitted to from | |
3204 | * group buckets. */ | |
3205 | ctx->xin->flow = old_flow; | |
3206 | ||
3207 | /* The group bucket popping MPLS should have no effect after bucket | |
3208 | * execution. */ | |
3209 | ctx->was_mpls = old_was_mpls; | |
3210 | ||
3211 | /* The fact that the group bucket exits (for any reason) does not mean that | |
3212 | * the translation after the group action should exit. Specifically, if | |
3213 | * the group bucket recirculates (which typically modifies the packet), the | |
3214 | * actions after the group action must continue processing with the | |
3215 | * original, not the recirculated packet! */ | |
3216 | ctx->exit = false; | |
f4fb341b SH |
3217 | } |
3218 | ||
3219 | static void | |
3220 | xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3221 | { | |
1e684d7d | 3222 | struct ofputil_bucket *bucket; |
ca6ba700 | 3223 | const struct ovs_list *buckets; |
f4fb341b SH |
3224 | |
3225 | group_dpif_get_buckets(group, &buckets); | |
3226 | ||
3227 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
3228 | xlate_group_bucket(ctx, bucket); | |
f4fb341b | 3229 | } |
1e684d7d | 3230 | xlate_group_stats(ctx, group, NULL); |
f4fb341b SH |
3231 | } |
3232 | ||
dd8cd4b4 SH |
3233 | static void |
3234 | xlate_ff_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3235 | { | |
1e684d7d | 3236 | struct ofputil_bucket *bucket; |
dd8cd4b4 SH |
3237 | |
3238 | bucket = group_first_live_bucket(ctx, group, 0); | |
3239 | if (bucket) { | |
3240 | xlate_group_bucket(ctx, bucket); | |
1e684d7d | 3241 | xlate_group_stats(ctx, group, bucket); |
dd8cd4b4 SH |
3242 | } |
3243 | } | |
3244 | ||
fe7e5749 | 3245 | static void |
7565c3e4 | 3246 | xlate_default_select_group(struct xlate_ctx *ctx, struct group_dpif *group) |
fe7e5749 | 3247 | { |
49a73e0c | 3248 | struct flow_wildcards *wc = ctx->wc; |
1e684d7d | 3249 | struct ofputil_bucket *bucket; |
fe7e5749 SH |
3250 | uint32_t basis; |
3251 | ||
1d1aae0b | 3252 | basis = flow_hash_symmetric_l4(&ctx->xin->flow, 0); |
80e3509d | 3253 | flow_mask_hash_fields(&ctx->xin->flow, wc, NX_HASH_FIELDS_SYMMETRIC_L4); |
fe7e5749 SH |
3254 | bucket = group_best_live_bucket(ctx, group, basis); |
3255 | if (bucket) { | |
fe7e5749 | 3256 | xlate_group_bucket(ctx, bucket); |
1e684d7d | 3257 | xlate_group_stats(ctx, group, bucket); |
fe7e5749 SH |
3258 | } |
3259 | } | |
3260 | ||
0c4b9393 SH |
3261 | static void |
3262 | xlate_hash_fields_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3263 | { | |
3264 | struct mf_bitmap hash_fields = MF_BITMAP_INITIALIZER; | |
0c4b9393 SH |
3265 | const struct field_array *fields; |
3266 | struct ofputil_bucket *bucket; | |
3267 | uint32_t basis; | |
3268 | int i; | |
3269 | ||
3270 | fields = group_dpif_get_fields(group); | |
3271 | basis = hash_uint64(group_dpif_get_selection_method_param(group)); | |
3272 | ||
3273 | /* Determine which fields to hash */ | |
3274 | for (i = 0; i < MFF_N_IDS; i++) { | |
3275 | if (bitmap_is_set(fields->used.bm, i)) { | |
3276 | const struct mf_field *mf; | |
3277 | ||
3278 | /* If the field is already present in 'hash_fields' then | |
3279 | * this loop has already checked that it and its pre-requisites | |
3280 | * are present in the flow and its pre-requisites have | |
3281 | * already been added to 'hash_fields'. There is nothing more | |
3282 | * to do here and as an optimisation the loop can continue. */ | |
3283 | if (bitmap_is_set(hash_fields.bm, i)) { | |
3284 | continue; | |
3285 | } | |
3286 | ||
3287 | mf = mf_from_id(i); | |
3288 | ||
3289 | /* Only hash a field if it and its pre-requisites are present | |
3290 | * in the flow. */ | |
3291 | if (!mf_are_prereqs_ok(mf, &ctx->xin->flow)) { | |
3292 | continue; | |
3293 | } | |
3294 | ||
3295 | /* Hash both the field and its pre-requisites */ | |
3296 | mf_bitmap_set_field_and_prereqs(mf, &hash_fields); | |
3297 | } | |
3298 | } | |
3299 | ||
3300 | /* Hash the fields */ | |
3301 | for (i = 0; i < MFF_N_IDS; i++) { | |
3302 | if (bitmap_is_set(hash_fields.bm, i)) { | |
3303 | const struct mf_field *mf = mf_from_id(i); | |
3304 | union mf_value value; | |
3305 | int j; | |
3306 | ||
3307 | mf_get_value(mf, &ctx->xin->flow, &value); | |
3308 | /* This seems inefficient but so does apply_mask() */ | |
3309 | for (j = 0; j < mf->n_bytes; j++) { | |
3310 | ((uint8_t *) &value)[j] &= ((uint8_t *) &fields->value[i])[j]; | |
3311 | } | |
3312 | basis = hash_bytes(&value, mf->n_bytes, basis); | |
3313 | ||
49a73e0c | 3314 | mf_mask_field(mf, &ctx->wc->masks); |
0c4b9393 SH |
3315 | } |
3316 | } | |
3317 | ||
3318 | bucket = group_best_live_bucket(ctx, group, basis); | |
3319 | if (bucket) { | |
3320 | xlate_group_bucket(ctx, bucket); | |
3321 | xlate_group_stats(ctx, group, bucket); | |
3322 | } | |
3323 | } | |
3324 | ||
7565c3e4 SH |
3325 | static void |
3326 | xlate_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3327 | { | |
3328 | const char *selection_method = group_dpif_get_selection_method(group); | |
3329 | ||
3330 | if (selection_method[0] == '\0') { | |
3331 | xlate_default_select_group(ctx, group); | |
0c4b9393 SH |
3332 | } else if (!strcasecmp("hash", selection_method)) { |
3333 | xlate_hash_fields_select_group(ctx, group); | |
7565c3e4 SH |
3334 | } else { |
3335 | /* Parsing of groups should ensure this never happens */ | |
3336 | OVS_NOT_REACHED(); | |
3337 | } | |
3338 | } | |
3339 | ||
f4fb341b SH |
3340 | static void |
3341 | xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group) | |
3342 | { | |
0eb48fe1 | 3343 | bool was_in_group = ctx->in_group; |
5a070238 BP |
3344 | ctx->in_group = true; |
3345 | ||
f4fb341b SH |
3346 | switch (group_dpif_get_type(group)) { |
3347 | case OFPGT11_ALL: | |
3348 | case OFPGT11_INDIRECT: | |
3349 | xlate_all_group(ctx, group); | |
3350 | break; | |
3351 | case OFPGT11_SELECT: | |
fe7e5749 | 3352 | xlate_select_group(ctx, group); |
f4fb341b | 3353 | break; |
dd8cd4b4 SH |
3354 | case OFPGT11_FF: |
3355 | xlate_ff_group(ctx, group); | |
3356 | break; | |
f4fb341b | 3357 | default: |
428b2edd | 3358 | OVS_NOT_REACHED(); |
f4fb341b | 3359 | } |
809c7548 | 3360 | group_dpif_unref(group); |
5a070238 | 3361 | |
0eb48fe1 | 3362 | ctx->in_group = was_in_group; |
f4fb341b SH |
3363 | } |
3364 | ||
3365 | static bool | |
3366 | xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id) | |
3367 | { | |
0eb48fe1 | 3368 | if (xlate_resubmit_resource_check(ctx)) { |
f4fb341b SH |
3369 | struct group_dpif *group; |
3370 | bool got_group; | |
3371 | ||
3372 | got_group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group); | |
3373 | if (got_group) { | |
3374 | xlate_group_action__(ctx, group); | |
3375 | } else { | |
3376 | return true; | |
3377 | } | |
3378 | } | |
3379 | ||
3380 | return false; | |
3381 | } | |
3382 | ||
9583bc14 EJ |
3383 | static void |
3384 | xlate_ofpact_resubmit(struct xlate_ctx *ctx, | |
3385 | const struct ofpact_resubmit *resubmit) | |
3386 | { | |
4e022ec0 | 3387 | ofp_port_t in_port; |
9583bc14 | 3388 | uint8_t table_id; |
adcf00ba AZ |
3389 | bool may_packet_in = false; |
3390 | bool honor_table_miss = false; | |
3391 | ||
3392 | if (ctx->rule && rule_dpif_is_internal(ctx->rule)) { | |
3393 | /* Still allow missed packets to be sent to the controller | |
3394 | * if resubmitting from an internal table. */ | |
3395 | may_packet_in = true; | |
3396 | honor_table_miss = true; | |
3397 | } | |
9583bc14 EJ |
3398 | |
3399 | in_port = resubmit->in_port; | |
3400 | if (in_port == OFPP_IN_PORT) { | |
4e022ec0 | 3401 | in_port = ctx->xin->flow.in_port.ofp_port; |
9583bc14 EJ |
3402 | } |
3403 | ||
3404 | table_id = resubmit->table_id; | |
3405 | if (table_id == 255) { | |
3406 | table_id = ctx->table_id; | |
3407 | } | |
3408 | ||
adcf00ba AZ |
3409 | xlate_table_action(ctx, in_port, table_id, may_packet_in, |
3410 | honor_table_miss); | |
9583bc14 EJ |
3411 | } |
3412 | ||
3413 | static void | |
3414 | flood_packets(struct xlate_ctx *ctx, bool all) | |
3415 | { | |
46c88433 | 3416 | const struct xport *xport; |
9583bc14 | 3417 | |
46c88433 EJ |
3418 | HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) { |
3419 | if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3420 | continue; |
3421 | } | |
3422 | ||
3423 | if (all) { | |
e93ef1c7 | 3424 | compose_output_action__(ctx, xport->ofp_port, NULL, false); |
46c88433 | 3425 | } else if (!(xport->config & OFPUTIL_PC_NO_FLOOD)) { |
e93ef1c7 | 3426 | compose_output_action(ctx, xport->ofp_port, NULL); |
9583bc14 EJ |
3427 | } |
3428 | } | |
3429 | ||
3430 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
3431 | } | |
3432 | ||
3433 | static void | |
3434 | execute_controller_action(struct xlate_ctx *ctx, int len, | |
3435 | enum ofp_packet_in_reason reason, | |
3436 | uint16_t controller_id) | |
3437 | { | |
0fb7792a | 3438 | struct ofproto_packet_in *pin; |
e14deea0 | 3439 | struct dp_packet *packet; |
b440dd8c | 3440 | bool use_masked; |
9583bc14 | 3441 | |
04594cd5 | 3442 | ctx->xout->slow |= SLOW_CONTROLLER; |
9583bc14 EJ |
3443 | if (!ctx->xin->packet) { |
3444 | return; | |
3445 | } | |
3446 | ||
cf62fa4c | 3447 | packet = dp_packet_clone(ctx->xin->packet); |
9583bc14 | 3448 | |
b440dd8c | 3449 | use_masked = ctx->xbridge->support.masked_set_action; |
7fd91025 | 3450 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, |
1520ef4f | 3451 | ctx->odp_actions, |
49a73e0c | 3452 | ctx->wc, use_masked); |
9583bc14 | 3453 | |
41ccaa24 | 3454 | odp_execute_actions(NULL, &packet, 1, false, |
1520ef4f | 3455 | ctx->odp_actions->data, ctx->odp_actions->size, NULL); |
9583bc14 | 3456 | |
ada3a58d | 3457 | pin = xmalloc(sizeof *pin); |
cf62fa4c PS |
3458 | pin->up.packet_len = dp_packet_size(packet); |
3459 | pin->up.packet = dp_packet_steal_data(packet); | |
0fb7792a | 3460 | pin->up.reason = reason; |
0fb7792a | 3461 | pin->up.table_id = ctx->table_id; |
8b1e5560 | 3462 | pin->up.cookie = ctx->rule_cookie; |
0fb7792a | 3463 | |
50dcbd8e | 3464 | flow_get_metadata(&ctx->xin->flow, &pin->up.flow_metadata); |
9583bc14 | 3465 | |
f11c7538 | 3466 | pin->controller_id = controller_id; |
d38a3c7b | 3467 | pin->send_len = len; |
32260212 SH |
3468 | /* If a rule is a table-miss rule then this is |
3469 | * a table-miss handled by a table-miss rule. | |
3470 | * | |
3471 | * Else, if rule is internal and has a controller action, | |
