]>
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". */ | |
178 | union mf_subvalue init_stack[1024 / sizeof(union mf_subvalue)]; | |
179 | struct ofpbuf stack; | |
180 | ||
181 | /* The rule that we are currently translating, or NULL. */ | |
182 | struct rule_dpif *rule; | |
183 | ||
98b07853 BP |
184 | /* Resubmit statistics, via xlate_table_action(). */ |
185 | int recurse; /* Current resubmit nesting depth. */ | |
186 | int resubmits; /* Total number of resubmits. */ | |
5a070238 | 187 | bool in_group; /* Currently translating ofgroup, if true. */ |
029ca940 | 188 | bool in_action_set; /* Currently translating action_set, if true. */ |
98b07853 | 189 | |
4d0acc70 | 190 | uint8_t table_id; /* OpenFlow table ID where flow was found. */ |
8b1e5560 JR |
191 | ovs_be64 rule_cookie; /* Cookie of the rule being translated. */ |
192 | uint32_t orig_skb_priority; /* Priority when packet arrived. */ | |
4d0acc70 | 193 | uint32_t sflow_n_outputs; /* Number of output ports. */ |
4e022ec0 | 194 | odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */ |
4d0acc70 EJ |
195 | uint16_t user_cookie_offset;/* Used for user_action_cookie fixup. */ |
196 | bool exit; /* No further actions should be processed. */ | |
7fdb60a7 | 197 | |
e672ff9b JR |
198 | /* These are used for non-bond recirculation. The recirculation IDs are |
199 | * stored in xout and must be associated with a datapath flow (ukey), | |
200 | * otherwise they will be freed when the xout is uninitialized. | |
201 | * | |
202 | * | |
203 | * Steps in Recirculation Translation | |
204 | * ================================== | |
205 | * | |
206 | * At some point during translation, the code recognizes the need for | |
207 | * recirculation. For example, recirculation is necessary when, after | |
208 | * popping the last MPLS label, an action or a match tries to examine or | |
209 | * modify a field that has been newly revealed following the MPLS label. | |
210 | * | |
211 | * The simplest part of the work to be done is to commit existing changes to | |
212 | * the packet, which produces datapath actions corresponding to the changes, | |
213 | * and after this, add an OVS_ACTION_ATTR_RECIRC datapath action. | |
214 | * | |
215 | * The main problem here is preserving state. When the datapath executes | |
216 | * OVS_ACTION_ATTR_RECIRC, it will upcall to userspace to get a translation | |
217 | * for the post-recirculation actions. At this point userspace has to | |
218 | * resume the translation where it left off, which means that it has to | |
219 | * execute the following: | |
220 | * | |
221 | * - The action that prompted recirculation, and any actions following | |
222 | * it within the same flow. | |
223 | * | |
224 | * - If the action that prompted recirculation was invoked within a | |
225 | * NXAST_RESUBMIT, then any actions following the resubmit. These | |
226 | * "resubmit"s can be nested, so this has to go all the way up the | |
227 | * control stack. | |
228 | * | |
229 | * - The OpenFlow 1.1+ action set. | |
230 | * | |
231 | * State that actions and flow table lookups can depend on, such as the | |
232 | * following, must also be preserved: | |
233 | * | |
234 | * - Metadata fields (input port, registers, OF1.1+ metadata, ...). | |
235 | * | |
236 | * - Action set, stack | |
237 | * | |
238 | * - The table ID and cookie of the flow being translated at each level | |
239 | * of the control stack (since OFPAT_CONTROLLER actions send these to | |
240 | * the controller). | |
241 | * | |
242 | * Translation allows for the control of this state preservation via these | |
243 | * members. When a need for recirculation is identified, the translation | |
244 | * process: | |
245 | * | |
246 | * 1. Sets 'recirc_action_offset' to the current size of 'action_set'. The | |
247 | * action set is part of what needs to be preserved, so this allows the | |
248 | * action set and the additional state to share the 'action_set' buffer. | |
249 | * Later steps can tell that setup for recirculation is in progress from | |
250 | * the nonnegative value of 'recirc_action_offset'. | |
251 | * | |
252 | * 2. Sets 'exit' to true to tell later steps that we're exiting from the | |
253 | * translation process. | |
254 | * | |
255 | * 3. Adds an OFPACT_UNROLL_XLATE action to 'action_set'. This action | |
256 | * holds the current table ID and cookie so that they can be restored | |
257 | * during a post-recirculation upcall translation. | |
258 | * | |
259 | * 4. Adds the action that prompted recirculation and any actions following | |
260 | * it within the same flow to 'action_set', so that they can be executed | |
261 | * during a post-recirculation upcall translation. | |
262 | * | |
263 | * 5. Returns. | |
264 | * | |
265 | * 6. The action that prompted recirculation might be nested in a stack of | |
266 | * nested "resubmit"s that have actions remaining. Each of these notices | |
267 | * that we're exiting (from 'exit') and that recirculation setup is in | |
268 | * progress (from 'recirc_action_offset') and responds by adding more | |
269 | * OFPACT_UNROLL_XLATE actions to 'action_set', as necessary, and any | |
270 | * actions that were yet unprocessed. | |
271 | * | |
272 | * The caller stores all the state produced by this process associated with | |
273 | * the recirculation ID. For post-recirculation upcall translation, the | |
274 | * caller passes it back in for the new translation to execute. The | |
275 | * process yielded a set of ofpacts that can be translated directly, so it | |
276 | * is not much of a special case at that point. | |
277 | */ | |
278 | int recirc_action_offset; /* Offset in 'action_set' to actions to be | |
279 | * executed after recirculation, or -1. */ | |
280 | int last_unroll_offset; /* Offset in 'action_set' to the latest unroll | |
281 | * action, or -1. */ | |
282 | ||
7bbdd84f SH |
283 | /* True if a packet was but is no longer MPLS (due to an MPLS pop action). |
284 | * This is a trigger for recirculation in cases where translating an action | |
285 | * or looking up a flow requires access to the fields of the packet after | |
1d741d6d | 286 | * the MPLS label stack that was originally present. */ |
7bbdd84f SH |
287 | bool was_mpls; |
288 | ||
7fdb60a7 SH |
289 | /* OpenFlow 1.1+ action set. |
290 | * | |
291 | * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS. | |
292 | * When translation is otherwise complete, ofpacts_execute_action_set() | |
293 | * converts it to a set of "struct ofpact"s that can be translated into | |
ed9c9e3e | 294 | * datapath actions. */ |
c61f3870 | 295 | bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ |
7fdb60a7 SH |
296 | struct ofpbuf action_set; /* Action set. */ |
297 | uint64_t action_set_stub[1024 / 8]; | |
4d0acc70 EJ |
298 | }; |
299 | ||
ed9c9e3e JR |
300 | static void xlate_action_set(struct xlate_ctx *ctx); |
301 | ||
1d741d6d JR |
302 | static void |
303 | ctx_trigger_recirculation(struct xlate_ctx *ctx) | |
304 | { | |
305 | ctx->exit = true; | |
306 | ctx->recirc_action_offset = ctx->action_set.size; | |
307 | } | |
308 | ||
309 | static bool | |
310 | ctx_first_recirculation_action(const struct xlate_ctx *ctx) | |
311 | { | |
312 | return ctx->recirc_action_offset == ctx->action_set.size; | |
313 | } | |
314 | ||
e672ff9b JR |
315 | static inline bool |
316 | exit_recirculates(const struct xlate_ctx *ctx) | |
317 | { | |
318 | /* When recirculating the 'recirc_action_offset' has a non-negative value. | |
319 | */ | |
320 | return ctx->recirc_action_offset >= 0; | |
321 | } | |
322 | ||
323 | static void compose_recirculate_action(struct xlate_ctx *ctx); | |
324 | ||
9583bc14 EJ |
325 | /* A controller may use OFPP_NONE as the ingress port to indicate that |
326 | * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for | |
327 | * when an input bundle is needed for validation (e.g., mirroring or | |
328 | * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have | |
3548d242 BP |
329 | * any 'port' structs, so care must be taken when dealing with it. */ |
330 | static struct xbundle ofpp_none_bundle = { | |
331 | .name = "OFPP_NONE", | |
332 | .vlan_mode = PORT_VLAN_TRUNK | |
333 | }; | |
9583bc14 | 334 | |
55954f6e EJ |
335 | /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from |
336 | * 'priority' (the datapath's term for QoS queue) to the dscp bits which all | |
337 | * traffic egressing the 'ofport' with that priority should be marked with. */ | |
338 | struct skb_priority_to_dscp { | |
339 | struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */ | |
340 | uint32_t skb_priority; /* Priority of this queue (see struct flow). */ | |
341 | ||
342 | uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */ | |
343 | }; | |
344 | ||
b256dc52 JS |
345 | enum xc_type { |
346 | XC_RULE, | |
347 | XC_BOND, | |
348 | XC_NETDEV, | |
349 | XC_NETFLOW, | |
350 | XC_MIRROR, | |
351 | XC_LEARN, | |
352 | XC_NORMAL, | |
353 | XC_FIN_TIMEOUT, | |
1e684d7d | 354 | XC_GROUP, |
a36de779 | 355 | XC_TNL_ARP, |
b256dc52 JS |
356 | }; |
357 | ||
358 | /* xlate_cache entries hold enough information to perform the side effects of | |
359 | * xlate_actions() for a rule, without needing to perform rule translation | |
360 | * from scratch. The primary usage of these is to submit statistics to objects | |
361 | * that a flow relates to, although they may be used for other effects as well | |
362 | * (for instance, refreshing hard timeouts for learned flows). */ | |
363 | struct xc_entry { | |
364 | enum xc_type type; | |
365 | union { | |
366 | struct rule_dpif *rule; | |
367 | struct { | |
368 | struct netdev *tx; | |
369 | struct netdev *rx; | |
370 | struct bfd *bfd; | |
371 | } dev; | |
372 | struct { | |
373 | struct netflow *netflow; | |
374 | struct flow *flow; | |
375 | ofp_port_t iface; | |
376 | } nf; | |
377 | struct { | |
378 | struct mbridge *mbridge; | |
379 | mirror_mask_t mirrors; | |
380 | } mirror; | |
381 | struct { | |
382 | struct bond *bond; | |
383 | struct flow *flow; | |
384 | uint16_t vid; | |
385 | } bond; | |
386 | struct { | |
4165b5e0 JS |
387 | struct ofproto_dpif *ofproto; |
388 | struct ofputil_flow_mod *fm; | |
389 | struct ofpbuf *ofpacts; | |
b256dc52 JS |
390 | } learn; |
391 | struct { | |
392 | struct ofproto_dpif *ofproto; | |
393 | struct flow *flow; | |
394 | int vlan; | |
395 | } normal; | |
396 | struct { | |
397 | struct rule_dpif *rule; | |
398 | uint16_t idle; | |
399 | uint16_t hard; | |
400 | } fin; | |
1e684d7d RW |
401 | struct { |
402 | struct group_dpif *group; | |
403 | struct ofputil_bucket *bucket; | |
404 | } group; | |
a36de779 PS |
405 | struct { |
406 | char br_name[IFNAMSIZ]; | |
407 | ovs_be32 d_ip; | |
408 | } tnl_arp_cache; | |
b256dc52 JS |
409 | } u; |
410 | }; | |
411 | ||
412 | #define XC_ENTRY_FOR_EACH(entry, entries, xcache) \ | |
413 | entries = xcache->entries; \ | |
414 | for (entry = ofpbuf_try_pull(&entries, sizeof *entry); \ | |
415 | entry; \ | |
416 | entry = ofpbuf_try_pull(&entries, sizeof *entry)) | |
417 | ||
418 | struct xlate_cache { | |
419 | struct ofpbuf entries; | |
420 | }; | |
421 | ||
84f0f298 RW |
422 | /* Xlate config contains hash maps of all bridges, bundles and ports. |
423 | * Xcfgp contains the pointer to the current xlate configuration. | |
424 | * When the main thread needs to change the configuration, it copies xcfgp to | |
425 | * new_xcfg and edits new_xcfg. This enables the use of RCU locking which | |
426 | * does not block handler and revalidator threads. */ | |
427 | struct xlate_cfg { | |
428 | struct hmap xbridges; | |
429 | struct hmap xbundles; | |
430 | struct hmap xports; | |
431 | }; | |
b1b72f2d | 432 | static OVSRCU_TYPE(struct xlate_cfg *) xcfgp = OVSRCU_INITIALIZER(NULL); |
f439f23b | 433 | static struct xlate_cfg *new_xcfg = NULL; |
46c88433 EJ |
434 | |
435 | static bool may_receive(const struct xport *, struct xlate_ctx *); | |
9583bc14 EJ |
436 | static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
437 | struct xlate_ctx *); | |
adcf00ba | 438 | static void xlate_normal(struct xlate_ctx *); |
c1b3756c BP |
439 | static inline void xlate_report(struct xlate_ctx *, const char *, ...) |
440 | OVS_PRINTF_FORMAT(2, 3); | |
6d328fa2 SH |
441 | static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port, |
442 | uint8_t table_id, bool may_packet_in, | |
443 | bool honor_table_miss); | |
46c88433 EJ |
444 | static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn); |
445 | static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid); | |
446 | static void output_normal(struct xlate_ctx *, const struct xbundle *, | |
9583bc14 | 447 | uint16_t vlan); |
e93ef1c7 JR |
448 | |
449 | /* Optional bond recirculation parameter to compose_output_action(). */ | |
450 | struct xlate_bond_recirc { | |
451 | uint32_t recirc_id; /* !0 Use recirculation instead of output. */ | |
452 | uint8_t hash_alg; /* !0 Compute hash for recirc before. */ | |
453 | uint32_t hash_basis; /* Compute hash for recirc before. */ | |
454 | }; | |
455 | ||
456 | static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port, | |
457 | const struct xlate_bond_recirc *xr); | |
9583bc14 | 458 | |
84f0f298 RW |
459 | static struct xbridge *xbridge_lookup(struct xlate_cfg *, |
460 | const struct ofproto_dpif *); | |
461 | static struct xbundle *xbundle_lookup(struct xlate_cfg *, | |
462 | const struct ofbundle *); | |
463 | static struct xport *xport_lookup(struct xlate_cfg *, | |
464 | const struct ofport_dpif *); | |
46c88433 | 465 | static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port); |
55954f6e EJ |
466 | static struct skb_priority_to_dscp *get_skb_priority(const struct xport *, |
467 | uint32_t skb_priority); | |
468 | static void clear_skb_priorities(struct xport *); | |
16194afd | 469 | static size_t count_skb_priorities(const struct xport *); |
55954f6e EJ |
470 | static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, |
471 | uint8_t *dscp); | |
46c88433 | 472 | |
b256dc52 JS |
473 | static struct xc_entry *xlate_cache_add_entry(struct xlate_cache *xc, |
474 | enum xc_type type); | |
84f0f298 RW |
475 | static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *); |
476 | static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *); | |
477 | static void xlate_xport_init(struct xlate_cfg *, struct xport *); | |
9efd308e | 478 | static void xlate_xbridge_set(struct xbridge *, struct dpif *, |
9efd308e DV |
479 | const struct mac_learning *, struct stp *, |
480 | struct rstp *, const struct mcast_snooping *, | |
481 | const struct mbridge *, | |
482 | const struct dpif_sflow *, | |
483 | const struct dpif_ipfix *, | |
2f47cdf4 | 484 | const struct netflow *, |
84f0f298 | 485 | bool forward_bpdu, bool has_in_band, |
b440dd8c | 486 | const struct dpif_backer_support *); |
84f0f298 RW |
487 | static void xlate_xbundle_set(struct xbundle *xbundle, |
488 | enum port_vlan_mode vlan_mode, int vlan, | |
489 | unsigned long *trunks, bool use_priority_tags, | |
490 | const struct bond *bond, const struct lacp *lacp, | |
491 | bool floodable); | |
492 | static void xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
493 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 DF |
494 | const struct bfd *bfd, const struct lldp *lldp, |
495 | int stp_port_no, const struct rstp_port *rstp_port, | |
84f0f298 RW |
496 | enum ofputil_port_config config, |
497 | enum ofputil_port_state state, bool is_tunnel, | |
498 | bool may_enable); | |
499 | static void xlate_xbridge_remove(struct xlate_cfg *, struct xbridge *); | |
500 | static void xlate_xbundle_remove(struct xlate_cfg *, struct xbundle *); | |
501 | static void xlate_xport_remove(struct xlate_cfg *, struct xport *); | |
502 | static void xlate_xbridge_copy(struct xbridge *); | |
503 | static void xlate_xbundle_copy(struct xbridge *, struct xbundle *); | |
504 | static void xlate_xport_copy(struct xbridge *, struct xbundle *, | |
505 | struct xport *); | |
506 | static void xlate_xcfg_free(struct xlate_cfg *); | |
b256dc52 | 507 | |
34dd0d78 | 508 | static inline void |
c1b3756c | 509 | xlate_report(struct xlate_ctx *ctx, const char *format, ...) |
34dd0d78 JR |
510 | { |
511 | if (OVS_UNLIKELY(ctx->xin->report_hook)) { | |
c1b3756c BP |
512 | va_list args; |
513 | ||
514 | va_start(args, format); | |
515 | ctx->xin->report_hook(ctx->xin, ctx->recurse, format, args); | |
516 | va_end(args); | |
34dd0d78 JR |
517 | } |
518 | } | |
84f0f298 RW |
519 | |
520 | static void | |
521 | xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
522 | { | |
523 | list_init(&xbridge->xbundles); | |
524 | hmap_init(&xbridge->xports); | |
525 | hmap_insert(&xcfg->xbridges, &xbridge->hmap_node, | |
526 | hash_pointer(xbridge->ofproto, 0)); | |
527 | } | |
528 | ||
529 | static void | |
530 | xlate_xbundle_init(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
531 | { | |
532 | list_init(&xbundle->xports); | |
533 | list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); | |
534 | hmap_insert(&xcfg->xbundles, &xbundle->hmap_node, | |
535 | hash_pointer(xbundle->ofbundle, 0)); | |
536 | } | |
537 | ||
538 | static void | |
539 | xlate_xport_init(struct xlate_cfg *xcfg, struct xport *xport) | |
540 | { | |
541 | hmap_init(&xport->skb_priorities); | |
542 | hmap_insert(&xcfg->xports, &xport->hmap_node, | |
543 | hash_pointer(xport->ofport, 0)); | |
544 | hmap_insert(&xport->xbridge->xports, &xport->ofp_node, | |
545 | hash_ofp_port(xport->ofp_port)); | |
546 | } | |
547 | ||
548 | static void | |
549 | xlate_xbridge_set(struct xbridge *xbridge, | |
550 | struct dpif *dpif, | |
ec89fc6f | 551 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 552 | struct rstp *rstp, const struct mcast_snooping *ms, |
ec89fc6f | 553 | const struct mbridge *mbridge, |
46c88433 | 554 | const struct dpif_sflow *sflow, |
ce3955be | 555 | const struct dpif_ipfix *ipfix, |
2f47cdf4 | 556 | const struct netflow *netflow, |
4b97b70d | 557 | bool forward_bpdu, bool has_in_band, |
b440dd8c | 558 | const struct dpif_backer_support *support) |
46c88433 | 559 | { |
46c88433 EJ |
560 | if (xbridge->ml != ml) { |
561 | mac_learning_unref(xbridge->ml); | |
562 | xbridge->ml = mac_learning_ref(ml); | |
563 | } | |
564 | ||
6d95c4e8 FL |
565 | if (xbridge->ms != ms) { |
566 | mcast_snooping_unref(xbridge->ms); | |
567 | xbridge->ms = mcast_snooping_ref(ms); | |
568 | } | |
569 | ||
46c88433 EJ |
570 | if (xbridge->mbridge != mbridge) { |
571 | mbridge_unref(xbridge->mbridge); | |
572 | xbridge->mbridge = mbridge_ref(mbridge); | |
573 | } | |
574 | ||
575 | if (xbridge->sflow != sflow) { | |
576 | dpif_sflow_unref(xbridge->sflow); | |
577 | xbridge->sflow = dpif_sflow_ref(sflow); | |
578 | } | |
579 | ||
580 | if (xbridge->ipfix != ipfix) { | |
581 | dpif_ipfix_unref(xbridge->ipfix); | |
582 | xbridge->ipfix = dpif_ipfix_ref(ipfix); | |
583 | } | |
584 | ||
9d189a50 EJ |
585 | if (xbridge->stp != stp) { |
586 | stp_unref(xbridge->stp); | |
587 | xbridge->stp = stp_ref(stp); | |
588 | } | |
589 | ||
9efd308e DV |
590 | if (xbridge->rstp != rstp) { |
591 | rstp_unref(xbridge->rstp); | |
592 | xbridge->rstp = rstp_ref(rstp); | |
593 | } | |
594 | ||
ce3955be EJ |
595 | if (xbridge->netflow != netflow) { |
596 | netflow_unref(xbridge->netflow); | |
597 | xbridge->netflow = netflow_ref(netflow); | |
598 | } | |
599 | ||
89a8a7f0 | 600 | xbridge->dpif = dpif; |
46c88433 EJ |
601 | xbridge->forward_bpdu = forward_bpdu; |
602 | xbridge->has_in_band = has_in_band; | |
b440dd8c | 603 | xbridge->support = *support; |
46c88433 EJ |
604 | } |
605 | ||
84f0f298 RW |
606 | static void |
607 | xlate_xbundle_set(struct xbundle *xbundle, | |
608 | enum port_vlan_mode vlan_mode, int vlan, | |
609 | unsigned long *trunks, bool use_priority_tags, | |
610 | const struct bond *bond, const struct lacp *lacp, | |
611 | bool floodable) | |
612 | { | |
613 | ovs_assert(xbundle->xbridge); | |
614 | ||
615 | xbundle->vlan_mode = vlan_mode; | |
616 | xbundle->vlan = vlan; | |
617 | xbundle->trunks = trunks; | |
618 | xbundle->use_priority_tags = use_priority_tags; | |
619 | xbundle->floodable = floodable; | |
620 | ||
621 | if (xbundle->bond != bond) { | |
622 | bond_unref(xbundle->bond); | |
623 | xbundle->bond = bond_ref(bond); | |
624 | } | |
625 | ||
626 | if (xbundle->lacp != lacp) { | |
627 | lacp_unref(xbundle->lacp); | |
628 | xbundle->lacp = lacp_ref(lacp); | |
629 | } | |
630 | } | |
631 | ||
632 | static void | |
633 | xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
634 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 | 635 | const struct bfd *bfd, const struct lldp *lldp, int stp_port_no, |
f025bcb7 | 636 | const struct rstp_port* rstp_port, |
84f0f298 RW |
637 | enum ofputil_port_config config, enum ofputil_port_state state, |
638 | bool is_tunnel, bool may_enable) | |
639 | { | |
640 | xport->config = config; | |
641 | xport->state = state; | |
642 | xport->stp_port_no = stp_port_no; | |
643 | xport->is_tunnel = is_tunnel; | |
644 | xport->may_enable = may_enable; | |
645 | xport->odp_port = odp_port; | |
646 | ||
f025bcb7 JR |
647 | if (xport->rstp_port != rstp_port) { |
648 | rstp_port_unref(xport->rstp_port); | |
649 | xport->rstp_port = rstp_port_ref(rstp_port); | |
650 | } | |
651 | ||
84f0f298 RW |
652 | if (xport->cfm != cfm) { |
653 | cfm_unref(xport->cfm); | |
654 | xport->cfm = cfm_ref(cfm); | |
655 | } | |
656 | ||
657 | if (xport->bfd != bfd) { | |
658 | bfd_unref(xport->bfd); | |
659 | xport->bfd = bfd_ref(bfd); | |
660 | } | |
661 | ||
0477baa9 DF |
662 | if (xport->lldp != lldp) { |
663 | lldp_unref(xport->lldp); | |
664 | xport->lldp = lldp_ref(lldp); | |
665 | } | |
666 | ||
84f0f298 RW |
667 | if (xport->netdev != netdev) { |
668 | netdev_close(xport->netdev); | |
669 | xport->netdev = netdev_ref(netdev); | |
670 | } | |
671 | } | |
672 | ||
673 | static void | |
674 | xlate_xbridge_copy(struct xbridge *xbridge) | |
675 | { | |
676 | struct xbundle *xbundle; | |
677 | struct xport *xport; | |
678 | struct xbridge *new_xbridge = xzalloc(sizeof *xbridge); | |
679 | new_xbridge->ofproto = xbridge->ofproto; | |
680 | new_xbridge->name = xstrdup(xbridge->name); | |
681 | xlate_xbridge_init(new_xcfg, new_xbridge); | |
682 | ||
683 | xlate_xbridge_set(new_xbridge, | |
34dd0d78 | 684 | xbridge->dpif, xbridge->ml, xbridge->stp, |
9efd308e DV |
685 | xbridge->rstp, xbridge->ms, xbridge->mbridge, |
686 | xbridge->sflow, xbridge->ipfix, xbridge->netflow, | |
b440dd8c JS |
687 | xbridge->forward_bpdu, xbridge->has_in_band, |
688 | &xbridge->support); | |
84f0f298 RW |
689 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
690 | xlate_xbundle_copy(new_xbridge, xbundle); | |
691 | } | |
692 | ||
693 | /* Copy xports which are not part of a xbundle */ | |
694 | HMAP_FOR_EACH (xport, ofp_node, &xbridge->xports) { | |
695 | if (!xport->xbundle) { | |
696 | xlate_xport_copy(new_xbridge, NULL, xport); | |
697 | } | |
698 | } | |
699 | } | |
700 | ||
701 | static void | |
702 | xlate_xbundle_copy(struct xbridge *xbridge, struct xbundle *xbundle) | |
703 | { | |
704 | struct xport *xport; | |
705 | struct xbundle *new_xbundle = xzalloc(sizeof *xbundle); | |
706 | new_xbundle->ofbundle = xbundle->ofbundle; | |
707 | new_xbundle->xbridge = xbridge; | |
708 | new_xbundle->name = xstrdup(xbundle->name); | |
709 | xlate_xbundle_init(new_xcfg, new_xbundle); | |
710 | ||
711 | xlate_xbundle_set(new_xbundle, xbundle->vlan_mode, | |
712 | xbundle->vlan, xbundle->trunks, | |
713 | xbundle->use_priority_tags, xbundle->bond, xbundle->lacp, | |
714 | xbundle->floodable); | |
715 | LIST_FOR_EACH (xport, bundle_node, &xbundle->xports) { | |
716 | xlate_xport_copy(xbridge, new_xbundle, xport); | |
717 | } | |
718 | } | |
719 | ||
720 | static void | |
721 | xlate_xport_copy(struct xbridge *xbridge, struct xbundle *xbundle, | |
722 | struct xport *xport) | |
723 | { | |
724 | struct skb_priority_to_dscp *pdscp, *new_pdscp; | |
725 | struct xport *new_xport = xzalloc(sizeof *xport); | |
726 | new_xport->ofport = xport->ofport; | |
727 | new_xport->ofp_port = xport->ofp_port; | |
728 | new_xport->xbridge = xbridge; | |
729 | xlate_xport_init(new_xcfg, new_xport); | |
730 | ||
731 | xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm, | |
0477baa9 DF |
732 | xport->bfd, xport->lldp, xport->stp_port_no, |
733 | xport->rstp_port, xport->config, xport->state, | |
734 | xport->is_tunnel, xport->may_enable); | |
84f0f298 RW |
735 | |
736 | if (xport->peer) { | |
737 | struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport); | |
738 | if (peer) { | |
739 | new_xport->peer = peer; | |
740 | new_xport->peer->peer = new_xport; | |
741 | } | |
742 | } | |
743 | ||
744 | if (xbundle) { | |
745 | new_xport->xbundle = xbundle; | |
746 | list_insert(&new_xport->xbundle->xports, &new_xport->bundle_node); | |
747 | } | |
748 | ||
749 | HMAP_FOR_EACH (pdscp, hmap_node, &xport->skb_priorities) { | |
750 | new_pdscp = xmalloc(sizeof *pdscp); | |
751 | new_pdscp->skb_priority = pdscp->skb_priority; | |
752 | new_pdscp->dscp = pdscp->dscp; | |
753 | hmap_insert(&new_xport->skb_priorities, &new_pdscp->hmap_node, | |
754 | hash_int(new_pdscp->skb_priority, 0)); | |
755 | } | |
756 | } | |
757 | ||
758 | /* Sets the current xlate configuration to new_xcfg and frees the old xlate | |
759 | * configuration in xcfgp. | |
760 | * | |
761 | * This needs to be called after editing the xlate configuration. | |
762 | * | |
763 | * Functions that edit the new xlate configuration are | |
764 | * xlate_<ofport/bundle/ofport>_set and xlate_<ofport/bundle/ofport>_remove. | |
765 | * | |
766 | * A sample workflow: | |
767 | * | |
768 | * xlate_txn_start(); | |
769 | * ... | |
770 | * edit_xlate_configuration(); | |
771 | * ... | |
772 | * xlate_txn_commit(); */ | |
46c88433 | 773 | void |
84f0f298 RW |
774 | xlate_txn_commit(void) |
775 | { | |
776 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
777 | ||
778 | ovsrcu_set(&xcfgp, new_xcfg); | |
40a9c4c2 AW |
779 | ovsrcu_synchronize(); |
780 | xlate_xcfg_free(xcfg); | |
84f0f298 RW |
781 | new_xcfg = NULL; |
782 | } | |
783 | ||
784 | /* Copies the current xlate configuration in xcfgp to new_xcfg. | |
785 | * | |
786 | * This needs to be called prior to editing the xlate configuration. */ | |
787 | void | |
788 | xlate_txn_start(void) | |
789 | { | |
790 | struct xbridge *xbridge; | |
791 | struct xlate_cfg *xcfg; | |
