]>
Commit | Line | Data |
---|---|---|
b1986b08 | 1 | /* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2019 Nicira, Inc. |
9583bc14 EJ |
2 | * |
3 | * Licensed under the Apache License, Version 2.0 (the "License"); | |
4 | * you may not use this file except in compliance with the License. | |
5 | * You may obtain a copy of the License at: | |
6 | * | |
7 | * http://www.apache.org/licenses/LICENSE-2.0 | |
8 | * | |
9 | * Unless required by applicable law or agreed to in writing, software | |
10 | * distributed under the License is distributed on an "AS IS" BASIS, | |
11 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |
12 | * See the License for the specific language governing permissions and | |
13 | * limitations under the License. */ | |
14 | ||
15 | #include <config.h> | |
16 | ||
17 | #include "ofproto/ofproto-dpif-xlate.h" | |
18 | ||
8449c4d6 | 19 | #include <errno.h> |
b2befd5b BP |
20 | #include <sys/types.h> |
21 | #include <netinet/in.h> | |
a36de779 PS |
22 | #include <arpa/inet.h> |
23 | #include <net/if.h> | |
24 | #include <sys/socket.h> | |
8449c4d6 | 25 | |
db7d4e46 | 26 | #include "bfd.h" |
9583bc14 EJ |
27 | #include "bitmap.h" |
28 | #include "bond.h" | |
29 | #include "bundle.h" | |
30 | #include "byte-order.h" | |
db7d4e46 | 31 | #include "cfm.h" |
9583bc14 EJ |
32 | #include "connmgr.h" |
33 | #include "coverage.h" | |
46445c63 | 34 | #include "csum.h" |
e14deea0 | 35 | #include "dp-packet.h" |
9583bc14 | 36 | #include "dpif.h" |
f7f1ea29 | 37 | #include "in-band.h" |
db7d4e46 | 38 | #include "lacp.h" |
9583bc14 EJ |
39 | #include "learn.h" |
40 | #include "mac-learning.h" | |
6d95c4e8 | 41 | #include "mcast-snooping.h" |
9583bc14 EJ |
42 | #include "multipath.h" |
43 | #include "netdev-vport.h" | |
44 | #include "netlink.h" | |
45 | #include "nx-match.h" | |
46 | #include "odp-execute.h" | |
9583bc14 | 47 | #include "ofproto/ofproto-dpif-ipfix.h" |
ec7ceaed | 48 | #include "ofproto/ofproto-dpif-mirror.h" |
60d02c72 | 49 | #include "ofproto/ofproto-dpif-monitor.h" |
9583bc14 | 50 | #include "ofproto/ofproto-dpif-sflow.h" |
2d9b49dd | 51 | #include "ofproto/ofproto-dpif-trace.h" |
901a517e | 52 | #include "ofproto/ofproto-dpif-xlate-cache.h" |
9583bc14 | 53 | #include "ofproto/ofproto-dpif.h" |
6f00e29b | 54 | #include "ofproto/ofproto-provider.h" |
b598f214 BW |
55 | #include "openvswitch/dynamic-string.h" |
56 | #include "openvswitch/meta-flow.h" | |
57 | #include "openvswitch/list.h" | |
58 | #include "openvswitch/ofp-actions.h" | |
1fc11c59 | 59 | #include "openvswitch/ofp-ed-props.h" |
b598f214 BW |
60 | #include "openvswitch/vlog.h" |
61 | #include "ovs-lldp.h" | |
a36de779 | 62 | #include "ovs-router.h" |
b598f214 BW |
63 | #include "packets.h" |
64 | #include "tnl-neigh-cache.h" | |
a36de779 | 65 | #include "tnl-ports.h" |
9583bc14 | 66 | #include "tunnel.h" |
ee89ea7b | 67 | #include "util.h" |
c3594cc3 | 68 | #include "uuid.h" |
9583bc14 | 69 | |
46c88433 | 70 | COVERAGE_DEFINE(xlate_actions); |
0f032e95 | 71 | COVERAGE_DEFINE(xlate_actions_oversize); |
7d031d7e | 72 | COVERAGE_DEFINE(xlate_actions_too_many_output); |
9583bc14 EJ |
73 | |
74 | VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate); | |
75 | ||
8a553e9a | 76 | /* Maximum depth of flow table recursion (due to resubmit actions) in a |
790c5d26 BP |
77 | * flow translation. |
78 | * | |
79 | * The goal of limiting the depth of resubmits is to ensure that flow | |
80 | * translation eventually terminates. Only resubmits to the same table or an | |
81 | * earlier table count against the maximum depth. This is because resubmits to | |
82 | * strictly monotonically increasing table IDs will eventually terminate, since | |
83 | * any OpenFlow switch has a finite number of tables. OpenFlow tables are most | |
84 | * commonly traversed in numerically increasing order, so this limit has little | |
85 | * effect on conventionally designed OpenFlow pipelines. | |
86 | * | |
87 | * Outputs to patch ports and to groups also count against the depth limit. */ | |
88 | #define MAX_DEPTH 64 | |
8a553e9a | 89 | |
98b07853 BP |
90 | /* Maximum number of resubmit actions in a flow translation, whether they are |