3472 | * the later being implied by the rule being processed here, | |
3473 | * then this is a table-miss handled without a table-miss rule. | |
3474 | * | |
3475 | * Otherwise this is not a table-miss. */ | |
3476 | pin->miss_type = OFPROTO_PACKET_IN_NO_MISS; | |
3477 | if (ctx->rule) { | |
3478 | if (rule_dpif_is_table_miss(ctx->rule)) { | |
3479 | pin->miss_type = OFPROTO_PACKET_IN_MISS_FLOW; | |
3480 | } else if (rule_dpif_is_internal(ctx->rule)) { | |
3481 | pin->miss_type = OFPROTO_PACKET_IN_MISS_WITHOUT_FLOW; | |
3482 | } | |
3483 | } | |
ada3a58d | 3484 | ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, pin); |
e14deea0 | 3485 | dp_packet_delete(packet); |
9583bc14 EJ |
3486 | } |
3487 | ||
e672ff9b | 3488 | /* Called only when ctx->recirc_action_offset is set. */ |
7bbdd84f | 3489 | static void |
e672ff9b | 3490 | compose_recirculate_action(struct xlate_ctx *ctx) |
7bbdd84f | 3491 | { |
e672ff9b | 3492 | struct recirc_metadata md; |
b440dd8c | 3493 | bool use_masked; |
7bbdd84f | 3494 | uint32_t id; |
7bbdd84f | 3495 | |
b440dd8c | 3496 | use_masked = ctx->xbridge->support.masked_set_action; |
e672ff9b | 3497 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, |
1520ef4f | 3498 | ctx->odp_actions, |
49a73e0c | 3499 | ctx->wc, use_masked); |
7bbdd84f | 3500 | |
e672ff9b | 3501 | recirc_metadata_from_flow(&md, &ctx->xin->flow); |
7bbdd84f | 3502 | |
e672ff9b | 3503 | ovs_assert(ctx->recirc_action_offset >= 0); |
7bbdd84f | 3504 | |
e672ff9b JR |
3505 | /* Only allocate recirculation ID if we have a packet. */ |
3506 | if (ctx->xin->packet) { | |
3507 | /* Allocate a unique recirc id for the given metadata state in the | |
3508 | * flow. The life-cycle of this recirc id is managed by associating it | |
3509 | * with the udpif key ('ukey') created for each new datapath flow. */ | |
3510 | id = recirc_alloc_id_ctx(ctx->xbridge->ofproto, 0, &md, &ctx->stack, | |
3511 | ctx->recirc_action_offset, | |
3512 | ctx->action_set.size, ctx->action_set.data); | |
3513 | if (!id) { | |
3514 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3515 | VLOG_ERR_RL(&rl, "Failed to allocate recirculation id"); | |
3516 | return; | |
e2676b44 | 3517 | } |
e672ff9b JR |
3518 | xlate_out_add_recirc(ctx->xout, id); |
3519 | } else { | |
3520 | /* Look up an existing recirc id for the given metadata state in the | |
3521 | * flow. No new reference is taken, as the ID is RCU protected and is | |
3522 | * only required temporarily for verification. */ | |
3523 | id = recirc_find_id(ctx->xbridge->ofproto, 0, &md, &ctx->stack, | |
3524 | ctx->recirc_action_offset, | |
3525 | ctx->action_set.size, ctx->action_set.data); | |
3526 | /* We let zero 'id' to be used in the RECIRC action below, which will | |
3527 | * fail all revalidations as zero is not a valid recirculation ID. */ | |
7bbdd84f SH |
3528 | } |
3529 | ||
1520ef4f | 3530 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, id); |
e672ff9b JR |
3531 | |
3532 | /* Undo changes done by recirculation. */ | |
3533 | ctx->action_set.size = ctx->recirc_action_offset; | |
3534 | ctx->recirc_action_offset = -1; | |
3535 | ctx->last_unroll_offset = -1; | |
7bbdd84f SH |
3536 | } |
3537 | ||
8bfd0fda BP |
3538 | static void |
3539 | compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) | |
9583bc14 | 3540 | { |
33bf9176 | 3541 | struct flow *flow = &ctx->xin->flow; |
8bfd0fda | 3542 | int n; |
33bf9176 | 3543 | |
8bfd0fda | 3544 | ovs_assert(eth_type_mpls(mpls->ethertype)); |
b0a17866 | 3545 | |
49a73e0c | 3546 | n = flow_count_mpls_labels(flow, ctx->wc); |
8bfd0fda | 3547 | if (!n) { |
b440dd8c JS |
3548 | bool use_masked = ctx->xbridge->support.masked_set_action; |
3549 | ||
8bfd0fda | 3550 | ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow, |
1520ef4f | 3551 | ctx->odp_actions, |
49a73e0c | 3552 | ctx->wc, use_masked); |
8bfd0fda BP |
3553 | } else if (n >= FLOW_MAX_MPLS_LABELS) { |
3554 | if (ctx->xin->packet != NULL) { | |
3555 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3556 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an " | |
3557 | "MPLS push action can't be performed as it would " | |
3558 | "have more MPLS LSEs than the %d supported.", | |
3559 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
9583bc14 | 3560 | } |
8bfd0fda BP |
3561 | ctx->exit = true; |
3562 | return; | |
9583bc14 | 3563 | } |
b0a17866 | 3564 | |
49a73e0c | 3565 | flow_push_mpls(flow, n, mpls->ethertype, ctx->wc); |
9583bc14 EJ |
3566 | } |
3567 | ||
8bfd0fda | 3568 | static void |
9cfef3d0 | 3569 | compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) |
9583bc14 | 3570 | { |
8bfd0fda | 3571 | struct flow *flow = &ctx->xin->flow; |
49a73e0c | 3572 | int n = flow_count_mpls_labels(flow, ctx->wc); |
33bf9176 | 3573 | |
49a73e0c | 3574 | if (flow_pop_mpls(flow, n, eth_type, ctx->wc)) { |
2494ccd7 | 3575 | if (ctx->xbridge->support.odp.recirc) { |
7bbdd84f SH |
3576 | ctx->was_mpls = true; |
3577 | } | |
3578 | } else if (n >= FLOW_MAX_MPLS_LABELS) { | |
8bfd0fda BP |
3579 | if (ctx->xin->packet != NULL) { |
3580 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3581 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an " | |
3582 | "MPLS pop action can't be performed as it has " | |
3583 | "more MPLS LSEs than the %d supported.", | |
3584 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
3585 | } | |
3586 | ctx->exit = true; | |
1520ef4f | 3587 | ofpbuf_clear(ctx->odp_actions); |
9583bc14 EJ |
3588 | } |
3589 | } | |
3590 | ||
3591 | static bool | |
3592 | compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids) | |
3593 | { | |
33bf9176 BP |
3594 | struct flow *flow = &ctx->xin->flow; |
3595 | ||
3596 | if (!is_ip_any(flow)) { | |
9583bc14 EJ |
3597 | return false; |
3598 | } | |
3599 | ||
49a73e0c | 3600 | ctx->wc->masks.nw_ttl = 0xff; |
33bf9176 BP |
3601 | if (flow->nw_ttl > 1) { |
3602 | flow->nw_ttl--; | |
9583bc14 EJ |
3603 | return false; |
3604 | } else { | |
3605 | size_t i; | |
3606 | ||
3607 | for (i = 0; i < ids->n_controllers; i++) { | |
3608 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, | |
3609 | ids->cnt_ids[i]); | |
3610 | } | |
3611 | ||
3612 | /* Stop processing for current table. */ | |
3613 | return true; | |
3614 | } | |
3615 | } | |
3616 | ||
8bfd0fda | 3617 | static void |
097d4939 JR |
3618 | compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label) |
3619 | { | |
8bfd0fda | 3620 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3621 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK); |
8bfd0fda | 3622 | set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label); |
097d4939 | 3623 | } |
097d4939 JR |
3624 | } |
3625 | ||
8bfd0fda | 3626 | static void |
097d4939 JR |
3627 | compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc) |
3628 | { | |
8bfd0fda | 3629 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3630 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TC_MASK); |
8bfd0fda | 3631 | set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc); |
097d4939 | 3632 | } |
097d4939 JR |
3633 | } |
3634 | ||
8bfd0fda | 3635 | static void |
9cfef3d0 | 3636 | compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl) |
9583bc14 | 3637 | { |
8bfd0fda | 3638 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 3639 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda | 3640 | set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl); |
b0a17866 | 3641 | } |
9583bc14 EJ |
3642 | } |
3643 | ||
3644 | static bool | |
9cfef3d0 | 3645 | compose_dec_mpls_ttl_action(struct xlate_ctx *ctx) |
9583bc14 | 3646 | { |
33bf9176 | 3647 | struct flow *flow = &ctx->xin->flow; |
1dd35f8a | 3648 | |
8bfd0fda | 3649 | if (eth_type_mpls(flow->dl_type)) { |
22d38fca JR |
3650 | uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]); |
3651 | ||
49a73e0c | 3652 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda BP |
3653 | if (ttl > 1) { |
3654 | ttl--; | |
3655 | set_mpls_lse_ttl(&flow->mpls_lse[0], ttl); | |
3656 | return false; | |
3657 | } else { | |
3658 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0); | |
8bfd0fda | 3659 | } |
9583bc14 | 3660 | } |
22d38fca JR |
3661 | |
3662 | /* Stop processing for current table. */ | |
3663 | return true; | |
9583bc14 EJ |
3664 | } |
3665 | ||
3666 | static void | |
3667 | xlate_output_action(struct xlate_ctx *ctx, | |
4e022ec0 | 3668 | ofp_port_t port, uint16_t max_len, bool may_packet_in) |
9583bc14 | 3669 | { |
4e022ec0 | 3670 | ofp_port_t prev_nf_output_iface = ctx->xout->nf_output_iface; |
9583bc14 EJ |
3671 | |
3672 | ctx->xout->nf_output_iface = NF_OUT_DROP; | |
3673 | ||
3674 | switch (port) { | |
3675 | case OFPP_IN_PORT: | |
e93ef1c7 | 3676 | compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port, NULL); |
9583bc14 EJ |
3677 | break; |
3678 | case OFPP_TABLE: | |
4e022ec0 | 3679 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 3680 | 0, may_packet_in, true); |
9583bc14 EJ |
3681 | break; |
3682 | case OFPP_NORMAL: | |
3683 | xlate_normal(ctx); | |
3684 | break; | |
3685 | case OFPP_FLOOD: | |
3686 | flood_packets(ctx, false); | |
3687 | break; | |
3688 | case OFPP_ALL: | |
3689 | flood_packets(ctx, true); | |
3690 | break; | |
3691 | case OFPP_CONTROLLER: | |
3a11fd5b | 3692 | execute_controller_action(ctx, max_len, |
029ca940 SS |
3693 | (ctx->in_group ? OFPR_GROUP |
3694 | : ctx->in_action_set ? OFPR_ACTION_SET | |
3695 | : OFPR_ACTION), | |
3696 | 0); | |
9583bc14 EJ |
3697 | break; |
3698 | case OFPP_NONE: | |
3699 | break; | |
3700 | case OFPP_LOCAL: | |
3701 | default: | |
4e022ec0 | 3702 | if (port != ctx->xin->flow.in_port.ofp_port) { |
e93ef1c7 | 3703 | compose_output_action(ctx, port, NULL); |
9583bc14 EJ |
3704 | } else { |
3705 | xlate_report(ctx, "skipping output to input port"); | |
3706 | } | |
3707 | break; | |
3708 | } | |
3709 | ||
3710 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
3711 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
3712 | } else if (ctx->xout->nf_output_iface == NF_OUT_DROP) { | |
3713 | ctx->xout->nf_output_iface = prev_nf_output_iface; | |
3714 | } else if (prev_nf_output_iface != NF_OUT_DROP && | |
3715 | ctx->xout->nf_output_iface != NF_OUT_FLOOD) { | |
3716 | ctx->xout->nf_output_iface = NF_OUT_MULTI; | |
3717 | } | |
3718 | } | |
3719 | ||
3720 | static void | |