792 | ||
793 | ovs_assert(!new_xcfg); | |
794 | ||
795 | new_xcfg = xmalloc(sizeof *new_xcfg); | |
796 | hmap_init(&new_xcfg->xbridges); | |
797 | hmap_init(&new_xcfg->xbundles); | |
798 | hmap_init(&new_xcfg->xports); | |
799 | ||
800 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
801 | if (!xcfg) { | |
802 | return; | |
803 | } | |
804 | ||
805 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
806 | xlate_xbridge_copy(xbridge); | |
807 | } | |
808 | } | |
809 | ||
810 | ||
811 | static void | |
812 | xlate_xcfg_free(struct xlate_cfg *xcfg) | |
813 | { | |
814 | struct xbridge *xbridge, *next_xbridge; | |
815 | ||
816 | if (!xcfg) { | |
817 | return; | |
818 | } | |
819 | ||
820 | HMAP_FOR_EACH_SAFE (xbridge, next_xbridge, hmap_node, &xcfg->xbridges) { | |
821 | xlate_xbridge_remove(xcfg, xbridge); | |
822 | } | |
823 | ||
824 | hmap_destroy(&xcfg->xbridges); | |
825 | hmap_destroy(&xcfg->xbundles); | |
826 | hmap_destroy(&xcfg->xports); | |
827 | free(xcfg); | |
828 | } | |
829 | ||
830 | void | |
831 | xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, | |
34dd0d78 | 832 | struct dpif *dpif, |
84f0f298 | 833 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 834 | struct rstp *rstp, const struct mcast_snooping *ms, |
84f0f298 RW |
835 | const struct mbridge *mbridge, |
836 | const struct dpif_sflow *sflow, | |
837 | const struct dpif_ipfix *ipfix, | |
2f47cdf4 | 838 | const struct netflow *netflow, |
b440dd8c JS |
839 | bool forward_bpdu, bool has_in_band, |
840 | const struct dpif_backer_support *support) | |
84f0f298 RW |
841 | { |
842 | struct xbridge *xbridge; | |
843 | ||
844 | ovs_assert(new_xcfg); | |
845 | ||
846 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
847 | if (!xbridge) { | |
848 | xbridge = xzalloc(sizeof *xbridge); | |
849 | xbridge->ofproto = ofproto; | |
850 | ||
851 | xlate_xbridge_init(new_xcfg, xbridge); | |
852 | } | |
853 | ||
854 | free(xbridge->name); | |
855 | xbridge->name = xstrdup(name); | |
856 | ||
34dd0d78 | 857 | xlate_xbridge_set(xbridge, dpif, ml, stp, rstp, ms, mbridge, sflow, ipfix, |
b440dd8c | 858 | netflow, forward_bpdu, has_in_band, support); |
84f0f298 RW |
859 | } |
860 | ||
861 | static void | |
862 | xlate_xbridge_remove(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
46c88433 | 863 | { |
46c88433 EJ |
864 | struct xbundle *xbundle, *next_xbundle; |
865 | struct xport *xport, *next_xport; | |
866 | ||
867 | if (!xbridge) { | |
868 | return; | |
869 | } | |
870 | ||
871 | HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) { | |
84f0f298 | 872 | xlate_xport_remove(xcfg, xport); |
46c88433 EJ |
873 | } |
874 | ||
875 | LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) { | |
84f0f298 | 876 | xlate_xbundle_remove(xcfg, xbundle); |
46c88433 EJ |
877 | } |
878 | ||
84f0f298 | 879 | hmap_remove(&xcfg->xbridges, &xbridge->hmap_node); |
795cc5c1 | 880 | mac_learning_unref(xbridge->ml); |
6d95c4e8 | 881 | mcast_snooping_unref(xbridge->ms); |
795cc5c1 EJ |
882 | mbridge_unref(xbridge->mbridge); |
883 | dpif_sflow_unref(xbridge->sflow); | |
884 | dpif_ipfix_unref(xbridge->ipfix); | |
885 | stp_unref(xbridge->stp); | |
9efd308e | 886 | rstp_unref(xbridge->rstp); |
795cc5c1 | 887 | hmap_destroy(&xbridge->xports); |
46c88433 EJ |
888 | free(xbridge->name); |
889 | free(xbridge); | |
890 | } | |
891 | ||
84f0f298 RW |
892 | void |
893 | xlate_remove_ofproto(struct ofproto_dpif *ofproto) | |
894 | { | |
895 | struct xbridge *xbridge; | |
896 | ||
897 | ovs_assert(new_xcfg); | |
898 | ||
899 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
900 | xlate_xbridge_remove(new_xcfg, xbridge); | |
901 | } | |
902 | ||
46c88433 EJ |
903 | void |
904 | xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
905 | const char *name, enum port_vlan_mode vlan_mode, int vlan, | |
906 | unsigned long *trunks, bool use_priority_tags, | |
907 | const struct bond *bond, const struct lacp *lacp, | |
908 | bool floodable) | |
909 | { | |
84f0f298 | 910 | struct xbundle *xbundle; |
46c88433 | 911 | |
84f0f298 RW |
912 | ovs_assert(new_xcfg); |
913 | ||
914 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
46c88433 EJ |
915 | if (!xbundle) { |
916 | xbundle = xzalloc(sizeof *xbundle); | |
917 | xbundle->ofbundle = ofbundle; | |
84f0f298 | 918 | xbundle->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 919 | |
84f0f298 | 920 | xlate_xbundle_init(new_xcfg, xbundle); |
46c88433 EJ |
921 | } |
922 | ||
46c88433 EJ |
923 | free(xbundle->name); |
924 | xbundle->name = xstrdup(name); | |
925 | ||
84f0f298 RW |
926 | xlate_xbundle_set(xbundle, vlan_mode, vlan, trunks, |
927 | use_priority_tags, bond, lacp, floodable); | |
46c88433 EJ |
928 | } |
929 | ||
84f0f298 RW |
930 | static void |
931 | xlate_xbundle_remove(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
46c88433 | 932 | { |
5f03c983 | 933 | struct xport *xport; |
46c88433 EJ |
934 | |
935 | if (!xbundle) { | |
936 | return; | |
937 | } | |
938 | ||
5f03c983 | 939 | LIST_FOR_EACH_POP (xport, bundle_node, &xbundle->xports) { |
46c88433 EJ |
940 | xport->xbundle = NULL; |
941 | } | |
942 | ||
84f0f298 | 943 | hmap_remove(&xcfg->xbundles, &xbundle->hmap_node); |
46c88433 EJ |
944 | list_remove(&xbundle->list_node); |
945 | bond_unref(xbundle->bond); | |
946 | lacp_unref(xbundle->lacp); | |
947 | free(xbundle->name); | |
948 | free(xbundle); | |
949 | } | |
950 | ||
84f0f298 RW |
951 | void |
952 | xlate_bundle_remove(struct ofbundle *ofbundle) | |
953 | { | |
954 | struct xbundle *xbundle; | |
955 | ||
956 | ovs_assert(new_xcfg); | |
957 | ||
958 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
959 | xlate_xbundle_remove(new_xcfg, xbundle); | |
960 | } | |
961 | ||
46c88433 EJ |
962 | void |
963 | xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
964 | struct ofport_dpif *ofport, ofp_port_t ofp_port, | |
965 | odp_port_t odp_port, const struct netdev *netdev, | |
966 | const struct cfm *cfm, const struct bfd *bfd, | |
0477baa9 DF |
967 | const struct lldp *lldp, struct ofport_dpif *peer, |
968 | int stp_port_no, const struct rstp_port *rstp_port, | |
55954f6e | 969 | const struct ofproto_port_queue *qdscp_list, size_t n_qdscp, |
dd8cd4b4 SH |
970 | enum ofputil_port_config config, |
971 | enum ofputil_port_state state, bool is_tunnel, | |
9d189a50 | 972 | bool may_enable) |
46c88433 | 973 | { |
55954f6e | 974 | size_t i; |
84f0f298 RW |
975 | struct xport *xport; |
976 | ||
977 | ovs_assert(new_xcfg); | |
46c88433 | 978 | |
84f0f298 | 979 | xport = xport_lookup(new_xcfg, ofport); |
46c88433 EJ |
980 | if (!xport) { |
981 | xport = xzalloc(sizeof *xport); | |
982 | xport->ofport = ofport; | |
84f0f298 | 983 | xport->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 EJ |
984 | xport->ofp_port = ofp_port; |
985 | ||
84f0f298 | 986 | xlate_xport_init(new_xcfg, xport); |
46c88433 EJ |
987 | } |
988 | ||
989 | ovs_assert(xport->ofp_port == ofp_port); | |
990 | ||
0477baa9 DF |
991 | xlate_xport_set(xport, odp_port, netdev, cfm, bfd, lldp, |
992 | stp_port_no, rstp_port, config, state, is_tunnel, | |
993 | may_enable); | |
46c88433 EJ |
994 | |
995 | if (xport->peer) { | |
996 | xport->peer->peer = NULL; | |
997 | } | |
84f0f298 | 998 | xport->peer = xport_lookup(new_xcfg, peer); |
46c88433 EJ |
999 | if (xport->peer) { |
1000 | xport->peer->peer = xport; | |
1001 | } | |
1002 | ||
1003 | if (xport->xbundle) { | |
1004 | list_remove(&xport->bundle_node); | |
1005 | } | |
84f0f298 | 1006 | xport->xbundle = xbundle_lookup(new_xcfg, ofbundle); |
46c88433 EJ |
1007 | if (xport->xbundle) { |
1008 | list_insert(&xport->xbundle->xports, &xport->bundle_node); | |
1009 | } | |
55954f6e EJ |
1010 | |
1011 | clear_skb_priorities(xport); | |
1012 | for (i = 0; i < n_qdscp; i++) { | |
1013 | struct skb_priority_to_dscp *pdscp; | |
1014 | uint32_t skb_priority; | |
1015 | ||
89a8a7f0 EJ |
1016 | if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue, |
1017 | &skb_priority)) { | |
55954f6e EJ |
1018 | continue; |
1019 | } | |
1020 | ||
1021 | pdscp = xmalloc(sizeof *pdscp); | |
1022 | pdscp->skb_priority = skb_priority; | |
1023 | pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK; | |
1024 | hmap_insert(&xport->skb_priorities, &pdscp->hmap_node, | |
1025 | hash_int(pdscp->skb_priority, 0)); | |
1026 | } | |
46c88433 EJ |
1027 | } |
1028 | ||
84f0f298 RW |
1029 | static void |
1030 | xlate_xport_remove(struct xlate_cfg *xcfg, struct xport *xport) | |
46c88433 | 1031 | { |
46c88433 EJ |
1032 | if (!xport) { |
1033 | return; | |
1034 | } | |
1035 | ||
1036 | if (xport->peer) { | |
1037 | xport->peer->peer = NULL; | |
1038 | xport->peer = NULL; | |
1039 | } | |
1040 | ||
e621a12d EJ |
1041 | if (xport->xbundle) { |
1042 | list_remove(&xport->bundle_node); | |
1043 | } | |
1044 | ||
55954f6e EJ |
1045 | clear_skb_priorities(xport); |
1046 | hmap_destroy(&xport->skb_priorities); | |
1047 | ||
84f0f298 | 1048 | hmap_remove(&xcfg->xports, &xport->hmap_node); |
46c88433 EJ |
1049 | hmap_remove(&xport->xbridge->xports, &xport->ofp_node); |
1050 | ||
1051 | netdev_close(xport->netdev); | |
f025bcb7 | 1052 | rstp_port_unref(xport->rstp_port); |
46c88433 EJ |
1053 | cfm_unref(xport->cfm); |
1054 | bfd_unref(xport->bfd); | |
0477baa9 | 1055 | lldp_unref(xport->lldp); |
46c88433 EJ |
1056 | free(xport); |
1057 | } | |
1058 | ||
84f0f298 RW |
1059 | void |
1060 | xlate_ofport_remove(struct ofport_dpif *ofport) | |
1061 | { | |
1062 | struct xport *xport; | |
1063 | ||
1064 | ovs_assert(new_xcfg); | |
1065 | ||
1066 | xport = xport_lookup(new_xcfg, ofport); | |
1067 | xlate_xport_remove(new_xcfg, xport); | |
1068 | } | |
1069 | ||
ef377a58 JR |
1070 | static struct ofproto_dpif * |
1071 | xlate_lookup_ofproto_(const struct dpif_backer *backer, const struct flow *flow, | |
1072 | ofp_port_t *ofp_in_port, const struct xport **xportp) | |
1073 | { | |
e672ff9b | 1074 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
ef377a58 | 1075 | const struct xport *xport; |
f9038ef6 | 1076 | |
e672ff9b JR |
1077 | xport = xport_lookup(xcfg, tnl_port_should_receive(flow) |
1078 | ? tnl_port_receive(flow) | |
1079 | : odp_port_to_ofport(backer, flow->in_port.odp_port)); | |
1080 | if (OVS_UNLIKELY(!xport)) { | |
1081 | return NULL; | |
ef377a58 | 1082 | } |
e672ff9b | 1083 | *xportp = xport; |
f9038ef6 | 1084 | if (ofp_in_port) { |
e672ff9b | 1085 | *ofp_in_port = xport->ofp_port; |
f9038ef6 | 1086 | } |
e672ff9b | 1087 | return xport->xbridge->ofproto; |
ef377a58 JR |
1088 | } |
1089 | ||
1090 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively) | |
1091 | * returns the corresponding struct ofproto_dpif and OpenFlow port number. */ | |
1092 | struct ofproto_dpif * | |
1093 | xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow, | |
1094 | ofp_port_t *ofp_in_port) | |
1095 | { | |
1096 | const struct xport *xport; | |
1097 | ||
1098 | return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); | |
1099 | } | |
1100 | ||
cc377352 | 1101 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively), |
ef377a58 | 1102 | * optionally populates 'ofproto' with the ofproto_dpif, 'ofp_in_port' with the |
cc377352 | 1103 | * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate |
dcc2c6cd JR |
1104 | * handles for those protocols if they're enabled. Caller may use the returned |
1105 | * pointers until quiescing, for longer term use additional references must | |
1106 | * be taken. | |
8449c4d6 | 1107 | * |
f9038ef6 | 1108 | * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto. |
ef377a58 | 1109 | */ |
8449c4d6 | 1110 | int |
5c476ea3 JR |
1111 | xlate_lookup(const struct dpif_backer *backer, const struct flow *flow, |
1112 | struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix, | |
1113 | struct dpif_sflow **sflow, struct netflow **netflow, | |
1114 | ofp_port_t *ofp_in_port) | |
8449c4d6 | 1115 | { |
ef377a58 | 1116 | struct ofproto_dpif *ofproto; |
84f0f298 | 1117 | const struct xport *xport; |
8449c4d6 | 1118 | |
ef377a58 | 1119 | ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); |
8449c4d6 | 1120 | |
f9038ef6 | 1121 | if (!ofproto) { |
cc377352 | 1122 | return ENODEV; |
8449c4d6 | 1123 | } |
8449c4d6 | 1124 | |
ef377a58 JR |
1125 | if (ofprotop) { |
1126 | *ofprotop = ofproto; | |
8449c4d6 EJ |
1127 | } |
1128 | ||
1dfdb9b3 | 1129 | if (ipfix) { |
f9038ef6 | 1130 | *ipfix = xport ? xport->xbridge->ipfix : NULL; |
1dfdb9b3 EJ |
1131 | } |
1132 | ||
1133 | if (sflow) { | |
f9038ef6 | 1134 | *sflow = xport ? xport->xbridge->sflow : NULL; |
1dfdb9b3 EJ |
1135 | } |
1136 | ||
1137 | if (netflow) { | |
f9038ef6 | 1138 | *netflow = xport ? xport->xbridge->netflow : NULL; |
1dfdb9b3 | 1139 | } |
f9038ef6 | 1140 | |
cc377352 | 1141 | return 0; |
8449c4d6 EJ |
1142 | } |
1143 | ||
46c88433 | 1144 | static struct xbridge * |
84f0f298 | 1145 | xbridge_lookup(struct xlate_cfg *xcfg, const struct ofproto_dpif *ofproto) |
46c88433 | 1146 | { |
84f0f298 | 1147 | struct hmap *xbridges; |
46c88433 EJ |
1148 | struct xbridge *xbridge; |
1149 | ||
84f0f298 | 1150 | if (!ofproto || !xcfg) { |
5e6af486 EJ |
1151 | return NULL; |
1152 | } | |
1153 | ||
84f0f298 RW |
1154 | xbridges = &xcfg->xbridges; |
1155 | ||
46c88433 | 1156 | HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0), |
84f0f298 | 1157 | xbridges) { |
46c88433 EJ |
1158 | if (xbridge->ofproto == ofproto) { |
1159 | return xbridge; | |
1160 | } | |
1161 | } | |
1162 | return NULL; | |
1163 | } | |
1164 | ||
1165 | static struct xbundle * | |
84f0f298 | 1166 | xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle) |
46c88433 | 1167 | { |
84f0f298 | 1168 | struct hmap *xbundles; |
46c88433 EJ |
1169 | struct xbundle *xbundle; |
1170 | ||
84f0f298 | 1171 | if (!ofbundle || !xcfg) { |
5e6af486 EJ |
1172 | return NULL; |
1173 | } | |
1174 | ||
84f0f298 RW |
1175 | xbundles = &xcfg->xbundles; |
1176 | ||
46c88433 | 1177 | HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), |
84f0f298 | 1178 | xbundles) { |
46c88433 EJ |
1179 | if (xbundle->ofbundle == ofbundle) { |
1180 | return xbundle; | |
1181 | } | |
1182 | } | |
1183 | return NULL; | |
1184 | } | |
1185 | ||
1186 | static struct xport * | |
84f0f298 | 1187 | xport_lookup(struct xlate_cfg *xcfg, const struct ofport_dpif *ofport) |
46c88433 | 1188 | { |
84f0f298 | 1189 | struct hmap *xports; |
46c88433 EJ |
1190 | struct xport *xport; |
1191 | ||
84f0f298 | 1192 | if (!ofport || !xcfg) { |
5e6af486 EJ |
1193 | return NULL; |
1194 | } | |
1195 | ||
84f0f298 RW |
1196 | xports = &xcfg->xports; |
1197 | ||
46c88433 | 1198 | HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0), |
84f0f298 | 1199 | xports) { |
46c88433 EJ |
1200 | if (xport->ofport == ofport) { |
1201 | return xport; | |
1202 | } | |
1203 | } | |
1204 | return NULL; | |
1205 | } | |
1206 | ||
40085e56 EJ |
1207 | static struct stp_port * |
1208 | xport_get_stp_port(const struct xport *xport) | |
1209 | { | |
92cf817b | 1210 | return xport->xbridge->stp && xport->stp_port_no != -1 |
40085e56 EJ |
1211 | ? stp_get_port(xport->xbridge->stp, xport->stp_port_no) |
1212 | : NULL; | |
1213 | } | |
9d189a50 | 1214 | |
0d1cee12 | 1215 | static bool |
9d189a50 EJ |
1216 | xport_stp_learn_state(const struct xport *xport) |
1217 | { | |
40085e56 | 1218 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1219 | return sp |
1220 | ? stp_learn_in_state(stp_port_get_state(sp)) | |
1221 | : true; | |
9d189a50 EJ |
1222 | } |
1223 | ||
1224 | static bool | |
1225 | xport_stp_forward_state(const struct xport *xport) | |
1226 | { | |
40085e56 | 1227 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1228 | return sp |
1229 | ? stp_forward_in_state(stp_port_get_state(sp)) | |
1230 | : true; | |
9d189a50 EJ |
1231 | } |
1232 | ||
0d1cee12 | 1233 | static bool |
bacdb85a | 1234 | xport_stp_should_forward_bpdu(const struct xport *xport) |
0d1cee12 K |
1235 | { |
1236 | struct stp_port *sp = xport_get_stp_port(xport); | |
bacdb85a | 1237 | return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED); |
0d1cee12 K |
1238 | } |
1239 | ||
9d189a50 EJ |
1240 | /* Returns true if STP should process 'flow'. Sets fields in 'wc' that |
1241 | * were used to make the determination.*/ | |
1242 | static bool | |
1243 | stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc) | |
1244 | { | |
bbbca389 | 1245 | /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */ |
9d189a50 | 1246 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
bbbca389 | 1247 | return is_stp(flow); |
9d189a50 EJ |
1248 | } |
1249 | ||
1250 | static void | |
cf62fa4c | 1251 | stp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9d189a50 | 1252 | { |
40085e56 | 1253 | struct stp_port *sp = xport_get_stp_port(xport); |
cf62fa4c PS |
1254 | struct dp_packet payload = *packet; |
1255 | struct eth_header *eth = dp_packet_data(&payload); | |
9d189a50 EJ |
1256 | |
1257 | /* Sink packets on ports that have STP disabled when the bridge has | |
1258 | * STP enabled. */ | |
1259 | if (!sp || stp_port_get_state(sp) == STP_DISABLED) { | |
1260 | return; | |
1261 | } | |
1262 | ||
1263 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1264 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1265 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9d189a50 EJ |
1266 | } |
1267 | ||
cf62fa4c PS |
1268 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1269 | stp_received_bpdu(sp, dp_packet_data(&payload), dp_packet_size(&payload)); | |
9d189a50 EJ |
1270 | } |
1271 | } | |
1272 | ||
f025bcb7 JR |
1273 | static enum rstp_state |
1274 | xport_get_rstp_port_state(const struct xport *xport) | |
9efd308e | 1275 | { |
f025bcb7 JR |
1276 | return xport->rstp_port |
1277 | ? rstp_port_get_state(xport->rstp_port) | |
1278 | : RSTP_DISABLED; | |
9efd308e DV |
1279 | } |
1280 | ||
1281 | static bool | |
1282 | xport_rstp_learn_state(const struct xport *xport) | |
1283 | { | |
4b5f1996 DV |
1284 | return xport->xbridge->rstp && xport->rstp_port |
1285 | ? rstp_learn_in_state(xport_get_rstp_port_state(xport)) | |
1286 | : true; | |
9efd308e DV |
1287 | } |
1288 | ||
1289 | static bool | |
1290 | xport_rstp_forward_state(const struct xport *xport) | |
1291 | { | |
4b5f1996 DV |
1292 | return xport->xbridge->rstp && xport->rstp_port |
1293 | ? rstp_forward_in_state(xport_get_rstp_port_state(xport)) | |
1294 | : true; | |
9efd308e DV |
1295 | } |
1296 | ||
1297 | static bool | |
1298 | xport_rstp_should_manage_bpdu(const struct xport *xport) | |
1299 | { | |
f025bcb7 | 1300 | return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport)); |
9efd308e DV |
1301 | } |
1302 | ||
1303 | static void | |
cf62fa4c | 1304 | rstp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9efd308e | 1305 | { |
cf62fa4c PS |
1306 | struct dp_packet payload = *packet; |
1307 | struct eth_header *eth = dp_packet_data(&payload); | |
9efd308e | 1308 | |
f025bcb7 JR |
1309 | /* Sink packets on ports that have no RSTP. */ |
1310 | if (!xport->rstp_port) { | |
9efd308e DV |
1311 | return; |
1312 | } | |
1313 | ||
1314 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1315 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1316 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9efd308e DV |
1317 | } |
1318 | ||
cf62fa4c PS |
1319 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1320 | rstp_port_received_bpdu(xport->rstp_port, dp_packet_data(&payload), | |
1321 | dp_packet_size(&payload)); | |
9efd308e DV |
1322 | } |
1323 | } | |
1324 | ||
46c88433 EJ |
1325 | static struct xport * |
1326 | get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1327 | { | |
1328 | struct xport *xport; | |
1329 | ||
1330 | HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port), | |
1331 | &xbridge->xports) { | |
1332 | if (xport->ofp_port == ofp_port) { | |
1333 | return xport; | |
1334 | } | |
1335 | } | |
1336 | return NULL; | |
1337 | } | |
1338 | ||
1339 | static odp_port_t | |
1340 | ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1341 | { | |
1342 | const struct xport *xport = get_ofp_port(xbridge, ofp_port); | |
1343 | return xport ? xport->odp_port : ODPP_NONE; | |
1344 | } | |
1345 | ||
dd8cd4b4 SH |
1346 | static bool |
1347 | odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port) | |
1348 | { | |
086fa873 BP |
1349 | struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1350 | return xport && xport->may_enable; | |
dd8cd4b4 SH |
1351 | } |
1352 | ||
1e684d7d | 1353 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1354 | group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *, |
1355 | int depth); | |
1356 | ||
1357 | static bool | |
1358 | group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) | |
1359 | { | |
1360 | struct group_dpif *group; | |
dd8cd4b4 | 1361 | |
dc25893e AZ |
1362 | if (group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group)) { |
1363 | struct ofputil_bucket *bucket; | |
dd8cd4b4 | 1364 | |
dc25893e AZ |
1365 | bucket = group_first_live_bucket(ctx, group, depth); |
1366 | group_dpif_unref(group); | |
1367 | return bucket == NULL; | |
1368 | } | |
dd8cd4b4 | 1369 | |
dc25893e | 1370 | return false; |
dd8cd4b4 SH |
1371 | } |
1372 | ||
1373 | #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ | |
1374 | ||
1375 | static bool | |
1376 | bucket_is_alive(const struct xlate_ctx *ctx, | |
1e684d7d | 1377 | struct ofputil_bucket *bucket, int depth) |
dd8cd4b4 SH |
1378 | { |
1379 | if (depth >= MAX_LIVENESS_RECURSION) { | |
1380 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
1381 | ||
1382 | VLOG_WARN_RL(&rl, "bucket chaining exceeded %d links", | |
1383 | MAX_LIVENESS_RECURSION); | |
1384 | return false; | |
1385 | } | |
1386 | ||
fdb1999b AZ |
1387 | return (!ofputil_bucket_has_liveness(bucket) |
1388 | || (bucket->watch_port != OFPP_ANY | |
1389 | && odp_port_is_alive(ctx, bucket->watch_port)) | |
1390 | || (bucket->watch_group != OFPG_ANY | |
1391 | && group_is_alive(ctx, bucket->watch_group, depth + 1))); | |
dd8cd4b4 SH |
1392 | } |
1393 | ||
1e684d7d | 1394 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1395 | group_first_live_bucket(const struct xlate_ctx *ctx, |
1396 | const struct group_dpif *group, int depth) | |
1397 | { | |
1398 | struct ofputil_bucket *bucket; | |
ca6ba700 | 1399 | const struct ovs_list *buckets; |
dd8cd4b4 SH |
1400 | |
1401 | group_dpif_get_buckets(group, &buckets); | |
1402 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1403 | if (bucket_is_alive(ctx, bucket, depth)) { | |
1404 | return bucket; | |
1405 | } | |
1406 | } | |
1407 | ||
1408 | return NULL; | |
1409 | } | |
1410 | ||
1e684d7d | 1411 | static struct ofputil_bucket * |
fe7e5749 SH |
1412 | group_best_live_bucket(const struct xlate_ctx *ctx, |
1413 | const struct group_dpif *group, | |
1414 | uint32_t basis) | |
1415 | { | |
1e684d7d | 1416 | struct ofputil_bucket *best_bucket = NULL; |
fe7e5749 SH |
1417 | uint32_t best_score = 0; |
1418 | int i = 0; | |
1419 | ||
1e684d7d | 1420 | struct ofputil_bucket *bucket; |
ca6ba700 | 1421 | const struct ovs_list *buckets; |
fe7e5749 SH |
1422 | |
1423 | group_dpif_get_buckets(group, &buckets); | |
1424 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1425 | if (bucket_is_alive(ctx, bucket, 0)) { | |
7cb279c2 | 1426 | uint32_t score = (hash_int(i, basis) & 0xffff) * bucket->weight; |
fe7e5749 SH |
1427 | if (score >= best_score) { |
1428 | best_bucket = bucket; | |
1429 | best_score = score; | |
1430 | } | |
1431 | } | |
1432 | i++; | |
1433 | } | |
1434 | ||
1435 | return best_bucket; | |
1436 | } | |
1437 | ||
9583bc14 | 1438 | static bool |
46c88433 | 1439 | xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan) |
9583bc14 EJ |
1440 | { |
1441 | return (bundle->vlan_mode != PORT_VLAN_ACCESS | |
1442 | && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan))); | |
1443 | } | |
1444 | ||
1445 | static bool | |
46c88433 EJ |
1446 | xbundle_includes_vlan(const struct xbundle *xbundle, uint16_t vlan) |
1447 | { | |
1448 | return vlan == xbundle->vlan || xbundle_trunks_vlan(xbundle, vlan); | |
1449 | } | |
1450 | ||
1451 | static mirror_mask_t | |
1452 | xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle) | |
1453 | { | |
1454 | return xbundle != &ofpp_none_bundle | |
1455 | ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle) | |
1456 | : 0; | |
1457 | } | |
1458 | ||
1459 | static mirror_mask_t | |
1460 | xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1461 | { |
46c88433 EJ |
1462 | return xbundle != &ofpp_none_bundle |
1463 | ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle) | |
1464 | : 0; | |
9583bc14 EJ |
1465 | } |
1466 | ||
46c88433 EJ |
1467 | static mirror_mask_t |
1468 | xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1469 | { |
46c88433 EJ |
1470 | return xbundle != &ofpp_none_bundle |
1471 | ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle) | |
1472 | : 0; | |
1473 | } | |
1474 | ||
1475 | static struct xbundle * | |
1476 | lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port, | |
1477 | bool warn, struct xport **in_xportp) | |
1478 | { | |
1479 | struct xport *xport; | |
9583bc14 EJ |
1480 | |
1481 | /* Find the port and bundle for the received packet. */ | |
46c88433 EJ |
1482 | xport = get_ofp_port(xbridge, in_port); |
1483 | if (in_xportp) { | |
1484 | *in_xportp = xport; | |
9583bc14 | 1485 | } |
46c88433 EJ |
1486 | if (xport && xport->xbundle) { |
1487 | return xport->xbundle; | |
9583bc14 EJ |
1488 | } |
1489 | ||
6362203b YT |
1490 | /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+), |
1491 | * which a controller may use as the ingress port for traffic that | |
1492 | * it is sourcing. */ | |
1493 | if (in_port == OFPP_CONTROLLER || in_port == OFPP_NONE) { | |
9583bc14 EJ |
1494 | return &ofpp_none_bundle; |
1495 | } | |
1496 | ||
1497 | /* Odd. A few possible reasons here: | |
1498 | * | |
1499 | * - We deleted a port but there are still a few packets queued up | |
1500 | * from it. | |
1501 | * | |
1502 | * - Someone externally added a port (e.g. "ovs-dpctl add-if") that | |
1503 | * we don't know about. | |
1504 | * | |
1505 | * - The ofproto client didn't configure the port as part of a bundle. | |
1506 | * This is particularly likely to happen if a packet was received on the | |
1507 | * port after it was created, but before the client had a chance to | |
1508 | * configure its bundle. | |
1509 | */ | |
1510 | if (warn) { | |
1511 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1512 | ||