91 | * recursive or not. */ | |
790c5d26 | 92 | #define MAX_RESUBMITS (MAX_DEPTH * MAX_DEPTH) |
98b07853 | 93 | |
83c2757b ZB |
94 | /* The structure holds an array of IP addresses assigned to a bridge and the |
95 | * number of elements in the array. These data are mutable and are evaluated | |
96 | * when ARP or Neighbor Advertisement packets received on a native tunnel | |
97 | * port are xlated. So 'ref_cnt' and RCU are used for synchronization. */ | |
98 | struct xbridge_addr { | |
99 | struct in6_addr *addr; /* Array of IP addresses of xbridge. */ | |
100 | int n_addr; /* Number of IP addresses. */ | |
101 | struct ovs_refcount ref_cnt; | |
102 | }; | |
103 | ||
46c88433 EJ |
104 | struct xbridge { |
105 | struct hmap_node hmap_node; /* Node in global 'xbridges' map. */ | |
106 | struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */ | |
107 | ||
ca6ba700 | 108 | struct ovs_list xbundles; /* Owned xbundles. */ |
46c88433 EJ |
109 | struct hmap xports; /* Indexed by ofp_port. */ |
110 | ||
111 | char *name; /* Name used in log messages. */ | |
89a8a7f0 | 112 | struct dpif *dpif; /* Datapath interface. */ |
46c88433 | 113 | struct mac_learning *ml; /* Mac learning handle. */ |
6d95c4e8 | 114 | struct mcast_snooping *ms; /* Multicast Snooping handle. */ |
46c88433 EJ |
115 | struct mbridge *mbridge; /* Mirroring. */ |
116 | struct dpif_sflow *sflow; /* SFlow handle, or null. */ | |
117 | struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */ | |
ce3955be | 118 | struct netflow *netflow; /* Netflow handle, or null. */ |
9d189a50 | 119 | struct stp *stp; /* STP or null if disabled. */ |
9efd308e | 120 | struct rstp *rstp; /* RSTP or null if disabled. */ |
46c88433 | 121 | |
46c88433 EJ |
122 | bool has_in_band; /* Bridge has in band control? */ |
123 | bool forward_bpdu; /* Bridge forwards STP BPDUs? */ | |
4b97b70d | 124 | |
b440dd8c JS |
125 | /* Datapath feature support. */ |
126 | struct dpif_backer_support support; | |
83c2757b ZB |
127 | |
128 | struct xbridge_addr *addr; | |
46c88433 EJ |
129 | }; |
130 | ||
131 | struct xbundle { | |
132 | struct hmap_node hmap_node; /* In global 'xbundles' map. */ | |
133 | struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */ | |
134 | ||
ca6ba700 | 135 | struct ovs_list list_node; /* In parent 'xbridges' list. */ |
46c88433 EJ |
136 | struct xbridge *xbridge; /* Parent xbridge. */ |
137 | ||
ca6ba700 | 138 | struct ovs_list xports; /* Contains "struct xport"s. */ |
46c88433 EJ |
139 | |
140 | char *name; /* Name used in log messages. */ | |
141 | struct bond *bond; /* Nonnull iff more than one port. */ | |
142 | struct lacp *lacp; /* LACP handle or null. */ | |
143 | ||
144 | enum port_vlan_mode vlan_mode; /* VLAN mode. */ | |
fed8962a EG |
145 | uint16_t qinq_ethtype; /* Ethertype of dot1q-tunnel interface |
146 | * either 0x8100 or 0x88a8. */ | |
46c88433 EJ |
147 | int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */ |
148 | unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1. | |
149 | * NULL if all VLANs are trunked. */ | |
fed8962a EG |
150 | unsigned long *cvlans; /* Bitmap of allowed customer vlans, |
151 | * NULL if all VLANs are allowed */ | |
88f52d7f EB |
152 | enum port_priority_tags_mode use_priority_tags; |
153 | /* Use 802.1p tag for frames in VLAN 0? */ | |
46c88433 | 154 | bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */ |
c005f976 | 155 | bool protected; /* Protected port mode */ |
46c88433 EJ |
156 | }; |
157 | ||
158 | struct xport { | |
159 | struct hmap_node hmap_node; /* Node in global 'xports' map. */ | |
160 | struct ofport_dpif *ofport; /* Key in global 'xports map. */ | |
161 | ||
162 | struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */ | |
163 | ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */ | |
164 | ||
c3594cc3 ZB |
165 | struct hmap_node uuid_node; /* Node in global 'xports_uuid' map. */ |
166 | struct uuid uuid; /* Key in global 'xports_uuid' map. */ | |
167 | ||
46c88433 EJ |
168 | odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */ |
169 | ||
ca6ba700 | 170 | struct ovs_list bundle_node; /* In parent xbundle (if it exists). */ |
46c88433 EJ |