3721 | xlate_output_reg_action(struct xlate_ctx *ctx, | |
3722 | const struct ofpact_output_reg *or) | |
3723 | { | |
3724 | uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow); | |
3725 | if (port <= UINT16_MAX) { | |
3726 | union mf_subvalue value; | |
3727 | ||
3728 | memset(&value, 0xff, sizeof value); | |
49a73e0c | 3729 | mf_write_subfield_flow(&or->src, &value, &ctx->wc->masks); |
4e022ec0 AW |
3730 | xlate_output_action(ctx, u16_to_ofp(port), |
3731 | or->max_len, false); | |
9583bc14 EJ |
3732 | } |
3733 | } | |
3734 | ||
3735 | static void | |
3736 | xlate_enqueue_action(struct xlate_ctx *ctx, | |
3737 | const struct ofpact_enqueue *enqueue) | |
3738 | { | |
4e022ec0 | 3739 | ofp_port_t ofp_port = enqueue->port; |
9583bc14 EJ |
3740 | uint32_t queue_id = enqueue->queue; |
3741 | uint32_t flow_priority, priority; | |
3742 | int error; | |
3743 | ||
3744 | /* Translate queue to priority. */ | |
89a8a7f0 | 3745 | error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority); |
9583bc14 EJ |
3746 | if (error) { |
3747 | /* Fall back to ordinary output action. */ | |
3748 | xlate_output_action(ctx, enqueue->port, 0, false); | |
3749 | return; | |
3750 | } | |
3751 | ||
3752 | /* Check output port. */ | |
3753 | if (ofp_port == OFPP_IN_PORT) { | |
4e022ec0 AW |
3754 | ofp_port = ctx->xin->flow.in_port.ofp_port; |
3755 | } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3756 | return; |
3757 | } | |
3758 | ||
3759 | /* Add datapath actions. */ | |
3760 | flow_priority = ctx->xin->flow.skb_priority; | |
3761 | ctx->xin->flow.skb_priority = priority; | |
e93ef1c7 | 3762 | compose_output_action(ctx, ofp_port, NULL); |
9583bc14 EJ |
3763 | ctx->xin->flow.skb_priority = flow_priority; |
3764 | ||
3765 | /* Update NetFlow output port. */ | |
3766 | if (ctx->xout->nf_output_iface == NF_OUT_DROP) { | |
3767 | ctx->xout->nf_output_iface = ofp_port; | |
3768 | } else if (ctx->xout->nf_output_iface != NF_OUT_FLOOD) { | |
3769 | ctx->xout->nf_output_iface = NF_OUT_MULTI; | |
3770 | } | |
3771 | } | |
3772 | ||
3773 | static void | |
3774 | xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id) | |
3775 | { | |
3776 | uint32_t skb_priority; | |
3777 | ||
89a8a7f0 | 3778 | if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) { |
9583bc14 EJ |
3779 | ctx->xin->flow.skb_priority = skb_priority; |
3780 | } else { | |
3781 | /* Couldn't translate queue to a priority. Nothing to do. A warning | |
3782 | * has already been logged. */ | |
3783 | } | |
3784 | } | |
3785 | ||
3786 | static bool | |
46c88433 | 3787 | slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_) |
9583bc14 | 3788 | { |
46c88433 EJ |
3789 | const struct xbridge *xbridge = xbridge_; |
3790 | struct xport *port; | |
9583bc14 EJ |
3791 | |
3792 | switch (ofp_port) { | |
3793 | case OFPP_IN_PORT: | |
3794 | case OFPP_TABLE: | |
3795 | case OFPP_NORMAL: | |
3796 | case OFPP_FLOOD: | |
3797 | case OFPP_ALL: | |
3798 | case OFPP_NONE: | |
3799 | return true; | |
3800 | case OFPP_CONTROLLER: /* Not supported by the bundle action. */ | |
3801 | return false; | |
3802 | default: | |
46c88433 | 3803 | port = get_ofp_port(xbridge, ofp_port); |
9583bc14 EJ |
3804 | return port ? port->may_enable : false; |
3805 | } | |
3806 | } | |
3807 | ||
3808 | static void | |
3809 | xlate_bundle_action(struct xlate_ctx *ctx, | |
3810 | const struct ofpact_bundle *bundle) | |
3811 | { | |
4e022ec0 | 3812 | ofp_port_t port; |
9583bc14 | 3813 | |
49a73e0c | 3814 | port = bundle_execute(bundle, &ctx->xin->flow, ctx->wc, slave_enabled_cb, |
46c88433 | 3815 | CONST_CAST(struct xbridge *, ctx->xbridge)); |
9583bc14 | 3816 | if (bundle->dst.field) { |
49a73e0c | 3817 | nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow, ctx->wc); |
9583bc14 EJ |
3818 | } else { |
3819 | xlate_output_action(ctx, port, 0, false); | |
3820 | } | |
3821 | } | |
3822 | ||
3823 | static void | |
4165b5e0 JS |
3824 | xlate_learn_action__(struct xlate_ctx *ctx, const struct ofpact_learn *learn, |
3825 | struct ofputil_flow_mod *fm, struct ofpbuf *ofpacts) | |
9583bc14 | 3826 | { |
4165b5e0 JS |
3827 | learn_execute(learn, &ctx->xin->flow, fm, ofpacts); |
3828 | if (ctx->xin->may_learn) { | |
3829 | ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm); | |
3830 | } | |
3831 | } | |
9583bc14 | 3832 | |
4165b5e0 JS |
3833 | static void |
3834 | xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) | |
3835 | { | |
49a73e0c | 3836 | learn_mask(learn, ctx->wc); |
9583bc14 | 3837 | |
b256dc52 JS |
3838 | if (ctx->xin->xcache) { |
3839 | struct xc_entry *entry; | |
3840 | ||
3841 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN); | |
4165b5e0 JS |
3842 | entry->u.learn.ofproto = ctx->xbridge->ofproto; |
3843 | entry->u.learn.fm = xmalloc(sizeof *entry->u.learn.fm); | |
3844 | entry->u.learn.ofpacts = ofpbuf_new(64); | |
3845 | xlate_learn_action__(ctx, learn, entry->u.learn.fm, | |
3846 | entry->u.learn.ofpacts); | |
3847 | } else if (ctx->xin->may_learn) { | |
3848 | uint64_t ofpacts_stub[1024 / 8]; | |
3849 | struct ofputil_flow_mod fm; | |
3850 | struct ofpbuf ofpacts; | |
3851 | ||
3852 | ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); | |
3853 | xlate_learn_action__(ctx, learn, &fm, &ofpacts); | |
3854 | ofpbuf_uninit(&ofpacts); | |
b256dc52 JS |
3855 | } |
3856 | } | |
3857 | ||
3858 | static void | |
3859 | xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags, | |
3860 | uint16_t idle_timeout, uint16_t hard_timeout) | |
3861 | { | |
3862 | if (tcp_flags & (TCP_FIN | TCP_RST)) { | |
3863 | rule_dpif_reduce_timeouts(rule, idle_timeout, hard_timeout); | |
3864 | } | |
9583bc14 EJ |
3865 | } |
3866 | ||
9583bc14 EJ |
3867 | static void |
3868 | xlate_fin_timeout(struct xlate_ctx *ctx, | |
3869 | const struct ofpact_fin_timeout *oft) | |
3870 | { | |
b256dc52 JS |
3871 | if (ctx->rule) { |
3872 | xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags, | |
3873 | oft->fin_idle_timeout, oft->fin_hard_timeout); | |
3874 | if (ctx->xin->xcache) { | |
3875 | struct xc_entry *entry; | |
3876 | ||
3877 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT); | |
83709dfa JR |
3878 | /* XC_RULE already holds a reference on the rule, none is taken |
3879 | * here. */ | |
b256dc52 JS |
3880 | entry->u.fin.rule = ctx->rule; |
3881 | entry->u.fin.idle = oft->fin_idle_timeout; | |
3882 | entry->u.fin.hard = oft->fin_hard_timeout; | |
b256dc52 | 3883 | } |
9583bc14 EJ |
3884 | } |
3885 | } | |
3886 | ||
3887 | static void | |
3888 | xlate_sample_action(struct xlate_ctx *ctx, | |
3889 | const struct ofpact_sample *os) | |
3890 | { | |
e824d78d JR |
3891 | /* Scale the probability from 16-bit to 32-bit while representing |
3892 | * the same percentage. */ | |
3893 | uint32_t probability = (os->probability << 16) | os->probability; | |
b440dd8c | 3894 | bool use_masked; |
e824d78d | 3895 | |
b440dd8c | 3896 | if (!ctx->xbridge->support.variable_length_userdata) { |
e824d78d JR |
3897 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); |
3898 | ||
3899 | VLOG_ERR_RL(&rl, "ignoring NXAST_SAMPLE action because datapath " | |
3900 | "lacks support (needs Linux 3.10+ or kernel module from " | |
3901 | "OVS 1.11+)"); | |
3902 | return; | |
3903 | } | |
3904 | ||
b440dd8c | 3905 | use_masked = ctx->xbridge->support.masked_set_action; |
e824d78d | 3906 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, |
1520ef4f | 3907 | ctx->odp_actions, |
49a73e0c | 3908 | ctx->wc, use_masked); |
e824d78d | 3909 | |
a6092018 BP |
3910 | union user_action_cookie cookie = { |
3911 | .flow_sample = { | |
3912 | .type = USER_ACTION_COOKIE_FLOW_SAMPLE, | |
3913 | .probability = os->probability, | |
3914 | .collector_set_id = os->collector_set_id, | |
3915 | .obs_domain_id = os->obs_domain_id, | |
3916 | .obs_point_id = os->obs_point_id, | |
3917 | } | |
3918 | }; | |
3919 | compose_sample_action(ctx, probability, &cookie, sizeof cookie.flow_sample, | |
3920 | ODPP_NONE, false); | |
9583bc14 EJ |
3921 | } |
3922 | ||
3923 | static bool | |
46c88433 | 3924 | may_receive(const struct xport *xport, struct xlate_ctx *ctx) |
9583bc14 | 3925 | { |
bbbca389 | 3926 | if (xport->config & (is_stp(&ctx->xin->flow) |
46c88433 EJ |
3927 | ? OFPUTIL_PC_NO_RECV_STP |
3928 | : OFPUTIL_PC_NO_RECV)) { | |
9583bc14 EJ |
3929 | return false; |
3930 | } | |
3931 | ||
3932 | /* Only drop packets here if both forwarding and learning are | |
3933 | * disabled. If just learning is enabled, we need to have | |
3934 | * OFPP_NORMAL and the learning action have a look at the packet | |
3935 | * before we can drop it. */ | |
9efd308e DV |
3936 | if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) || |
3937 | (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) { | |
9583bc14 EJ |
3938 | return false; |
3939 | } | |
3940 | ||
3941 | return true; | |
3942 | } | |
3943 | ||
7fdb60a7 SH |
3944 | static void |
3945 | xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact *a) | |
3946 | { | |
c61f3870 BP |
3947 | const struct ofpact_nest *on = ofpact_get_WRITE_ACTIONS(a); |
3948 | size_t on_len = ofpact_nest_get_action_len(on); | |
3949 | const struct ofpact *inner; | |
3950 | ||
3951 | /* Maintain actset_output depending on the contents of the action set: | |
3952 | * | |
3953 | * - OFPP_UNSET, if there is no "output" action. | |
3954 | * | |
3955 | * - The output port, if there is an "output" action and no "group" | |
3956 | * action. | |
3957 | * | |
3958 | * - OFPP_UNSET, if there is a "group" action. | |
3959 | */ | |
3960 | if (!ctx->action_set_has_group) { | |
3961 | OFPACT_FOR_EACH (inner, on->actions, on_len) { | |
3962 | if (inner->type == OFPACT_OUTPUT) { | |
3963 | ctx->xin->flow.actset_output = ofpact_get_OUTPUT(inner)->port; | |
3964 | } else if (inner->type == OFPACT_GROUP) { | |
3965 | ctx->xin->flow.actset_output = OFPP_UNSET; | |
3966 | ctx->action_set_has_group = true; | |
9055ca9a | 3967 | break; |
c61f3870 BP |
3968 | } |
3969 | } | |
3970 | } | |
3971 | ||
3972 | ofpbuf_put(&ctx->action_set, on->actions, on_len); | |
7fdb60a7 SH |
3973 | ofpact_pad(&ctx->action_set); |
3974 | } | |
3975 | ||
3976 | static void | |
3977 | xlate_action_set(struct xlate_ctx *ctx) | |
3978 | { | |
3979 | uint64_t action_list_stub[1024 / 64]; | |
3980 | struct ofpbuf action_list; | |
3981 | ||
029ca940 | 3982 | ctx->in_action_set = true; |