1513 | VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown " | |
46c88433 | 1514 | "port %"PRIu16, xbridge->name, in_port); |
9583bc14 EJ |
1515 | } |
1516 | return NULL; | |
1517 | } | |
1518 | ||
1519 | static void | |
1520 | add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) | |
1521 | { | |
46c88433 | 1522 | const struct xbridge *xbridge = ctx->xbridge; |
9583bc14 | 1523 | mirror_mask_t mirrors; |
46c88433 | 1524 | struct xbundle *in_xbundle; |
9583bc14 EJ |
1525 | uint16_t vlan; |
1526 | uint16_t vid; | |
cdf5d3a5 EJ |
1527 | |
1528 | mirrors = ctx->xout->mirrors; | |
1529 | ctx->xout->mirrors = 0; | |
9583bc14 | 1530 | |
46c88433 EJ |
1531 | in_xbundle = lookup_input_bundle(xbridge, orig_flow->in_port.ofp_port, |
1532 | ctx->xin->packet != NULL, NULL); | |
1533 | if (!in_xbundle) { | |
9583bc14 EJ |
1534 | return; |
1535 | } | |
46c88433 | 1536 | mirrors |= xbundle_mirror_src(xbridge, in_xbundle); |
9583bc14 EJ |
1537 | |
1538 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 1539 | if (xbundle_mirror_out(xbridge, in_xbundle)) { |
9583bc14 EJ |
1540 | if (ctx->xin->packet != NULL) { |
1541 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1542 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
1543 | "%s, which is reserved exclusively for mirroring", | |
46c88433 | 1544 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 | 1545 | } |
cc377352 | 1546 | ofpbuf_clear(ctx->xout->odp_actions); |
9583bc14 EJ |
1547 | return; |
1548 | } | |
1549 | ||
1550 | /* Check VLAN. */ | |
1551 | vid = vlan_tci_to_vid(orig_flow->vlan_tci); | |
46c88433 | 1552 | if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { |
9583bc14 EJ |
1553 | return; |
1554 | } | |
46c88433 | 1555 | vlan = input_vid_to_vlan(in_xbundle, vid); |
9583bc14 | 1556 | |
9583bc14 EJ |
1557 | if (!mirrors) { |
1558 | return; | |
1559 | } | |
1560 | ||
1561 | /* Restore the original packet before adding the mirror actions. */ | |
1562 | ctx->xin->flow = *orig_flow; | |
1563 | ||
1564 | while (mirrors) { | |
ec7ceaed EJ |
1565 | mirror_mask_t dup_mirrors; |
1566 | struct ofbundle *out; | |
1567 | unsigned long *vlans; | |
1568 | bool vlan_mirrored; | |
1569 | bool has_mirror; | |
1570 | int out_vlan; | |
1571 | ||
4b6ab2b0 | 1572 | has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors), |
ec7ceaed EJ |
1573 | &vlans, &dup_mirrors, &out, &out_vlan); |
1574 | ovs_assert(has_mirror); | |
1575 | ||
1576 | if (vlans) { | |
9583bc14 EJ |
1577 | ctx->xout->wc.masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK); |
1578 | } | |
ec7ceaed EJ |
1579 | vlan_mirrored = !vlans || bitmap_is_set(vlans, vlan); |
1580 | free(vlans); | |
9583bc14 | 1581 | |
ec7ceaed | 1582 | if (!vlan_mirrored) { |
9583bc14 EJ |
1583 | mirrors = zero_rightmost_1bit(mirrors); |
1584 | continue; | |
1585 | } | |
1586 | ||
ec7ceaed EJ |
1587 | mirrors &= ~dup_mirrors; |
1588 | ctx->xout->mirrors |= dup_mirrors; | |
1589 | if (out) { | |
84f0f298 RW |
1590 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
1591 | struct xbundle *out_xbundle = xbundle_lookup(xcfg, out); | |
46c88433 EJ |
1592 | if (out_xbundle) { |
1593 | output_normal(ctx, out_xbundle, vlan); | |
1594 | } | |
ec7ceaed | 1595 | } else if (vlan != out_vlan |
9583bc14 | 1596 | && !eth_addr_is_reserved(orig_flow->dl_dst)) { |
46c88433 | 1597 | struct xbundle *xbundle; |
9583bc14 | 1598 | |
46c88433 EJ |
1599 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
1600 | if (xbundle_includes_vlan(xbundle, out_vlan) | |
1601 | && !xbundle_mirror_out(xbridge, xbundle)) { | |
1602 | output_normal(ctx, xbundle, out_vlan); | |
9583bc14 EJ |
1603 | } |
1604 | } | |
1605 | } | |
1606 | } | |
1607 | } | |
1608 | ||
1609 | /* Given 'vid', the VID obtained from the 802.1Q header that was received as | |
46c88433 | 1610 | * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle', |
9583bc14 EJ |
1611 | * the bundle on which the packet was received, returns the VLAN to which the |
1612 | * packet belongs. | |
1613 | * | |
1614 | * Both 'vid' and the return value are in the range 0...4095. */ | |
1615 | static uint16_t | |
46c88433 | 1616 | input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid) |
9583bc14 | 1617 | { |
46c88433 | 1618 | switch (in_xbundle->vlan_mode) { |
9583bc14 | 1619 | case PORT_VLAN_ACCESS: |
46c88433 | 1620 | return in_xbundle->vlan; |
9583bc14 EJ |
1621 | break; |
1622 | ||
1623 | case PORT_VLAN_TRUNK: | |
1624 | return vid; | |
1625 | ||
1626 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1627 | case PORT_VLAN_NATIVE_TAGGED: | |
46c88433 | 1628 | return vid ? vid : in_xbundle->vlan; |
9583bc14 EJ |
1629 | |
1630 | default: | |
428b2edd | 1631 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1632 | } |
1633 | } | |
1634 | ||
46c88433 | 1635 | /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'. |
9583bc14 EJ |
1636 | * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs |
1637 | * a warning. | |
1638 | * | |
1639 | * 'vid' should be the VID obtained from the 802.1Q header that was received as | |
1640 | * part of a packet (specify 0 if there was no 802.1Q header), in the range | |
1641 | * 0...4095. */ | |
1642 | static bool | |
46c88433 | 1643 | input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn) |
9583bc14 EJ |
1644 | { |
1645 | /* Allow any VID on the OFPP_NONE port. */ | |
46c88433 | 1646 | if (in_xbundle == &ofpp_none_bundle) { |
9583bc14 EJ |
1647 | return true; |
1648 | } | |
1649 | ||
46c88433 | 1650 | switch (in_xbundle->vlan_mode) { |
9583bc14 EJ |
1651 | case PORT_VLAN_ACCESS: |
1652 | if (vid) { | |
1653 | if (warn) { | |
1654 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1655 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" tagged " |
9583bc14 | 1656 | "packet received on port %s configured as VLAN " |
46c88433 EJ |
1657 | "%"PRIu16" access port", vid, in_xbundle->name, |
1658 | in_xbundle->vlan); | |
9583bc14 EJ |
1659 | } |
1660 | return false; | |
1661 | } | |
1662 | return true; | |
1663 | ||
1664 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1665 | case PORT_VLAN_NATIVE_TAGGED: | |
1666 | if (!vid) { | |
1667 | /* Port must always carry its native VLAN. */ | |
1668 | return true; | |
1669 | } | |
1670 | /* Fall through. */ | |
1671 | case PORT_VLAN_TRUNK: | |
46c88433 | 1672 | if (!xbundle_includes_vlan(in_xbundle, vid)) { |
9583bc14 EJ |
1673 | if (warn) { |
1674 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1675 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" packet " |
9583bc14 | 1676 | "received on port %s not configured for trunking " |
46c88433 | 1677 | "VLAN %"PRIu16, vid, in_xbundle->name, vid); |
9583bc14 EJ |
1678 | } |
1679 | return false; | |
1680 | } | |
1681 | return true; | |
1682 | ||
1683 | default: | |
428b2edd | 1684 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1685 | } |
1686 | ||
1687 | } | |
1688 | ||
1689 | /* Given 'vlan', the VLAN that a packet belongs to, and | |
46c88433 | 1690 | * 'out_xbundle', a bundle on which the packet is to be output, returns the VID |
9583bc14 EJ |
1691 | * that should be included in the 802.1Q header. (If the return value is 0, |
1692 | * then the 802.1Q header should only be included in the packet if there is a | |
1693 | * nonzero PCP.) | |
1694 | * | |
1695 | * Both 'vlan' and the return value are in the range 0...4095. */ | |
1696 | static uint16_t | |
46c88433 | 1697 | output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan) |
9583bc14 | 1698 | { |
46c88433 | 1699 | switch (out_xbundle->vlan_mode) { |
9583bc14 EJ |
1700 | case PORT_VLAN_ACCESS: |
1701 | return 0; | |
1702 | ||
1703 | case PORT_VLAN_TRUNK: | |
1704 | case PORT_VLAN_NATIVE_TAGGED: | |
1705 | return vlan; | |
1706 | ||
1707 | case PORT_VLAN_NATIVE_UNTAGGED: | |
46c88433 | 1708 | return vlan == out_xbundle->vlan ? 0 : vlan; |
9583bc14 EJ |
1709 | |
1710 | default: | |
428b2edd | 1711 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1712 | } |
1713 | } | |
1714 | ||
1715 | static void | |
46c88433 | 1716 | output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, |
9583bc14 EJ |
1717 | uint16_t vlan) |
1718 | { | |
33bf9176 | 1719 | ovs_be16 *flow_tci = &ctx->xin->flow.vlan_tci; |
9583bc14 EJ |
1720 | uint16_t vid; |
1721 | ovs_be16 tci, old_tci; | |
46c88433 | 1722 | struct xport *xport; |
e93ef1c7 JR |
1723 | struct xlate_bond_recirc xr; |
1724 | bool use_recirc = false; | |
9583bc14 | 1725 | |
46c88433 EJ |
1726 | vid = output_vlan_to_vid(out_xbundle, vlan); |
1727 | if (list_is_empty(&out_xbundle->xports)) { | |
1728 | /* Partially configured bundle with no slaves. Drop the packet. */ | |
1729 | return; | |
1730 | } else if (!out_xbundle->bond) { | |
1731 | xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport, | |
1732 | bundle_node); | |
9583bc14 | 1733 | } else { |
84f0f298 | 1734 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
54ecb5a2 | 1735 | struct flow_wildcards *wc = &ctx->xout->wc; |
84f0f298 | 1736 | struct ofport_dpif *ofport; |
adcf00ba | 1737 | |
2494ccd7 | 1738 | if (ctx->xbridge->support.odp.recirc) { |
e93ef1c7 JR |
1739 | use_recirc = bond_may_recirc( |
1740 | out_xbundle->bond, &xr.recirc_id, &xr.hash_basis); | |
adcf00ba | 1741 | |
e93ef1c7 | 1742 | if (use_recirc) { |
adcf00ba | 1743 | /* Only TCP mode uses recirculation. */ |
e93ef1c7 | 1744 | xr.hash_alg = OVS_HASH_ALG_L4; |
adcf00ba | 1745 | bond_update_post_recirc_rules(out_xbundle->bond, false); |
54ecb5a2 AZ |
1746 | |
1747 | /* Recirculation does not require unmasking hash fields. */ | |
1748 | wc = NULL; | |
adcf00ba AZ |
1749 | } |
1750 | } | |
46c88433 | 1751 | |
54ecb5a2 AZ |
1752 | ofport = bond_choose_output_slave(out_xbundle->bond, |
1753 | &ctx->xin->flow, wc, vid); | |
84f0f298 | 1754 | xport = xport_lookup(xcfg, ofport); |
46c88433 EJ |
1755 | |
1756 | if (!xport) { | |
9583bc14 EJ |
1757 | /* No slaves enabled, so drop packet. */ |
1758 | return; | |
1759 | } | |
d6fc5f57 | 1760 | |
e93ef1c7 | 1761 | /* If use_recirc is set, the main thread will handle stats |
b256dc52 | 1762 | * accounting for this bond. */ |
e93ef1c7 | 1763 | if (!use_recirc) { |
b256dc52 JS |
1764 | if (ctx->xin->resubmit_stats) { |
1765 | bond_account(out_xbundle->bond, &ctx->xin->flow, vid, | |
1766 | ctx->xin->resubmit_stats->n_bytes); | |
1767 | } | |
1768 | if (ctx->xin->xcache) { | |
1769 | struct xc_entry *entry; | |
1770 | struct flow *flow; | |
1771 | ||
1772 | flow = &ctx->xin->flow; | |
1773 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND); | |
1774 | entry->u.bond.bond = bond_ref(out_xbundle->bond); | |
1775 | entry->u.bond.flow = xmemdup(flow, sizeof *flow); | |
1776 | entry->u.bond.vid = vid; | |
1777 | } | |
d6fc5f57 | 1778 | } |
9583bc14 EJ |
1779 | } |
1780 | ||
33bf9176 | 1781 | old_tci = *flow_tci; |
9583bc14 | 1782 | tci = htons(vid); |
46c88433 | 1783 | if (tci || out_xbundle->use_priority_tags) { |
33bf9176 | 1784 | tci |= *flow_tci & htons(VLAN_PCP_MASK); |
9583bc14 EJ |
1785 | if (tci) { |
1786 | tci |= htons(VLAN_CFI); | |
1787 | } | |
1788 | } | |
33bf9176 | 1789 | *flow_tci = tci; |
9583bc14 | 1790 | |
e93ef1c7 | 1791 | compose_output_action(ctx, xport->ofp_port, use_recirc ? &xr : NULL); |
33bf9176 | 1792 | *flow_tci = old_tci; |
9583bc14 EJ |
1793 | } |
1794 | ||
1795 | /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after | |
1796 | * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to | |
1797 | * indicate this; newer upstream kernels use gratuitous ARP requests. */ | |
1798 | static bool | |
1799 | is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) | |
1800 | { | |
1801 | if (flow->dl_type != htons(ETH_TYPE_ARP)) { | |
1802 | return false; | |
1803 | } | |
1804 | ||
1805 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1806 | if (!eth_addr_is_broadcast(flow->dl_dst)) { | |
1807 | return false; | |
1808 | } | |
1809 | ||
1810 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
1811 | if (flow->nw_proto == ARP_OP_REPLY) { | |
1812 | return true; | |
1813 | } else if (flow->nw_proto == ARP_OP_REQUEST) { | |
1814 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
1815 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
1816 | ||
1817 | return flow->nw_src == flow->nw_dst; | |
1818 | } else { | |
1819 | return false; | |
1820 | } | |
1821 | } | |
1822 | ||
ff69c24a FL |
1823 | /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or |
1824 | * dropped. Returns true if they may be forwarded, false if they should be | |
1825 | * dropped. | |
1826 | * | |
1827 | * 'in_port' must be the xport that corresponds to flow->in_port. | |
1828 | * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). | |
1829 | * | |
1830 | * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as | |
1831 | * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as | |
1832 | * checked by input_vid_is_valid(). | |
1833 | * | |
1834 | * May also add tags to '*tags', although the current implementation only does | |
1835 | * so in one special case. | |
1836 | */ | |
1837 | static bool | |
1838 | is_admissible(struct xlate_ctx *ctx, struct xport *in_port, | |
1839 | uint16_t vlan) | |
1840 | { | |
1841 | struct xbundle *in_xbundle = in_port->xbundle; | |
1842 | const struct xbridge *xbridge = ctx->xbridge; | |
1843 | struct flow *flow = &ctx->xin->flow; | |
1844 | ||
1845 | /* Drop frames for reserved multicast addresses | |
1846 | * only if forward_bpdu option is absent. */ | |
1847 | if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { | |
1848 | xlate_report(ctx, "packet has reserved destination MAC, dropping"); | |
1849 | return false; | |
1850 | } | |
1851 | ||
1852 | if (in_xbundle->bond) { | |
1853 | struct mac_entry *mac; | |
1854 | ||
1855 | switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, | |
1856 | flow->dl_dst)) { | |
1857 | case BV_ACCEPT: | |
1858 | break; | |
1859 | ||
1860 | case BV_DROP: | |
1861 | xlate_report(ctx, "bonding refused admissibility, dropping"); | |
1862 | return false; | |
1863 | ||
1864 | case BV_DROP_IF_MOVED: | |
1865 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
1866 | mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan); | |
9d078ec2 BP |
1867 | if (mac |
1868 | && mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle | |
1869 | && (!is_gratuitous_arp(flow, &ctx->xout->wc) | |
1870 | || mac_entry_is_grat_arp_locked(mac))) { | |
ff69c24a FL |
1871 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
1872 | xlate_report(ctx, "SLB bond thinks this packet looped back, " | |
1873 | "dropping"); | |
1874 | return false; | |
1875 | } | |
1876 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
1877 | break; | |
1878 | } | |
1879 | } | |
1880 | ||
1881 | return true; | |
1882 | } | |
1883 | ||
ee047520 BP |
1884 | /* Checks whether a MAC learning update is necessary for MAC learning table |
1885 | * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in | |
1886 | * 'vlan'. | |
1887 | * | |
1888 | * Most packets processed through the MAC learning table do not actually | |
1889 | * change it in any way. This function requires only a read lock on the MAC | |
1890 | * learning table, so it is much cheaper in this common case. | |
1891 | * | |
1892 | * Keep the code here synchronized with that in update_learning_table__() | |
1893 | * below. */ | |
1894 | static bool | |
1895 | is_mac_learning_update_needed(const struct mac_learning *ml, | |
1896 | const struct flow *flow, | |
1897 | struct flow_wildcards *wc, | |
1898 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 1899 | OVS_REQ_RDLOCK(ml->rwlock) |
9583bc14 EJ |
1900 | { |
1901 | struct mac_entry *mac; | |
1902 | ||
ee047520 BP |
1903 | if (!mac_learning_may_learn(ml, flow->dl_src, vlan)) { |
1904 | return false; | |
1905 | } | |
1906 | ||
1907 | mac = mac_learning_lookup(ml, flow->dl_src, vlan); | |
1908 | if (!mac || mac_entry_age(ml, mac)) { | |
1909 | return true; | |
9583bc14 EJ |
1910 | } |
1911 | ||
ee047520 BP |
1912 | if (is_gratuitous_arp(flow, wc)) { |
1913 | /* We don't want to learn from gratuitous ARP packets that are | |
1914 | * reflected back over bond slaves so we lock the learning table. */ | |
1915 | if (!in_xbundle->bond) { | |
1916 | return true; | |
1917 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
1918 | return false; | |
1919 | } | |
1920 | } | |
1921 | ||
9d078ec2 | 1922 | return mac_entry_get_port(ml, mac) != in_xbundle->ofbundle; |
ee047520 BP |
1923 | } |
1924 | ||
1925 | ||
1926 | /* Updates MAC learning table 'ml' given that a packet matching 'flow' was | |
1927 | * received on 'in_xbundle' in 'vlan'. | |
1928 | * | |
1929 | * This code repeats all the checks in is_mac_learning_update_needed() because | |
1930 | * the lock was released between there and here and thus the MAC learning state | |
1931 | * could have changed. | |
1932 | * | |
1933 | * Keep the code here synchronized with that in is_mac_learning_update_needed() | |
1934 | * above. */ | |
1935 | static void | |
1936 | update_learning_table__(const struct xbridge *xbridge, | |
1937 | const struct flow *flow, struct flow_wildcards *wc, | |
1938 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 1939 | OVS_REQ_WRLOCK(xbridge->ml->rwlock) |
ee047520 BP |
1940 | { |
1941 | struct mac_entry *mac; | |
1942 | ||
46c88433 | 1943 | if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) { |
ee047520 | 1944 | return; |
9583bc14 EJ |
1945 | } |
1946 | ||
46c88433 | 1947 | mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan); |
9583bc14 EJ |
1948 | if (is_gratuitous_arp(flow, wc)) { |
1949 | /* We don't want to learn from gratuitous ARP packets that are | |
1950 | * reflected back over bond slaves so we lock the learning table. */ | |
46c88433 | 1951 | if (!in_xbundle->bond) { |
9583bc14 EJ |
1952 | mac_entry_set_grat_arp_lock(mac); |
1953 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
ee047520 | 1954 | return; |
9583bc14 EJ |
1955 | } |
1956 | } | |
1957 | ||
9d078ec2 | 1958 | if (mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle) { |
9583bc14 EJ |
1959 | /* The log messages here could actually be useful in debugging, |
1960 | * so keep the rate limit relatively high. */ | |
1961 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
ee047520 | 1962 | |
9583bc14 EJ |
1963 | VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is " |
1964 | "on port %s in VLAN %d", | |
46c88433 EJ |
1965 | xbridge->name, ETH_ADDR_ARGS(flow->dl_src), |
1966 | in_xbundle->name, vlan); | |
9583bc14 | 1967 | |
9d078ec2 | 1968 | mac_entry_set_port(xbridge->ml, mac, in_xbundle->ofbundle); |
9583bc14 | 1969 | } |
ee047520 BP |
1970 | } |
1971 | ||
1972 | static void | |
1973 | update_learning_table(const struct xbridge *xbridge, | |
1974 | const struct flow *flow, struct flow_wildcards *wc, | |
1975 | int vlan, struct xbundle *in_xbundle) | |
1976 | { | |
1977 | bool need_update; | |
1978 | ||
1979 | /* Don't learn the OFPP_NONE port. */ | |
1980 | if (in_xbundle == &ofpp_none_bundle) { | |
1981 | return; | |
1982 | } | |
1983 | ||
1984 | /* First try the common case: no change to MAC learning table. */ | |
1985 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
1986 | need_update = is_mac_learning_update_needed(xbridge->ml, flow, wc, vlan, | |
1987 | in_xbundle); | |
509c0149 | 1988 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
ee047520 BP |
1989 | |
1990 | if (need_update) { | |
1991 | /* Slow path: MAC learning table might need an update. */ | |
1992 | ovs_rwlock_wrlock(&xbridge->ml->rwlock); | |
1993 | update_learning_table__(xbridge, flow, wc, vlan, in_xbundle); | |
1994 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
1995 | } | |
9583bc14 EJ |
1996 | } |
1997 | ||
86e2dcdd FL |
1998 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
1999 | * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */ | |
2000 | static void | |
06994f87 TLSC |
2001 | update_mcast_snooping_table4__(const struct xbridge *xbridge, |
2002 | const struct flow *flow, | |
2003 | struct mcast_snooping *ms, int vlan, | |
2004 | struct xbundle *in_xbundle, | |
2005 | const struct dp_packet *packet) | |
86e2dcdd FL |
2006 | OVS_REQ_WRLOCK(ms->rwlock) |
2007 | { | |
2008 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
e3102e42 | 2009 | int count; |
06994f87 | 2010 | ovs_be32 ip4 = flow->igmp_group_ip4; |
86e2dcdd FL |
2011 | |
2012 | switch (ntohs(flow->tp_src)) { | |
2013 | case IGMP_HOST_MEMBERSHIP_REPORT: | |
2014 | case IGMPV2_HOST_MEMBERSHIP_REPORT: | |
964a4d5f | 2015 | if (mcast_snooping_add_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
86e2dcdd FL |
2016 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping learned that " |
2017 | IP_FMT" is on port %s in VLAN %d", | |
2018 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
2019 | } | |
2020 | break; | |
2021 | case IGMP_HOST_LEAVE_MESSAGE: | |
964a4d5f | 2022 | if (mcast_snooping_leave_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
86e2dcdd FL |
2023 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping leaving " |
2024 | IP_FMT" is on port %s in VLAN %d", | |
2025 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
2026 | } | |
2027 | break; | |
2028 | case IGMP_HOST_MEMBERSHIP_QUERY: | |
2029 | if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan, | |
2030 | in_xbundle->ofbundle)) { | |
2031 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query from " | |
2032 | IP_FMT" is on port %s in VLAN %d", | |
2033 | xbridge->name, IP_ARGS(flow->nw_src), | |
2034 | in_xbundle->name, vlan); | |
2035 | } | |
2036 | break; | |
e3102e42 TLSC |
2037 | case IGMPV3_HOST_MEMBERSHIP_REPORT: |
2038 | if ((count = mcast_snooping_add_report(ms, packet, vlan, | |
2039 | in_xbundle->ofbundle))) { | |
2040 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping processed %d " | |
2041 | "addresses on port %s in VLAN %d", | |
2042 | xbridge->name, count, in_xbundle->name, vlan); | |
2043 | } | |
2044 | break; | |
86e2dcdd FL |
2045 | } |
2046 | } | |
2047 | ||
06994f87 TLSC |
2048 | static void |
2049 | update_mcast_snooping_table6__(const struct xbridge *xbridge, | |
2050 | const struct flow *flow, | |
2051 | struct mcast_snooping *ms, int vlan, | |
2052 | struct xbundle *in_xbundle, | |
2053 | const struct dp_packet *packet) | |
2054 | OVS_REQ_WRLOCK(ms->rwlock) | |
2055 | { | |
2056 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
2057 | int count; | |
2058 | ||
2059 | switch (ntohs(flow->tp_src)) { | |
2060 | case MLD_QUERY: | |
2061 | if (!ipv6_addr_equals(&flow->ipv6_src, &in6addr_any) | |
2062 | && mcast_snooping_add_mrouter(ms, vlan, in_xbundle->ofbundle)) { | |
2063 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query on port %s" | |
2064 | "in VLAN %d", | |
2065 | xbridge->name, in_xbundle->name, vlan); | |
2066 | } | |
2067 | break; | |
2068 | case MLD_REPORT: | |
2069 | case MLD_DONE: | |
2070 | case MLD2_REPORT: | |
2071 | count = mcast_snooping_add_mld(ms, packet, vlan, in_xbundle->ofbundle); | |
2072 | if (count) { | |
2073 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping processed %d " | |
2074 | "addresses on port %s in VLAN %d", | |
2075 | xbridge->name, count, in_xbundle->name, vlan); | |
2076 | } | |
2077 | break; | |
2078 | } | |
2079 | } | |
2080 | ||
86e2dcdd FL |
2081 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2082 | * was received on 'in_xbundle' in 'vlan'. */ | |
2083 | static void | |
2084 | update_mcast_snooping_table(const struct xbridge *xbridge, | |
2085 | const struct flow *flow, int vlan, | |
e3102e42 TLSC |
2086 | struct xbundle *in_xbundle, |
2087 | const struct dp_packet *packet) | |
86e2dcdd FL |
2088 | { |
2089 | struct mcast_snooping *ms = xbridge->ms; | |
2090 | struct xlate_cfg *xcfg; | |
2091 | struct xbundle *mcast_xbundle; | |
f4ae6e23 | 2092 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2093 | |
2094 | /* Don't learn the OFPP_NONE port. */ | |
2095 | if (in_xbundle == &ofpp_none_bundle) { | |
2096 | return; | |
2097 | } | |
2098 | ||
2099 | /* Don't learn from flood ports */ | |
2100 | mcast_xbundle = NULL; | |
2101 | ovs_rwlock_wrlock(&ms->rwlock); | |
2102 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2103 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2104 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2105 | if (mcast_xbundle == in_xbundle) { | |
2106 | break; | |
2107 | } | |
2108 | } | |
2109 | ||
2110 | if (!mcast_xbundle || mcast_xbundle != in_xbundle) { | |
06994f87 TLSC |
2111 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2112 | update_mcast_snooping_table4__(xbridge, flow, ms, vlan, | |
2113 | in_xbundle, packet); | |
2114 | } else { | |
2115 | update_mcast_snooping_table6__(xbridge, flow, ms, vlan, | |
2116 | in_xbundle, packet); | |
2117 | } | |
86e2dcdd FL |
2118 | } |
2119 | ovs_rwlock_unlock(&ms->rwlock); | |
2120 | } | |
2121 | ||
2122 | /* send the packet to ports having the multicast group learned */ | |
2123 | static void | |
2124 | xlate_normal_mcast_send_group(struct xlate_ctx *ctx, | |
2125 | struct mcast_snooping *ms OVS_UNUSED, | |
2126 | struct mcast_group *grp, | |
2127 | struct xbundle *in_xbundle, uint16_t vlan) | |
2128 | OVS_REQ_RDLOCK(ms->rwlock) | |
2129 | { | |
2130 | struct xlate_cfg *xcfg; | |
2131 | struct mcast_group_bundle *b; | |
2132 | struct xbundle *mcast_xbundle; | |
2133 | ||
2134 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2135 | LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) { | |
2136 | mcast_xbundle = xbundle_lookup(xcfg, b->port); | |
2137 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2138 | xlate_report(ctx, "forwarding to mcast group port"); | |
2139 | output_normal(ctx, mcast_xbundle, vlan); | |
2140 | } else if (!mcast_xbundle) { | |
2141 | xlate_report(ctx, "mcast group port is unknown, dropping"); | |
2142 | } else { | |
2143 | xlate_report(ctx, "mcast group port is input port, dropping"); | |
2144 | } | |
2145 | } | |
2146 | } | |
2147 | ||
2148 | /* send the packet to ports connected to multicast routers */ | |
2149 | static void | |
2150 | xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx, | |
2151 | struct mcast_snooping *ms, | |
2152 | struct xbundle *in_xbundle, uint16_t vlan) | |
2153 | OVS_REQ_RDLOCK(ms->rwlock) | |
2154 | { | |
2155 | struct xlate_cfg *xcfg; | |
2156 | struct mcast_mrouter_bundle *mrouter; | |
2157 | struct xbundle *mcast_xbundle; | |
2158 | ||