171 | struct xbundle *xbundle; /* Parent xbundle or null. */ |
172 | ||
173 | struct netdev *netdev; /* 'ofport''s netdev. */ | |
174 | ||
175 | struct xbridge *xbridge; /* Parent bridge. */ | |
176 | struct xport *peer; /* Patch port peer or null. */ | |
177 | ||
178 | enum ofputil_port_config config; /* OpenFlow port configuration. */ | |
dd8cd4b4 | 179 | enum ofputil_port_state state; /* OpenFlow port state. */ |
92cf817b | 180 | int stp_port_no; /* STP port number or -1 if not in use. */ |
f025bcb7 | 181 | struct rstp_port *rstp_port; /* RSTP port or null. */ |
46c88433 | 182 | |
55954f6e EJ |
183 | struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */ |
184 | ||
46c88433 EJ |
185 | bool may_enable; /* May be enabled in bonds. */ |
186 | bool is_tunnel; /* Is a tunnel port. */ | |
875ab130 | 187 | enum netdev_pt_mode pt_mode; /* packet_type handling. */ |
46c88433 EJ |
188 | |
189 | struct cfm *cfm; /* CFM handle or null. */ | |
190 | struct bfd *bfd; /* BFD handle or null. */ | |
0477baa9 | 191 | struct lldp *lldp; /* LLDP handle or null. */ |
46c88433 EJ |
192 | }; |
193 | ||
4d0acc70 EJ |
194 | struct xlate_ctx { |
195 | struct xlate_in *xin; | |
196 | struct xlate_out *xout; | |
197 | ||
0506f184 | 198 | struct xlate_cfg *xcfg; |
46c88433 | 199 | const struct xbridge *xbridge; |
4d0acc70 EJ |
200 | |
201 | /* Flow at the last commit. */ | |
202 | struct flow base_flow; | |
203 | ||
204 | /* Tunnel IP destination address as received. This is stored separately | |
205 | * as the base_flow.tunnel is cleared on init to reflect the datapath | |
206 | * behavior. Used to make sure not to send tunneled output to ourselves, | |
207 | * which might lead to an infinite loop. This could happen easily | |
208 | * if a tunnel is marked as 'ip_remote=flow', and the flow does not | |
209 | * actually set the tun_dst field. */ | |
e4d3706c | 210 | struct in6_addr orig_tunnel_ipv6_dst; |
4d0acc70 | 211 | |
84cf3c1f JR |
212 | /* Stack for the push and pop actions. See comment above nx_stack_push() |
213 | * in nx-match.c for info on how the stack is stored. */ | |
4d0acc70 EJ |
214 | struct ofpbuf stack; |
215 | ||
216 | /* The rule that we are currently translating, or NULL. */ | |
217 | struct rule_dpif *rule; | |
218 | ||
49a73e0c BP |
219 | /* Flow translation populates this with wildcards relevant in translation. |
220 | * When 'xin->wc' is nonnull, this is the same pointer. When 'xin->wc' is | |
c0e638aa | 221 | * null, this is a pointer to a temporary buffer. */ |
49a73e0c BP |
222 | struct flow_wildcards *wc; |
223 | ||
1520ef4f BP |
224 | /* Output buffer for datapath actions. When 'xin->odp_actions' is nonnull, |
225 | * this is the same pointer. When 'xin->odp_actions' is null, this points | |
226 | * to a scratch ofpbuf. This allows code to add actions to | |
227 | * 'ctx->odp_actions' without worrying about whether the caller really | |
228 | * wants actions. */ | |
229 | struct ofpbuf *odp_actions; | |
230 | ||
790c5d26 BP |
231 | /* Statistics maintained by xlate_table_action(). |
232 | * | |
2d9b49dd | 233 | * These statistics limit the amount of work that a single flow |
790c5d26 BP |
234 | * translation can perform. The goal of the first of these, 'depth', is |
235 | * primarily to prevent translation from performing an infinite amount of | |
236 | * work. It counts the current depth of nested "resubmit"s (and a few | |
237 | * other activities); when a resubmit returns, it decreases. Resubmits to | |
238 | * tables in strictly monotonically increasing order don't contribute to | |
239 | * 'depth' because they cannot cause a flow translation to take an infinite | |
240 | * amount of time (because the number of tables is finite). Translation | |
241 | * aborts when 'depth' exceeds MAX_DEPTH. | |
242 | * | |
243 | * 'resubmits', on the other hand, prevents flow translation from | |
244 | * performing an extraordinarily large while still finite amount of work. | |
245 | * It counts the total number of resubmits (and a few other activities) | |
246 | * that have been executed. Returning from a resubmit does not affect this | |
247 | * counter. Thus, this limits the amount of work that a particular | |