7fdb60a7 SH |
3983 | ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub); |
3984 | ofpacts_execute_action_set(&action_list, &ctx->action_set); | |
ed9c9e3e JR |
3985 | /* Clear the action set, as it is not needed any more. */ |
3986 | ofpbuf_clear(&ctx->action_set); | |
6fd6ed71 | 3987 | do_xlate_actions(action_list.data, action_list.size, ctx); |
029ca940 | 3988 | ctx->in_action_set = false; |
7fdb60a7 SH |
3989 | ofpbuf_uninit(&action_list); |
3990 | } | |
3991 | ||
e672ff9b JR |
3992 | static void |
3993 | recirc_put_unroll_xlate(struct xlate_ctx *ctx) | |
3994 | { | |
3995 | struct ofpact_unroll_xlate *unroll; | |
3996 | ||
3997 | unroll = ctx->last_unroll_offset < 0 | |
3998 | ? NULL | |
3999 | : ALIGNED_CAST(struct ofpact_unroll_xlate *, | |
4000 | (char *)ctx->action_set.data + ctx->last_unroll_offset); | |
4001 | ||
4002 | /* Restore the table_id and rule cookie for a potential PACKET | |
4003 | * IN if needed. */ | |
4004 | if (!unroll || | |
4005 | (ctx->table_id != unroll->rule_table_id | |
4006 | || ctx->rule_cookie != unroll->rule_cookie)) { | |
4007 | ||
4008 | ctx->last_unroll_offset = ctx->action_set.size; | |
4009 | unroll = ofpact_put_UNROLL_XLATE(&ctx->action_set); | |
4010 | unroll->rule_table_id = ctx->table_id; | |
4011 | unroll->rule_cookie = ctx->rule_cookie; | |
4012 | } | |
4013 | } | |
4014 | ||
4015 | ||
4016 | /* Copy remaining actions to the action_set to be executed after recirculation. | |
4017 | * UNROLL_XLATE action is inserted, if not already done so, before actions that | |
4018 | * may generate PACKET_INs from the current table and without matching another | |
4019 | * rule. */ | |
4020 | static void | |
4021 | recirc_unroll_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
4022 | struct xlate_ctx *ctx) | |
4023 | { | |
4024 | const struct ofpact *a; | |
4025 | ||
4026 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { | |
4027 | switch (a->type) { | |
4028 | /* May generate PACKET INs. */ | |
4029 | case OFPACT_OUTPUT_REG: | |
4030 | case OFPACT_GROUP: | |
4031 | case OFPACT_OUTPUT: | |
4032 | case OFPACT_CONTROLLER: | |
4033 | case OFPACT_DEC_MPLS_TTL: | |
4034 | case OFPACT_DEC_TTL: | |
4035 | recirc_put_unroll_xlate(ctx); | |
4036 | break; | |
4037 | ||
4038 | /* These may not generate PACKET INs. */ | |
4039 | case OFPACT_SET_TUNNEL: | |
4040 | case OFPACT_REG_MOVE: | |
4041 | case OFPACT_SET_FIELD: | |
4042 | case OFPACT_STACK_PUSH: | |
4043 | case OFPACT_STACK_POP: | |
4044 | case OFPACT_LEARN: | |
4045 | case OFPACT_WRITE_METADATA: | |
4046 | case OFPACT_RESUBMIT: /* May indirectly generate PACKET INs, */ | |
4047 | case OFPACT_GOTO_TABLE: /* but from a different table and rule. */ | |
4048 | case OFPACT_ENQUEUE: | |
4049 | case OFPACT_SET_VLAN_VID: | |
4050 | case OFPACT_SET_VLAN_PCP: | |
4051 | case OFPACT_STRIP_VLAN: | |
4052 | case OFPACT_PUSH_VLAN: | |
4053 | case OFPACT_SET_ETH_SRC: | |
4054 | case OFPACT_SET_ETH_DST: | |
4055 | case OFPACT_SET_IPV4_SRC: | |
4056 | case OFPACT_SET_IPV4_DST: | |
4057 | case OFPACT_SET_IP_DSCP: | |
4058 | case OFPACT_SET_IP_ECN: | |
4059 | case OFPACT_SET_IP_TTL: | |
4060 | case OFPACT_SET_L4_SRC_PORT: | |
4061 | case OFPACT_SET_L4_DST_PORT: | |
4062 | case OFPACT_SET_QUEUE: | |
4063 | case OFPACT_POP_QUEUE: | |
4064 | case OFPACT_PUSH_MPLS: | |
4065 | case OFPACT_POP_MPLS: | |
4066 | case OFPACT_SET_MPLS_LABEL: | |
4067 | case OFPACT_SET_MPLS_TC: | |
4068 | case OFPACT_SET_MPLS_TTL: | |
4069 | case OFPACT_MULTIPATH: | |
4070 | case OFPACT_BUNDLE: | |
4071 | case OFPACT_EXIT: | |
4072 | case OFPACT_UNROLL_XLATE: | |
4073 | case OFPACT_FIN_TIMEOUT: | |
4074 | case OFPACT_CLEAR_ACTIONS: | |
4075 | case OFPACT_WRITE_ACTIONS: | |
4076 | case OFPACT_METER: | |
4077 | case OFPACT_SAMPLE: | |
4078 | break; | |
4079 | ||
4080 | /* These need not be copied for restoration. */ | |
4081 | case OFPACT_NOTE: | |
4082 | case OFPACT_CONJUNCTION: | |
4083 | continue; | |
4084 | } | |
4085 | /* Copy the action over. */ | |
4086 | ofpbuf_put(&ctx->action_set, a, OFPACT_ALIGN(a->len)); | |
4087 | } | |
4088 | } | |
4089 | ||
1d741d6d JR |
4090 | #define CHECK_MPLS_RECIRCULATION() \ |
4091 | if (ctx->was_mpls) { \ | |
4092 | ctx_trigger_recirculation(ctx); \ | |
4093 | break; \ | |
4094 | } | |
4095 | #define CHECK_MPLS_RECIRCULATION_IF(COND) \ | |
4096 | if (COND) { \ | |
4097 | CHECK_MPLS_RECIRCULATION(); \ | |
4098 | } | |
4099 | ||
9583bc14 EJ |
4100 | static void |
4101 | do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
4102 | struct xlate_ctx *ctx) | |
4103 | { | |
49a73e0c | 4104 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 4105 | struct flow *flow = &ctx->xin->flow; |
9583bc14 EJ |
4106 | const struct ofpact *a; |
4107 | ||
a36de779 PS |
4108 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
4109 | tnl_arp_snoop(flow, wc, ctx->xbridge->name); | |
4110 | } | |
f47ea021 JR |
4111 | /* dl_type already in the mask, not set below. */ |
4112 | ||
9583bc14 EJ |
4113 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
4114 | struct ofpact_controller *controller; | |
4115 | const struct ofpact_metadata *metadata; | |
b2dd70be JR |
4116 | const struct ofpact_set_field *set_field; |
4117 | const struct mf_field *mf; | |
9583bc14 | 4118 | |
e672ff9b JR |
4119 | if (ctx->exit) { |
4120 | /* Check if need to store the remaining actions for later | |
4121 | * execution. */ | |
4122 | if (exit_recirculates(ctx)) { | |
4123 | recirc_unroll_actions(a, OFPACT_ALIGN(ofpacts_len - | |
4124 | ((uint8_t *)a - | |
4125 | (uint8_t *)ofpacts)), | |
4126 | ctx); | |
4127 | } | |
4128 | break; | |
7bbdd84f SH |
4129 | } |
4130 | ||
9583bc14 EJ |
4131 | switch (a->type) { |
4132 | case OFPACT_OUTPUT: | |
4133 | xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port, | |
4134 | ofpact_get_OUTPUT(a)->max_len, true); | |
4135 | break; | |
4136 | ||
7395c052 | 4137 | case OFPACT_GROUP: |
f4fb341b | 4138 | if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id)) { |
1d741d6d | 4139 | /* Group could not be found. */ |
f4fb341b SH |
4140 | return; |
4141 | } | |
7395c052 NZ |
4142 | break; |
4143 | ||
9583bc14 EJ |
4144 | case OFPACT_CONTROLLER: |
4145 | controller = ofpact_get_CONTROLLER(a); | |
4146 | execute_controller_action(ctx, controller->max_len, | |
4147 | controller->reason, | |
4148 | controller->controller_id); | |
4149 | break; | |
4150 | ||
4151 | case OFPACT_ENQUEUE: | |
16194afd DDP |
4152 | memset(&wc->masks.skb_priority, 0xff, |
4153 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4154 | xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a)); |
4155 | break; | |
4156 | ||
4157 | case OFPACT_SET_VLAN_VID: | |
f74e7df7 | 4158 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
ca287d20 JR |
4159 | if (flow->vlan_tci & htons(VLAN_CFI) || |
4160 | ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) { | |
4161 | flow->vlan_tci &= ~htons(VLAN_VID_MASK); | |
4162 | flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid) | |
4163 | | htons(VLAN_CFI)); | |
4164 | } | |
9583bc14 EJ |
4165 | break; |
4166 | ||
4167 | case OFPACT_SET_VLAN_PCP: | |
f74e7df7 | 4168 | wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI); |
ca287d20 JR |
4169 | if (flow->vlan_tci & htons(VLAN_CFI) || |
4170 | ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) { | |
4171 | flow->vlan_tci &= ~htons(VLAN_PCP_MASK); | |
4172 | flow->vlan_tci |= htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp | |
4173 | << VLAN_PCP_SHIFT) | VLAN_CFI); | |
4174 | } | |
9583bc14 EJ |
4175 | break; |
4176 | ||
4177 | case OFPACT_STRIP_VLAN: | |
f74e7df7 | 4178 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 4179 | flow->vlan_tci = htons(0); |
9583bc14 EJ |
4180 | break; |
4181 | ||
4182 | case OFPACT_PUSH_VLAN: | |
4183 | /* XXX 802.1AD(QinQ) */ | |
f74e7df7 | 4184 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 4185 | flow->vlan_tci = htons(VLAN_CFI); |
9583bc14 EJ |
4186 | break; |
4187 | ||
4188 | case OFPACT_SET_ETH_SRC: | |
f74e7df7 | 4189 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
33bf9176 | 4190 | memcpy(flow->dl_src, ofpact_get_SET_ETH_SRC(a)->mac, ETH_ADDR_LEN); |
9583bc14 EJ |
4191 | break; |
4192 | ||
4193 | case OFPACT_SET_ETH_DST: | |
f74e7df7 | 4194 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
33bf9176 | 4195 | memcpy(flow->dl_dst, ofpact_get_SET_ETH_DST(a)->mac, ETH_ADDR_LEN); |
9583bc14 EJ |
4196 | break; |
4197 | ||
4198 | case OFPACT_SET_IPV4_SRC: | |
1d741d6d | 4199 | CHECK_MPLS_RECIRCULATION(); |
33bf9176 | 4200 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 4201 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); |
33bf9176 | 4202 | flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4; |
9583bc14 EJ |
4203 | } |
4204 | break; | |
4205 | ||
4206 | case OFPACT_SET_IPV4_DST: | |
1d741d6d | 4207 | CHECK_MPLS_RECIRCULATION(); |
33bf9176 | 4208 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 4209 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); |
33bf9176 | 4210 | flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4; |
9583bc14 EJ |
4211 | } |
4212 | break; | |
4213 | ||
04f01c24 | 4214 | case OFPACT_SET_IP_DSCP: |
1d741d6d | 4215 | CHECK_MPLS_RECIRCULATION(); |
04f01c24 | 4216 | if (is_ip_any(flow)) { |
f47ea021 | 4217 | wc->masks.nw_tos |= IP_DSCP_MASK; |
33bf9176 | 4218 | flow->nw_tos &= ~IP_DSCP_MASK; |
04f01c24 | 4219 | flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp; |
9583bc14 EJ |
4220 | } |
4221 | break; | |
4222 | ||
ff14eb7a | 4223 | case OFPACT_SET_IP_ECN: |
1d741d6d | 4224 | CHECK_MPLS_RECIRCULATION(); |
ff14eb7a JR |
4225 | if (is_ip_any(flow)) { |
4226 | wc->masks.nw_tos |= IP_ECN_MASK; | |
4227 | flow->nw_tos &= ~IP_ECN_MASK; | |
4228 | flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn; | |
4229 | } | |
4230 | break; | |
4231 | ||
0c20dbe4 | 4232 | case OFPACT_SET_IP_TTL: |
1d741d6d | 4233 | CHECK_MPLS_RECIRCULATION(); |
0c20dbe4 JR |
4234 | if (is_ip_any(flow)) { |
4235 | wc->masks.nw_ttl = 0xff; | |
4236 | flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl; | |
4237 | } | |
4238 | break; | |
4239 | ||
9583bc14 | 4240 | case OFPACT_SET_L4_SRC_PORT: |
1d741d6d | 4241 | CHECK_MPLS_RECIRCULATION(); |
b8778a0d | 4242 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
4243 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