2159 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2160 | LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) { | |
2161 | mcast_xbundle = xbundle_lookup(xcfg, mrouter->port); | |
2162 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2163 | xlate_report(ctx, "forwarding to mcast router port"); | |
2164 | output_normal(ctx, mcast_xbundle, vlan); | |
2165 | } else if (!mcast_xbundle) { | |
2166 | xlate_report(ctx, "mcast router port is unknown, dropping"); | |
2167 | } else { | |
2168 | xlate_report(ctx, "mcast router port is input port, dropping"); | |
2169 | } | |
2170 | } | |
2171 | } | |
2172 | ||
2173 | /* send the packet to ports flagged to be flooded */ | |
2174 | static void | |
2175 | xlate_normal_mcast_send_fports(struct xlate_ctx *ctx, | |
2176 | struct mcast_snooping *ms, | |
2177 | struct xbundle *in_xbundle, uint16_t vlan) | |
2178 | OVS_REQ_RDLOCK(ms->rwlock) | |
2179 | { | |
2180 | struct xlate_cfg *xcfg; | |
f4ae6e23 | 2181 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2182 | struct xbundle *mcast_xbundle; |
2183 | ||
2184 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2185 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2186 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2187 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2188 | xlate_report(ctx, "forwarding to mcast flood port"); | |
2189 | output_normal(ctx, mcast_xbundle, vlan); | |
2190 | } else if (!mcast_xbundle) { | |
2191 | xlate_report(ctx, "mcast flood port is unknown, dropping"); | |
2192 | } else { | |
2193 | xlate_report(ctx, "mcast flood port is input port, dropping"); | |
2194 | } | |
2195 | } | |
2196 | } | |
2197 | ||
8e04a33f FL |
2198 | /* forward the Reports to configured ports */ |
2199 | static void | |
2200 | xlate_normal_mcast_send_rports(struct xlate_ctx *ctx, | |
2201 | struct mcast_snooping *ms, | |
2202 | struct xbundle *in_xbundle, uint16_t vlan) | |
2203 | OVS_REQ_RDLOCK(ms->rwlock) | |
2204 | { | |
2205 | struct xlate_cfg *xcfg; | |
2206 | struct mcast_port_bundle *rport; | |
2207 | struct xbundle *mcast_xbundle; | |
2208 | ||
2209 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2210 | LIST_FOR_EACH(rport, node, &ms->rport_list) { | |
2211 | mcast_xbundle = xbundle_lookup(xcfg, rport->port); | |
2212 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2213 | xlate_report(ctx, "forwarding Report to mcast flagged port"); | |
2214 | output_normal(ctx, mcast_xbundle, vlan); | |
2215 | } else if (!mcast_xbundle) { | |
2216 | xlate_report(ctx, "mcast port is unknown, dropping the Report"); | |
2217 | } else { | |
2218 | xlate_report(ctx, "mcast port is input port, dropping the Report"); | |
2219 | } | |
2220 | } | |
2221 | } | |
2222 | ||
682800a4 FL |
2223 | static void |
2224 | xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, | |
2225 | uint16_t vlan) | |
2226 | { | |
2227 | struct xbundle *xbundle; | |
2228 | ||
2229 | LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { | |
2230 | if (xbundle != in_xbundle | |
2231 | && xbundle_includes_vlan(xbundle, vlan) | |
2232 | && xbundle->floodable | |
2233 | && !xbundle_mirror_out(ctx->xbridge, xbundle)) { | |
2234 | output_normal(ctx, xbundle, vlan); | |
2235 | } | |
2236 | } | |
2237 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
2238 | } | |
2239 | ||
9583bc14 EJ |
2240 | static void |
2241 | xlate_normal(struct xlate_ctx *ctx) | |
2242 | { | |
33bf9176 BP |
2243 | struct flow_wildcards *wc = &ctx->xout->wc; |
2244 | struct flow *flow = &ctx->xin->flow; | |
46c88433 EJ |
2245 | struct xbundle *in_xbundle; |
2246 | struct xport *in_port; | |
9583bc14 | 2247 | struct mac_entry *mac; |
d6d5bbc9 | 2248 | void *mac_port; |
9583bc14 EJ |
2249 | uint16_t vlan; |
2250 | uint16_t vid; | |
2251 | ||
2252 | ctx->xout->has_normal = true; | |
2253 | ||
33bf9176 BP |
2254 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
2255 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1dd35f8a | 2256 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
9583bc14 | 2257 | |
46c88433 EJ |
2258 | in_xbundle = lookup_input_bundle(ctx->xbridge, flow->in_port.ofp_port, |
2259 | ctx->xin->packet != NULL, &in_port); | |
2260 | if (!in_xbundle) { | |
9583bc14 EJ |
2261 | xlate_report(ctx, "no input bundle, dropping"); |
2262 | return; | |
2263 | } | |
2264 | ||
2265 | /* Drop malformed frames. */ | |
33bf9176 BP |
2266 | if (flow->dl_type == htons(ETH_TYPE_VLAN) && |
2267 | !(flow->vlan_tci & htons(VLAN_CFI))) { | |
9583bc14 EJ |
2268 | if (ctx->xin->packet != NULL) { |
2269 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2270 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet with partial " | |
2271 | "VLAN tag received on port %s", | |
46c88433 | 2272 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2273 | } |
2274 | xlate_report(ctx, "partial VLAN tag, dropping"); | |
2275 | return; | |
2276 | } | |
2277 | ||
2278 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 2279 | if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) { |
9583bc14 EJ |
2280 | if (ctx->xin->packet != NULL) { |
2281 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2282 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
2283 | "%s, which is reserved exclusively for mirroring", | |
46c88433 | 2284 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2285 | } |
2286 | xlate_report(ctx, "input port is mirror output port, dropping"); | |
2287 | return; | |
2288 | } | |
2289 | ||
2290 | /* Check VLAN. */ | |
33bf9176 | 2291 | vid = vlan_tci_to_vid(flow->vlan_tci); |
46c88433 | 2292 | if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { |
9583bc14 EJ |
2293 | xlate_report(ctx, "disallowed VLAN VID for this input port, dropping"); |
2294 | return; | |
2295 | } | |
46c88433 | 2296 | vlan = input_vid_to_vlan(in_xbundle, vid); |
9583bc14 EJ |
2297 | |
2298 | /* Check other admissibility requirements. */ | |
2299 | if (in_port && !is_admissible(ctx, in_port, vlan)) { | |
2300 | return; | |
2301 | } | |
2302 | ||
2303 | /* Learn source MAC. */ | |
2304 | if (ctx->xin->may_learn) { | |
46c88433 | 2305 | update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle); |
9583bc14 | 2306 | } |
b256dc52 JS |
2307 | if (ctx->xin->xcache) { |
2308 | struct xc_entry *entry; | |
2309 | ||
2310 | /* Save enough info to update mac learning table later. */ | |
2311 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL); | |
9edf6b48 | 2312 | entry->u.normal.ofproto = ctx->xbridge->ofproto; |
b256dc52 JS |
2313 | entry->u.normal.flow = xmemdup(flow, sizeof *flow); |
2314 | entry->u.normal.vlan = vlan; | |
2315 | } | |
9583bc14 EJ |
2316 | |
2317 | /* Determine output bundle. */ | |
86e2dcdd FL |
2318 | if (mcast_snooping_enabled(ctx->xbridge->ms) |
2319 | && !eth_addr_is_broadcast(flow->dl_dst) | |
2320 | && eth_addr_is_multicast(flow->dl_dst) | |
06994f87 | 2321 | && is_ip_any(flow)) { |
86e2dcdd | 2322 | struct mcast_snooping *ms = ctx->xbridge->ms; |
06994f87 | 2323 | struct mcast_group *grp = NULL; |
86e2dcdd | 2324 | |
06994f87 | 2325 | if (is_igmp(flow)) { |
d29f137b TLSC |
2326 | if (mcast_snooping_is_membership(flow->tp_src) || |
2327 | mcast_snooping_is_query(flow->tp_src)) { | |
2328 | if (ctx->xin->may_learn) { | |
86e2dcdd | 2329 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, |
e3102e42 | 2330 | in_xbundle, ctx->xin->packet); |
d29f137b TLSC |
2331 | } |
2332 | /* | |
2333 | * IGMP packets need to take the slow path, in order to be | |
2334 | * processed for mdb updates. That will prevent expires | |
2335 | * firing off even after hosts have sent reports. | |
2336 | */ | |
2337 | ctx->xout->slow |= SLOW_ACTION; | |
86e2dcdd | 2338 | } |
d6d5bbc9 | 2339 | |
86e2dcdd FL |
2340 | if (mcast_snooping_is_membership(flow->tp_src)) { |
2341 | ovs_rwlock_rdlock(&ms->rwlock); | |
2342 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
8e04a33f FL |
2343 | /* RFC4541: section 2.1.1, item 1: A snooping switch should |
2344 | * forward IGMP Membership Reports only to those ports where | |
2345 | * multicast routers are attached. Alternatively stated: a | |
2346 | * snooping switch should not forward IGMP Membership Reports | |
2347 | * to ports on which only hosts are attached. | |
2348 | * An administrative control may be provided to override this | |
2349 | * restriction, allowing the report messages to be flooded to | |
2350 | * other ports. */ | |
2351 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
86e2dcdd FL |
2352 | ovs_rwlock_unlock(&ms->rwlock); |
2353 | } else { | |
2354 | xlate_report(ctx, "multicast traffic, flooding"); | |
2355 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2356 | } | |
2357 | return; | |
06994f87 TLSC |
2358 | } else if (is_mld(flow)) { |
2359 | ctx->xout->slow |= SLOW_ACTION; | |
2360 | if (ctx->xin->may_learn) { | |
2361 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, | |
2362 | in_xbundle, ctx->xin->packet); | |
2363 | } | |
2364 | if (is_mld_report(flow)) { | |
2365 | ovs_rwlock_rdlock(&ms->rwlock); | |
2366 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2367 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
2368 | ovs_rwlock_unlock(&ms->rwlock); | |
2369 | } else { | |
2370 | xlate_report(ctx, "MLD query, flooding"); | |
2371 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2372 | } | |
86e2dcdd | 2373 | } else { |
06994f87 TLSC |
2374 | if ((flow->dl_type == htons(ETH_TYPE_IP) |
2375 | && ip_is_local_multicast(flow->nw_dst)) | |
2376 | || (flow->dl_type == htons(ETH_TYPE_IPV6) | |
2377 | && ipv6_is_all_hosts(&flow->ipv6_dst))) { | |
86e2dcdd FL |
2378 | /* RFC4541: section 2.1.2, item 2: Packets with a dst IP |
2379 | * address in the 224.0.0.x range which are not IGMP must | |
2380 | * be forwarded on all ports */ | |
2381 | xlate_report(ctx, "RFC4541: section 2.1.2, item 2, flooding"); | |
2382 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2383 | return; | |
2384 | } | |
2385 | } | |
2386 | ||
2387 | /* forwarding to group base ports */ | |
2388 | ovs_rwlock_rdlock(&ms->rwlock); | |
06994f87 TLSC |
2389 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2390 | grp = mcast_snooping_lookup4(ms, flow->nw_dst, vlan); | |
2391 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
2392 | grp = mcast_snooping_lookup(ms, &flow->ipv6_dst, vlan); | |
2393 | } | |
86e2dcdd FL |
2394 | if (grp) { |
2395 | xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, vlan); | |
2396 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2397 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
9583bc14 | 2398 | } else { |
86e2dcdd FL |
2399 | if (mcast_snooping_flood_unreg(ms)) { |
2400 | xlate_report(ctx, "unregistered multicast, flooding"); | |
2401 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2402 | } else { | |
2403 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2404 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2405 | } | |
9583bc14 | 2406 | } |
86e2dcdd | 2407 | ovs_rwlock_unlock(&ms->rwlock); |
9583bc14 | 2408 | } else { |
86e2dcdd FL |
2409 | ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); |
2410 | mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); | |
9d078ec2 | 2411 | mac_port = mac ? mac_entry_get_port(ctx->xbridge->ml, mac) : NULL; |
86e2dcdd FL |
2412 | ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); |
2413 | ||
2414 | if (mac_port) { | |
2415 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2416 | struct xbundle *mac_xbundle = xbundle_lookup(xcfg, mac_port); | |
2417 | if (mac_xbundle && mac_xbundle != in_xbundle) { | |
2418 | xlate_report(ctx, "forwarding to learned port"); | |
2419 | output_normal(ctx, mac_xbundle, vlan); | |
2420 | } else if (!mac_xbundle) { | |
2421 | xlate_report(ctx, "learned port is unknown, dropping"); | |
2422 | } else { | |
2423 | xlate_report(ctx, "learned port is input port, dropping"); | |
2424 | } | |
2425 | } else { | |
2426 | xlate_report(ctx, "no learned MAC for destination, flooding"); | |
2427 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2428 | } | |
9583bc14 EJ |
2429 | } |
2430 | } | |
2431 | ||
2432 | /* Compose SAMPLE action for sFlow or IPFIX. The given probability is | |
2433 | * the number of packets out of UINT32_MAX to sample. The given | |
2434 | * cookie is passed back in the callback for each sampled packet. | |
2435 | */ | |
2436 | static size_t | |
46c88433 | 2437 | compose_sample_action(const struct xbridge *xbridge, |
9583bc14 EJ |
2438 | struct ofpbuf *odp_actions, |
2439 | const struct flow *flow, | |
2440 | const uint32_t probability, | |
2441 | const union user_action_cookie *cookie, | |
8b7ea2d4 WZ |
2442 | const size_t cookie_size, |
2443 | const odp_port_t tunnel_out_port) | |
9583bc14 EJ |
2444 | { |
2445 | size_t sample_offset, actions_offset; | |
89a8a7f0 | 2446 | odp_port_t odp_port; |
9583bc14 | 2447 | int cookie_offset; |
89a8a7f0 | 2448 | uint32_t pid; |
9583bc14 EJ |
2449 | |
2450 | sample_offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SAMPLE); | |
2451 | ||
2452 | nl_msg_put_u32(odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability); | |
2453 | ||
2454 | actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS); | |
89a8a7f0 EJ |
2455 | |
2456 | odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port); | |
9a159f74 AW |
2457 | pid = dpif_port_get_pid(xbridge->dpif, odp_port, |
2458 | flow_hash_5tuple(flow, 0)); | |
2459 | cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, | |
8b7ea2d4 | 2460 | tunnel_out_port, odp_actions); |
9583bc14 EJ |
2461 | |
2462 | nl_msg_end_nested(odp_actions, actions_offset); | |
2463 | nl_msg_end_nested(odp_actions, sample_offset); | |
2464 | return cookie_offset; | |
2465 | } | |
2466 | ||
2467 | static void | |
46c88433 EJ |
2468 | compose_sflow_cookie(const struct xbridge *xbridge, ovs_be16 vlan_tci, |
2469 | odp_port_t odp_port, unsigned int n_outputs, | |
2470 | union user_action_cookie *cookie) | |
9583bc14 EJ |
2471 | { |
2472 | int ifindex; | |
2473 | ||
2474 | cookie->type = USER_ACTION_COOKIE_SFLOW; | |
2475 | cookie->sflow.vlan_tci = vlan_tci; | |
2476 | ||
2477 | /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output | |
2478 | * port information") for the interpretation of cookie->output. */ | |
2479 | switch (n_outputs) { | |
2480 | case 0: | |
2481 | /* 0x40000000 | 256 means "packet dropped for unknown reason". */ | |
2482 | cookie->sflow.output = 0x40000000 | 256; | |
2483 | break; | |
2484 | ||
2485 | case 1: | |
46c88433 | 2486 | ifindex = dpif_sflow_odp_port_to_ifindex(xbridge->sflow, odp_port); |
9583bc14 EJ |
2487 | if (ifindex) { |
2488 | cookie->sflow.output = ifindex; | |
2489 | break; | |
2490 | } | |
2491 | /* Fall through. */ | |
2492 | default: | |
2493 | /* 0x80000000 means "multiple output ports. */ | |
2494 | cookie->sflow.output = 0x80000000 | n_outputs; | |
2495 | break; | |
2496 | } | |
2497 | } | |
2498 | ||
2499 | /* Compose SAMPLE action for sFlow bridge sampling. */ | |
2500 | static size_t | |
46c88433 | 2501 | compose_sflow_action(const struct xbridge *xbridge, |
9583bc14 EJ |
2502 | struct ofpbuf *odp_actions, |
2503 | const struct flow *flow, | |
4e022ec0 | 2504 | odp_port_t odp_port) |
9583bc14 EJ |
2505 | { |
2506 | uint32_t probability; | |
2507 | union user_action_cookie cookie; | |
2508 | ||
46c88433 | 2509 | if (!xbridge->sflow || flow->in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
2510 | return 0; |
2511 | } | |
2512 | ||
46c88433 EJ |
2513 | probability = dpif_sflow_get_probability(xbridge->sflow); |
2514 | compose_sflow_cookie(xbridge, htons(0), odp_port, | |
4e022ec0 | 2515 | odp_port == ODPP_NONE ? 0 : 1, &cookie); |
9583bc14 | 2516 | |
46c88433 | 2517 | return compose_sample_action(xbridge, odp_actions, flow, probability, |
8b7ea2d4 | 2518 | &cookie, sizeof cookie.sflow, ODPP_NONE); |
9583bc14 EJ |
2519 | } |
2520 | ||
2521 | static void | |
2522 | compose_flow_sample_cookie(uint16_t probability, uint32_t collector_set_id, | |
2523 | uint32_t obs_domain_id, uint32_t obs_point_id, | |
2524 | union user_action_cookie *cookie) | |
2525 | { | |
2526 | cookie->type = USER_ACTION_COOKIE_FLOW_SAMPLE; | |
2527 | cookie->flow_sample.probability = probability; | |
2528 | cookie->flow_sample.collector_set_id = collector_set_id; | |
2529 | cookie->flow_sample.obs_domain_id = obs_domain_id; | |
2530 | cookie->flow_sample.obs_point_id = obs_point_id; | |
2531 | } | |
2532 | ||
2533 | static void | |
8b7ea2d4 WZ |
2534 | compose_ipfix_cookie(union user_action_cookie *cookie, |
2535 | odp_port_t output_odp_port) | |
9583bc14 EJ |
2536 | { |
2537 | cookie->type = USER_ACTION_COOKIE_IPFIX; | |
8b7ea2d4 | 2538 | cookie->ipfix.output_odp_port = output_odp_port; |
9583bc14 EJ |
2539 | } |
2540 | ||
2541 | /* Compose SAMPLE action for IPFIX bridge sampling. */ | |
2542 | static void | |
46c88433 | 2543 | compose_ipfix_action(const struct xbridge *xbridge, |
9583bc14 | 2544 | struct ofpbuf *odp_actions, |
8b7ea2d4 WZ |
2545 | const struct flow *flow, |
2546 | odp_port_t output_odp_port) | |
9583bc14 EJ |
2547 | { |
2548 | uint32_t probability; | |
2549 | union user_action_cookie cookie; | |
8b7ea2d4 | 2550 | odp_port_t tunnel_out_port = ODPP_NONE; |
9583bc14 | 2551 | |
46c88433 | 2552 | if (!xbridge->ipfix || flow->in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
2553 | return; |
2554 | } | |
2555 | ||
8b7ea2d4 WZ |
2556 | /* For input case, output_odp_port is ODPP_NONE, which is an invalid port |
2557 | * number. */ | |
2558 | if (output_odp_port == ODPP_NONE && | |
2559 | !dpif_ipfix_get_bridge_exporter_input_sampling(xbridge->ipfix)) { | |
2560 | return; | |
2561 | } | |
2562 | ||
2563 | /* For output case, output_odp_port is valid*/ | |
2564 | if (output_odp_port != ODPP_NONE) { | |
2565 | if (!dpif_ipfix_get_bridge_exporter_output_sampling(xbridge->ipfix)) { | |
2566 | return; | |
2567 | } | |
2568 | /* If tunnel sampling is enabled, put an additional option attribute: | |
2569 | * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT | |
2570 | */ | |
2571 | if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(xbridge->ipfix) && | |
2572 | dpif_ipfix_get_tunnel_port(xbridge->ipfix, output_odp_port) ) { | |
2573 | tunnel_out_port = output_odp_port; | |
2574 | } | |
2575 | } | |
2576 | ||
46c88433 | 2577 | probability = dpif_ipfix_get_bridge_exporter_probability(xbridge->ipfix); |
8b7ea2d4 | 2578 | compose_ipfix_cookie(&cookie, output_odp_port); |
9583bc14 | 2579 | |
46c88433 | 2580 | compose_sample_action(xbridge, odp_actions, flow, probability, |
8b7ea2d4 | 2581 | &cookie, sizeof cookie.ipfix, tunnel_out_port); |
9583bc14 EJ |
2582 | } |
2583 | ||
2584 | /* SAMPLE action for sFlow must be first action in any given list of | |
2585 | * actions. At this point we do not have all information required to | |
2586 | * build it. So try to build sample action as complete as possible. */ | |
2587 | static void | |
2588 | add_sflow_action(struct xlate_ctx *ctx) | |
2589 | { | |
46c88433 | 2590 | ctx->user_cookie_offset = compose_sflow_action(ctx->xbridge, |
cc377352 | 2591 | ctx->xout->odp_actions, |
4e022ec0 | 2592 | &ctx->xin->flow, ODPP_NONE); |
9583bc14 EJ |
2593 | ctx->sflow_odp_port = 0; |
2594 | ctx->sflow_n_outputs = 0; | |
2595 | } | |
2596 | ||
2597 | /* SAMPLE action for IPFIX must be 1st or 2nd action in any given list | |
2598 | * of actions, eventually after the SAMPLE action for sFlow. */ | |
2599 | static void | |
2600 | add_ipfix_action(struct xlate_ctx *ctx) | |
2601 | { | |
cc377352 | 2602 | compose_ipfix_action(ctx->xbridge, ctx->xout->odp_actions, |
8b7ea2d4 WZ |
2603 | &ctx->xin->flow, ODPP_NONE); |
2604 | } | |
2605 | ||
2606 | static void | |
2607 | add_ipfix_output_action(struct xlate_ctx *ctx, odp_port_t port) | |
2608 | { | |
2609 | compose_ipfix_action(ctx->xbridge, ctx->xout->odp_actions, | |
2610 | &ctx->xin->flow, port); | |
9583bc14 EJ |
2611 | } |
2612 | ||
2613 | /* Fix SAMPLE action according to data collected while composing ODP actions. | |
2614 | * We need to fix SAMPLE actions OVS_SAMPLE_ATTR_ACTIONS attribute, i.e. nested | |
2615 | * USERSPACE action's user-cookie which is required for sflow. */ | |
2616 | static void | |
2617 | fix_sflow_action(struct xlate_ctx *ctx) | |
2618 | { | |
2619 | const struct flow *base = &ctx->base_flow; | |
2620 | union user_action_cookie *cookie; | |
2621 | ||
2622 | if (!ctx->user_cookie_offset) { | |
2623 | return; | |
2624 | } | |
2625 | ||
cc377352 | 2626 | cookie = ofpbuf_at(ctx->xout->odp_actions, ctx->user_cookie_offset, |
9583bc14 EJ |
2627 | sizeof cookie->sflow); |
2628 | ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); | |
2629 | ||
46c88433 | 2630 | compose_sflow_cookie(ctx->xbridge, base->vlan_tci, |
9583bc14 EJ |
2631 | ctx->sflow_odp_port, ctx->sflow_n_outputs, cookie); |
2632 | } | |
2633 | ||
db7d4e46 | 2634 | static enum slow_path_reason |
642dc74d | 2635 | process_special(struct xlate_ctx *ctx, const struct flow *flow, |
cf62fa4c | 2636 | const struct xport *xport, const struct dp_packet *packet) |
db7d4e46 | 2637 | { |
642dc74d | 2638 | struct flow_wildcards *wc = &ctx->xout->wc; |
46c88433 | 2639 | const struct xbridge *xbridge = ctx->xbridge; |
642dc74d | 2640 | |
46c88433 | 2641 | if (!xport) { |
db7d4e46 | 2642 | return 0; |
46c88433 | 2643 | } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) { |
db7d4e46 | 2644 | if (packet) { |
46c88433 | 2645 | cfm_process_heartbeat(xport->cfm, packet); |
db7d4e46 JP |
2646 | } |
2647 | return SLOW_CFM; | |
fab52e16 | 2648 | } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) { |
db7d4e46 | 2649 | if (packet) { |
46c88433 | 2650 | bfd_process_packet(xport->bfd, flow, packet); |
60d02c72 AW |
2651 | /* If POLL received, immediately sends FINAL back. */ |
2652 | if (bfd_should_send_packet(xport->bfd)) { | |
6d308b28 | 2653 | ofproto_dpif_monitor_port_send_soon(xport->ofport); |
60d02c72 | 2654 | } |
db7d4e46 JP |
2655 | } |
2656 | return SLOW_BFD; | |
46c88433 | 2657 | } else if (xport->xbundle && xport->xbundle->lacp |
db7d4e46 JP |
2658 | && flow->dl_type == htons(ETH_TYPE_LACP)) { |
2659 | if (packet) { | |
46c88433 | 2660 | lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet); |
db7d4e46 JP |
2661 | } |
2662 | return SLOW_LACP; | |
9efd308e DV |
2663 | } else if ((xbridge->stp || xbridge->rstp) && |
2664 | stp_should_process_flow(flow, wc)) { | |
db7d4e46 | 2665 | if (packet) { |
f025bcb7 JR |
2666 | xbridge->stp |
2667 | ? stp_process_packet(xport, packet) | |
2668 | : rstp_process_packet(xport, packet); | |
db7d4e46 JP |
2669 | } |
2670 | return SLOW_STP; | |
19aef6ef | 2671 | } else if (xport->lldp && lldp_should_process_flow(xport->lldp, flow)) { |
0477baa9 DF |
2672 | if (packet) { |
2673 | lldp_process_packet(xport->lldp, packet); | |
2674 | } | |
2675 | return SLOW_LLDP; | |
db7d4e46 JP |
2676 | } else { |
2677 | return 0; | |
2678 | } | |
2679 | } | |
2680 | ||
a36de779 PS |
2681 | static int |
2682 | tnl_route_lookup_flow(const struct flow *oflow, | |
2683 | ovs_be32 *ip, struct xport **out_port) | |
2684 | { | |
2685 | char out_dev[IFNAMSIZ]; | |
2686 | struct xbridge *xbridge; | |
2687 | struct xlate_cfg *xcfg; | |
2688 | ovs_be32 gw; | |
2689 | ||
2690 | if (!ovs_router_lookup(oflow->tunnel.ip_dst, out_dev, &gw)) { | |
2691 | return -ENOENT; | |
2692 | } | |
2693 | ||
2694 | if (gw) { | |
2695 | *ip = gw; | |
2696 | } else { | |
2697 | *ip = oflow->tunnel.ip_dst; | |
2698 | } | |
2699 | ||
2700 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2701 | ovs_assert(xcfg); | |
2702 | ||
2703 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
2704 | if (!strncmp(xbridge->name, out_dev, IFNAMSIZ)) { | |
2705 | struct xport *port; | |
2706 | ||
2707 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
2708 | if (!strncmp(netdev_get_name(port->netdev), out_dev, IFNAMSIZ)) { | |
2709 | *out_port = port; | |
2710 | return 0; | |
2711 | } | |
2712 | } | |
2713 | } | |
2714 | } | |
2715 | return -ENOENT; | |
2716 | } | |
2717 | ||
2718 | static int | |
cf62fa4c | 2719 | xlate_flood_packet(struct xbridge *xbridge, struct dp_packet *packet) |
a36de779 PS |
2720 | { |
2721 | struct ofpact_output output; | |
2722 | struct flow flow; | |
2723 | ||
2724 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); | |
2725 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ | |
cf62fa4c | 2726 | flow_extract(packet, &flow); |
a36de779 PS |
2727 | flow.in_port.ofp_port = OFPP_NONE; |
2728 | output.port = OFPP_FLOOD; | |
2729 | output.max_len = 0; | |
2730 | ||
2731 | return ofproto_dpif_execute_actions(xbridge->ofproto, &flow, NULL, | |
2732 | &output.ofpact, sizeof output, | |
2733 | packet); | |
2734 | } | |
2735 | ||
2736 | static void | |
2737 | tnl_send_arp_request(const struct xport *out_dev, const uint8_t eth_src[ETH_ADDR_LEN], | |
2738 | ovs_be32 ip_src, ovs_be32 ip_dst) | |
2739 | { | |
2740 | struct xbridge *xbridge = out_dev->xbridge; | |
cf62fa4c | 2741 | struct dp_packet packet; |
a36de779 | 2742 | |
cf62fa4c | 2743 | dp_packet_init(&packet, 0); |
eb0b295e BP |
2744 | compose_arp(&packet, ARP_OP_REQUEST, |
2745 | eth_src, eth_addr_zero, true, ip_src, ip_dst); | |
a36de779 PS |
2746 | |
2747 | xlate_flood_packet(xbridge, &packet); | |
cf62fa4c | 2748 | dp_packet_uninit(&packet); |
a36de779 PS |
2749 | } |
2750 | ||
2751 | static int | |
81de18ec | 2752 | build_tunnel_send(struct xlate_ctx *ctx, const struct xport *xport, |
a36de779 PS |
2753 | const struct flow *flow, odp_port_t tunnel_odp_port) |
2754 | { | |
2755 | struct ovs_action_push_tnl tnl_push_data; | |
2756 | struct xport *out_dev = NULL; | |
2757 | ovs_be32 s_ip, d_ip = 0; | |
2758 | uint8_t smac[ETH_ADDR_LEN]; | |
2759 | uint8_t dmac[ETH_ADDR_LEN]; | |
2760 | int err; | |
2761 | ||
2762 | err = tnl_route_lookup_flow(flow, &d_ip, &out_dev); | |
2763 | if (err) { | |
81de18ec | 2764 | xlate_report(ctx, "native tunnel routing failed"); |
a36de779 PS |
2765 | return err; |
2766 | } | |
81de18ec BP |
2767 | xlate_report(ctx, "tunneling to "IP_FMT" via %s", |
2768 | IP_ARGS(d_ip), netdev_get_name(out_dev->netdev)); | |
a36de779 PS |
2769 | |
2770 | /* Use mac addr of bridge port of the peer. */ | |
2771 | err = netdev_get_etheraddr(out_dev->netdev, smac); | |
2772 | if (err) { | |
81de18ec | 2773 | xlate_report(ctx, "tunnel output device lacks Ethernet address"); |
a36de779 PS |
2774 | return err; |
2775 | } | |
2776 | ||
2777 | err = netdev_get_in4(out_dev->netdev, (struct in_addr *) &s_ip, NULL); | |