4244 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
33bf9176 | 4245 | flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port); |
9583bc14 EJ |
4246 | } |
4247 | break; | |
4248 | ||
4249 | case OFPACT_SET_L4_DST_PORT: | |
1d741d6d | 4250 | CHECK_MPLS_RECIRCULATION(); |
b8778a0d | 4251 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
4252 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
4253 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
33bf9176 | 4254 | flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port); |
9583bc14 EJ |
4255 | } |
4256 | break; | |
4257 | ||
4258 | case OFPACT_RESUBMIT: | |
4259 | xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a)); | |
4260 | break; | |
4261 | ||
4262 | case OFPACT_SET_TUNNEL: | |
33bf9176 | 4263 | flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id); |
9583bc14 EJ |
4264 | break; |
4265 | ||
4266 | case OFPACT_SET_QUEUE: | |
16194afd DDP |
4267 | memset(&wc->masks.skb_priority, 0xff, |
4268 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4269 | xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id); |
4270 | break; | |
4271 | ||
4272 | case OFPACT_POP_QUEUE: | |
16194afd DDP |
4273 | memset(&wc->masks.skb_priority, 0xff, |
4274 | sizeof wc->masks.skb_priority); | |
33bf9176 | 4275 | flow->skb_priority = ctx->orig_skb_priority; |
9583bc14 EJ |
4276 | break; |
4277 | ||
4278 | case OFPACT_REG_MOVE: | |
1d741d6d JR |
4279 | CHECK_MPLS_RECIRCULATION_IF( |
4280 | mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->dst.field) || | |
4281 | mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->src.field)); | |
33bf9176 | 4282 | nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc); |
9583bc14 EJ |
4283 | break; |
4284 | ||
b2dd70be | 4285 | case OFPACT_SET_FIELD: |
1d741d6d JR |
4286 | CHECK_MPLS_RECIRCULATION_IF( |
4287 | mf_is_l3_or_higher(ofpact_get_SET_FIELD(a)->field)); | |
b2dd70be JR |
4288 | set_field = ofpact_get_SET_FIELD(a); |
4289 | mf = set_field->field; | |
b2dd70be JR |
4290 | |
4291 | /* Set field action only ever overwrites packet's outermost | |
4292 | * applicable header fields. Do nothing if no header exists. */ | |
4479846e JR |
4293 | if (mf->id == MFF_VLAN_VID) { |
4294 | wc->masks.vlan_tci |= htons(VLAN_CFI); | |
4295 | if (!(flow->vlan_tci & htons(VLAN_CFI))) { | |
4296 | break; | |
4297 | } | |
4298 | } else if ((mf->id == MFF_MPLS_LABEL || mf->id == MFF_MPLS_TC) | |
4299 | /* 'dl_type' is already unwildcarded. */ | |
4300 | && !eth_type_mpls(flow->dl_type)) { | |
4301 | break; | |
b2dd70be | 4302 | } |
b8778a0d JR |
4303 | /* A flow may wildcard nw_frag. Do nothing if setting a trasport |
4304 | * header field on a packet that does not have them. */ | |
4479846e | 4305 | mf_mask_field_and_prereqs(mf, &wc->masks); |
b8778a0d JR |
4306 | if (mf_are_prereqs_ok(mf, flow)) { |
4307 | mf_set_flow_value_masked(mf, &set_field->value, | |
4308 | &set_field->mask, flow); | |
4309 | } | |
b2dd70be JR |
4310 | break; |
4311 | ||
9583bc14 | 4312 | case OFPACT_STACK_PUSH: |
1d741d6d JR |
4313 | CHECK_MPLS_RECIRCULATION_IF( |
4314 | mf_is_l3_or_higher(ofpact_get_STACK_PUSH(a)->subfield.field)); | |
33bf9176 BP |
4315 | nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc, |
4316 | &ctx->stack); | |
9583bc14 EJ |
4317 | break; |
4318 | ||
4319 | case OFPACT_STACK_POP: | |
1d741d6d JR |
4320 | CHECK_MPLS_RECIRCULATION_IF( |
4321 | mf_is_l3_or_higher(ofpact_get_STACK_POP(a)->subfield.field)); | |
f74e7df7 JP |
4322 | nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc, |
4323 | &ctx->stack); | |
9583bc14 EJ |
4324 | break; |
4325 | ||
4326 | case OFPACT_PUSH_MPLS: | |
1d741d6d JR |
4327 | /* Recirculate if it is an IP packet with a zero ttl. This may |
4328 | * indicate that the packet was previously MPLS and an MPLS pop | |
4329 | * action converted it to IP. In this case recirculating should | |
4330 | * reveal the IP TTL which is used as the basis for a new MPLS | |
4331 | * LSE. */ | |
4332 | CHECK_MPLS_RECIRCULATION_IF( | |
4333 | !flow_count_mpls_labels(flow, wc) | |
4334 | && flow->nw_ttl == 0 | |
4335 | && is_ip_any(flow)); | |
8bfd0fda | 4336 | compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)); |
9583bc14 EJ |
4337 | break; |
4338 | ||
4339 | case OFPACT_POP_MPLS: | |
1d741d6d | 4340 | CHECK_MPLS_RECIRCULATION(); |
8bfd0fda | 4341 | compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype); |
9583bc14 EJ |
4342 | break; |
4343 | ||
097d4939 | 4344 | case OFPACT_SET_MPLS_LABEL: |
1d741d6d | 4345 | CHECK_MPLS_RECIRCULATION(); |
8bfd0fda BP |
4346 | compose_set_mpls_label_action( |
4347 | ctx, ofpact_get_SET_MPLS_LABEL(a)->label); | |
1d741d6d | 4348 | break; |
097d4939 JR |
4349 | |
4350 | case OFPACT_SET_MPLS_TC: | |
1d741d6d | 4351 | CHECK_MPLS_RECIRCULATION(); |
8bfd0fda | 4352 | compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc); |
097d4939 JR |
4353 | break; |
4354 | ||
9583bc14 | 4355 | case OFPACT_SET_MPLS_TTL: |
1d741d6d | 4356 | CHECK_MPLS_RECIRCULATION(); |
8bfd0fda | 4357 | compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl); |
9583bc14 EJ |
4358 | break; |
4359 | ||
4360 | case OFPACT_DEC_MPLS_TTL: | |
1d741d6d | 4361 | CHECK_MPLS_RECIRCULATION(); |
9cfef3d0 | 4362 | if (compose_dec_mpls_ttl_action(ctx)) { |
ad3efdcb | 4363 | return; |
9583bc14 EJ |
4364 | } |
4365 | break; | |
4366 | ||
4367 | case OFPACT_DEC_TTL: | |
1d741d6d | 4368 | CHECK_MPLS_RECIRCULATION(); |
f74e7df7 | 4369 | wc->masks.nw_ttl = 0xff; |
9583bc14 | 4370 | if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) { |
ad3efdcb | 4371 | return; |
9583bc14 EJ |
4372 | } |
4373 | break; | |
4374 | ||
4375 | case OFPACT_NOTE: | |
4376 | /* Nothing to do. */ | |
4377 | break; | |
4378 | ||
4379 | case OFPACT_MULTIPATH: | |
1d741d6d | 4380 | CHECK_MPLS_RECIRCULATION(); |
33bf9176 | 4381 | multipath_execute(ofpact_get_MULTIPATH(a), flow, wc); |
9583bc14 EJ |
4382 | break; |
4383 | ||
4384 | case OFPACT_BUNDLE: | |
1d741d6d | 4385 | CHECK_MPLS_RECIRCULATION(); |
9583bc14 EJ |
4386 | xlate_bundle_action(ctx, ofpact_get_BUNDLE(a)); |
4387 | break; | |
4388 | ||
4389 | case OFPACT_OUTPUT_REG: | |
4390 | xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a)); | |
4391 | break; | |
4392 | ||
4393 | case OFPACT_LEARN: | |
1d741d6d | 4394 | CHECK_MPLS_RECIRCULATION(); |
9583bc14 EJ |
4395 | xlate_learn_action(ctx, ofpact_get_LEARN(a)); |
4396 | break; | |
4397 | ||
afc3987b BP |
4398 | case OFPACT_CONJUNCTION: { |
4399 | /* A flow with a "conjunction" action represents part of a special | |
4400 | * kind of "set membership match". Such a flow should not actually | |
4401 | * get executed, but it could via, say, a "packet-out", even though | |
4402 | * that wouldn't be useful. Log it to help debugging. */ | |
4403 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4404 | VLOG_INFO_RL(&rl, "executing no-op conjunction action"); | |
18080541 | 4405 | break; |
afc3987b | 4406 | } |
18080541 | 4407 | |
9583bc14 EJ |
4408 | case OFPACT_EXIT: |
4409 | ctx->exit = true; | |
4410 | break; | |
4411 | ||
e672ff9b JR |
4412 | case OFPACT_UNROLL_XLATE: { |
4413 | struct ofpact_unroll_xlate *unroll = ofpact_get_UNROLL_XLATE(a); | |
4414 | ||
4415 | /* Restore translation context data that was stored earlier. */ | |
4416 | ctx->table_id = unroll->rule_table_id; | |
4417 | ctx->rule_cookie = unroll->rule_cookie; | |
4418 | break; | |
4419 | } | |
9583bc14 | 4420 | case OFPACT_FIN_TIMEOUT: |
1d741d6d | 4421 | CHECK_MPLS_RECIRCULATION(); |
33bf9176 | 4422 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
9583bc14 EJ |
4423 | xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a)); |
4424 | break; | |
4425 | ||
4426 | case OFPACT_CLEAR_ACTIONS: | |
7fdb60a7 | 4427 | ofpbuf_clear(&ctx->action_set); |
c61f3870 BP |
4428 | ctx->xin->flow.actset_output = OFPP_UNSET; |
4429 | ctx->action_set_has_group = false; | |
7fdb60a7 SH |
4430 | break; |
4431 | ||
4432 | case OFPACT_WRITE_ACTIONS: | |
4433 | xlate_write_actions(ctx, a); | |
9583bc14 EJ |
4434 | break; |
4435 | ||
4436 | case OFPACT_WRITE_METADATA: | |
4437 | metadata = ofpact_get_WRITE_METADATA(a); | |
33bf9176 BP |
4438 | flow->metadata &= ~metadata->mask; |
4439 | flow->metadata |= metadata->metadata & metadata->mask; | |
9583bc14 EJ |
4440 | break; |
4441 | ||
638a19b0 JR |
4442 | case OFPACT_METER: |
4443 | /* Not implemented yet. */ | |
4444 | break; | |
4445 | ||
9583bc14 | 4446 | case OFPACT_GOTO_TABLE: { |
9583bc14 | 4447 | struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a); |
9583bc14 | 4448 | |
7bbdd84f SH |
4449 | /* Allow ctx->table_id == TBL_INTERNAL, which will be greater |
4450 | * than ogt->table_id. This is to allow goto_table actions that | |
4451 | * triggered recirculation: ctx->table_id will be TBL_INTERNAL | |
4452 | * after recirculation. */ | |
4453 | ovs_assert(ctx->table_id == TBL_INTERNAL | |
4454 | || ctx->table_id < ogt->table_id); | |
4468099e | 4455 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 4456 | ogt->table_id, true, true); |
9583bc14 EJ |
4457 | break; |
4458 | } | |
4459 | ||
4460 | case OFPACT_SAMPLE: | |
4461 | xlate_sample_action(ctx, ofpact_get_SAMPLE(a)); | |
4462 | break; | |
4463 | } | |
1d741d6d JR |
4464 | |
4465 | /* Check if need to store this and the remaining actions for later | |
4466 | * execution. */ | |
4467 | if (ctx->exit && ctx_first_recirculation_action(ctx)) { | |
4468 | recirc_unroll_actions(a, OFPACT_ALIGN(ofpacts_len - | |
4469 | ((uint8_t *)a - | |
4470 | (uint8_t *)ofpacts)), | |
4471 | ctx); | |
4472 | break; | |
4473 | } | |
9583bc14 | 4474 | } |
9583bc14 EJ |
4475 | } |
4476 | ||
4477 | void | |
4478 | xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, | |
cc377352 EJ |
4479 | const struct flow *flow, ofp_port_t in_port, |
4480 | struct rule_dpif *rule, uint16_t tcp_flags, | |
1520ef4f BP |
4481 | const struct dp_packet *packet, struct flow_wildcards *wc, |
4482 | struct ofpbuf *odp_actions) | |
9583bc14 EJ |
4483 | { |
4484 | xin->ofproto = ofproto; | |
4485 | xin->flow = *flow; | |
cc377352 | 4486 | xin->flow.in_port.ofp_port = in_port; |
c61f3870 | 4487 | xin->flow.actset_output = OFPP_UNSET; |
9583bc14 EJ |
4488 | xin->packet = packet; |
4489 | xin->may_learn = packet != NULL; | |
4490 | xin->rule = rule; | |
b256dc52 | 4491 | xin->xcache = NULL; |
9583bc14 EJ |
4492 | xin->ofpacts = NULL; |
4493 | xin->ofpacts_len = 0; | |
4494 | xin->tcp_flags = tcp_flags; | |
4495 | xin->resubmit_hook = NULL; | |
4496 | xin->report_hook = NULL; | |
4497 | xin->resubmit_stats = NULL; | |
49a73e0c | 4498 | xin->wc = wc; |
1520ef4f | 4499 | xin->odp_actions = odp_actions; |
e672ff9b JR |
4500 | |
4501 | /* Do recirc lookup. */ | |
4502 | xin->recirc = flow->recirc_id | |
4503 | ? recirc_id_node_find(flow->recirc_id) | |
4504 | : NULL; | |
9583bc14 EJ |
4505 | } |
4506 | ||
4507 | void | |
4508 | xlate_out_uninit(struct xlate_out *xout) | |
4509 | { | |
e672ff9b | 4510 | if (xout) { |
e672ff9b | 4511 | xlate_out_free_recircs(xout); |
9583bc14 EJ |
4512 | } |
4513 | } | |
4514 | ||
4515 | /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts' | |
4516 | * into datapath actions, using 'ctx', and discards the datapath actions. */ | |
4517 | void | |
4518 | xlate_actions_for_side_effects(struct xlate_in *xin) | |
4519 | { | |
4520 | struct xlate_out xout; | |
4521 | ||
4522 | xlate_actions(xin, &xout); | |
4523 | xlate_out_uninit(&xout); | |
4524 | } | |
9583bc14 | 4525 | \f |
55954f6e EJ |
4526 | static struct skb_priority_to_dscp * |
4527 | get_skb_priority(const struct xport *xport, uint32_t skb_priority) | |
4528 | { | |
4529 | struct skb_priority_to_dscp *pdscp; | |
4530 | uint32_t hash; | |
4531 | ||
4532 | hash = hash_int(skb_priority, 0); | |
4533 | HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) { | |
4534 | if (pdscp->skb_priority == skb_priority) { | |
4535 | return pdscp; | |
4536 | } | |
4537 | } | |
4538 | return NULL; | |
4539 | } | |
4540 | ||
4541 | static bool | |
4542 | dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority, | |
4543 | uint8_t *dscp) | |
4544 | { | |
4545 | struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority); | |
4546 | *dscp = pdscp ? pdscp->dscp : 0; | |
4547 | return pdscp != NULL; | |
4548 | } | |
4549 | ||
16194afd DDP |
4550 | static size_t |
4551 | count_skb_priorities(const struct xport *xport) | |
4552 | { | |
4553 | return hmap_count(&xport->skb_priorities); | |
4554 | } | |
4555 | ||
55954f6e EJ |
4556 | static void |
4557 | clear_skb_priorities(struct xport *xport) | |
4558 | { | |
4559 | struct skb_priority_to_dscp *pdscp, *next; | |
4560 | ||
4561 | HMAP_FOR_EACH_SAFE (pdscp, next, hmap_node, &xport->skb_priorities) { | |
4562 | hmap_remove(&xport->skb_priorities, &pdscp->hmap_node); | |
4563 | free(pdscp); | |
4564 | } | |
4565 | } | |
4566 | ||
ce4a6b76 BP |
4567 | static bool |
4568 | actions_output_to_local_port(const struct xlate_ctx *ctx) | |
4569 | { | |
46c88433 | 4570 | odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL); |
ce4a6b76 BP |
4571 | const struct nlattr *a; |
4572 | unsigned int left; | |
4573 | ||
1520ef4f BP |
4574 | NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->odp_actions->data, |
4575 | ctx->odp_actions->size) { | |
ce4a6b76 BP |
4576 | if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT |
4577 | && nl_attr_get_odp_port(a) == local_odp_port) { | |
4578 | return true; | |
4579 | } | |
4580 | } | |
4581 | return false; | |
4582 | } | |
9583bc14 | 4583 | |
5e2a6702 | 4584 | #if defined(__linux__) |
7d031d7e BP |
4585 | /* Returns the maximum number of packets that the Linux kernel is willing to |
4586 | * queue up internally to certain kinds of software-implemented ports, or the | |
4587 | * default (and rarely modified) value if it cannot be determined. */ | |
4588 | static int | |
4589 | netdev_max_backlog(void) | |
4590 | { | |
4591 | static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; | |
4592 | static int max_backlog = 1000; /* The normal default value. */ | |
4593 | ||
4594 | if (ovsthread_once_start(&once)) { | |
4595 | static const char filename[] = "/proc/sys/net/core/netdev_max_backlog"; | |
4596 | FILE *stream; | |
4597 | int n; | |
4598 | ||
4599 | stream = fopen(filename, "r"); | |
4600 | if (!stream) { | |
4601 | VLOG_WARN("%s: open failed (%s)", filename, ovs_strerror(errno)); | |
4602 | } else { | |
4603 | if (fscanf(stream, "%d", &n) != 1) { | |
4604 | VLOG_WARN("%s: read error", filename); | |
4605 | } else if (n <= 100) { | |
4606 | VLOG_WARN("%s: unexpectedly small value %d", filename, n); | |
4607 | } else { | |
4608 | max_backlog = n; | |
4609 | } | |
4610 | fclose(stream); | |
4611 | } | |
4612 | ovsthread_once_done(&once); | |
4613 | ||
4614 | VLOG_DBG("%s: using %d max_backlog", filename, max_backlog); | |
4615 | } | |
4616 | ||
4617 | return max_backlog; | |
4618 | } | |
4619 | ||
4620 | /* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in | |
4621 | * 'odp_actions'. */ | |
4622 | static int | |
4623 | count_output_actions(const struct ofpbuf *odp_actions) | |
4624 | { | |
4625 | const struct nlattr *a; | |
4626 | size_t left; | |
4627 | int n = 0; | |
4628 | ||
6fd6ed71 | 4629 | NL_ATTR_FOR_EACH_UNSAFE (a, left, odp_actions->data, odp_actions->size) { |
7d031d7e BP |
4630 | if (a->nla_type == OVS_ACTION_ATTR_OUTPUT) { |
4631 | n++; | |
4632 | } | |
4633 | } | |
4634 | return n; | |
4635 | } | |
5e2a6702 | 4636 | #endif /* defined(__linux__) */ |
7d031d7e BP |
4637 | |
4638 | /* Returns true if 'odp_actions' contains more output actions than the datapath | |
4639 | * can reliably handle in one go. On Linux, this is the value of the | |
4640 | * net.core.netdev_max_backlog sysctl, which limits the maximum number of | |
4641 | * packets that the kernel is willing to queue up for processing while the | |
4642 | * datapath is processing a set of actions. */ | |
4643 | static bool | |
5e2a6702 | 4644 | too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) |
7d031d7e BP |
4645 | { |
4646 | #ifdef __linux__ | |
6fd6ed71 | 4647 | return (odp_actions->size / NL_A_U32_SIZE > netdev_max_backlog() |
7d031d7e BP |
4648 | && count_output_actions(odp_actions) > netdev_max_backlog()); |
4649 | #else | |
4650 | /* OSes other than Linux might have similar limits, but we don't know how | |
4651 | * to determine them.*/ | |
4652 | return false; | |
4653 | #endif | |
4654 | } | |
4655 | ||
234c3da9 BP |
4656 | static void |
4657 | xlate_wc_init(struct xlate_ctx *ctx) | |
4658 | { | |
4659 | flow_wildcards_init_catchall(ctx->wc); | |
4660 | ||
4661 | /* Some fields we consider to always be examined. */ | |
4662 | memset(&ctx->wc->masks.in_port, 0xff, sizeof ctx->wc->masks.in_port); | |
4663 | memset(&ctx->wc->masks.dl_type, 0xff, sizeof ctx->wc->masks.dl_type); | |
4664 | if (is_ip_any(&ctx->xin->flow)) { | |
4665 | ctx->wc->masks.nw_frag |= FLOW_NW_FRAG_MASK; | |
4666 | } | |
4667 | ||
4668 | if (ctx->xbridge->support.odp.recirc) { | |
4669 | /* Always exactly match recirc_id when datapath supports | |
4670 | * recirculation. */ | |
4671 | ctx->wc->masks.recirc_id = UINT32_MAX; | |
4672 | } | |
4673 | ||
4674 | if (ctx->xbridge->netflow) { | |
4675 | netflow_mask_wc(&ctx->xin->flow, ctx->wc); | |
4676 | } | |
4677 | ||
4678 | tnl_wc_init(&ctx->xin->flow, ctx->wc); | |
4679 | } | |
4680 | ||
4681 | static void | |
4682 | xlate_wc_finish(struct xlate_ctx *ctx) | |
4683 | { | |
4684 | /* Clear the metadata and register wildcard masks, because we won't | |
4685 | * use non-header fields as part of the cache. */ | |
4686 | flow_wildcards_clear_non_packet_fields(ctx->wc); | |
4687 | ||
4688 | /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow | |
4689 | * uses the low 8 bits of the 16-bit tp_src and tp_dst members to | |
4690 | * represent these fields. The datapath interface, on the other hand, | |
4691 | * represents them with just 8 bits each. This means that if the high | |
4692 | * 8 bits of the masks for these fields somehow become set, then they | |
4693 | * will get chopped off by a round trip through the datapath, and | |
4694 | * revalidation will spot that as an inconsistency and delete the flow. | |
4695 | * Avoid the problem here by making sure that only the low 8 bits of | |
4696 | * either field can be unwildcarded for ICMP. | |
4697 | */ | |
4698 | if (is_icmpv4(&ctx->xin->flow) || is_icmpv6(&ctx->xin->flow)) { | |
4699 | ctx->wc->masks.tp_src &= htons(UINT8_MAX); | |
4700 | ctx->wc->masks.tp_dst &= htons(UINT8_MAX); | |
4701 | } | |
4702 | /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */ | |
4703 | if (ctx->wc->masks.vlan_tci) { | |
4704 | ctx->wc->masks.vlan_tci |= htons(VLAN_CFI); | |
4705 | } | |
4706 | } | |
4707 | ||
e672ff9b JR |
4708 | /* Translates the flow, actions, or rule in 'xin' into datapath actions in |
4709 | * 'xout'. | |
56450a41 BP |
4710 | * The caller must take responsibility for eventually freeing 'xout', with |
4711 | * xlate_out_uninit(). */ | |
84f0f298 RW |
4712 | void |
4713 | xlate_actions(struct xlate_in *xin, struct xlate_out *xout) | |
9583bc14 | 4714 | { |
e467ea42 BP |
4715 | *xout = (struct xlate_out) { |
4716 | .slow = 0, | |
4717 | .fail_open = false, | |
e467ea42 | 4718 | .nf_output_iface = NF_OUT_DROP, |
e467ea42 BP |
4719 | .n_recircs = 0, |
4720 | }; | |
4721 | ||
84f0f298 | 4722 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
bb00fdef BP |
4723 | struct xbridge *xbridge = xbridge_lookup(xcfg, xin->ofproto); |
4724 | if (!xbridge) { | |
4725 | return; | |
4726 | } | |
4727 | ||
33bf9176 BP |
4728 | struct flow *flow = &xin->flow; |
4729 | ||
bb00fdef BP |
4730 | union mf_subvalue stack_stub[1024 / sizeof(union mf_subvalue)]; |
4731 | uint64_t action_set_stub[1024 / 8]; | |
49a73e0c | 4732 | struct flow_wildcards scratch_wc; |
1520ef4f BP |
4733 | uint64_t actions_stub[256 / 8]; |
4734 | struct ofpbuf scratch_actions = OFPBUF_STUB_INITIALIZER(actions_stub); | |
bb00fdef BP |
4735 | struct xlate_ctx ctx = { |
4736 | .xin = xin, | |
4737 | .xout = xout, | |
4738 | .base_flow = *flow, | |
4739 | .orig_tunnel_ip_dst = flow->tunnel.ip_dst, | |
4740 | .xbridge = xbridge, | |
4741 | .stack = OFPBUF_STUB_INITIALIZER(stack_stub), | |
4742 | .rule = xin->rule, | |
49a73e0c | 4743 | .wc = xin->wc ? xin->wc : &scratch_wc, |
1520ef4f | 4744 | .odp_actions = xin->odp_actions ? xin->odp_actions : &scratch_actions, |
bb00fdef BP |
4745 | |
4746 | .recurse = 0, | |
4747 | .resubmits = 0, | |
4748 | .in_group = false, | |
4749 | .in_action_set = false, | |
4750 | ||
4751 | .table_id = 0, | |
4752 | .rule_cookie = OVS_BE64_MAX, | |
4753 | .orig_skb_priority = flow->skb_priority, | |
4754 | .sflow_n_outputs = 0, | |