2778 | if (err) { | |
81de18ec | 2779 | xlate_report(ctx, "tunnel output device lacks IPv4 address"); |
a36de779 PS |
2780 | return err; |
2781 | } | |
2782 | ||
2783 | err = tnl_arp_lookup(out_dev->xbridge->name, d_ip, dmac); | |
2784 | if (err) { | |
81de18ec BP |
2785 | xlate_report(ctx, "ARP cache miss for "IP_FMT" on bridge %s, " |
2786 | "sending ARP request", | |
2787 | IP_ARGS(d_ip), out_dev->xbridge->name); | |
a36de779 PS |
2788 | tnl_send_arp_request(out_dev, smac, s_ip, d_ip); |
2789 | return err; | |
2790 | } | |
2791 | if (ctx->xin->xcache) { | |
2792 | struct xc_entry *entry; | |
2793 | ||
2794 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_ARP); | |
8742957c BP |
2795 | ovs_strlcpy(entry->u.tnl_arp_cache.br_name, out_dev->xbridge->name, |
2796 | sizeof entry->u.tnl_arp_cache.br_name); | |
a36de779 PS |
2797 | entry->u.tnl_arp_cache.d_ip = d_ip; |
2798 | } | |
81de18ec BP |
2799 | |
2800 | xlate_report(ctx, "tunneling from "ETH_ADDR_FMT" "IP_FMT | |
2801 | " to "ETH_ADDR_FMT" "IP_FMT, | |
2802 | ETH_ADDR_ARGS(smac), IP_ARGS(s_ip), | |
2803 | ETH_ADDR_ARGS(dmac), IP_ARGS(d_ip)); | |
a36de779 PS |
2804 | err = tnl_port_build_header(xport->ofport, flow, |
2805 | dmac, smac, s_ip, &tnl_push_data); | |
2806 | if (err) { | |
2807 | return err; | |
2808 | } | |
2809 | tnl_push_data.tnl_port = odp_to_u32(tunnel_odp_port); | |
2810 | tnl_push_data.out_port = odp_to_u32(out_dev->odp_port); | |
2811 | odp_put_tnl_push_action(ctx->xout->odp_actions, &tnl_push_data); | |
2812 | return 0; | |
2813 | } | |
2814 | ||
9583bc14 | 2815 | static void |
4e022ec0 | 2816 | compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
e93ef1c7 | 2817 | const struct xlate_bond_recirc *xr, bool check_stp) |
9583bc14 | 2818 | { |
46c88433 | 2819 | const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1dd35f8a | 2820 | struct flow_wildcards *wc = &ctx->xout->wc; |
33bf9176 | 2821 | struct flow *flow = &ctx->xin->flow; |
a36de779 | 2822 | struct flow_tnl flow_tnl; |
9583bc14 | 2823 | ovs_be16 flow_vlan_tci; |
1362e248 | 2824 | uint32_t flow_pkt_mark; |
9583bc14 | 2825 | uint8_t flow_nw_tos; |
4e022ec0 | 2826 | odp_port_t out_port, odp_port; |
a36de779 | 2827 | bool tnl_push_pop_send = false; |
ca077186 | 2828 | uint8_t dscp; |
9583bc14 EJ |
2829 | |
2830 | /* If 'struct flow' gets additional metadata, we'll need to zero it out | |
2831 | * before traversing a patch port. */ | |
b666962b | 2832 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 33); |
a36de779 | 2833 | memset(&flow_tnl, 0, sizeof flow_tnl); |
9583bc14 | 2834 | |
46c88433 | 2835 | if (!xport) { |
9583bc14 EJ |
2836 | xlate_report(ctx, "Nonexistent output port"); |
2837 | return; | |
46c88433 | 2838 | } else if (xport->config & OFPUTIL_PC_NO_FWD) { |
9583bc14 EJ |
2839 | xlate_report(ctx, "OFPPC_NO_FWD set, skipping output"); |
2840 | return; | |
0d1cee12 | 2841 | } else if (check_stp) { |
bbbca389 | 2842 | if (is_stp(&ctx->base_flow)) { |
9efd308e DV |
2843 | if (!xport_stp_should_forward_bpdu(xport) && |
2844 | !xport_rstp_should_manage_bpdu(xport)) { | |
2845 | if (ctx->xbridge->stp != NULL) { | |
2846 | xlate_report(ctx, "STP not in listening state, " | |
2847 | "skipping bpdu output"); | |
2848 | } else if (ctx->xbridge->rstp != NULL) { | |
2849 | xlate_report(ctx, "RSTP not managing BPDU in this state, " | |
2850 | "skipping bpdu output"); | |
2851 | } | |
0d1cee12 K |
2852 | return; |
2853 | } | |
9efd308e DV |
2854 | } else if (!xport_stp_forward_state(xport) || |
2855 | !xport_rstp_forward_state(xport)) { | |
2856 | if (ctx->xbridge->stp != NULL) { | |
2857 | xlate_report(ctx, "STP not in forwarding state, " | |
2858 | "skipping output"); | |
2859 | } else if (ctx->xbridge->rstp != NULL) { | |
2860 | xlate_report(ctx, "RSTP not in forwarding state, " | |
2861 | "skipping output"); | |
2862 | } | |
0d1cee12 K |
2863 | return; |
2864 | } | |
9583bc14 EJ |
2865 | } |
2866 | ||
46c88433 EJ |
2867 | if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { |
2868 | ctx->xout->mirrors |= xbundle_mirror_dst(xport->xbundle->xbridge, | |
2869 | xport->xbundle); | |
cdf5d3a5 EJ |
2870 | } |
2871 | ||
46c88433 EJ |
2872 | if (xport->peer) { |
2873 | const struct xport *peer = xport->peer; | |
9583bc14 | 2874 | struct flow old_flow = ctx->xin->flow; |
7f5fe537 | 2875 | bool old_was_mpls = ctx->was_mpls; |
18721c4a | 2876 | cls_version_t old_version = ctx->tables_version; |
9583bc14 | 2877 | enum slow_path_reason special; |
1774d762 JR |
2878 | struct ofpbuf old_stack = ctx->stack; |
2879 | union mf_subvalue new_stack[1024 / sizeof(union mf_subvalue)]; | |
ed9c9e3e JR |
2880 | struct ofpbuf old_action_set = ctx->action_set; |
2881 | uint64_t actset_stub[1024 / 8]; | |
9583bc14 | 2882 | |
1774d762 | 2883 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); |
ed9c9e3e | 2884 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); |
46c88433 EJ |
2885 | ctx->xbridge = peer->xbridge; |
2886 | flow->in_port.ofp_port = peer->ofp_port; | |
33bf9176 BP |
2887 | flow->metadata = htonll(0); |
2888 | memset(&flow->tunnel, 0, sizeof flow->tunnel); | |
2889 | memset(flow->regs, 0, sizeof flow->regs); | |
c61f3870 | 2890 | flow->actset_output = OFPP_UNSET; |
9583bc14 | 2891 | |
621b8064 JR |
2892 | /* The bridge is now known so obtain its table version. */ |
2893 | ctx->tables_version | |
2894 | = ofproto_dpif_get_tables_version(ctx->xbridge->ofproto); | |
2895 | ||
642dc74d | 2896 | special = process_special(ctx, &ctx->xin->flow, peer, |
9583bc14 EJ |
2897 | ctx->xin->packet); |
2898 | if (special) { | |
04594cd5 | 2899 | ctx->xout->slow |= special; |
ddd3c975 | 2900 | } else if (may_receive(peer, ctx)) { |
9efd308e | 2901 | if (xport_stp_forward_state(peer) && xport_rstp_forward_state(peer)) { |
e672ff9b | 2902 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); |
ed9c9e3e | 2903 | if (ctx->action_set.size) { |
e672ff9b JR |
2904 | /* Translate action set only if not dropping the packet and |
2905 | * not recirculating. */ | |
2906 | if (!exit_recirculates(ctx)) { | |
2907 | xlate_action_set(ctx); | |
2908 | } | |
2909 | } | |
2910 | /* Check if need to recirculate. */ | |
2911 | if (exit_recirculates(ctx)) { | |
2912 | compose_recirculate_action(ctx); | |
ed9c9e3e | 2913 | } |
9583bc14 | 2914 | } else { |
9efd308e DV |
2915 | /* Forwarding is disabled by STP and RSTP. Let OFPP_NORMAL and |
2916 | * the learning action look at the packet, then drop it. */ | |
9583bc14 | 2917 | struct flow old_base_flow = ctx->base_flow; |
6fd6ed71 | 2918 | size_t old_size = ctx->xout->odp_actions->size; |
cdf5d3a5 | 2919 | mirror_mask_t old_mirrors = ctx->xout->mirrors; |
f3d5b473 | 2920 | |
e672ff9b | 2921 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true); |
cdf5d3a5 | 2922 | ctx->xout->mirrors = old_mirrors; |
9583bc14 | 2923 | ctx->base_flow = old_base_flow; |
6fd6ed71 | 2924 | ctx->xout->odp_actions->size = old_size; |
e672ff9b JR |
2925 | |
2926 | /* Undo changes that may have been done for recirculation. */ | |
2927 | if (exit_recirculates(ctx)) { | |
2928 | ctx->action_set.size = ctx->recirc_action_offset; | |
2929 | ctx->recirc_action_offset = -1; | |
2930 | ctx->last_unroll_offset = -1; | |
2931 | } | |
9583bc14 EJ |
2932 | } |
2933 | } | |
2934 | ||
2935 | ctx->xin->flow = old_flow; | |
832554e3 | 2936 | ctx->xbridge = xport->xbridge; |
ed9c9e3e JR |
2937 | ofpbuf_uninit(&ctx->action_set); |
2938 | ctx->action_set = old_action_set; | |
1774d762 JR |
2939 | ofpbuf_uninit(&ctx->stack); |
2940 | ctx->stack = old_stack; | |
9583bc14 | 2941 | |
621b8064 JR |
2942 | /* Restore calling bridge's lookup version. */ |
2943 | ctx->tables_version = old_version; | |
2944 | ||
7f5fe537 JR |
2945 | /* The peer bridge popping MPLS should have no effect on the original |
2946 | * bridge. */ | |
2947 | ctx->was_mpls = old_was_mpls; | |
2948 | ||
f3d5b473 JR |
2949 | /* The fact that the peer bridge exits (for any reason) does not mean |
2950 | * that the original bridge should exit. Specifically, if the peer | |
2951 | * bridge recirculates (which typically modifies the packet), the | |
2952 | * original bridge must continue processing with the original, not the | |
2953 | * recirculated packet! */ | |
2954 | ctx->exit = false; | |
2955 | ||
9583bc14 | 2956 | if (ctx->xin->resubmit_stats) { |
46c88433 EJ |
2957 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
2958 | netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats); | |
a1aeea86 AW |
2959 | if (peer->bfd) { |
2960 | bfd_account_rx(peer->bfd, ctx->xin->resubmit_stats); | |
2961 | } | |
9583bc14 | 2962 | } |
b256dc52 JS |
2963 | if (ctx->xin->xcache) { |
2964 | struct xc_entry *entry; | |
2965 | ||
2966 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
2967 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
2968 | entry->u.dev.rx = netdev_ref(peer->netdev); | |
2969 | entry->u.dev.bfd = bfd_ref(peer->bfd); | |
2970 | } | |
9583bc14 EJ |
2971 | return; |
2972 | } | |
2973 | ||
33bf9176 | 2974 | flow_vlan_tci = flow->vlan_tci; |
1362e248 | 2975 | flow_pkt_mark = flow->pkt_mark; |
33bf9176 | 2976 | flow_nw_tos = flow->nw_tos; |
9583bc14 | 2977 | |
16194afd DDP |
2978 | if (count_skb_priorities(xport)) { |
2979 | memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority); | |
2980 | if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { | |
2981 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
2982 | flow->nw_tos &= ~IP_DSCP_MASK; | |
2983 | flow->nw_tos |= dscp; | |
2984 | } | |
9583bc14 EJ |
2985 | } |
2986 | ||
46c88433 | 2987 | if (xport->is_tunnel) { |
9583bc14 EJ |
2988 | /* Save tunnel metadata so that changes made due to |
2989 | * the Logical (tunnel) Port are not visible for any further | |
2990 | * matches, while explicit set actions on tunnel metadata are. | |
2991 | */ | |
a36de779 | 2992 | flow_tnl = flow->tunnel; |
46c88433 | 2993 | odp_port = tnl_port_send(xport->ofport, flow, &ctx->xout->wc); |
4e022ec0 | 2994 | if (odp_port == ODPP_NONE) { |
9583bc14 EJ |
2995 | xlate_report(ctx, "Tunneling decided against output"); |
2996 | goto out; /* restore flow_nw_tos */ | |
2997 | } | |
33bf9176 | 2998 | if (flow->tunnel.ip_dst == ctx->orig_tunnel_ip_dst) { |
9583bc14 EJ |
2999 | xlate_report(ctx, "Not tunneling to our own address"); |
3000 | goto out; /* restore flow_nw_tos */ | |
3001 | } | |
3002 | if (ctx->xin->resubmit_stats) { | |
46c88433 | 3003 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
9583bc14 | 3004 | } |
b256dc52 JS |
3005 | if (ctx->xin->xcache) { |
3006 | struct xc_entry *entry; | |
3007 | ||
3008 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
3009 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
3010 | } | |
9583bc14 | 3011 | out_port = odp_port; |
a36de779 | 3012 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
81de18ec | 3013 | xlate_report(ctx, "output to native tunnel"); |
a36de779 PS |
3014 | tnl_push_pop_send = true; |
3015 | } else { | |
81de18ec | 3016 | xlate_report(ctx, "output to kernel tunnel"); |
a36de779 PS |
3017 | commit_odp_tunnel_action(flow, &ctx->base_flow, |
3018 | ctx->xout->odp_actions); | |
3019 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ | |
3020 | } | |
9583bc14 | 3021 | } else { |
46c88433 | 3022 | odp_port = xport->odp_port; |
7614e5d0 | 3023 | out_port = odp_port; |
46c88433 | 3024 | if (ofproto_has_vlan_splinters(ctx->xbridge->ofproto)) { |
7614e5d0 JR |
3025 | ofp_port_t vlandev_port; |
3026 | ||
1dd35f8a | 3027 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
7614e5d0 JR |
3028 | vlandev_port = vsp_realdev_to_vlandev(ctx->xbridge->ofproto, |
3029 | ofp_port, flow->vlan_tci); | |
3030 | if (vlandev_port != ofp_port) { | |
3031 | out_port = ofp_port_to_odp_port(ctx->xbridge, vlandev_port); | |
3032 | flow->vlan_tci = htons(0); | |
3033 | } | |
9583bc14 | 3034 | } |
9583bc14 | 3035 | } |
9583bc14 | 3036 | |
4e022ec0 | 3037 | if (out_port != ODPP_NONE) { |
b440dd8c JS |
3038 | bool use_masked = ctx->xbridge->support.masked_set_action; |
3039 | ||
7fd91025 | 3040 | ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow, |
cc377352 | 3041 | ctx->xout->odp_actions, |
b440dd8c | 3042 | wc, use_masked); |
adcf00ba | 3043 | |
e93ef1c7 | 3044 | if (xr) { |
347bf289 | 3045 | struct ovs_action_hash *act_hash; |
adcf00ba | 3046 | |
347bf289 | 3047 | /* Hash action. */ |
cc377352 | 3048 | act_hash = nl_msg_put_unspec_uninit(ctx->xout->odp_actions, |
347bf289 AZ |
3049 | OVS_ACTION_ATTR_HASH, |
3050 | sizeof *act_hash); | |
3051 | act_hash->hash_alg = xr->hash_alg; | |
62ac1f20 | 3052 | act_hash->hash_basis = xr->hash_basis; |
347bf289 AZ |
3053 | |
3054 | /* Recirc action. */ | |
cc377352 | 3055 | nl_msg_put_u32(ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC, |
347bf289 | 3056 | xr->recirc_id); |
adcf00ba | 3057 | } else { |
a36de779 PS |
3058 | |
3059 | if (tnl_push_pop_send) { | |
3060 | build_tunnel_send(ctx, xport, flow, odp_port); | |
3061 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ | |
3062 | } else { | |
3063 | odp_port_t odp_tnl_port = ODPP_NONE; | |
3064 | ||
3065 | /* XXX: Write better Filter for tunnel port. We can use inport | |
3066 | * int tunnel-port flow to avoid these checks completely. */ | |
3067 | if (ofp_port == OFPP_LOCAL && | |
3068 | ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { | |
3069 | ||
3070 | odp_tnl_port = tnl_port_map_lookup(flow, wc); | |
3071 | } | |
3072 | ||
3073 | if (odp_tnl_port != ODPP_NONE) { | |
3074 | nl_msg_put_odp_port(ctx->xout->odp_actions, | |
3075 | OVS_ACTION_ATTR_TUNNEL_POP, | |
3076 | odp_tnl_port); | |
3077 | } else { | |
3078 | /* Tunnel push-pop action is not compatible with | |
3079 | * IPFIX action. */ | |
3080 | add_ipfix_output_action(ctx, out_port); | |
3081 | nl_msg_put_odp_port(ctx->xout->odp_actions, | |
3082 | OVS_ACTION_ATTR_OUTPUT, | |
3083 | out_port); | |
3084 | } | |
3085 | } | |
adcf00ba | 3086 | } |
9583bc14 | 3087 | |
6cbbf4fa EJ |
3088 | ctx->sflow_odp_port = odp_port; |
3089 | ctx->sflow_n_outputs++; | |
3090 | ctx->xout->nf_output_iface = ofp_port; | |
3091 | } | |
3092 | ||
3093 | out: | |
9583bc14 | 3094 | /* Restore flow */ |
33bf9176 | 3095 | flow->vlan_tci = flow_vlan_tci; |
1362e248 | 3096 | flow->pkt_mark = flow_pkt_mark; |
33bf9176 | 3097 | flow->nw_tos = flow_nw_tos; |
9583bc14 EJ |
3098 | } |
3099 | ||
3100 | static void | |
e93ef1c7 JR |
3101 | compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
3102 | const struct xlate_bond_recirc *xr) | |
9583bc14 | 3103 | { |
e93ef1c7 | 3104 | compose_output_action__(ctx, ofp_port, xr, true); |
9583bc14 EJ |
3105 | } |
3106 | ||
bb61b33d BP |
3107 | static void |
3108 | xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule) | |
bb61b33d BP |
3109 | { |
3110 | struct rule_dpif *old_rule = ctx->rule; | |
8b1e5560 | 3111 | ovs_be64 old_cookie = ctx->rule_cookie; |
dc723c44 | 3112 | const struct rule_actions *actions; |
bb61b33d BP |
3113 | |
3114 | if (ctx->xin->resubmit_stats) { | |
70742c7f | 3115 | rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats); |
bb61b33d BP |
3116 | } |
3117 | ||
98b07853 | 3118 | ctx->resubmits++; |
bb61b33d BP |
3119 | ctx->recurse++; |
3120 | ctx->rule = rule; | |
8b1e5560 | 3121 | ctx->rule_cookie = rule_dpif_get_flow_cookie(rule); |
6f00e29b BP |
3122 | actions = rule_dpif_get_actions(rule); |
3123 | do_xlate_actions(actions->ofpacts, actions->ofpacts_len, ctx); | |
8b1e5560 | 3124 | ctx->rule_cookie = old_cookie; |
bb61b33d BP |
3125 | ctx->rule = old_rule; |
3126 | ctx->recurse--; | |
bb61b33d BP |
3127 | } |
3128 | ||
bd3240ba SH |
3129 | static bool |
3130 | xlate_resubmit_resource_check(struct xlate_ctx *ctx) | |
9583bc14 | 3131 | { |
98b07853 BP |
3132 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); |
3133 | ||
adcf00ba | 3134 | if (ctx->recurse >= MAX_RESUBMIT_RECURSION + MAX_INTERNAL_RESUBMITS) { |
98b07853 BP |
3135 | VLOG_ERR_RL(&rl, "resubmit actions recursed over %d times", |
3136 | MAX_RESUBMIT_RECURSION); | |
adcf00ba | 3137 | } else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) { |
98b07853 | 3138 | VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS); |
6fd6ed71 | 3139 | } else if (ctx->xout->odp_actions->size > UINT16_MAX) { |
98b07853 | 3140 | VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of actions"); |
6fd6ed71 | 3141 | } else if (ctx->stack.size >= 65536) { |
98b07853 BP |
3142 | VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of stack"); |
3143 | } else { | |
bd3240ba SH |
3144 | return true; |
3145 | } | |
3146 | ||
3147 | return false; | |
3148 | } | |
3149 | ||
3150 | static void | |
6d328fa2 SH |
3151 | xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, |
3152 | bool may_packet_in, bool honor_table_miss) | |
bd3240ba | 3153 | { |
1d741d6d JR |
3154 | /* Check if we need to recirculate before matching in a table. */ |
3155 | if (ctx->was_mpls) { | |
3156 | ctx_trigger_recirculation(ctx); | |
3157 | return; | |
3158 | } | |
bd3240ba | 3159 | if (xlate_resubmit_resource_check(ctx)) { |
34dd0d78 | 3160 | struct flow_wildcards *wc; |
9583bc14 | 3161 | uint8_t old_table_id = ctx->table_id; |
3f207910 | 3162 | struct rule_dpif *rule; |
9583bc14 EJ |
3163 | |
3164 | ctx->table_id = table_id; | |
34dd0d78 | 3165 | wc = (ctx->xin->skip_wildcards) ? NULL : &ctx->xout->wc; |
9583bc14 | 3166 | |
34dd0d78 | 3167 | rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, |
621b8064 | 3168 | ctx->tables_version, |
34dd0d78 JR |
3169 | &ctx->xin->flow, wc, |
3170 | ctx->xin->xcache != NULL, | |
3171 | ctx->xin->resubmit_stats, | |
3172 | &ctx->table_id, in_port, | |
3173 | may_packet_in, honor_table_miss); | |
9583bc14 | 3174 | |
a8c31348 BP |
3175 | if (OVS_UNLIKELY(ctx->xin->resubmit_hook)) { |
3176 | ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse + 1); | |
ad3efdcb EJ |
3177 | } |
3178 | ||
a2143702 | 3179 | if (rule) { |
83709dfa JR |
3180 | /* Fill in the cache entry here instead of xlate_recursively |
3181 | * to make the reference counting more explicit. We take a | |
3182 | * reference in the lookups above if we are going to cache the | |
3183 | * rule. */ | |
3184 | if (ctx->xin->xcache) { | |
3185 | struct xc_entry *entry; | |
3186 | ||
3187 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE); | |
3188 | entry->u.rule = rule; | |
3189 | } | |
bb61b33d | 3190 | xlate_recursively(ctx, rule); |
ad3efdcb EJ |
3191 | } |
3192 | ||
9583bc14 | 3193 | ctx->table_id = old_table_id; |
98b07853 | 3194 | return; |
9583bc14 | 3195 | } |
98b07853 BP |
3196 | |
3197 | ctx->exit = true; | |
9583bc14 EJ |
3198 | } |
3199 | ||
f4fb341b | 3200 | static void |
1e684d7d RW |
3201 | xlate_group_stats(struct xlate_ctx *ctx, struct group_dpif *group, |
3202 | struct ofputil_bucket *bucket) | |
3203 | { | |
3204 | if (ctx->xin->resubmit_stats) { | |
3205 | group_dpif_credit_stats(group, bucket, ctx->xin->resubmit_stats); | |
3206 | } | |
3207 | if (ctx->xin->xcache) { | |
3208 | struct xc_entry *entry; | |
3209 | ||
3210 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_GROUP); | |
3211 | entry->u.group.group = group_dpif_ref(group); | |
3212 | entry->u.group.bucket = bucket; | |
3213 | } | |
3214 | } | |
3215 | ||
3216 | static void | |
3217 | xlate_group_bucket(struct xlate_ctx *ctx, struct ofputil_bucket *bucket) | |
f4fb341b SH |
3218 | { |
3219 | uint64_t action_list_stub[1024 / 8]; | |
3220 | struct ofpbuf action_list, action_set; | |
5b09e569 JR |
3221 | struct flow old_flow = ctx->xin->flow; |
3222 | bool old_was_mpls = ctx->was_mpls; | |
f4fb341b SH |
3223 | |
3224 | ofpbuf_use_const(&action_set, bucket->ofpacts, bucket->ofpacts_len); | |
3225 | ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub); | |
3226 | ||
3227 | ofpacts_execute_action_set(&action_list, &action_set); | |
3228 | ctx->recurse++; | |
6fd6ed71 | 3229 | do_xlate_actions(action_list.data, action_list.size, ctx); |
f4fb341b SH |
3230 | ctx->recurse--; |
3231 | ||
3232 | ofpbuf_uninit(&action_set); | |
3233 | ofpbuf_uninit(&action_list); | |
5b09e569 | 3234 | |
e672ff9b JR |
3235 | /* Check if need to recirculate. */ |
3236 | if (exit_recirculates(ctx)) { | |
3237 | compose_recirculate_action(ctx); | |
3238 | } | |
3239 | ||
5b09e569 JR |
3240 | /* Roll back flow to previous state. |
3241 | * This is equivalent to cloning the packet for each bucket. | |
3242 | * | |
3243 | * As a side effect any subsequently applied actions will | |
3244 | * also effectively be applied to a clone of the packet taken | |
3245 | * just before applying the all or indirect group. | |
3246 | * | |
3247 | * Note that group buckets are action sets, hence they cannot modify the | |
3248 | * main action set. Also any stack actions are ignored when executing an | |
3249 | * action set, so group buckets cannot change the stack either. | |
3250 | * However, we do allow resubmit actions in group buckets, which could | |
3251 | * break the above assumptions. It is up to the controller to not mess up | |
3252 | * with the action_set and stack in the tables resubmitted to from | |
3253 | * group buckets. */ | |
3254 | ctx->xin->flow = old_flow; | |
3255 | ||
3256 | /* The group bucket popping MPLS should have no effect after bucket | |
3257 | * execution. */ | |
3258 | ctx->was_mpls = old_was_mpls; | |
3259 | ||
3260 | /* The fact that the group bucket exits (for any reason) does not mean that | |
3261 | * the translation after the group action should exit. Specifically, if | |
3262 | * the group bucket recirculates (which typically modifies the packet), the | |
3263 | * actions after the group action must continue processing with the | |
3264 | * original, not the recirculated packet! */ | |
3265 | ctx->exit = false; | |
f4fb341b SH |
3266 | } |
3267 | ||
3268 | static void | |
3269 | xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3270 | { | |
1e684d7d | 3271 | struct ofputil_bucket *bucket; |
ca6ba700 | 3272 | const struct ovs_list *buckets; |
f4fb341b SH |
3273 | |
3274 | group_dpif_get_buckets(group, &buckets); | |
3275 | ||
3276 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
3277 | xlate_group_bucket(ctx, bucket); | |
f4fb341b | 3278 | } |
1e684d7d | 3279 | xlate_group_stats(ctx, group, NULL); |
f4fb341b SH |
3280 | } |
3281 | ||
dd8cd4b4 SH |
3282 | static void |
3283 | xlate_ff_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3284 | { | |
1e684d7d | 3285 | struct ofputil_bucket *bucket; |
dd8cd4b4 SH |
3286 | |
3287 | bucket = group_first_live_bucket(ctx, group, 0); | |
3288 | if (bucket) { | |
3289 | xlate_group_bucket(ctx, bucket); | |
1e684d7d | 3290 | xlate_group_stats(ctx, group, bucket); |
dd8cd4b4 SH |
3291 | } |
3292 | } | |
3293 | ||
fe7e5749 | 3294 | static void |
7565c3e4 | 3295 | xlate_default_select_group(struct xlate_ctx *ctx, struct group_dpif *group) |
fe7e5749 SH |
3296 | { |
3297 | struct flow_wildcards *wc = &ctx->xout->wc; | |
1e684d7d | 3298 | struct ofputil_bucket *bucket; |
fe7e5749 SH |
3299 | uint32_t basis; |
3300 | ||
1d1aae0b | 3301 | basis = flow_hash_symmetric_l4(&ctx->xin->flow, 0); |
80e3509d | 3302 | flow_mask_hash_fields(&ctx->xin->flow, wc, NX_HASH_FIELDS_SYMMETRIC_L4); |
fe7e5749 SH |
3303 | bucket = group_best_live_bucket(ctx, group, basis); |
3304 | if (bucket) { | |
fe7e5749 | 3305 | xlate_group_bucket(ctx, bucket); |
1e684d7d | 3306 | xlate_group_stats(ctx, group, bucket); |
fe7e5749 SH |
3307 | } |
3308 | } | |
3309 | ||
0c4b9393 SH |
3310 | static void |
3311 | xlate_hash_fields_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3312 | { | |
3313 | struct mf_bitmap hash_fields = MF_BITMAP_INITIALIZER; | |
3314 | struct flow_wildcards *wc = &ctx->xout->wc; | |
3315 | const struct field_array *fields; | |
3316 | struct ofputil_bucket *bucket; | |
3317 | uint32_t basis; | |
3318 | int i; | |
3319 | ||
3320 | fields = group_dpif_get_fields(group); | |
3321 | basis = hash_uint64(group_dpif_get_selection_method_param(group)); | |
3322 | ||
3323 | /* Determine which fields to hash */ | |
3324 | for (i = 0; i < MFF_N_IDS; i++) { | |
3325 | if (bitmap_is_set(fields->used.bm, i)) { | |
3326 | const struct mf_field *mf; | |
3327 | ||
3328 | /* If the field is already present in 'hash_fields' then | |
3329 | * this loop has already checked that it and its pre-requisites | |
3330 | * are present in the flow and its pre-requisites have | |
3331 | * already been added to 'hash_fields'. There is nothing more | |
3332 | * to do here and as an optimisation the loop can continue. */ | |
3333 | if (bitmap_is_set(hash_fields.bm, i)) { | |
3334 | continue; | |
3335 | } | |
3336 | ||
3337 | mf = mf_from_id(i); | |
3338 | ||
3339 | /* Only hash a field if it and its pre-requisites are present | |
3340 | * in the flow. */ | |
3341 | if (!mf_are_prereqs_ok(mf, &ctx->xin->flow)) { | |
3342 | continue; | |
3343 | } | |
3344 | ||
3345 | /* Hash both the field and its pre-requisites */ | |
3346 | mf_bitmap_set_field_and_prereqs(mf, &hash_fields); | |
3347 | } | |
3348 | } | |
3349 | ||
3350 | /* Hash the fields */ | |
3351 | for (i = 0; i < MFF_N_IDS; i++) { | |
3352 | if (bitmap_is_set(hash_fields.bm, i)) { | |
3353 | const struct mf_field *mf = mf_from_id(i); | |
3354 | union mf_value value; | |
3355 | int j; | |
3356 | ||
3357 | mf_get_value(mf, &ctx->xin->flow, &value); | |
3358 | /* This seems inefficient but so does apply_mask() */ | |
3359 | for (j = 0; j < mf->n_bytes; j++) { | |
3360 | ((uint8_t *) &value)[j] &= ((uint8_t *) &fields->value[i])[j]; | |
3361 | } | |
3362 | basis = hash_bytes(&value, mf->n_bytes, basis); | |
3363 | ||
3364 | mf_mask_field(mf, &wc->masks); | |
3365 | } | |
3366 | } | |
3367 | ||
3368 | bucket = group_best_live_bucket(ctx, group, basis); | |
3369 | if (bucket) { | |
3370 | xlate_group_bucket(ctx, bucket); | |
3371 | xlate_group_stats(ctx, group, bucket); | |
3372 | } | |
3373 | } | |
3374 | ||
7565c3e4 SH |
3375 | static void |
3376 | xlate_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3377 | { | |
3378 | const char *selection_method = group_dpif_get_selection_method(group); | |
3379 | ||
3380 | if (selection_method[0] == '\0') { | |
3381 | xlate_default_select_group(ctx, group); | |
0c4b9393 SH |
3382 | } else if (!strcasecmp("hash", selection_method)) { |
3383 | xlate_hash_fields_select_group(ctx, group); | |
7565c3e4 SH |
3384 | } else { |
3385 | /* Parsing of groups should ensure this never happens */ | |
3386 | OVS_NOT_REACHED(); | |
3387 | } | |
3388 | } | |
3389 | ||
f4fb341b SH |