4755 | .sflow_odp_port = 0, | |
bb00fdef | 4756 | .exit = false, |
3d6151f3 | 4757 | .mirrors = 0, |
bb00fdef BP |
4758 | |
4759 | .recirc_action_offset = -1, | |
4760 | .last_unroll_offset = -1, | |
4761 | ||
4762 | .was_mpls = false, | |
4763 | ||
4764 | .action_set_has_group = false, | |
4765 | .action_set = OFPBUF_STUB_INITIALIZER(action_set_stub), | |
4766 | }; | |
865ca6cf BP |
4767 | |
4768 | /* 'base_flow' reflects the packet as it came in, but we need it to reflect | |
4769 | * the packet as the datapath will treat it for output actions: | |
4770 | * | |
4771 | * - Our datapath doesn't retain tunneling information without us | |
4772 | * re-setting it, so clear the tunnel data. | |
4773 | * | |
4774 | * - For VLAN splinters, a higher layer may pretend that the packet | |
4775 | * came in on 'flow->in_port.ofp_port' with 'flow->vlan_tci' | |
4776 | * attached, because that's how we want to treat it from an OpenFlow | |
4777 | * perspective. But from the datapath's perspective it actually came | |
4778 | * in on a VLAN device without any VLAN attached. So here we put the | |
4779 | * datapath's view of the VLAN information in 'base_flow' to ensure | |
4780 | * correct treatment. | |
4781 | */ | |
bb00fdef | 4782 | memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); |
865ca6cf BP |
4783 | if (flow->in_port.ofp_port |
4784 | != vsp_realdev_to_vlandev(xbridge->ofproto, | |
4785 | flow->in_port.ofp_port, | |
4786 | flow->vlan_tci)) { | |
4787 | ctx.base_flow.vlan_tci = 0; | |
4788 | } | |
4789 | ||
1520ef4f | 4790 | ofpbuf_reserve(ctx.odp_actions, NL_A_U32_SIZE); |
234c3da9 BP |
4791 | if (xin->wc) { |
4792 | xlate_wc_init(&ctx); | |
4793 | } | |
bb00fdef | 4794 | |
46c88433 | 4795 | struct xport *in_port; |
9583bc14 | 4796 | |
46c88433 | 4797 | COVERAGE_INC(xlate_actions); |
9583bc14 | 4798 | |
e672ff9b JR |
4799 | /* The in_port of the original packet before recirculation. */ |
4800 | in_port = get_ofp_port(xbridge, flow->in_port.ofp_port); | |
4801 | ||
4802 | if (xin->recirc) { | |
4803 | const struct recirc_id_node *recirc = xin->recirc; | |
4804 | ||
4805 | if (xin->ofpacts_len > 0 || ctx.rule) { | |
4806 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4807 | ||
4808 | VLOG_WARN_RL(&rl, "Recirculation conflict (%s)!", | |
4809 | xin->ofpacts_len > 0 | |
4810 | ? "actions" | |
4811 | : "rule"); | |
1520ef4f | 4812 | goto exit; |
e672ff9b JR |
4813 | } |
4814 | ||
4815 | /* Set the bridge for post-recirculation processing if needed. */ | |
4816 | if (ctx.xbridge->ofproto != recirc->ofproto) { | |
4817 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
4818 | const struct xbridge *new_bridge = xbridge_lookup(xcfg, | |
4819 | recirc->ofproto); | |
4820 | ||
4821 | if (OVS_UNLIKELY(!new_bridge)) { | |
4822 | /* Drop the packet if the bridge cannot be found. */ | |
4823 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4824 | VLOG_WARN_RL(&rl, "Recirculation bridge no longer exists."); | |
1520ef4f | 4825 | goto exit; |
e672ff9b JR |
4826 | } |
4827 | ctx.xbridge = new_bridge; | |
4828 | } | |
4829 | ||
4830 | /* Set the post-recirculation table id. Note: A table lookup is done | |
4831 | * only if there are no post-recirculation actions. */ | |
4832 | ctx.table_id = recirc->table_id; | |
4833 | ||
4834 | /* Restore pipeline metadata. May change flow's in_port and other | |
4835 | * metadata to the values that existed when recirculation was | |
4836 | * triggered. */ | |
4837 | recirc_metadata_to_flow(&recirc->metadata, flow); | |
4838 | ||
4839 | /* Restore stack, if any. */ | |
4840 | if (recirc->stack) { | |
4841 | ofpbuf_put(&ctx.stack, recirc->stack->data, recirc->stack->size); | |
4842 | } | |
4843 | ||
4844 | /* Restore action set, if any. */ | |
4845 | if (recirc->action_set_len) { | |
4846 | const struct ofpact *a; | |
4847 | ||
4848 | ofpbuf_put(&ctx.action_set, recirc->ofpacts, | |
4849 | recirc->action_set_len); | |
4850 | ||
4851 | OFPACT_FOR_EACH(a, recirc->ofpacts, recirc->action_set_len) { | |
4852 | if (a->type == OFPACT_GROUP) { | |
4853 | ctx.action_set_has_group = true; | |
4854 | break; | |
4855 | } | |
4856 | } | |
4857 | } | |
4858 | ||
4859 | /* Restore recirculation actions. If there are no actions, processing | |
4860 | * will start with a lookup in the table set above. */ | |
4861 | if (recirc->ofpacts_len > recirc->action_set_len) { | |
4862 | xin->ofpacts_len = recirc->ofpacts_len - recirc->action_set_len; | |
4863 | xin->ofpacts = recirc->ofpacts + | |
4864 | recirc->action_set_len / sizeof *recirc->ofpacts; | |
4865 | } | |
4866 | } else if (OVS_UNLIKELY(flow->recirc_id)) { | |
4867 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4868 | ||
4869 | VLOG_WARN_RL(&rl, "Recirculation context not found for ID %"PRIx32, | |
4870 | flow->recirc_id); | |
1520ef4f | 4871 | goto exit; |
e672ff9b | 4872 | } |
621b8064 JR |
4873 | /* The bridge is now known so obtain its table version. */ |
4874 | ctx.tables_version = ofproto_dpif_get_tables_version(ctx.xbridge->ofproto); | |
9583bc14 | 4875 | |
10c44245 | 4876 | if (!xin->ofpacts && !ctx.rule) { |
b2e89cc9 BP |
4877 | ctx.rule = rule_dpif_lookup_from_table( |
4878 | ctx.xbridge->ofproto, ctx.tables_version, flow, xin->wc, | |
4879 | ctx.xin->xcache != NULL, ctx.xin->resubmit_stats, &ctx.table_id, | |
4880 | flow->in_port.ofp_port, true, true); | |
10c44245 | 4881 | if (ctx.xin->resubmit_stats) { |
b2e89cc9 | 4882 | rule_dpif_credit_stats(ctx.rule, ctx.xin->resubmit_stats); |
10c44245 | 4883 | } |
b256dc52 JS |
4884 | if (ctx.xin->xcache) { |
4885 | struct xc_entry *entry; | |
4886 | ||
4887 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE); | |
b2e89cc9 | 4888 | entry->u.rule = ctx.rule; |
b256dc52 | 4889 | } |
a8c31348 BP |
4890 | |
4891 | if (OVS_UNLIKELY(ctx.xin->resubmit_hook)) { | |
b2e89cc9 | 4892 | ctx.xin->resubmit_hook(ctx.xin, ctx.rule, 0); |
a8c31348 | 4893 | } |
10c44245 | 4894 | } |
ee1afdd5 | 4895 | xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule); |
10c44245 | 4896 | |
2f739af0 | 4897 | struct flow orig_flow; |
e672ff9b | 4898 | if (mbridge_has_mirrors(xbridge->mbridge)) { |
9583bc14 EJ |
4899 | /* Do this conditionally because the copy is expensive enough that it |
4900 | * shows up in profiles. */ | |
33bf9176 | 4901 | orig_flow = *flow; |
9583bc14 EJ |
4902 | } |
4903 | ||
e672ff9b JR |
4904 | /* Tunnel stats only for non-recirculated packets. */ |
4905 | if (!xin->recirc && in_port && in_port->is_tunnel) { | |
b256dc52 JS |
4906 | if (ctx.xin->resubmit_stats) { |
4907 | netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats); | |
4908 | if (in_port->bfd) { | |
4909 | bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats); | |
4910 | } | |
4911 | } | |
4912 | if (ctx.xin->xcache) { | |
4913 | struct xc_entry *entry; | |
4914 | ||
4915 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV); | |
4916 | entry->u.dev.rx = netdev_ref(in_port->netdev); | |
4917 | entry->u.dev.bfd = bfd_ref(in_port->bfd); | |
d6fc5f57 EJ |
4918 | } |
4919 | } | |
4920 | ||
e672ff9b JR |
4921 | /* Do not perform special processing on recirculated packets, |
4922 | * as recirculated packets are not really received by the bridge. */ | |
515793d5 | 4923 | if (xin->recirc || !process_special(&ctx, in_port)) { |
e672ff9b | 4924 | /* Sampling is done only for packets really received by the bridge. */ |
a6092018 | 4925 | unsigned int user_cookie_offset = 0; |
e672ff9b | 4926 | if (!xin->recirc) { |
a6092018 BP |
4927 | user_cookie_offset = compose_sflow_action(&ctx); |
4928 | compose_ipfix_action(&ctx, ODPP_NONE); | |
e672ff9b | 4929 | } |
0731abc5 | 4930 | size_t sample_actions_len = ctx.odp_actions->size; |
9583bc14 | 4931 | |
234c3da9 BP |
4932 | if (tnl_process_ecn(flow) |
4933 | && (!in_port || may_receive(in_port, &ctx))) { | |
1806291d BP |
4934 | const struct ofpact *ofpacts; |
4935 | size_t ofpacts_len; | |
4936 | ||
4937 | if (xin->ofpacts) { | |
4938 | ofpacts = xin->ofpacts; | |
4939 | ofpacts_len = xin->ofpacts_len; | |
4940 | } else if (ctx.rule) { | |
4941 | const struct rule_actions *actions | |
4942 | = rule_dpif_get_actions(ctx.rule); | |
4943 | ofpacts = actions->ofpacts; | |
4944 | ofpacts_len = actions->ofpacts_len; | |
4945 | ctx.rule_cookie = rule_dpif_get_flow_cookie(ctx.rule); | |
4946 | } else { | |
4947 | OVS_NOT_REACHED(); | |
4948 | } | |
4949 | ||
9583bc14 EJ |
4950 | do_xlate_actions(ofpacts, ofpacts_len, &ctx); |
4951 | ||
4952 | /* We've let OFPP_NORMAL and the learning action look at the | |
4953 | * packet, so drop it now if forwarding is disabled. */ | |
9efd308e DV |
4954 | if (in_port && (!xport_stp_forward_state(in_port) || |
4955 | !xport_rstp_forward_state(in_port))) { | |
8ec90813 | 4956 | /* Drop all actions added by do_xlate_actions() above. */ |
1520ef4f | 4957 | ctx.odp_actions->size = sample_actions_len; |
e672ff9b JR |
4958 | |
4959 | /* Undo changes that may have been done for recirculation. */ | |
4960 | if (exit_recirculates(&ctx)) { | |
4961 | ctx.action_set.size = ctx.recirc_action_offset; | |
4962 | ctx.recirc_action_offset = -1; | |
4963 | ctx.last_unroll_offset = -1; | |
4964 | } | |
8ec90813 | 4965 | } else if (ctx.action_set.size) { |
e672ff9b JR |
4966 | /* Translate action set only if not dropping the packet and |
4967 | * not recirculating. */ | |
4968 | if (!exit_recirculates(&ctx)) { | |
4969 | xlate_action_set(&ctx); | |
4970 | } | |
4971 | } | |
4972 | /* Check if need to recirculate. */ | |
4973 | if (exit_recirculates(&ctx)) { | |
4974 | compose_recirculate_action(&ctx); | |
9583bc14 EJ |
4975 | } |
4976 | } | |
4977 | ||
e672ff9b JR |
4978 | /* Output only fully processed packets. */ |
4979 | if (!exit_recirculates(&ctx) | |
4980 | && xbridge->has_in_band | |
ce4a6b76 BP |
4981 | && in_band_must_output_to_local_port(flow) |
4982 | && !actions_output_to_local_port(&ctx)) { | |
e93ef1c7 | 4983 | compose_output_action(&ctx, OFPP_LOCAL, NULL); |
9583bc14 | 4984 | } |
aaa0fbae | 4985 | |
a6092018 BP |
4986 | if (user_cookie_offset) { |
4987 | fix_sflow_action(&ctx, user_cookie_offset); | |
e672ff9b JR |
4988 | } |
4989 | /* Only mirror fully processed packets. */ | |
4990 | if (!exit_recirculates(&ctx) | |
4991 | && mbridge_has_mirrors(xbridge->mbridge)) { | |
9583bc14 EJ |
4992 | add_mirror_actions(&ctx, &orig_flow); |
4993 | } | |
9583bc14 EJ |
4994 | } |
4995 | ||
1520ef4f | 4996 | if (nl_attr_oversized(ctx.odp_actions->size)) { |
542024c4 | 4997 | /* These datapath actions are too big for a Netlink attribute, so we |
0f032e95 BP |
4998 | * can't hand them to the kernel directly. dpif_execute() can execute |
4999 | * them one by one with help, so just mark the result as SLOW_ACTION to | |
5000 | * prevent the flow from being installed. */ | |
5001 | COVERAGE_INC(xlate_actions_oversize); | |
5002 | ctx.xout->slow |= SLOW_ACTION; | |
1520ef4f | 5003 | } else if (too_many_output_actions(ctx.odp_actions)) { |
7d031d7e BP |
5004 | COVERAGE_INC(xlate_actions_too_many_output); |
5005 | ctx.xout->slow |= SLOW_ACTION; | |
542024c4 BP |
5006 | } |
5007 | ||
e672ff9b JR |
5008 | /* Update mirror stats only for packets really received by the bridge. */ |
5009 | if (!xin->recirc && mbridge_has_mirrors(xbridge->mbridge)) { | |
b256dc52 | 5010 | if (ctx.xin->resubmit_stats) { |
3d6151f3 | 5011 | mirror_update_stats(xbridge->mbridge, ctx.mirrors, |
b256dc52 JS |
5012 | ctx.xin->resubmit_stats->n_packets, |
5013 | ctx.xin->resubmit_stats->n_bytes); | |
5014 | } | |
5015 | if (ctx.xin->xcache) { | |
5016 | struct xc_entry *entry; | |
5017 | ||
5018 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_MIRROR); | |
e672ff9b | 5019 | entry->u.mirror.mbridge = mbridge_ref(xbridge->mbridge); |
3d6151f3 | 5020 | entry->u.mirror.mirrors = ctx.mirrors; |
b256dc52 JS |
5021 | } |
5022 | } | |
5023 | ||
1806291d BP |
5024 | /* Do netflow only for packets really received by the bridge and not sent |
5025 | * to the controller. We consider packets sent to the controller to be | |
5026 | * part of the control plane rather than the data plane. */ | |
5027 | if (!xin->recirc && xbridge->netflow && !(xout->slow & SLOW_CONTROLLER)) { | |
5028 | if (ctx.xin->resubmit_stats) { | |
5029 | netflow_flow_update(xbridge->netflow, flow, | |
5030 | xout->nf_output_iface, | |
5031 | ctx.xin->resubmit_stats); | |
5032 | } | |
5033 | if (ctx.xin->xcache) { | |
5034 | struct xc_entry *entry; | |
b256dc52 | 5035 | |
1806291d BP |
5036 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW); |
5037 | entry->u.nf.netflow = netflow_ref(xbridge->netflow); | |
5038 | entry->u.nf.flow = xmemdup(flow, sizeof *flow); | |
5039 | entry->u.nf.iface = xout->nf_output_iface; | |
d6fc5f57 EJ |
5040 | } |
5041 | } | |
5042 | ||
49a73e0c | 5043 | if (xin->wc) { |
234c3da9 | 5044 | xlate_wc_finish(&ctx); |
c56fac1b | 5045 | } |
1520ef4f BP |
5046 | |
5047 | exit: | |
5048 | ofpbuf_uninit(&ctx.stack); | |
5049 | ofpbuf_uninit(&ctx.action_set); | |
5050 | ofpbuf_uninit(&scratch_actions); | |
91d6cd12 AW |
5051 | } |
5052 | ||
5053 | /* Sends 'packet' out 'ofport'. | |
5054 | * May modify 'packet'. | |
5055 | * Returns 0 if successful, otherwise a positive errno value. */ | |
5056 | int | |
cf62fa4c | 5057 | xlate_send_packet(const struct ofport_dpif *ofport, struct dp_packet *packet) |
91d6cd12 | 5058 | { |
84f0f298 | 5059 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
91d6cd12 | 5060 | struct xport *xport; |
91d6cd12 | 5061 | struct ofpact_output output; |
91d6cd12 | 5062 | struct flow flow; |
91d6cd12 | 5063 | |
91d6cd12 AW |
5064 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); |
5065 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ | |
cf62fa4c | 5066 | flow_extract(packet, &flow); |
b5e7e61a | 5067 | flow.in_port.ofp_port = OFPP_NONE; |
91d6cd12 | 5068 | |
84f0f298 | 5069 | xport = xport_lookup(xcfg, ofport); |
91d6cd12 | 5070 | if (!xport) { |
02ea2703 | 5071 | return EINVAL; |
91d6cd12 | 5072 | } |
91d6cd12 AW |
5073 | output.port = xport->ofp_port; |
5074 | output.max_len = 0; | |
e491a67a YT |
5075 | |
5076 | return ofproto_dpif_execute_actions(xport->xbridge->ofproto, &flow, NULL, | |
5077 | &output.ofpact, sizeof output, | |
5078 | packet); | |
9583bc14 | 5079 | } |
b256dc52 JS |
5080 | |
5081 | struct xlate_cache * | |
5082 | xlate_cache_new(void) | |
5083 | { | |
5084 | struct xlate_cache *xcache = xmalloc(sizeof *xcache); | |
5085 | ||
5086 | ofpbuf_init(&xcache->entries, 512); | |
5087 | return xcache; | |
5088 | } | |
5089 | ||
5090 | static struct xc_entry * | |
5091 | xlate_cache_add_entry(struct xlate_cache *xcache, enum xc_type type) | |
5092 | { | |
5093 | struct xc_entry *entry; | |
5094 | ||
5095 | entry = ofpbuf_put_zeros(&xcache->entries, sizeof *entry); | |
5096 | entry->type = type; | |
5097 | ||
5098 | return entry; | |
5099 | } | |
5100 | ||
5101 | static void | |
5102 | xlate_cache_netdev(struct xc_entry *entry, const struct dpif_flow_stats *stats) | |
5103 | { | |
5104 | if (entry->u.dev.tx) { | |
5105 | netdev_vport_inc_tx(entry->u.dev.tx, stats); | |
5106 | } | |
5107 | if (entry->u.dev.rx) { | |
5108 | netdev_vport_inc_rx(entry->u.dev.rx, stats); | |
5109 | } | |
5110 | if (entry->u.dev.bfd) { | |
5111 | bfd_account_rx(entry->u.dev.bfd, stats); | |
5112 | } | |
5113 | } | |
5114 | ||
5115 | static void | |
5116 | xlate_cache_normal(struct ofproto_dpif *ofproto, struct flow *flow, int vlan) | |
5117 | { | |
84f0f298 | 5118 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
b256dc52 JS |
5119 | struct xbridge *xbridge; |
5120 | struct xbundle *xbundle; | |
5121 | struct flow_wildcards wc; | |
5122 | ||
84f0f298 | 5123 | xbridge = xbridge_lookup(xcfg, ofproto); |
b256dc52 JS |
5124 | if (!xbridge) { |
5125 | return; | |
5126 | } | |
5127 | ||
5128 | xbundle = lookup_input_bundle(xbridge, flow->in_port.ofp_port, false, | |
5129 | NULL); | |
5130 | if (!xbundle) { | |
5131 | return; | |
5132 | } | |
5133 | ||
5134 | update_learning_table(xbridge, flow, &wc, vlan, xbundle); | |
5135 | } | |
5136 | ||
5137 | /* Push stats and perform side effects of flow translation. */ | |
5138 | void | |
0725e747 | 5139 | xlate_push_stats(struct xlate_cache *xcache, |
b256dc52 JS |
5140 | const struct dpif_flow_stats *stats) |
5141 | { | |
5142 | struct xc_entry *entry; | |
5143 | struct ofpbuf entries = xcache->entries; | |
a36de779 | 5144 | uint8_t dmac[ETH_ADDR_LEN]; |
b256dc52 | 5145 | |
bc388583 BP |
5146 | if (!stats->n_packets) { |
5147 | return; | |
5148 | } | |
5149 | ||
b256dc52 JS |
5150 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { |
5151 | switch (entry->type) { | |
5152 | case XC_RULE: | |
5153 | rule_dpif_credit_stats(entry->u.rule, stats); | |
5154 | break; | |
5155 | case XC_BOND: | |
5156 | bond_account(entry->u.bond.bond, entry->u.bond.flow, | |
5157 | entry->u.bond.vid, stats->n_bytes); | |
5158 | break; | |
5159 | case XC_NETDEV: | |
5160 | xlate_cache_netdev(entry, stats); | |
5161 | break; | |
5162 | case XC_NETFLOW: | |
5163 | netflow_flow_update(entry->u.nf.netflow, entry->u.nf.flow, | |
5164 | entry->u.nf.iface, stats); | |
5165 | break; | |
5166 | case XC_MIRROR: | |
5167 | mirror_update_stats(entry->u.mirror.mbridge, | |
5168 | entry->u.mirror.mirrors, | |
5169 | stats->n_packets, stats->n_bytes); | |
5170 | break; | |
5171 | case XC_LEARN: | |
0725e747 | 5172 | ofproto_dpif_flow_mod(entry->u.learn.ofproto, entry->u.learn.fm); |
b256dc52 JS |
5173 | break; |
5174 | case XC_NORMAL: | |
0725e747 BP |
5175 | xlate_cache_normal(entry->u.normal.ofproto, entry->u.normal.flow, |
5176 | entry->u.normal.vlan); | |
b256dc52 JS |
5177 | break; |
5178 | case XC_FIN_TIMEOUT: | |
5179 | xlate_fin_timeout__(entry->u.fin.rule, stats->tcp_flags, | |
5180 | entry->u.fin.idle, entry->u.fin.hard); | |
5181 | break; | |
1e684d7d RW |
5182 | case XC_GROUP: |
5183 | group_dpif_credit_stats(entry->u.group.group, entry->u.group.bucket, | |
5184 | stats); | |
5185 | break; | |
a36de779 PS |
5186 | case XC_TNL_ARP: |
5187 | /* Lookup arp to avoid arp timeout. */ | |
5188 | tnl_arp_lookup(entry->u.tnl_arp_cache.br_name, entry->u.tnl_arp_cache.d_ip, dmac); | |
5189 | break; | |
b256dc52 JS |
5190 | default: |
5191 | OVS_NOT_REACHED(); | |
5192 | } | |
5193 | } | |
5194 | } | |
5195 | ||
5196 | static void | |
5197 | xlate_dev_unref(struct xc_entry *entry) | |
5198 | { | |
5199 | if (entry->u.dev.tx) { | |
5200 | netdev_close(entry->u.dev.tx); | |
5201 | } | |
5202 | if (entry->u.dev.rx) { | |
5203 | netdev_close(entry->u.dev.rx); | |
5204 | } | |
5205 | if (entry->u.dev.bfd) { | |
5206 | bfd_unref(entry->u.dev.bfd); | |
5207 | } | |
5208 | } | |
5209 | ||
5210 | static void | |
5211 | xlate_cache_clear_netflow(struct netflow *netflow, struct flow *flow) | |
5212 | { | |
b256dc52 JS |
5213 | netflow_flow_clear(netflow, flow); |
5214 | netflow_unref(netflow); | |
5215 | free(flow); | |
5216 | } | |
5217 | ||
5218 | void | |
5219 | xlate_cache_clear(struct xlate_cache *xcache) | |
5220 | { | |
5221 | struct xc_entry *entry; | |
5222 | struct ofpbuf entries; | |
5223 | ||
5224 | if (!xcache) { | |
5225 | return; | |
5226 | } | |
5227 | ||
5228 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { | |
5229 | switch (entry->type) { | |
5230 | case XC_RULE: | |
5231 | rule_dpif_unref(entry->u.rule); | |
5232 | break; | |
5233 | case XC_BOND: | |
5234 | free(entry->u.bond.flow); | |
5235 | bond_unref(entry->u.bond.bond); | |
5236 | break; | |
5237 | case XC_NETDEV: | |
5238 | xlate_dev_unref(entry); | |
5239 | break; | |
5240 | case XC_NETFLOW: | |
5241 | xlate_cache_clear_netflow(entry->u.nf.netflow, entry->u.nf.flow); | |
5242 | break; | |
5243 | case XC_MIRROR: | |
5244 | mbridge_unref(entry->u.mirror.mbridge); | |
5245 | break; | |
5246 | case XC_LEARN: | |
4165b5e0 JS |
5247 | free(entry->u.learn.fm); |
5248 | ofpbuf_delete(entry->u.learn.ofpacts); | |
b256dc52 JS |
5249 | break; |
5250 | case XC_NORMAL: | |
5251 | free(entry->u.normal.flow); | |
5252 | break; | |
5253 | case XC_FIN_TIMEOUT: | |
83709dfa JR |
5254 | /* 'u.fin.rule' is always already held as a XC_RULE, which |
5255 | * has already released it's reference above. */ | |
b256dc52 | 5256 | break; |
1e684d7d RW |
5257 | case XC_GROUP: |
5258 | group_dpif_unref(entry->u.group.group); | |
5259 | break; | |
a36de779 PS |
5260 | case XC_TNL_ARP: |
5261 | break; | |
b256dc52 JS |
5262 | default: |
5263 | OVS_NOT_REACHED(); | |
5264 | } | |
5265 | } | |
5266 | ||
5267 | ofpbuf_clear(&xcache->entries); | |
5268 | } | |
5269 | ||
5270 | void | |
5271 | xlate_cache_delete(struct xlate_cache *xcache) | |
5272 | { | |
5273 | xlate_cache_clear(xcache); | |
5274 | ofpbuf_uninit(&xcache->entries); | |
5275 | free(xcache); | |
5276 | } |