3390 | static void |
3391 | xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group) | |
3392 | { | |
0eb48fe1 | 3393 | bool was_in_group = ctx->in_group; |
5a070238 BP |
3394 | ctx->in_group = true; |
3395 | ||
f4fb341b SH |
3396 | switch (group_dpif_get_type(group)) { |
3397 | case OFPGT11_ALL: | |
3398 | case OFPGT11_INDIRECT: | |
3399 | xlate_all_group(ctx, group); | |
3400 | break; | |
3401 | case OFPGT11_SELECT: | |
fe7e5749 | 3402 | xlate_select_group(ctx, group); |
f4fb341b | 3403 | break; |
dd8cd4b4 SH |
3404 | case OFPGT11_FF: |
3405 | xlate_ff_group(ctx, group); | |
3406 | break; | |
f4fb341b | 3407 | default: |
428b2edd | 3408 | OVS_NOT_REACHED(); |
f4fb341b | 3409 | } |
809c7548 | 3410 | group_dpif_unref(group); |
5a070238 | 3411 | |
0eb48fe1 | 3412 | ctx->in_group = was_in_group; |
f4fb341b SH |
3413 | } |
3414 | ||
3415 | static bool | |
3416 | xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id) | |
3417 | { | |
0eb48fe1 | 3418 | if (xlate_resubmit_resource_check(ctx)) { |
f4fb341b SH |
3419 | struct group_dpif *group; |
3420 | bool got_group; | |
3421 | ||
3422 | got_group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group); | |
3423 | if (got_group) { | |
3424 | xlate_group_action__(ctx, group); | |
3425 | } else { | |
3426 | return true; | |
3427 | } | |
3428 | } | |
3429 | ||
3430 | return false; | |
3431 | } | |
3432 | ||
9583bc14 EJ |
3433 | static void |
3434 | xlate_ofpact_resubmit(struct xlate_ctx *ctx, | |
3435 | const struct ofpact_resubmit *resubmit) | |
3436 | { | |
4e022ec0 | 3437 | ofp_port_t in_port; |
9583bc14 | 3438 | uint8_t table_id; |
adcf00ba AZ |
3439 | bool may_packet_in = false; |
3440 | bool honor_table_miss = false; | |
3441 | ||
3442 | if (ctx->rule && rule_dpif_is_internal(ctx->rule)) { | |
3443 | /* Still allow missed packets to be sent to the controller | |
3444 | * if resubmitting from an internal table. */ | |
3445 | may_packet_in = true; | |
3446 | honor_table_miss = true; | |
3447 | } | |
9583bc14 EJ |
3448 | |
3449 | in_port = resubmit->in_port; | |
3450 | if (in_port == OFPP_IN_PORT) { | |
4e022ec0 | 3451 | in_port = ctx->xin->flow.in_port.ofp_port; |
9583bc14 EJ |
3452 | } |
3453 | ||
3454 | table_id = resubmit->table_id; | |
3455 | if (table_id == 255) { | |
3456 | table_id = ctx->table_id; | |
3457 | } | |
3458 | ||
adcf00ba AZ |
3459 | xlate_table_action(ctx, in_port, table_id, may_packet_in, |
3460 | honor_table_miss); | |
9583bc14 EJ |
3461 | } |
3462 | ||
3463 | static void | |
3464 | flood_packets(struct xlate_ctx *ctx, bool all) | |
3465 | { | |
46c88433 | 3466 | const struct xport *xport; |
9583bc14 | 3467 | |
46c88433 EJ |
3468 | HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) { |
3469 | if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3470 | continue; |
3471 | } | |
3472 | ||
3473 | if (all) { | |
e93ef1c7 | 3474 | compose_output_action__(ctx, xport->ofp_port, NULL, false); |
46c88433 | 3475 | } else if (!(xport->config & OFPUTIL_PC_NO_FLOOD)) { |
e93ef1c7 | 3476 | compose_output_action(ctx, xport->ofp_port, NULL); |
9583bc14 EJ |
3477 | } |
3478 | } | |
3479 | ||
3480 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
3481 | } | |
3482 | ||
3483 | static void | |
3484 | execute_controller_action(struct xlate_ctx *ctx, int len, | |
3485 | enum ofp_packet_in_reason reason, | |
3486 | uint16_t controller_id) | |
3487 | { | |
0fb7792a | 3488 | struct ofproto_packet_in *pin; |
e14deea0 | 3489 | struct dp_packet *packet; |
b440dd8c | 3490 | bool use_masked; |
9583bc14 | 3491 | |
04594cd5 | 3492 | ctx->xout->slow |= SLOW_CONTROLLER; |
9583bc14 EJ |
3493 | if (!ctx->xin->packet) { |
3494 | return; | |
3495 | } | |
3496 | ||
cf62fa4c | 3497 | packet = dp_packet_clone(ctx->xin->packet); |
9583bc14 | 3498 | |
b440dd8c | 3499 | use_masked = ctx->xbridge->support.masked_set_action; |
7fd91025 | 3500 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, |
cc377352 | 3501 | ctx->xout->odp_actions, |
b440dd8c | 3502 | &ctx->xout->wc, use_masked); |
9583bc14 | 3503 | |
41ccaa24 | 3504 | odp_execute_actions(NULL, &packet, 1, false, |
6fd6ed71 PS |
3505 | ctx->xout->odp_actions->data, |
3506 | ctx->xout->odp_actions->size, NULL); | |
9583bc14 | 3507 | |
ada3a58d | 3508 | pin = xmalloc(sizeof *pin); |
cf62fa4c PS |
3509 | pin->up.packet_len = dp_packet_size(packet); |
3510 | pin->up.packet = dp_packet_steal_data(packet); | |
0fb7792a | 3511 | pin->up.reason = reason; |
0fb7792a | 3512 | pin->up.table_id = ctx->table_id; |
8b1e5560 | 3513 | pin->up.cookie = ctx->rule_cookie; |
0fb7792a | 3514 | |
50dcbd8e | 3515 | flow_get_metadata(&ctx->xin->flow, &pin->up.flow_metadata); |
9583bc14 | 3516 | |
f11c7538 | 3517 | pin->controller_id = controller_id; |
d38a3c7b | 3518 | pin->send_len = len; |
32260212 SH |
3519 | /* If a rule is a table-miss rule then this is |
3520 | * a table-miss handled by a table-miss rule. | |
3521 | * | |
3522 | * Else, if rule is internal and has a controller action, | |
3523 | * the later being implied by the rule being processed here, | |
3524 | * then this is a table-miss handled without a table-miss rule. | |
3525 | * | |
3526 | * Otherwise this is not a table-miss. */ | |
3527 | pin->miss_type = OFPROTO_PACKET_IN_NO_MISS; | |
3528 | if (ctx->rule) { | |
3529 | if (rule_dpif_is_table_miss(ctx->rule)) { | |
3530 | pin->miss_type = OFPROTO_PACKET_IN_MISS_FLOW; | |
3531 | } else if (rule_dpif_is_internal(ctx->rule)) { | |
3532 | pin->miss_type = OFPROTO_PACKET_IN_MISS_WITHOUT_FLOW; | |
3533 | } | |
3534 | } | |
ada3a58d | 3535 | ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, pin); |
e14deea0 | 3536 | dp_packet_delete(packet); |
9583bc14 EJ |
3537 | } |
3538 | ||
e672ff9b | 3539 | /* Called only when ctx->recirc_action_offset is set. */ |
7bbdd84f | 3540 | static void |
e672ff9b | 3541 | compose_recirculate_action(struct xlate_ctx *ctx) |
7bbdd84f | 3542 | { |
e672ff9b | 3543 | struct recirc_metadata md; |
b440dd8c | 3544 | bool use_masked; |
7bbdd84f | 3545 | uint32_t id; |
7bbdd84f | 3546 | |
b440dd8c | 3547 | use_masked = ctx->xbridge->support.masked_set_action; |
e672ff9b JR |
3548 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, |
3549 | ctx->xout->odp_actions, | |
b440dd8c | 3550 | &ctx->xout->wc, use_masked); |
7bbdd84f | 3551 | |
e672ff9b | 3552 | recirc_metadata_from_flow(&md, &ctx->xin->flow); |
7bbdd84f | 3553 | |
e672ff9b | 3554 | ovs_assert(ctx->recirc_action_offset >= 0); |
7bbdd84f | 3555 | |
e672ff9b JR |
3556 | /* Only allocate recirculation ID if we have a packet. */ |
3557 | if (ctx->xin->packet) { | |
3558 | /* Allocate a unique recirc id for the given metadata state in the | |
3559 | * flow. The life-cycle of this recirc id is managed by associating it | |
3560 | * with the udpif key ('ukey') created for each new datapath flow. */ | |
3561 | id = recirc_alloc_id_ctx(ctx->xbridge->ofproto, 0, &md, &ctx->stack, | |
3562 | ctx->recirc_action_offset, | |
3563 | ctx->action_set.size, ctx->action_set.data); | |
3564 | if (!id) { | |
3565 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3566 | VLOG_ERR_RL(&rl, "Failed to allocate recirculation id"); | |
3567 | return; | |
e2676b44 | 3568 | } |
e672ff9b JR |
3569 | xlate_out_add_recirc(ctx->xout, id); |
3570 | } else { | |
3571 | /* Look up an existing recirc id for the given metadata state in the | |
3572 | * flow. No new reference is taken, as the ID is RCU protected and is | |
3573 | * only required temporarily for verification. */ | |
3574 | id = recirc_find_id(ctx->xbridge->ofproto, 0, &md, &ctx->stack, | |
3575 | ctx->recirc_action_offset, | |
3576 | ctx->action_set.size, ctx->action_set.data); | |
3577 | /* We let zero 'id' to be used in the RECIRC action below, which will | |
3578 | * fail all revalidations as zero is not a valid recirculation ID. */ | |
7bbdd84f SH |
3579 | } |
3580 | ||
cc377352 | 3581 | nl_msg_put_u32(ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC, id); |
e672ff9b JR |
3582 | |
3583 | /* Undo changes done by recirculation. */ | |
3584 | ctx->action_set.size = ctx->recirc_action_offset; | |
3585 | ctx->recirc_action_offset = -1; | |
3586 | ctx->last_unroll_offset = -1; | |
7bbdd84f SH |
3587 | } |
3588 | ||
8bfd0fda BP |
3589 | static void |
3590 | compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) | |
9583bc14 | 3591 | { |
33bf9176 BP |
3592 | struct flow_wildcards *wc = &ctx->xout->wc; |
3593 | struct flow *flow = &ctx->xin->flow; | |
8bfd0fda | 3594 | int n; |
33bf9176 | 3595 | |
8bfd0fda | 3596 | ovs_assert(eth_type_mpls(mpls->ethertype)); |
b0a17866 | 3597 | |
8bfd0fda BP |
3598 | n = flow_count_mpls_labels(flow, wc); |
3599 | if (!n) { | |
b440dd8c JS |
3600 | bool use_masked = ctx->xbridge->support.masked_set_action; |
3601 | ||
8bfd0fda | 3602 | ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow, |
cc377352 | 3603 | ctx->xout->odp_actions, |
b440dd8c | 3604 | &ctx->xout->wc, use_masked); |
8bfd0fda BP |
3605 | } else if (n >= FLOW_MAX_MPLS_LABELS) { |
3606 | if (ctx->xin->packet != NULL) { | |
3607 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3608 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an " | |
3609 | "MPLS push action can't be performed as it would " | |
3610 | "have more MPLS LSEs than the %d supported.", | |
3611 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
9583bc14 | 3612 | } |
8bfd0fda BP |
3613 | ctx->exit = true; |
3614 | return; | |
9583bc14 | 3615 | } |
b0a17866 | 3616 | |
8bfd0fda | 3617 | flow_push_mpls(flow, n, mpls->ethertype, wc); |
9583bc14 EJ |
3618 | } |
3619 | ||
8bfd0fda | 3620 | static void |
9cfef3d0 | 3621 | compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) |
9583bc14 | 3622 | { |
33bf9176 | 3623 | struct flow_wildcards *wc = &ctx->xout->wc; |
8bfd0fda BP |
3624 | struct flow *flow = &ctx->xin->flow; |
3625 | int n = flow_count_mpls_labels(flow, wc); | |
33bf9176 | 3626 | |
7bbdd84f | 3627 | if (flow_pop_mpls(flow, n, eth_type, wc)) { |
2494ccd7 | 3628 | if (ctx->xbridge->support.odp.recirc) { |
7bbdd84f SH |
3629 | ctx->was_mpls = true; |
3630 | } | |
3631 | } else if (n >= FLOW_MAX_MPLS_LABELS) { | |
8bfd0fda BP |
3632 | if (ctx->xin->packet != NULL) { |
3633 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3634 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an " | |
3635 | "MPLS pop action can't be performed as it has " | |
3636 | "more MPLS LSEs than the %d supported.", | |
3637 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
3638 | } | |
3639 | ctx->exit = true; | |
cc377352 | 3640 | ofpbuf_clear(ctx->xout->odp_actions); |
9583bc14 EJ |
3641 | } |
3642 | } | |
3643 | ||
3644 | static bool | |
3645 | compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids) | |
3646 | { | |
33bf9176 BP |
3647 | struct flow *flow = &ctx->xin->flow; |
3648 | ||
3649 | if (!is_ip_any(flow)) { | |
9583bc14 EJ |
3650 | return false; |
3651 | } | |
3652 | ||
1dd35f8a | 3653 | ctx->xout->wc.masks.nw_ttl = 0xff; |
33bf9176 BP |
3654 | if (flow->nw_ttl > 1) { |
3655 | flow->nw_ttl--; | |
9583bc14 EJ |
3656 | return false; |
3657 | } else { | |
3658 | size_t i; | |
3659 | ||
3660 | for (i = 0; i < ids->n_controllers; i++) { | |
3661 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, | |
3662 | ids->cnt_ids[i]); | |
3663 | } | |
3664 | ||
3665 | /* Stop processing for current table. */ | |
3666 | return true; | |
3667 | } | |
3668 | } | |
3669 | ||
8bfd0fda | 3670 | static void |
097d4939 JR |
3671 | compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label) |
3672 | { | |
8bfd0fda BP |
3673 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
3674 | ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK); | |
3675 | set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label); | |
097d4939 | 3676 | } |
097d4939 JR |
3677 | } |
3678 | ||
8bfd0fda | 3679 | static void |
097d4939 JR |
3680 | compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc) |
3681 | { | |
8bfd0fda BP |
3682 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
3683 | ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_TC_MASK); | |
3684 | set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc); | |
097d4939 | 3685 | } |
097d4939 JR |
3686 | } |
3687 | ||
8bfd0fda | 3688 | static void |
9cfef3d0 | 3689 | compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl) |
9583bc14 | 3690 | { |
8bfd0fda BP |
3691 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
3692 | ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); | |
3693 | set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl); | |
b0a17866 | 3694 | } |
9583bc14 EJ |
3695 | } |
3696 | ||
3697 | static bool | |
9cfef3d0 | 3698 | compose_dec_mpls_ttl_action(struct xlate_ctx *ctx) |
9583bc14 | 3699 | { |
33bf9176 | 3700 | struct flow *flow = &ctx->xin->flow; |
1dd35f8a JP |
3701 | struct flow_wildcards *wc = &ctx->xout->wc; |
3702 | ||
8bfd0fda | 3703 | if (eth_type_mpls(flow->dl_type)) { |
22d38fca JR |
3704 | uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]); |
3705 | ||
3706 | wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); | |
8bfd0fda BP |
3707 | if (ttl > 1) { |
3708 | ttl--; | |
3709 | set_mpls_lse_ttl(&flow->mpls_lse[0], ttl); | |
3710 | return false; | |
3711 | } else { | |
3712 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0); | |
8bfd0fda | 3713 | } |
9583bc14 | 3714 | } |
22d38fca JR |
3715 | |
3716 | /* Stop processing for current table. */ | |
3717 | return true; | |
9583bc14 EJ |
3718 | } |
3719 | ||
3720 | static void | |
3721 | xlate_output_action(struct xlate_ctx *ctx, | |
4e022ec0 | 3722 | ofp_port_t port, uint16_t max_len, bool may_packet_in) |
9583bc14 | 3723 | { |
4e022ec0 | 3724 | ofp_port_t prev_nf_output_iface = ctx->xout->nf_output_iface; |
9583bc14 EJ |
3725 | |
3726 | ctx->xout->nf_output_iface = NF_OUT_DROP; | |
3727 | ||
3728 | switch (port) { | |
3729 | case OFPP_IN_PORT: | |
e93ef1c7 | 3730 | compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port, NULL); |
9583bc14 EJ |
3731 | break; |
3732 | case OFPP_TABLE: | |
4e022ec0 | 3733 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 3734 | 0, may_packet_in, true); |
9583bc14 EJ |
3735 | break; |
3736 | case OFPP_NORMAL: | |
3737 | xlate_normal(ctx); | |
3738 | break; | |
3739 | case OFPP_FLOOD: | |
3740 | flood_packets(ctx, false); | |
3741 | break; | |
3742 | case OFPP_ALL: | |
3743 | flood_packets(ctx, true); | |
3744 | break; | |
3745 | case OFPP_CONTROLLER: | |
3a11fd5b | 3746 | execute_controller_action(ctx, max_len, |
029ca940 SS |
3747 | (ctx->in_group ? OFPR_GROUP |
3748 | : ctx->in_action_set ? OFPR_ACTION_SET | |
3749 | : OFPR_ACTION), | |
3750 | 0); | |
9583bc14 EJ |
3751 | break; |
3752 | case OFPP_NONE: | |
3753 | break; | |
3754 | case OFPP_LOCAL: | |
3755 | default: | |
4e022ec0 | 3756 | if (port != ctx->xin->flow.in_port.ofp_port) { |
e93ef1c7 | 3757 | compose_output_action(ctx, port, NULL); |
9583bc14 EJ |
3758 | } else { |
3759 | xlate_report(ctx, "skipping output to input port"); | |
3760 | } | |
3761 | break; | |
3762 | } | |
3763 | ||
3764 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
3765 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
3766 | } else if (ctx->xout->nf_output_iface == NF_OUT_DROP) { | |
3767 | ctx->xout->nf_output_iface = prev_nf_output_iface; | |
3768 | } else if (prev_nf_output_iface != NF_OUT_DROP && | |
3769 | ctx->xout->nf_output_iface != NF_OUT_FLOOD) { | |
3770 | ctx->xout->nf_output_iface = NF_OUT_MULTI; | |
3771 | } | |
3772 | } | |
3773 | ||
3774 | static void | |
3775 | xlate_output_reg_action(struct xlate_ctx *ctx, | |
3776 | const struct ofpact_output_reg *or) | |
3777 | { | |
3778 | uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow); | |
3779 | if (port <= UINT16_MAX) { | |
3780 | union mf_subvalue value; | |
3781 | ||
3782 | memset(&value, 0xff, sizeof value); | |
3783 | mf_write_subfield_flow(&or->src, &value, &ctx->xout->wc.masks); | |
4e022ec0 AW |
3784 | xlate_output_action(ctx, u16_to_ofp(port), |
3785 | or->max_len, false); | |
9583bc14 EJ |
3786 | } |
3787 | } | |
3788 | ||
3789 | static void | |
3790 | xlate_enqueue_action(struct xlate_ctx *ctx, | |
3791 | const struct ofpact_enqueue *enqueue) | |
3792 | { | |
4e022ec0 | 3793 | ofp_port_t ofp_port = enqueue->port; |
9583bc14 EJ |
3794 | uint32_t queue_id = enqueue->queue; |
3795 | uint32_t flow_priority, priority; | |
3796 | int error; | |
3797 | ||
3798 | /* Translate queue to priority. */ | |
89a8a7f0 | 3799 | error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority); |
9583bc14 EJ |
3800 | if (error) { |
3801 | /* Fall back to ordinary output action. */ | |
3802 | xlate_output_action(ctx, enqueue->port, 0, false); | |
3803 | return; | |
3804 | } | |
3805 | ||
3806 | /* Check output port. */ | |
3807 | if (ofp_port == OFPP_IN_PORT) { | |
4e022ec0 AW |
3808 | ofp_port = ctx->xin->flow.in_port.ofp_port; |
3809 | } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3810 | return; |
3811 | } | |
3812 | ||
3813 | /* Add datapath actions. */ | |
3814 | flow_priority = ctx->xin->flow.skb_priority; | |
3815 | ctx->xin->flow.skb_priority = priority; | |
e93ef1c7 | 3816 | compose_output_action(ctx, ofp_port, NULL); |
9583bc14 EJ |
3817 | ctx->xin->flow.skb_priority = flow_priority; |
3818 | ||
3819 | /* Update NetFlow output port. */ | |
3820 | if (ctx->xout->nf_output_iface == NF_OUT_DROP) { | |
3821 | ctx->xout->nf_output_iface = ofp_port; | |
3822 | } else if (ctx->xout->nf_output_iface != NF_OUT_FLOOD) { | |
3823 | ctx->xout->nf_output_iface = NF_OUT_MULTI; | |
3824 | } | |
3825 | } | |
3826 | ||
3827 | static void | |
3828 | xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id) | |
3829 | { | |
3830 | uint32_t skb_priority; | |
3831 | ||
89a8a7f0 | 3832 | if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) { |
9583bc14 EJ |
3833 | ctx->xin->flow.skb_priority = skb_priority; |
3834 | } else { | |
3835 | /* Couldn't translate queue to a priority. Nothing to do. A warning | |
3836 | * has already been logged. */ | |
3837 | } | |
3838 | } | |
3839 | ||
3840 | static bool | |
46c88433 | 3841 | slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_) |
9583bc14 | 3842 | { |
46c88433 EJ |
3843 | const struct xbridge *xbridge = xbridge_; |
3844 | struct xport *port; | |
9583bc14 EJ |
3845 | |
3846 | switch (ofp_port) { | |
3847 | case OFPP_IN_PORT: | |
3848 | case OFPP_TABLE: | |
3849 | case OFPP_NORMAL: | |
3850 | case OFPP_FLOOD: | |
3851 | case OFPP_ALL: | |
3852 | case OFPP_NONE: | |
3853 | return true; | |
3854 | case OFPP_CONTROLLER: /* Not supported by the bundle action. */ | |
3855 | return false; | |
3856 | default: | |
46c88433 | 3857 | port = get_ofp_port(xbridge, ofp_port); |
9583bc14 EJ |
3858 | return port ? port->may_enable : false; |
3859 | } | |
3860 | } | |
3861 | ||
3862 | static void | |
3863 | xlate_bundle_action(struct xlate_ctx *ctx, | |
3864 | const struct ofpact_bundle *bundle) | |
3865 | { | |
4e022ec0 | 3866 | ofp_port_t port; |
9583bc14 EJ |
3867 | |
3868 | port = bundle_execute(bundle, &ctx->xin->flow, &ctx->xout->wc, | |
46c88433 EJ |
3869 | slave_enabled_cb, |
3870 | CONST_CAST(struct xbridge *, ctx->xbridge)); | |
9583bc14 | 3871 | if (bundle->dst.field) { |
f74e7df7 JP |
3872 | nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow, |
3873 | &ctx->xout->wc); | |
9583bc14 EJ |
3874 | } else { |
3875 | xlate_output_action(ctx, port, 0, false); | |
3876 | } | |
3877 | } | |
3878 | ||
3879 | static void | |
4165b5e0 JS |
3880 | xlate_learn_action__(struct xlate_ctx *ctx, const struct ofpact_learn *learn, |
3881 | struct ofputil_flow_mod *fm, struct ofpbuf *ofpacts) | |
9583bc14 | 3882 | { |
4165b5e0 JS |
3883 | learn_execute(learn, &ctx->xin->flow, fm, ofpacts); |
3884 | if (ctx->xin->may_learn) { | |
3885 | ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm); | |
3886 | } | |
3887 | } | |
9583bc14 | 3888 | |
4165b5e0 JS |
3889 | static void |
3890 | xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) | |
3891 | { | |
9583bc14 | 3892 | ctx->xout->has_learn = true; |
9583bc14 EJ |
3893 | learn_mask(learn, &ctx->xout->wc); |
3894 | ||
b256dc52 JS |
3895 | if (ctx->xin->xcache) { |
3896 | struct xc_entry *entry; | |
3897 | ||
3898 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN); | |
4165b5e0 JS |
3899 | entry->u.learn.ofproto = ctx->xbridge->ofproto; |
3900 | entry->u.learn.fm = xmalloc(sizeof *entry->u.learn.fm); | |
3901 | entry->u.learn.ofpacts = ofpbuf_new(64); | |
3902 | xlate_learn_action__(ctx, learn, entry->u.learn.fm, | |
3903 | entry->u.learn.ofpacts); | |
3904 | } else if (ctx->xin->may_learn) { | |
3905 | uint64_t ofpacts_stub[1024 / 8]; | |
3906 | struct ofputil_flow_mod fm; | |
3907 | struct ofpbuf ofpacts; | |
3908 | ||
3909 | ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); | |
3910 | xlate_learn_action__(ctx, learn, &fm, &ofpacts); | |
3911 | ofpbuf_uninit(&ofpacts); | |
b256dc52 JS |
3912 | } |
3913 | } | |
3914 | ||
3915 | static void | |
3916 | xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags, | |
3917 | uint16_t idle_timeout, uint16_t hard_timeout) | |
3918 | { | |
3919 | if (tcp_flags & (TCP_FIN | TCP_RST)) { | |
3920 | rule_dpif_reduce_timeouts(rule, idle_timeout, hard_timeout); | |
3921 | } | |
9583bc14 EJ |
3922 | } |
3923 | ||
9583bc14 EJ |
3924 | static void |
3925 | xlate_fin_timeout(struct xlate_ctx *ctx, | |
3926 | const struct ofpact_fin_timeout *oft) | |
3927 | { | |
b256dc52 JS |
3928 | if (ctx->rule) { |
3929 | xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags, | |
3930 | oft->fin_idle_timeout, oft->fin_hard_timeout); | |
3931 | if (ctx->xin->xcache) { | |
3932 | struct xc_entry *entry; | |
3933 | ||
3934 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT); | |
83709dfa JR |
3935 | /* XC_RULE already holds a reference on the rule, none is taken |
3936 | * here. */ | |
b256dc52 JS |
3937 | entry->u.fin.rule = ctx->rule; |
3938 | entry->u.fin.idle = oft->fin_idle_timeout; | |
3939 | entry->u.fin.hard = oft->fin_hard_timeout; | |
b256dc52 | 3940 | } |
9583bc14 EJ |
3941 | } |
3942 | } | |
3943 | ||
3944 | static void | |
3945 | xlate_sample_action(struct xlate_ctx *ctx, | |
3946 | const struct ofpact_sample *os) | |
3947 | { | |
e824d78d JR |
3948 | union user_action_cookie cookie; |
3949 | /* Scale the probability from 16-bit to 32-bit while representing | |
3950 | * the same percentage. */ | |
3951 | uint32_t probability = (os->probability << 16) | os->probability; | |
b440dd8c | 3952 | bool use_masked; |
e824d78d | 3953 | |
b440dd8c | 3954 | if (!ctx->xbridge->support.variable_length_userdata) { |
e824d78d JR |
3955 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); |
3956 | ||
3957 | VLOG_ERR_RL(&rl, "ignoring NXAST_SAMPLE action because datapath " | |
3958 | "lacks support (needs Linux 3.10+ or kernel module from " | |
3959 | "OVS 1.11+)"); | |
3960 | return; | |
3961 | } | |
3962 | ||
b440dd8c | 3963 | use_masked = ctx->xbridge->support.masked_set_action; |
e824d78d JR |
3964 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, |
3965 | ctx->xout->odp_actions, | |
b440dd8c | 3966 | &ctx->xout->wc, use_masked); |
e824d78d JR |
3967 | |
3968 | compose_flow_sample_cookie(os->probability, os->collector_set_id, | |
3969 | os->obs_domain_id, os->obs_point_id, &cookie); | |
3970 | compose_sample_action(ctx->xbridge, ctx->xout->odp_actions, | |
3971 | &ctx->xin->flow, probability, &cookie, | |
3972 | sizeof cookie.flow_sample, ODPP_NONE); | |
9583bc14 EJ |
3973 | } |
3974 | ||
3975 | static bool | |
46c88433 | 3976 | may_receive(const struct xport *xport, struct xlate_ctx *ctx) |
9583bc14 | 3977 | { |
bbbca389 | 3978 | if (xport->config & (is_stp(&ctx->xin->flow) |
46c88433 EJ |
3979 | ? OFPUTIL_PC_NO_RECV_STP |
3980 | : OFPUTIL_PC_NO_RECV)) { | |
9583bc14 EJ |
3981 | return false; |
3982 | } | |
3983 | ||
3984 | /* Only drop packets here if both forwarding and learning are | |
3985 | * disabled. If just learning is enabled, we need to have | |
3986 | * OFPP_NORMAL and the learning action have a look at the packet | |
3987 | * before we can drop it. */ | |
9efd308e DV |
3988 | if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) || |
3989 | (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) { | |
9583bc14 EJ |
3990 | return false; |
3991 | } | |
3992 | ||
3993 | return true; | |
3994 | } | |
3995 | ||
7fdb60a7 SH |
3996 | static void |
3997 | xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact *a) | |
3998 | { | |
c61f3870 BP |
3999 | const struct ofpact_nest *on = ofpact_get_WRITE_ACTIONS(a); |
4000 | size_t on_len = ofpact_nest_get_action_len(on); | |
4001 | const struct ofpact *inner; | |
4002 | ||
4003 | /* Maintain actset_output depending on the contents of the action set: | |
4004 | * | |
4005 | * - OFPP_UNSET, if there is no "output" action. | |
4006 | * | |
4007 | * - The output port, if there is an "output" action and no "group" | |
4008 | * action. | |
4009 | * | |
4010 | * - OFPP_UNSET, if there is a "group" action. | |
4011 | */ | |
4012 | if (!ctx->action_set_has_group) { | |
4013 | OFPACT_FOR_EACH (inner, on->actions, on_len) { | |
4014 | if (inner->type == OFPACT_OUTPUT) { | |
4015 | ctx->xin->flow.actset_output = ofpact_get_OUTPUT(inner)->port; | |
4016 | } else if (inner->type == OFPACT_GROUP) { | |
4017 | ctx->xin->flow.actset_output = OFPP_UNSET; | |
4018 | ctx->action_set_has_group = true; | |
4019 | } | |
4020 | } | |
4021 | } | |
4022 | ||
4023 | ofpbuf_put(&ctx->action_set, on->actions, on_len); | |
7fdb60a7 SH |
4024 | ofpact_pad(&ctx->action_set); |
4025 | } | |
4026 | ||
4027 | static void | |
4028 | xlate_action_set(struct xlate_ctx *ctx) | |
4029 | { | |
4030 | uint64_t action_list_stub[1024 / 64]; | |
4031 | struct ofpbuf action_list; | |
4032 | ||
029ca940 | 4033 | ctx->in_action_set = true; |
7fdb60a7 SH |
4034 | ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub); |
4035 | ofpacts_execute_action_set(&action_list, &ctx->action_set); | |
ed9c9e3e JR |
4036 | /* Clear the action set, as it is not needed any more. */ |
4037 | ofpbuf_clear(&ctx->action_set); | |
6fd6ed71 | 4038 | do_xlate_actions(action_list.data, action_list.size, ctx); |
029ca940 | 4039 | ctx->in_action_set = false; |
7fdb60a7 SH |
4040 | ofpbuf_uninit(&action_list); |
4041 | } | |
4042 | ||
e672ff9b JR |
4043 | static void |
4044 | recirc_put_unroll_xlate(struct xlate_ctx *ctx) | |
4045 | { | |
4046 | struct ofpact_unroll_xlate *unroll; | |
4047 | ||
4048 | unroll = ctx->last_unroll_offset < 0 | |
4049 | ? NULL | |
4050 | : ALIGNED_CAST(struct ofpact_unroll_xlate *, | |
4051 | (char *)ctx->action_set.data + ctx->last_unroll_offset); | |
4052 | ||
4053 | /* Restore the table_id and rule cookie for a potential PACKET | |
4054 | * IN if needed. */ | |
4055 | if (!unroll || | |
4056 | (ctx->table_id != unroll->rule_table_id | |
4057 | || ctx->rule_cookie != unroll->rule_cookie)) { | |
4058 | ||
4059 | ctx->last_unroll_offset = ctx->action_set.size; | |
4060 | unroll = ofpact_put_UNROLL_XLATE(&ctx->action_set); | |
4061 | unroll->rule_table_id = ctx->table_id; | |
4062 | unroll->rule_cookie = ctx->rule_cookie; | |
4063 | } | |
4064 | } | |
4065 | ||
4066 | ||
4067 | /* Copy remaining actions to the action_set to be executed after recirculation. | |
4068 | * UNROLL_XLATE action is inserted, if not already done so, before actions that | |
4069 | * may generate PACKET_INs from the current table and without matching another | |
4070 | * rule. */ | |
4071 | static void | |
4072 | recirc_unroll_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
4073 | struct xlate_ctx *ctx) | |
4074 | { | |
4075 | const struct ofpact *a; | |
4076 | ||
4077 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { | |
4078 | switch (a->type) { | |
4079 | /* May generate PACKET INs. */ | |
4080 | case OFPACT_OUTPUT_REG: | |
4081 | case OFPACT_GROUP: | |
4082 | case OFPACT_OUTPUT: | |
4083 | case OFPACT_CONTROLLER: | |
4084 | case OFPACT_DEC_MPLS_TTL: | |
4085 | case OFPACT_DEC_TTL: | |
4086 | recirc_put_unroll_xlate(ctx); | |
4087 | break; | |
4088 | ||
4089 | /* These may not generate PACKET INs. */ | |
4090 | case OFPACT_SET_TUNNEL: | |
4091 | case OFPACT_REG_MOVE: | |
4092 | case OFPACT_SET_FIELD: | |
4093 | case OFPACT_STACK_PUSH: | |
4094 | case OFPACT_STACK_POP: | |
4095 | case OFPACT_LEARN: | |
4096 | case OFPACT_WRITE_METADATA: | |
4097 | case OFPACT_RESUBMIT: /* May indirectly generate PACKET INs, */ | |
4098 | case OFPACT_GOTO_TABLE: /* but from a different table and rule. */ | |
4099 | case OFPACT_ENQUEUE: | |
4100 | case OFPACT_SET_VLAN_VID: | |
4101 | case OFPACT_SET_VLAN_PCP: | |
4102 | case OFPACT_STRIP_VLAN: | |
4103 | case OFPACT_PUSH_VLAN: | |
4104 | case OFPACT_SET_ETH_SRC: | |
4105 | case OFPACT_SET_ETH_DST: | |
4106 | case OFPACT_SET_IPV4_SRC: | |
4107 | case OFPACT_SET_IPV4_DST: | |
4108 | case OFPACT_SET_IP_DSCP: | |
4109 | case OFPACT_SET_IP_ECN: | |
4110 | case OFPACT_SET_IP_TTL: | |
4111 | case OFPACT_SET_L4_SRC_PORT: | |
4112 | case OFPACT_SET_L4_DST_PORT: | |
4113 | case OFPACT_SET_QUEUE: | |
4114 | case OFPACT_POP_QUEUE: | |
4115 | case OFPACT_PUSH_MPLS: | |
4116 | case OFPACT_POP_MPLS: | |
4117 | case OFPACT_SET_MPLS_LABEL: | |
4118 | case OFPACT_SET_MPLS_TC: | |
4119 | case OFPACT_SET_MPLS_TTL: | |
4120 | case OFPACT_MULTIPATH: | |
4121 | case OFPACT_BUNDLE: | |
4122 | case OFPACT_EXIT: | |
4123 | case OFPACT_UNROLL_XLATE: | |
4124 | case OFPACT_FIN_TIMEOUT: | |
4125 | case OFPACT_CLEAR_ACTIONS: | |
4126 | case OFPACT_WRITE_ACTIONS: | |
4127 | case OFPACT_METER: | |
4128 | case OFPACT_SAMPLE: | |
4129 | break; | |
4130 | ||
4131 | /* These need not be copied for restoration. */ | |
4132 | case OFPACT_NOTE: | |
4133 | case OFPACT_CONJUNCTION: | |
4134 | continue; | |
4135 | } | |
4136 | /* Copy the action over. */ | |
4137 | ofpbuf_put(&ctx->action_set, a, OFPACT_ALIGN(a->len)); | |
4138 | } | |
4139 | } | |
4140 | ||
1d741d6d JR |
4141 | #define CHECK_MPLS_RECIRCULATION() \ |
4142 | if (ctx->was_mpls) { \ | |
4143 | ctx_trigger_recirculation(ctx); \ | |
4144 | break; \ | |
4145 | } | |
4146 | #define CHECK_MPLS_RECIRCULATION_IF(COND) \ | |
4147 | if (COND) { \ | |
4148 | CHECK_MPLS_RECIRCULATION(); \ | |
4149 | } | |
4150 | ||
9583bc14 EJ |
4151 | static void |
4152 | do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
4153 | struct xlate_ctx *ctx) | |
4154 | { | |
33bf9176 BP |
4155 | struct flow_wildcards *wc = &ctx->xout->wc; |
4156 | struct flow *flow = &ctx->xin->flow; | |
9583bc14 EJ |
4157 | const struct ofpact *a; |
4158 | ||
a36de779 PS |
4159 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
4160 | tnl_arp_snoop(flow, wc, ctx->xbridge->name); | |
4161 | } | |
f47ea021 JR |
4162 | /* dl_type already in the mask, not set below. */ |
4163 | ||
9583bc14 EJ |
4164 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
4165 | struct ofpact_controller *controller; | |
4166 | const struct ofpact_metadata *metadata; | |
b2dd70be JR |
4167 | const struct ofpact_set_field *set_field; |
4168 | const struct mf_field *mf; | |
9583bc14 | 4169 | |
e672ff9b JR |
4170 | if (ctx->exit) { |
4171 | /* Check if need to store the remaining actions for later | |
4172 | * execution. */ | |
4173 | if (exit_recirculates(ctx)) { | |
4174 | recirc_unroll_actions(a, OFPACT_ALIGN(ofpacts_len - | |
4175 | ((uint8_t *)a - | |
4176 | (uint8_t *)ofpacts)), | |
4177 | ctx); | |
4178 | } | |
4179 | break; | |
7bbdd84f SH |
4180 | } |
4181 | ||
9583bc14 EJ |
4182 | switch (a->type) { |
4183 | case OFPACT_OUTPUT: | |
4184 | xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port, | |
4185 | ofpact_get_OUTPUT(a)->max_len, true); | |
4186 | break; | |
4187 | ||
7395c052 | 4188 | case OFPACT_GROUP: |
f4fb341b | 4189 | if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id)) { |
1d741d6d | 4190 | /* Group could not be found. */ |
f4fb341b SH |
4191 | return; |
4192 | } | |
7395c052 NZ |
4193 | break; |
4194 | ||
9583bc14 EJ |
4195 | case OFPACT_CONTROLLER: |
4196 | controller = ofpact_get_CONTROLLER(a); | |
4197 | execute_controller_action(ctx, controller->max_len, | |
4198 | controller->reason, | |
4199 | controller->controller_id); | |
4200 | break; | |
4201 | ||
4202 | case OFPACT_ENQUEUE: | |
16194afd DDP |
4203 | memset(&wc->masks.skb_priority, 0xff, |
4204 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4205 | xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a)); |
4206 | break; | |
4207 | ||
4208 | case OFPACT_SET_VLAN_VID: | |
f74e7df7 | 4209 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
ca287d20 JR |
4210 | if (flow->vlan_tci & htons(VLAN_CFI) || |
4211 | ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) { | |
4212 | flow->vlan_tci &= ~htons(VLAN_VID_MASK); | |
4213 | flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid) | |
4214 | | htons(VLAN_CFI)); | |
4215 | } | |
9583bc14 EJ |
4216 | break; |
4217 | ||
4218 | case OFPACT_SET_VLAN_PCP: | |
f74e7df7 | 4219 | wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI); |
ca287d20 JR |
4220 | if (flow->vlan_tci & htons(VLAN_CFI) || |
4221 | ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) { | |
4222 | flow->vlan_tci &= ~htons(VLAN_PCP_MASK); | |
4223 | flow->vlan_tci |= htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp | |
4224 | << VLAN_PCP_SHIFT) | VLAN_CFI); | |
4225 | } | |
9583bc14 EJ |
4226 | break; |
4227 | ||
4228 | case OFPACT_STRIP_VLAN: | |
f74e7df7 | 4229 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 4230 | flow->vlan_tci = htons(0); |
9583bc14 EJ |
4231 | break; |
4232 | ||
4233 | case OFPACT_PUSH_VLAN: | |
4234 | /* XXX 802.1AD(QinQ) */ | |
f74e7df7 | 4235 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 4236 | flow->vlan_tci = htons(VLAN_CFI); |
9583bc14 EJ |
4237 | break; |
4238 | ||
4239 | case OFPACT_SET_ETH_SRC: | |
f74e7df7 | 4240 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
33bf9176 | 4241 | memcpy(flow->dl_src, ofpact_get_SET_ETH_SRC(a)->mac, ETH_ADDR_LEN); |
9583bc14 EJ |
4242 | break; |
4243 | ||
4244 | case OFPACT_SET_ETH_DST: | |
f74e7df7 | 4245 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
33bf9176 | 4246 | memcpy(flow->dl_dst, ofpact_get_SET_ETH_DST(a)->mac, ETH_ADDR_LEN); |
9583bc14 EJ |
4247 | break; |
4248 | ||
4249 | case OFPACT_SET_IPV4_SRC: | |
1d741d6d | 4250 | CHECK_MPLS_RECIRCULATION(); |
33bf9176 | 4251 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 4252 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); |
33bf9176 | 4253 | flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4; |
9583bc14 EJ |
4254 | } |
4255 | break; | |
4256 | ||
4257 | case OFPACT_SET_IPV4_DST: | |
1d741d6d | 4258 | CHECK_MPLS_RECIRCULATION(); |
33bf9176 | 4259 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 4260 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); |
33bf9176 | 4261 | flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4; |
9583bc14 EJ |
4262 | } |
4263 | break; | |
4264 | ||
04f01c24 | 4265 | case OFPACT_SET_IP_DSCP: |
1d741d6d | 4266 | CHECK_MPLS_RECIRCULATION(); |
04f01c24 | 4267 | if (is_ip_any(flow)) { |
f47ea021 | 4268 | wc->masks.nw_tos |= IP_DSCP_MASK; |
33bf9176 | 4269 | flow->nw_tos &= ~IP_DSCP_MASK; |
04f01c24 | 4270 | flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp; |
9583bc14 EJ |
4271 | } |
4272 | break; | |
4273 | ||
ff14eb7a | 4274 | case OFPACT_SET_IP_ECN: |
1d741d6d | 4275 | CHECK_MPLS_RECIRCULATION(); |
ff14eb7a JR |
4276 | if (is_ip_any(flow)) { |
4277 | wc->masks.nw_tos |= IP_ECN_MASK; | |
4278 | flow->nw_tos &= ~IP_ECN_MASK; | |
4279 | flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn; | |
4280 | } | |
4281 | break; | |
4282 | ||
0c20dbe4 | 4283 | case OFPACT_SET_IP_TTL: |
1d741d6d | 4284 | CHECK_MPLS_RECIRCULATION(); |
0c20dbe4 JR |
4285 | if (is_ip_any(flow)) { |
4286 | wc->masks.nw_ttl = 0xff; | |
4287 | flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl; | |
4288 | } | |
4289 | break; | |
4290 | ||
9583bc14 | 4291 | case OFPACT_SET_L4_SRC_PORT: |
1d741d6d | 4292 | CHECK_MPLS_RECIRCULATION(); |
b8778a0d | 4293 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
4294 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
4295 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
33bf9176 | 4296 | flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port); |
9583bc14 EJ |
4297 | } |
4298 | break; | |
4299 | ||
4300 | case OFPACT_SET_L4_DST_PORT: | |
1d741d6d | 4301 | CHECK_MPLS_RECIRCULATION(); |
b8778a0d | 4302 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
4303 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
4304 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
33bf9176 | 4305 | flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port); |
9583bc14 EJ |
4306 | } |
4307 | break; | |
4308 | ||
4309 | case OFPACT_RESUBMIT: | |
4310 | xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a)); | |
4311 | break; | |
4312 | ||
4313 | case OFPACT_SET_TUNNEL: | |
33bf9176 | 4314 | flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id); |
9583bc14 EJ |
4315 | break; |
4316 | ||
4317 | case OFPACT_SET_QUEUE: | |
16194afd DDP |
4318 | memset(&wc->masks.skb_priority, 0xff, |
4319 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4320 | xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id); |
4321 | break; | |
4322 | ||
4323 | case OFPACT_POP_QUEUE: | |
16194afd DDP |
4324 | memset(&wc->masks.skb_priority, 0xff, |
4325 | sizeof wc->masks.skb_priority); | |
33bf9176 | 4326 | flow->skb_priority = ctx->orig_skb_priority; |
9583bc14 EJ |
4327 | break; |
4328 | ||
4329 | case OFPACT_REG_MOVE: | |
1d741d6d JR |
4330 | CHECK_MPLS_RECIRCULATION_IF( |
4331 | mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->dst.field) || | |
4332 | mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->src.field)); | |
33bf9176 | 4333 | nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc); |
9583bc14 EJ |
4334 | break; |
4335 | ||
b2dd70be | 4336 | case OFPACT_SET_FIELD: |
1d741d6d JR |
4337 | CHECK_MPLS_RECIRCULATION_IF( |
4338 | mf_is_l3_or_higher(ofpact_get_SET_FIELD(a)->field)); | |
b2dd70be JR |
4339 | set_field = ofpact_get_SET_FIELD(a); |
4340 | mf = set_field->field; | |
b2dd70be JR |
4341 | |
4342 | /* Set field action only ever overwrites packet's outermost | |
4343 | * applicable header fields. Do nothing if no header exists. */ | |
4479846e JR |
4344 | if (mf->id == MFF_VLAN_VID) { |
4345 | wc->masks.vlan_tci |= htons(VLAN_CFI); | |
4346 | if (!(flow->vlan_tci & htons(VLAN_CFI))) { | |
4347 | break; | |
4348 | } | |
4349 | } else if ((mf->id == MFF_MPLS_LABEL || mf->id == MFF_MPLS_TC) | |
4350 | /* 'dl_type' is already unwildcarded. */ | |
4351 | && !eth_type_mpls(flow->dl_type)) { | |
4352 | break; | |
b2dd70be | 4353 | } |
b8778a0d JR |
4354 | /* A flow may wildcard nw_frag. Do nothing if setting a trasport |
4355 | * header field on a packet that does not have them. */ | |
4479846e | 4356 | mf_mask_field_and_prereqs(mf, &wc->masks); |
b8778a0d JR |
4357 | if (mf_are_prereqs_ok(mf, flow)) { |
4358 | mf_set_flow_value_masked(mf, &set_field->value, | |
4359 | &set_field->mask, flow); | |
4360 | } | |
b2dd70be JR |
4361 | break; |
4362 | ||
9583bc14 | 4363 | case OFPACT_STACK_PUSH: |
1d741d6d JR |
4364 | CHECK_MPLS_RECIRCULATION_IF( |
4365 | mf_is_l3_or_higher(ofpact_get_STACK_PUSH(a)->subfield.field)); | |
33bf9176 BP |
4366 | nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc, |
4367 | &ctx->stack); | |
9583bc14 EJ |
4368 | break; |
4369 | ||
4370 | case OFPACT_STACK_POP: | |
1d741d6d JR |
4371 | CHECK_MPLS_RECIRCULATION_IF( |
4372 | mf_is_l3_or_higher(ofpact_get_STACK_POP(a)->subfield.field)); | |
f74e7df7 JP |
4373 | nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc, |
4374 | &ctx->stack); | |
9583bc14 EJ |
4375 | break; |
4376 | ||
4377 | case OFPACT_PUSH_MPLS: | |
1d741d6d JR |
4378 | /* Recirculate if it is an IP packet with a zero ttl. This may |
4379 | * indicate that the packet was previously MPLS and an MPLS pop | |
4380 | * action converted it to IP. In this case recirculating should | |
4381 | * reveal the IP TTL which is used as the basis for a new MPLS | |
4382 | * LSE. */ | |
4383 | CHECK_MPLS_RECIRCULATION_IF( | |
4384 | !flow_count_mpls_labels(flow, wc) | |
4385 | && flow->nw_ttl == 0 | |
4386 | && is_ip_any(flow)); | |
8bfd0fda | 4387 | compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)); |
9583bc14 EJ |
4388 | break; |
4389 | ||
4390 | case OFPACT_POP_MPLS: | |
1d741d6d | 4391 | CHECK_MPLS_RECIRCULATION(); |
8bfd0fda | 4392 | compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype); |
9583bc14 EJ |
4393 | break; |
4394 | ||
097d4939 | 4395 | case OFPACT_SET_MPLS_LABEL: |
1d741d6d | 4396 | CHECK_MPLS_RECIRCULATION(); |
8bfd0fda BP |
4397 | compose_set_mpls_label_action( |
4398 | ctx, ofpact_get_SET_MPLS_LABEL(a)->label); | |
1d741d6d | 4399 | break; |
097d4939 JR |
4400 | |
4401 | case OFPACT_SET_MPLS_TC: | |
1d741d6d | 4402 | CHECK_MPLS_RECIRCULATION(); |
8bfd0fda | 4403 | compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc); |
097d4939 JR |
4404 | break; |
4405 | ||
9583bc14 | 4406 | case OFPACT_SET_MPLS_TTL: |
1d741d6d | 4407 | CHECK_MPLS_RECIRCULATION(); |
8bfd0fda | 4408 | compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl); |
9583bc14 EJ |
4409 | break; |
4410 | ||
4411 | case OFPACT_DEC_MPLS_TTL: | |
1d741d6d | 4412 | CHECK_MPLS_RECIRCULATION(); |
9cfef3d0 | 4413 | if (compose_dec_mpls_ttl_action(ctx)) { |
ad3efdcb | 4414 | return; |
9583bc14 EJ |
4415 | } |
4416 | break; | |
4417 | ||
4418 | case OFPACT_DEC_TTL: | |
1d741d6d | 4419 | CHECK_MPLS_RECIRCULATION(); |
f74e7df7 | 4420 | wc->masks.nw_ttl = 0xff; |
9583bc14 | 4421 | if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) { |
ad3efdcb | 4422 | return; |
9583bc14 EJ |
4423 | } |
4424 | break; | |
4425 | ||
4426 | case OFPACT_NOTE: | |
4427 | /* Nothing to do. */ | |
4428 | break; | |
4429 | ||
4430 | case OFPACT_MULTIPATH: | |
1d741d6d | 4431 | CHECK_MPLS_RECIRCULATION(); |
33bf9176 | 4432 | multipath_execute(ofpact_get_MULTIPATH(a), flow, wc); |
9583bc14 EJ |
4433 | break; |
4434 | ||
4435 | case OFPACT_BUNDLE: | |
1d741d6d | 4436 | CHECK_MPLS_RECIRCULATION(); |
9583bc14 EJ |
4437 | xlate_bundle_action(ctx, ofpact_get_BUNDLE(a)); |
4438 | break; | |
4439 | ||
4440 | case OFPACT_OUTPUT_REG: | |
4441 | xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a)); | |
4442 | break; | |
4443 | ||
4444 | case OFPACT_LEARN: | |
1d741d6d | 4445 | CHECK_MPLS_RECIRCULATION(); |
9583bc14 EJ |
4446 | xlate_learn_action(ctx, ofpact_get_LEARN(a)); |
4447 | break; | |
4448 | ||
afc3987b BP |
4449 | case OFPACT_CONJUNCTION: { |
4450 | /* A flow with a "conjunction" action represents part of a special | |
4451 | * kind of "set membership match". Such a flow should not actually | |
4452 | * get executed, but it could via, say, a "packet-out", even though | |
4453 | * that wouldn't be useful. Log it to help debugging. */ | |
4454 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4455 | VLOG_INFO_RL(&rl, "executing no-op conjunction action"); | |
18080541 | 4456 | break; |
afc3987b | 4457 | } |
18080541 | 4458 | |
9583bc14 EJ |
4459 | case OFPACT_EXIT: |
4460 | ctx->exit = true; | |
4461 | break; | |
4462 | ||
e672ff9b JR |
4463 | case OFPACT_UNROLL_XLATE: { |
4464 | struct ofpact_unroll_xlate *unroll = ofpact_get_UNROLL_XLATE(a); | |
4465 | ||
4466 | /* Restore translation context data that was stored earlier. */ | |
4467 | ctx->table_id = unroll->rule_table_id; | |
4468 | ctx->rule_cookie = unroll->rule_cookie; | |
4469 | break; | |
4470 | } | |
9583bc14 | 4471 | case OFPACT_FIN_TIMEOUT: |
1d741d6d | 4472 | CHECK_MPLS_RECIRCULATION(); |
33bf9176 | 4473 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
9583bc14 EJ |
4474 | ctx->xout->has_fin_timeout = true; |
4475 | xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a)); | |
4476 | break; | |
4477 | ||
4478 | case OFPACT_CLEAR_ACTIONS: | |
7fdb60a7 | 4479 | ofpbuf_clear(&ctx->action_set); |
c61f3870 BP |
4480 | ctx->xin->flow.actset_output = OFPP_UNSET; |
4481 | ctx->action_set_has_group = false; | |
7fdb60a7 SH |
4482 | break; |
4483 | ||
4484 | case OFPACT_WRITE_ACTIONS: | |
4485 | xlate_write_actions(ctx, a); | |
9583bc14 EJ |
4486 | break; |
4487 | ||
4488 | case OFPACT_WRITE_METADATA: | |
4489 | metadata = ofpact_get_WRITE_METADATA(a); | |
33bf9176 BP |
4490 | flow->metadata &= ~metadata->mask; |
4491 | flow->metadata |= metadata->metadata & metadata->mask; | |
9583bc14 EJ |
4492 | break; |
4493 | ||
638a19b0 JR |
4494 | case OFPACT_METER: |
4495 | /* Not implemented yet. */ | |
4496 | break; | |
4497 | ||
9583bc14 | 4498 | case OFPACT_GOTO_TABLE: { |
9583bc14 | 4499 | struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a); |
9583bc14 | 4500 | |
7bbdd84f SH |
4501 | /* Allow ctx->table_id == TBL_INTERNAL, which will be greater |
4502 | * than ogt->table_id. This is to allow goto_table actions that | |
4503 | * triggered recirculation: ctx->table_id will be TBL_INTERNAL | |
4504 | * after recirculation. */ | |
4505 | ovs_assert(ctx->table_id == TBL_INTERNAL | |
4506 | || ctx->table_id < ogt->table_id); | |
4468099e | 4507 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 4508 | ogt->table_id, true, true); |
9583bc14 EJ |
4509 | break; |
4510 | } | |
4511 | ||
4512 | case OFPACT_SAMPLE: | |
4513 | xlate_sample_action(ctx, ofpact_get_SAMPLE(a)); | |
4514 | break; | |
4515 | } | |
1d741d6d JR |
4516 | |
4517 | /* Check if need to store this and the remaining actions for later | |
4518 | * execution. */ | |
4519 | if (ctx->exit && ctx_first_recirculation_action(ctx)) { | |
4520 | recirc_unroll_actions(a, OFPACT_ALIGN(ofpacts_len - | |
4521 | ((uint8_t *)a - | |
4522 | (uint8_t *)ofpacts)), | |
4523 | ctx); | |
4524 | break; | |
4525 | } | |
9583bc14 | 4526 | } |
9583bc14 EJ |
4527 | } |
4528 | ||
4529 | void | |
4530 | xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, | |
cc377352 EJ |
4531 | const struct flow *flow, ofp_port_t in_port, |
4532 | struct rule_dpif *rule, uint16_t tcp_flags, | |
cf62fa4c | 4533 | const struct dp_packet *packet) |
9583bc14 EJ |
4534 | { |
4535 | xin->ofproto = ofproto; | |
4536 | xin->flow = *flow; | |
cc377352 | 4537 | xin->flow.in_port.ofp_port = in_port; |
c61f3870 | 4538 | xin->flow.actset_output = OFPP_UNSET; |
9583bc14 EJ |
4539 | xin->packet = packet; |
4540 | xin->may_learn = packet != NULL; | |
4541 | xin->rule = rule; | |
b256dc52 | 4542 | xin->xcache = NULL; |
9583bc14 EJ |
4543 | xin->ofpacts = NULL; |
4544 | xin->ofpacts_len = 0; | |
4545 | xin->tcp_flags = tcp_flags; | |
4546 | xin->resubmit_hook = NULL; | |
4547 | xin->report_hook = NULL; | |
4548 | xin->resubmit_stats = NULL; | |
3f207910 | 4549 | xin->skip_wildcards = false; |
cc377352 | 4550 | xin->odp_actions = NULL; |
e672ff9b JR |
4551 | |
4552 | /* Do recirc lookup. */ | |
4553 | xin->recirc = flow->recirc_id | |
4554 | ? recirc_id_node_find(flow->recirc_id) | |
4555 | : NULL; | |
9583bc14 EJ |
4556 | } |
4557 | ||
4558 | void | |
4559 | xlate_out_uninit(struct xlate_out *xout) | |
4560 | { | |
e672ff9b JR |
4561 | if (xout) { |
4562 | if (xout->odp_actions == &xout->odp_actions_buf) { | |
4563 | ofpbuf_uninit(xout->odp_actions); | |
4564 | } | |
4565 | xlate_out_free_recircs(xout); | |
9583bc14 EJ |
4566 | } |
4567 | } | |
4568 | ||
4569 | /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts' | |
4570 | * into datapath actions, using 'ctx', and discards the datapath actions. */ | |
4571 | void | |
4572 | xlate_actions_for_side_effects(struct xlate_in *xin) | |
4573 | { | |
4574 | struct xlate_out xout; | |
4575 | ||
4576 | xlate_actions(xin, &xout); | |
4577 | xlate_out_uninit(&xout); | |
4578 | } | |
4579 | ||
9583bc14 EJ |
4580 | void |
4581 | xlate_out_copy(struct xlate_out *dst, const struct xlate_out *src) | |
4582 | { | |
4583 | dst->wc = src->wc; | |
9583bc14 EJ |
4584 | dst->slow = src->slow; |
4585 | dst->has_learn = src->has_learn; | |
4586 | dst->has_normal = src->has_normal; | |
4587 | dst->has_fin_timeout = src->has_fin_timeout; | |
4588 | dst->nf_output_iface = src->nf_output_iface; | |
4589 | dst->mirrors = src->mirrors; | |
4590 | ||
cc377352 EJ |
4591 | dst->odp_actions = &dst->odp_actions_buf; |
4592 | ofpbuf_use_stub(dst->odp_actions, dst->odp_actions_stub, | |
9583bc14 | 4593 | sizeof dst->odp_actions_stub); |
6fd6ed71 | 4594 | ofpbuf_put(dst->odp_actions, src->odp_actions->data, src->odp_actions->size); |
9583bc14 EJ |
4595 | } |
4596 | \f | |
55954f6e EJ |
4597 | static struct skb_priority_to_dscp * |
4598 | get_skb_priority(const struct xport *xport, uint32_t skb_priority) | |
4599 | { | |
4600 | struct skb_priority_to_dscp *pdscp; | |
4601 | uint32_t hash; | |
4602 | ||
4603 | hash = hash_int(skb_priority, 0); | |
4604 | HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) { | |
4605 | if (pdscp->skb_priority == skb_priority) { | |
4606 | return pdscp; | |
4607 | } | |
4608 | } | |
4609 | return NULL; | |
4610 | } | |
4611 | ||
4612 | static bool | |
4613 | dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority, | |
4614 | uint8_t *dscp) | |
4615 | { | |
4616 | struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority); | |
4617 | *dscp = pdscp ? pdscp->dscp : 0; | |
4618 | return pdscp != NULL; | |
4619 | } | |
4620 | ||
16194afd DDP |
4621 | static size_t |
4622 | count_skb_priorities(const struct xport *xport) | |
4623 | { | |
4624 | return hmap_count(&xport->skb_priorities); | |
4625 | } | |
4626 | ||
55954f6e EJ |
4627 | static void |
4628 | clear_skb_priorities(struct xport *xport) | |
4629 | { | |
4630 | struct skb_priority_to_dscp *pdscp, *next; | |
4631 | ||
4632 | HMAP_FOR_EACH_SAFE (pdscp, next, hmap_node, &xport->skb_priorities) { | |
4633 | hmap_remove(&xport->skb_priorities, &pdscp->hmap_node); | |
4634 | free(pdscp); | |
4635 | } | |
4636 | } | |
4637 | ||
ce4a6b76 BP |
4638 | static bool |
4639 | actions_output_to_local_port(const struct xlate_ctx *ctx) | |
4640 | { | |
46c88433 | 4641 | odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL); |
ce4a6b76 BP |
4642 | const struct nlattr *a; |
4643 | unsigned int left; | |
4644 | ||
6fd6ed71 PS |
4645 | NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->xout->odp_actions->data, |
4646 | ctx->xout->odp_actions->size) { | |
ce4a6b76 BP |
4647 | if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT |
4648 | && nl_attr_get_odp_port(a) == local_odp_port) { | |
4649 | return true; | |
4650 | } | |
4651 | } | |
4652 | return false; | |
4653 | } | |
9583bc14 | 4654 | |
5e2a6702 | 4655 | #if defined(__linux__) |
7d031d7e BP |
4656 | /* Returns the maximum number of packets that the Linux kernel is willing to |
4657 | * queue up internally to certain kinds of software-implemented ports, or the | |
4658 | * default (and rarely modified) value if it cannot be determined. */ | |
4659 | static int | |
4660 | netdev_max_backlog(void) | |
4661 | { | |
4662 | static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; | |
4663 | static int max_backlog = 1000; /* The normal default value. */ | |
4664 | ||
4665 | if (ovsthread_once_start(&once)) { | |
4666 | static const char filename[] = "/proc/sys/net/core/netdev_max_backlog"; | |
4667 | FILE *stream; | |
4668 | int n; | |
4669 | ||
4670 | stream = fopen(filename, "r"); | |
4671 | if (!stream) { | |
4672 | VLOG_WARN("%s: open failed (%s)", filename, ovs_strerror(errno)); | |
4673 | } else { | |
4674 | if (fscanf(stream, "%d", &n) != 1) { | |
4675 | VLOG_WARN("%s: read error", filename); | |
4676 | } else if (n <= 100) { | |
4677 | VLOG_WARN("%s: unexpectedly small value %d", filename, n); | |
4678 | } else { | |
4679 | max_backlog = n; | |
4680 | } | |
4681 | fclose(stream); | |
4682 | } | |
4683 | ovsthread_once_done(&once); | |
4684 | ||
4685 | VLOG_DBG("%s: using %d max_backlog", filename, max_backlog); | |
4686 | } | |
4687 | ||
4688 | return max_backlog; | |
4689 | } | |
4690 | ||
4691 | /* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in | |
4692 | * 'odp_actions'. */ | |
4693 | static int | |
4694 | count_output_actions(const struct ofpbuf *odp_actions) | |
4695 | { | |
4696 | const struct nlattr *a; | |
4697 | size_t left; | |
4698 | int n = 0; | |
4699 | ||
6fd6ed71 | 4700 | NL_ATTR_FOR_EACH_UNSAFE (a, left, odp_actions->data, odp_actions->size) { |
7d031d7e BP |
4701 | if (a->nla_type == OVS_ACTION_ATTR_OUTPUT) { |
4702 | n++; | |
4703 | } | |
4704 | } | |
4705 | return n; | |
4706 | } | |
5e2a6702 | 4707 | #endif /* defined(__linux__) */ |
7d031d7e BP |
4708 | |
4709 | /* Returns true if 'odp_actions' contains more output actions than the datapath | |
4710 | * can reliably handle in one go. On Linux, this is the value of the | |
4711 | * net.core.netdev_max_backlog sysctl, which limits the maximum number of | |
4712 | * packets that the kernel is willing to queue up for processing while the | |
4713 | * datapath is processing a set of actions. */ | |
4714 | static bool | |
5e2a6702 | 4715 | too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) |
7d031d7e BP |
4716 | { |
4717 | #ifdef __linux__ | |
6fd6ed71 | 4718 | return (odp_actions->size / NL_A_U32_SIZE > netdev_max_backlog() |
7d031d7e BP |
4719 | && count_output_actions(odp_actions) > netdev_max_backlog()); |
4720 | #else | |
4721 | /* OSes other than Linux might have similar limits, but we don't know how | |
4722 | * to determine them.*/ | |
4723 | return false; | |
4724 | #endif | |
4725 | } | |
4726 | ||
e672ff9b JR |
4727 | /* Translates the flow, actions, or rule in 'xin' into datapath actions in |
4728 | * 'xout'. | |
56450a41 BP |
4729 | * The caller must take responsibility for eventually freeing 'xout', with |
4730 | * xlate_out_uninit(). */ | |
84f0f298 RW |
4731 | void |
4732 | xlate_actions(struct xlate_in *xin, struct xlate_out *xout) | |
9583bc14 | 4733 | { |
84f0f298 | 4734 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
7b7dcc76 | 4735 | struct flow_wildcards *wc = NULL; |
33bf9176 | 4736 | struct flow *flow = &xin->flow; |
10c44245 | 4737 | struct rule_dpif *rule = NULL; |
33bf9176 | 4738 | |
9583bc14 EJ |
4739 | enum slow_path_reason special; |
4740 | const struct ofpact *ofpacts; | |
e672ff9b | 4741 | struct xbridge *xbridge; |
46c88433 | 4742 | struct xport *in_port; |
9583bc14 EJ |
4743 | struct flow orig_flow; |
4744 | struct xlate_ctx ctx; | |
4745 | size_t ofpacts_len; | |
62a7cc71 | 4746 | bool tnl_may_send; |
c56fac1b | 4747 | bool is_icmp; |
9583bc14 | 4748 | |
46c88433 | 4749 | COVERAGE_INC(xlate_actions); |
9583bc14 EJ |
4750 | |
4751 | /* Flow initialization rules: | |
4752 | * - 'base_flow' must match the kernel's view of the packet at the | |
4753 | * time that action processing starts. 'flow' represents any | |
4754 | * transformations we wish to make through actions. | |
4755 | * - By default 'base_flow' and 'flow' are the same since the input | |
4756 | * packet matches the output before any actions are applied. | |
4757 | * - When using VLAN splinters, 'base_flow''s VLAN is set to the value | |
4758 | * of the received packet as seen by the kernel. If we later output | |
4759 | * to another device without any modifications this will cause us to | |
4760 | * insert a new tag since the original one was stripped off by the | |
4761 | * VLAN device. | |
4762 | * - Tunnel metadata as received is retained in 'flow'. This allows | |
4763 | * tunnel metadata matching also in later tables. | |
4764 | * Since a kernel action for setting the tunnel metadata will only be | |
4765 | * generated with actual tunnel output, changing the tunnel metadata | |
4766 | * values in 'flow' (such as tun_id) will only have effect with a later | |
4767 | * tunnel output action. | |
4768 | * - Tunnel 'base_flow' is completely cleared since that is what the | |
4769 | * kernel does. If we wish to maintain the original values an action | |
4770 | * needs to be generated. */ | |
4771 | ||
4772 | ctx.xin = xin; | |
4773 | ctx.xout = xout; | |
46c88433 EJ |
4774 | ctx.xout->slow = 0; |
4775 | ctx.xout->has_learn = false; | |
4776 | ctx.xout->has_normal = false; | |
4777 | ctx.xout->has_fin_timeout = false; | |
4778 | ctx.xout->nf_output_iface = NF_OUT_DROP; | |
4779 | ctx.xout->mirrors = 0; | |
e672ff9b | 4780 | ctx.xout->n_recircs = 0; |
cc377352 EJ |
4781 | |
4782 | xout->odp_actions = xin->odp_actions; | |
4783 | if (!xout->odp_actions) { | |
4784 | xout->odp_actions = &xout->odp_actions_buf; | |
4785 | ofpbuf_use_stub(xout->odp_actions, xout->odp_actions_stub, | |
4786 | sizeof xout->odp_actions_stub); | |
4787 | } | |
4788 | ofpbuf_reserve(xout->odp_actions, NL_A_U32_SIZE); | |
46c88433 | 4789 | |
e672ff9b JR |
4790 | xbridge = xbridge_lookup(xcfg, xin->ofproto); |
4791 | if (!xbridge) { | |
83709dfa | 4792 | return; |
46c88433 | 4793 | } |
e672ff9b JR |
4794 | /* 'ctx.xbridge' may be changed by action processing, whereas 'xbridge' |
4795 | * will remain set on the original input bridge. */ | |
4796 | ctx.xbridge = xbridge; | |
9583bc14 EJ |
4797 | ctx.rule = xin->rule; |
4798 | ||
33bf9176 | 4799 | ctx.base_flow = *flow; |
9583bc14 | 4800 | memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); |
33bf9176 | 4801 | ctx.orig_tunnel_ip_dst = flow->tunnel.ip_dst; |
9583bc14 | 4802 | |
7b7dcc76 JR |
4803 | if (!xin->skip_wildcards) { |
4804 | wc = &xout->wc; | |
4805 | flow_wildcards_init_catchall(wc); | |
4806 | memset(&wc->masks.in_port, 0xff, sizeof wc->masks.in_port); | |
4807 | memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type); | |
4808 | if (is_ip_any(flow)) { | |
4809 | wc->masks.nw_frag |= FLOW_NW_FRAG_MASK; | |
4810 | } | |
2494ccd7 | 4811 | if (xbridge->support.odp.recirc) { |
7b7dcc76 JR |
4812 | /* Always exactly match recirc_id when datapath supports |
4813 | * recirculation. */ | |
4814 | wc->masks.recirc_id = UINT32_MAX; | |
4815 | } | |
e672ff9b | 4816 | if (xbridge->netflow) { |
7b7dcc76 JR |
4817 | netflow_mask_wc(flow, wc); |
4818 | } | |
0934ebba | 4819 | } |
c56fac1b | 4820 | is_icmp = is_icmpv4(flow) || is_icmpv6(flow); |
9583bc14 | 4821 | |
62a7cc71 | 4822 | tnl_may_send = tnl_xlate_init(&ctx.base_flow, flow, wc); |
9583bc14 | 4823 | |
9583bc14 | 4824 | ctx.recurse = 0; |
98b07853 | 4825 | ctx.resubmits = 0; |
5a070238 | 4826 | ctx.in_group = false; |
029ca940 | 4827 | ctx.in_action_set = false; |
33bf9176 | 4828 | ctx.orig_skb_priority = flow->skb_priority; |
9583bc14 | 4829 | ctx.table_id = 0; |
8b1e5560 | 4830 | ctx.rule_cookie = OVS_BE64_MAX; |
9583bc14 | 4831 | ctx.exit = false; |
7bbdd84f | 4832 | ctx.was_mpls = false; |
e672ff9b JR |
4833 | ctx.recirc_action_offset = -1; |
4834 | ctx.last_unroll_offset = -1; | |
4835 | ||
4836 | ctx.action_set_has_group = false; | |
4837 | ofpbuf_use_stub(&ctx.action_set, | |
4838 | ctx.action_set_stub, sizeof ctx.action_set_stub); | |
4839 | ||
4840 | ofpbuf_use_stub(&ctx.stack, ctx.init_stack, sizeof ctx.init_stack); | |
4841 | ||
4842 | /* The in_port of the original packet before recirculation. */ | |
4843 | in_port = get_ofp_port(xbridge, flow->in_port.ofp_port); | |
4844 | ||
4845 | if (xin->recirc) { | |
4846 | const struct recirc_id_node *recirc = xin->recirc; | |
4847 | ||
4848 | if (xin->ofpacts_len > 0 || ctx.rule) { | |
4849 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4850 | ||
4851 | VLOG_WARN_RL(&rl, "Recirculation conflict (%s)!", | |
4852 | xin->ofpacts_len > 0 | |
4853 | ? "actions" | |
4854 | : "rule"); | |
4855 | return; | |
4856 | } | |
4857 | ||
4858 | /* Set the bridge for post-recirculation processing if needed. */ | |
4859 | if (ctx.xbridge->ofproto != recirc->ofproto) { | |
4860 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
4861 | const struct xbridge *new_bridge = xbridge_lookup(xcfg, | |
4862 | recirc->ofproto); | |
4863 | ||
4864 | if (OVS_UNLIKELY(!new_bridge)) { | |
4865 | /* Drop the packet if the bridge cannot be found. */ | |
4866 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4867 | VLOG_WARN_RL(&rl, "Recirculation bridge no longer exists."); | |
4868 | return; | |
4869 | } | |
4870 | ctx.xbridge = new_bridge; | |
4871 | } | |
4872 | ||
4873 | /* Set the post-recirculation table id. Note: A table lookup is done | |
4874 | * only if there are no post-recirculation actions. */ | |
4875 | ctx.table_id = recirc->table_id; | |
4876 | ||
4877 | /* Restore pipeline metadata. May change flow's in_port and other | |
4878 | * metadata to the values that existed when recirculation was | |
4879 | * triggered. */ | |
4880 | recirc_metadata_to_flow(&recirc->metadata, flow); | |
4881 | ||
4882 | /* Restore stack, if any. */ | |
4883 | if (recirc->stack) { | |
4884 | ofpbuf_put(&ctx.stack, recirc->stack->data, recirc->stack->size); | |
4885 | } | |
4886 | ||
4887 | /* Restore action set, if any. */ | |
4888 | if (recirc->action_set_len) { | |
4889 | const struct ofpact *a; | |
4890 | ||
4891 | ofpbuf_put(&ctx.action_set, recirc->ofpacts, | |
4892 | recirc->action_set_len); | |
4893 | ||
4894 | OFPACT_FOR_EACH(a, recirc->ofpacts, recirc->action_set_len) { | |
4895 | if (a->type == OFPACT_GROUP) { | |
4896 | ctx.action_set_has_group = true; | |
4897 | break; | |
4898 | } | |
4899 | } | |
4900 | } | |
4901 | ||
4902 | /* Restore recirculation actions. If there are no actions, processing | |
4903 | * will start with a lookup in the table set above. */ | |
4904 | if (recirc->ofpacts_len > recirc->action_set_len) { | |
4905 | xin->ofpacts_len = recirc->ofpacts_len - recirc->action_set_len; | |
4906 | xin->ofpacts = recirc->ofpacts + | |
4907 | recirc->action_set_len / sizeof *recirc->ofpacts; | |
4908 | } | |
4909 | } else if (OVS_UNLIKELY(flow->recirc_id)) { | |
4910 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4911 | ||
4912 | VLOG_WARN_RL(&rl, "Recirculation context not found for ID %"PRIx32, | |
4913 | flow->recirc_id); | |
4914 | return; | |
4915 | } | |
621b8064 JR |
4916 | /* The bridge is now known so obtain its table version. */ |
4917 | ctx.tables_version = ofproto_dpif_get_tables_version(ctx.xbridge->ofproto); | |
9583bc14 | 4918 | |
10c44245 | 4919 | if (!xin->ofpacts && !ctx.rule) { |
621b8064 JR |
4920 | rule = rule_dpif_lookup_from_table(ctx.xbridge->ofproto, |
4921 | ctx.tables_version, flow, wc, | |
e672ff9b JR |
4922 | ctx.xin->xcache != NULL, |
4923 | ctx.xin->resubmit_stats, | |
4924 | &ctx.table_id, | |
4925 | flow->in_port.ofp_port, true, true); | |
10c44245 EJ |
4926 | if (ctx.xin->resubmit_stats) { |
4927 | rule_dpif_credit_stats(rule, ctx.xin->resubmit_stats); | |
4928 | } | |
b256dc52 JS |
4929 | if (ctx.xin->xcache) { |
4930 | struct xc_entry *entry; | |
4931 | ||
4932 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE); | |
b256dc52 JS |
4933 | entry->u.rule = rule; |
4934 | } | |
10c44245 | 4935 | ctx.rule = rule; |
a8c31348 BP |
4936 | |
4937 | if (OVS_UNLIKELY(ctx.xin->resubmit_hook)) { | |
4938 | ctx.xin->resubmit_hook(ctx.xin, rule, 0); | |
4939 | } | |
10c44245 | 4940 | } |
ee1afdd5 | 4941 | xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule); |
10c44245 | 4942 | |
9583bc14 EJ |
4943 | if (xin->ofpacts) { |
4944 | ofpacts = xin->ofpacts; | |
4945 | ofpacts_len = xin->ofpacts_len; | |
10c44245 | 4946 | } else if (ctx.rule) { |
d8227eba JR |
4947 | const struct rule_actions *actions = rule_dpif_get_actions(ctx.rule); |
4948 | ||
6f00e29b BP |
4949 | ofpacts = actions->ofpacts; |
4950 | ofpacts_len = actions->ofpacts_len; | |
8b1e5560 JR |
4951 | |
4952 | ctx.rule_cookie = rule_dpif_get_flow_cookie(ctx.rule); | |
9583bc14 | 4953 | } else { |
428b2edd | 4954 | OVS_NOT_REACHED(); |
9583bc14 EJ |
4955 | } |
4956 | ||
e672ff9b | 4957 | if (mbridge_has_mirrors(xbridge->mbridge)) { |
9583bc14 EJ |
4958 | /* Do this conditionally because the copy is expensive enough that it |
4959 | * shows up in profiles. */ | |
33bf9176 | 4960 | orig_flow = *flow; |
9583bc14 EJ |
4961 | } |
4962 | ||
e672ff9b JR |
4963 | /* Tunnel stats only for non-recirculated packets. */ |
4964 | if (!xin->recirc && in_port && in_port->is_tunnel) { | |
b256dc52 JS |
4965 | if (ctx.xin->resubmit_stats) { |
4966 | netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats); | |
4967 | if (in_port->bfd) { | |
4968 | bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats); | |
4969 | } | |
4970 | } | |
4971 | if (ctx.xin->xcache) { | |
4972 | struct xc_entry *entry; | |
4973 | ||
4974 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV); | |
4975 | entry->u.dev.rx = netdev_ref(in_port->netdev); | |
4976 | entry->u.dev.bfd = bfd_ref(in_port->bfd); | |
d6fc5f57 EJ |
4977 | } |
4978 | } | |
4979 | ||
e672ff9b JR |
4980 | /* Do not perform special processing on recirculated packets, |
4981 | * as recirculated packets are not really received by the bridge. */ | |
4982 | if (!xin->recirc && | |
4983 | (special = process_special(&ctx, flow, in_port, ctx.xin->packet))) { | |
04594cd5 | 4984 | ctx.xout->slow |= special; |
9583bc14 | 4985 | } else { |
9583bc14 | 4986 | size_t sample_actions_len; |
9583bc14 | 4987 | |
4e022ec0 | 4988 | if (flow->in_port.ofp_port |
e672ff9b | 4989 | != vsp_realdev_to_vlandev(xbridge->ofproto, |
46c88433 | 4990 | flow->in_port.ofp_port, |
33bf9176 | 4991 | flow->vlan_tci)) { |
9583bc14 EJ |
4992 | ctx.base_flow.vlan_tci = 0; |
4993 | } | |
4994 | ||
e672ff9b JR |
4995 | /* Sampling is done only for packets really received by the bridge. */ |
4996 | if (!xin->recirc) { | |
4997 | add_sflow_action(&ctx); | |
4998 | add_ipfix_action(&ctx); | |
4999 | sample_actions_len = ctx.xout->odp_actions->size; | |
5000 | } else { | |
5001 | sample_actions_len = 0; | |
5002 | } | |
9583bc14 | 5003 | |
62a7cc71 | 5004 | if (tnl_may_send && (!in_port || may_receive(in_port, &ctx))) { |
9583bc14 EJ |
5005 | do_xlate_actions(ofpacts, ofpacts_len, &ctx); |
5006 | ||
5007 | /* We've let OFPP_NORMAL and the learning action look at the | |
5008 | * packet, so drop it now if forwarding is disabled. */ | |
9efd308e DV |
5009 | if (in_port && (!xport_stp_forward_state(in_port) || |
5010 | !xport_rstp_forward_state(in_port))) { | |
8ec90813 | 5011 | /* Drop all actions added by do_xlate_actions() above. */ |
6fd6ed71 | 5012 | ctx.xout->odp_actions->size = sample_actions_len; |
e672ff9b JR |
5013 | |
5014 | /* Undo changes that may have been done for recirculation. */ | |
5015 | if (exit_recirculates(&ctx)) { | |
5016 | ctx.action_set.size = ctx.recirc_action_offset; | |
5017 | ctx.recirc_action_offset = -1; | |
5018 | ctx.last_unroll_offset = -1; | |
5019 | } | |
8ec90813 | 5020 | } else if (ctx.action_set.size) { |
e672ff9b JR |
5021 | /* Translate action set only if not dropping the packet and |
5022 | * not recirculating. */ | |
5023 | if (!exit_recirculates(&ctx)) { | |
5024 | xlate_action_set(&ctx); | |
5025 | } | |
5026 | } | |
5027 | /* Check if need to recirculate. */ | |
5028 | if (exit_recirculates(&ctx)) { | |
5029 | compose_recirculate_action(&ctx); | |
9583bc14 EJ |
5030 | } |
5031 | } | |
5032 | ||
e672ff9b JR |
5033 | /* Output only fully processed packets. */ |
5034 | if (!exit_recirculates(&ctx) | |
5035 | && xbridge->has_in_band | |
ce4a6b76 BP |
5036 | && in_band_must_output_to_local_port(flow) |
5037 | && !actions_output_to_local_port(&ctx)) { | |
e93ef1c7 | 5038 | compose_output_action(&ctx, OFPP_LOCAL, NULL); |
9583bc14 | 5039 | } |
aaa0fbae | 5040 | |
e672ff9b JR |
5041 | if (!xin->recirc) { |
5042 | fix_sflow_action(&ctx); | |
5043 | } | |
5044 | /* Only mirror fully processed packets. */ | |
5045 | if (!exit_recirculates(&ctx) | |
5046 | && mbridge_has_mirrors(xbridge->mbridge)) { | |
9583bc14 EJ |
5047 | add_mirror_actions(&ctx, &orig_flow); |
5048 | } | |
9583bc14 EJ |
5049 | } |
5050 | ||
6fd6ed71 | 5051 | if (nl_attr_oversized(ctx.xout->odp_actions->size)) { |
542024c4 | 5052 | /* These datapath actions are too big for a Netlink attribute, so we |
0f032e95 BP |
5053 | * can't hand them to the kernel directly. dpif_execute() can execute |
5054 | * them one by one with help, so just mark the result as SLOW_ACTION to | |
5055 | * prevent the flow from being installed. */ | |
5056 | COVERAGE_INC(xlate_actions_oversize); | |
5057 | ctx.xout->slow |= SLOW_ACTION; | |
7d031d7e BP |
5058 | } else if (too_many_output_actions(ctx.xout->odp_actions)) { |
5059 | COVERAGE_INC(xlate_actions_too_many_output); | |
5060 | ctx.xout->slow |= SLOW_ACTION; | |
542024c4 BP |
5061 | } |
5062 | ||
e672ff9b JR |
5063 | /* Update mirror stats only for packets really received by the bridge. */ |
5064 | if (!xin->recirc && mbridge_has_mirrors(xbridge->mbridge)) { | |
b256dc52 | 5065 | if (ctx.xin->resubmit_stats) { |
e672ff9b | 5066 | mirror_update_stats(xbridge->mbridge, xout->mirrors, |
b256dc52 JS |
5067 | ctx.xin->resubmit_stats->n_packets, |
5068 | ctx.xin->resubmit_stats->n_bytes); | |
5069 | } | |
5070 | if (ctx.xin->xcache) { | |
5071 | struct xc_entry *entry; | |
5072 | ||
5073 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_MIRROR); | |
e672ff9b | 5074 | entry->u.mirror.mbridge = mbridge_ref(xbridge->mbridge); |
b256dc52 JS |
5075 | entry->u.mirror.mirrors = xout->mirrors; |
5076 | } | |
5077 | } | |
5078 | ||
e672ff9b JR |
5079 | /* Do netflow only for packets really received by the bridge. */ |
5080 | if (!xin->recirc && xbridge->netflow) { | |
b256dc52 JS |
5081 | /* Only update netflow if we don't have controller flow. We don't |
5082 | * report NetFlow expiration messages for such facets because they | |
5083 | * are just part of the control logic for the network, not real | |
5084 | * traffic. */ | |
5085 | if (ofpacts_len == 0 | |
5086 | || ofpacts->type != OFPACT_CONTROLLER | |
5087 | || ofpact_next(ofpacts) < ofpact_end(ofpacts, ofpacts_len)) { | |
5088 | if (ctx.xin->resubmit_stats) { | |
e672ff9b | 5089 | netflow_flow_update(xbridge->netflow, flow, |
d6fc5f57 EJ |
5090 | xout->nf_output_iface, |
5091 | ctx.xin->resubmit_stats); | |
5092 | } | |
b256dc52 JS |
5093 | if (ctx.xin->xcache) { |
5094 | struct xc_entry *entry; | |
5095 | ||
5096 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW); | |
e672ff9b | 5097 | entry->u.nf.netflow = netflow_ref(xbridge->netflow); |
b256dc52 JS |
5098 | entry->u.nf.flow = xmemdup(flow, sizeof *flow); |
5099 | entry->u.nf.iface = xout->nf_output_iface; | |
5100 | } | |
d6fc5f57 EJ |
5101 | } |
5102 | } | |
5103 | ||
9583bc14 | 5104 | ofpbuf_uninit(&ctx.stack); |
7fdb60a7 | 5105 | ofpbuf_uninit(&ctx.action_set); |
9583bc14 | 5106 | |
7b7dcc76 JR |
5107 | if (wc) { |
5108 | /* Clear the metadata and register wildcard masks, because we won't | |
5109 | * use non-header fields as part of the cache. */ | |
5110 | flow_wildcards_clear_non_packet_fields(wc); | |
5111 | ||
5112 | /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow | |
5113 | * uses the low 8 bits of the 16-bit tp_src and tp_dst members to | |
5114 | * represent these fields. The datapath interface, on the other hand, | |
5115 | * represents them with just 8 bits each. This means that if the high | |
5116 | * 8 bits of the masks for these fields somehow become set, then they | |
5117 | * will get chopped off by a round trip through the datapath, and | |
5118 | * revalidation will spot that as an inconsistency and delete the flow. | |
5119 | * Avoid the problem here by making sure that only the low 8 bits of | |
5120 | * either field can be unwildcarded for ICMP. | |
5121 | */ | |
5122 | if (is_icmp) { | |
5123 | wc->masks.tp_src &= htons(UINT8_MAX); | |
5124 | wc->masks.tp_dst &= htons(UINT8_MAX); | |
5125 | } | |
f171fa88 AW |
5126 | /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */ |
5127 | if (wc->masks.vlan_tci) { | |
5128 | wc->masks.vlan_tci |= htons(VLAN_CFI); | |
5129 | } | |
c56fac1b | 5130 | } |
91d6cd12 AW |
5131 | } |
5132 | ||
5133 | /* Sends 'packet' out 'ofport'. | |
5134 | * May modify 'packet'. | |
5135 | * Returns 0 if successful, otherwise a positive errno value. */ | |
5136 | int | |
cf62fa4c | 5137 | xlate_send_packet(const struct ofport_dpif *ofport, struct dp_packet *packet) |
91d6cd12 | 5138 | { |
84f0f298 | 5139 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
91d6cd12 | 5140 | struct xport *xport; |
91d6cd12 | 5141 | struct ofpact_output output; |
91d6cd12 | 5142 | struct flow flow; |
91d6cd12 | 5143 | |
91d6cd12 AW |
5144 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); |
5145 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ | |
cf62fa4c | 5146 | flow_extract(packet, &flow); |
b5e7e61a | 5147 | flow.in_port.ofp_port = OFPP_NONE; |
91d6cd12 | 5148 | |
84f0f298 | 5149 | xport = xport_lookup(xcfg, ofport); |
91d6cd12 | 5150 | if (!xport) { |
02ea2703 | 5151 | return EINVAL; |
91d6cd12 | 5152 | } |
91d6cd12 AW |
5153 | output.port = xport->ofp_port; |
5154 | output.max_len = 0; | |
e491a67a YT |
5155 | |
5156 | return ofproto_dpif_execute_actions(xport->xbridge->ofproto, &flow, NULL, | |
5157 | &output.ofpact, sizeof output, | |
5158 | packet); | |
9583bc14 | 5159 | } |
b256dc52 JS |
5160 | |
5161 | struct xlate_cache * | |
5162 | xlate_cache_new(void) | |
5163 | { | |
5164 | struct xlate_cache *xcache = xmalloc(sizeof *xcache); | |
5165 | ||
5166 | ofpbuf_init(&xcache->entries, 512); | |
5167 | return xcache; | |
5168 | } | |
5169 | ||
5170 | static struct xc_entry * | |
5171 | xlate_cache_add_entry(struct xlate_cache *xcache, enum xc_type type) | |
5172 | { | |
5173 | struct xc_entry *entry; | |
5174 | ||
5175 | entry = ofpbuf_put_zeros(&xcache->entries, sizeof *entry); | |
5176 | entry->type = type; | |
5177 | ||
5178 | return entry; | |
5179 | } | |
5180 | ||
5181 | static void | |
5182 | xlate_cache_netdev(struct xc_entry *entry, const struct dpif_flow_stats *stats) | |
5183 | { | |
5184 | if (entry->u.dev.tx) { | |
5185 | netdev_vport_inc_tx(entry->u.dev.tx, stats); | |
5186 | } | |
5187 | if (entry->u.dev.rx) { | |
5188 | netdev_vport_inc_rx(entry->u.dev.rx, stats); | |
5189 | } | |
5190 | if (entry->u.dev.bfd) { | |
5191 | bfd_account_rx(entry->u.dev.bfd, stats); | |
5192 | } | |
5193 | } | |
5194 | ||
5195 | static void | |
5196 | xlate_cache_normal(struct ofproto_dpif *ofproto, struct flow *flow, int vlan) | |
5197 | { | |
84f0f298 | 5198 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
b256dc52 JS |
5199 | struct xbridge *xbridge; |
5200 | struct xbundle *xbundle; | |
5201 | struct flow_wildcards wc; | |
5202 | ||
84f0f298 | 5203 | xbridge = xbridge_lookup(xcfg, ofproto); |
b256dc52 JS |
5204 | if (!xbridge) { |
5205 | return; | |
5206 | } | |
5207 | ||
5208 | xbundle = lookup_input_bundle(xbridge, flow->in_port.ofp_port, false, | |
5209 | NULL); | |
5210 | if (!xbundle) { | |
5211 | return; | |
5212 | } | |
5213 | ||
5214 | update_learning_table(xbridge, flow, &wc, vlan, xbundle); | |
5215 | } | |
5216 | ||
5217 | /* Push stats and perform side effects of flow translation. */ | |
5218 | void | |
0725e747 | 5219 | xlate_push_stats(struct xlate_cache *xcache, |
b256dc52 JS |
5220 | const struct dpif_flow_stats *stats) |
5221 | { | |
5222 | struct xc_entry *entry; | |
5223 | struct ofpbuf entries = xcache->entries; | |
a36de779 | 5224 | uint8_t dmac[ETH_ADDR_LEN]; |
b256dc52 | 5225 | |
bc388583 BP |
5226 | if (!stats->n_packets) { |
5227 | return; | |
5228 | } | |
5229 | ||
b256dc52 JS |
5230 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { |
5231 | switch (entry->type) { | |
5232 | case XC_RULE: | |
5233 | rule_dpif_credit_stats(entry->u.rule, stats); | |
5234 | break; | |
5235 | case XC_BOND: | |
5236 | bond_account(entry->u.bond.bond, entry->u.bond.flow, | |
5237 | entry->u.bond.vid, stats->n_bytes); | |
5238 | break; | |
5239 | case XC_NETDEV: | |
5240 | xlate_cache_netdev(entry, stats); | |
5241 | break; | |
5242 | case XC_NETFLOW: | |
5243 | netflow_flow_update(entry->u.nf.netflow, entry->u.nf.flow, | |
5244 | entry->u.nf.iface, stats); | |
5245 | break; | |
5246 | case XC_MIRROR: | |
5247 | mirror_update_stats(entry->u.mirror.mbridge, | |
5248 | entry->u.mirror.mirrors, | |
5249 | stats->n_packets, stats->n_bytes); | |
5250 | break; | |
5251 | case XC_LEARN: | |
0725e747 | 5252 | ofproto_dpif_flow_mod(entry->u.learn.ofproto, entry->u.learn.fm); |
b256dc52 JS |
5253 | break; |
5254 | case XC_NORMAL: | |
0725e747 BP |
5255 | xlate_cache_normal(entry->u.normal.ofproto, entry->u.normal.flow, |
5256 | entry->u.normal.vlan); | |
b256dc52 JS |
5257 | break; |
5258 | case XC_FIN_TIMEOUT: | |
5259 | xlate_fin_timeout__(entry->u.fin.rule, stats->tcp_flags, | |
5260 | entry->u.fin.idle, entry->u.fin.hard); | |
5261 | break; | |
1e684d7d RW |
5262 | case XC_GROUP: |
5263 | group_dpif_credit_stats(entry->u.group.group, entry->u.group.bucket, | |
5264 | stats); | |
5265 | break; | |
a36de779 PS |
5266 | case XC_TNL_ARP: |
5267 | /* Lookup arp to avoid arp timeout. */ | |
5268 | tnl_arp_lookup(entry->u.tnl_arp_cache.br_name, entry->u.tnl_arp_cache.d_ip, dmac); | |
5269 | break; | |
b256dc52 JS |
5270 | default: |
5271 | OVS_NOT_REACHED(); | |
5272 | } | |
5273 | } | |
5274 | } | |
5275 | ||
5276 | static void | |
5277 | xlate_dev_unref(struct xc_entry *entry) | |
5278 | { | |
5279 | if (entry->u.dev.tx) { | |
5280 | netdev_close(entry->u.dev.tx); | |
5281 | } | |
5282 | if (entry->u.dev.rx) { | |
5283 | netdev_close(entry->u.dev.rx); | |
5284 | } | |
5285 | if (entry->u.dev.bfd) { | |
5286 | bfd_unref(entry->u.dev.bfd); | |
5287 | } | |
5288 | } | |
5289 | ||
5290 | static void | |
5291 | xlate_cache_clear_netflow(struct netflow *netflow, struct flow *flow) | |
5292 | { | |
b256dc52 JS |
5293 | netflow_flow_clear(netflow, flow); |
5294 | netflow_unref(netflow); | |
5295 | free(flow); | |
5296 | } | |
5297 | ||
5298 | void | |
5299 | xlate_cache_clear(struct xlate_cache *xcache) | |
5300 | { | |
5301 | struct xc_entry *entry; | |
5302 | struct ofpbuf entries; | |
5303 | ||
5304 | if (!xcache) { | |
5305 | return; | |
5306 | } | |
5307 | ||
5308 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { | |
5309 | switch (entry->type) { | |
5310 | case XC_RULE: | |
5311 | rule_dpif_unref(entry->u.rule); | |
5312 | break; | |
5313 | case XC_BOND: | |
5314 | free(entry->u.bond.flow); | |
5315 | bond_unref(entry->u.bond.bond); | |
5316 | break; | |
5317 | case XC_NETDEV: | |
5318 | xlate_dev_unref(entry); | |
5319 | break; | |
5320 | case XC_NETFLOW: | |
5321 | xlate_cache_clear_netflow(entry->u.nf.netflow, entry->u.nf.flow); | |
5322 | break; | |
5323 | case XC_MIRROR: | |
5324 | mbridge_unref(entry->u.mirror.mbridge); | |
5325 | break; | |
5326 | case XC_LEARN: | |
4165b5e0 JS |
5327 | free(entry->u.learn.fm); |
5328 | ofpbuf_delete(entry->u.learn.ofpacts); | |
b256dc52 JS |
5329 | break; |
5330 | case XC_NORMAL: | |
5331 | free(entry->u.normal.flow); | |
5332 | break; | |
5333 | case XC_FIN_TIMEOUT: | |
83709dfa JR |
5334 | /* 'u.fin.rule' is always already held as a XC_RULE, which |
5335 | * has already released it's reference above. */ | |
b256dc52 | 5336 | break; |
1e684d7d RW |
5337 | case XC_GROUP: |
5338 | group_dpif_unref(entry->u.group.group); | |
5339 | break; | |
a36de779 PS |
5340 | case XC_TNL_ARP: |
5341 | break; | |
b256dc52 JS |
5342 | default: |
5343 | OVS_NOT_REACHED(); | |
5344 | } | |
5345 | } | |
5346 | ||
5347 | ofpbuf_clear(&xcache->entries); | |
5348 | } | |
5349 | ||
5350 | void | |
5351 | xlate_cache_delete(struct xlate_cache *xcache) | |
5352 | { | |
5353 | xlate_cache_clear(xcache); | |
5354 | ofpbuf_uninit(&xcache->entries); | |
5355 | free(xcache); | |
5356 | } |