248 | * translation can perform. Translation aborts when 'resubmits' exceeds | |
249 | * MAX_RESUBMITS (which is much larger than MAX_DEPTH). | |
250 | */ | |
790c5d26 | 251 | int depth; /* Current resubmit nesting depth. */ |
98b07853 | 252 | int resubmits; /* Total number of resubmits. */ |
029ca940 | 253 | bool in_action_set; /* Currently translating action_set, if true. */ |
331c07ac YHW |
254 | bool in_packet_out; /* Currently translating a packet_out msg, if |
255 | * true. */ | |
1fc11c59 JS |
256 | bool pending_encap; /* True when waiting to commit a pending |
257 | * encap action. */ | |
88ec1e0a JS |
258 | bool pending_decap; /* True when waiting to commit a pending |
259 | * decap action. */ | |
1fc11c59 JS |
260 | struct ofpbuf *encap_data; /* May contain a pointer to an ofpbuf with |
261 | * context for the datapath encap action.*/ | |
98b07853 | 262 | |
4d0acc70 | 263 | uint8_t table_id; /* OpenFlow table ID where flow was found. */ |
8b1e5560 JR |
264 | ovs_be64 rule_cookie; /* Cookie of the rule being translated. */ |
265 | uint32_t orig_skb_priority; /* Priority when packet arrived. */ | |
4d0acc70 | 266 | uint32_t sflow_n_outputs; /* Number of output ports. */ |
4e022ec0 | 267 | odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */ |
2031ef97 | 268 | ofp_port_t nf_output_iface; /* Output interface index for NetFlow. */ |
4d0acc70 | 269 | bool exit; /* No further actions should be processed. */ |
3d6151f3 | 270 | mirror_mask_t mirrors; /* Bitmap of associated mirrors. */ |
1356dbd1 | 271 | int mirror_snaplen; /* Max size of a mirror packet in byte. */ |
7fdb60a7 | 272 | |
1d361a81 BP |
273 | /* Freezing Translation |
274 | * ==================== | |
e672ff9b | 275 | * |
1d361a81 BP |
276 | * At some point during translation, the code may recognize the need to halt |
277 | * and checkpoint the translation in a way that it can be restarted again | |
278 | * later. We call the checkpointing process "freezing" and the restarting | |
279 | * process "thawing". | |
e672ff9b | 280 | * |
1d361a81 | 281 | * The use cases for freezing are: |
e672ff9b | 282 | * |
1d361a81 BP |
283 | * - "Recirculation", where the translation process discovers that it |
284 | * doesn't have enough information to complete translation without | |
285 | * actually executing the actions that have already been translated, | |
286 | * which provides the additionally needed information. In these | |
287 | * situations, translation freezes translation and assigns the frozen | |
288 | * data a unique "recirculation ID", which it associates with the data | |
289 | * in a table in userspace (see ofproto-dpif-rid.h). It also adds a | |
290 | * OVS_ACTION_ATTR_RECIRC action specifying that ID to the datapath | |
291 | * actions. When a packet hits that action, the datapath looks its | |
292 | * flow up again using the ID. If there's a miss, it comes back to | |
293 | * userspace, which find the recirculation table entry for the ID, | |
294 | * thaws the associated frozen data, and continues translation from | |
295 | * that point given the additional information that is now known. | |
e672ff9b | 296 | * |
1d361a81 BP |
297 | * The archetypal example is MPLS. As MPLS is implemented in |
298 | * OpenFlow, the protocol that follows the last MPLS label becomes | |
299 | * known only when that label is popped by an OpenFlow action. That | |
300 | * means that Open vSwitch can't extract the headers beyond the MPLS | |
301 | * labels until the pop action is executed. Thus, at that point | |
302 | * translation uses the recirculation process to extract the headers | |
303 | * beyond the MPLS labels. | |
e672ff9b | 304 | * |
1d361a81 BP |
305 | * (OVS also uses OVS_ACTION_ATTR_RECIRC to implement hashing for |
306 | * output to bonds. OVS pre-populates all the datapath flows for bond | |
307 | * output in the datapath, though, which means that the elaborate | |
308 | * process of coming back to userspace for a second round of | |
309 | * translation isn't needed, and so bonds don't follow the above | |
310 | * process.) | |
e672ff9b | 311 | * |
77ab5fd2 BP |
312 | * - "Continuation". A continuation is a way for an OpenFlow controller |
313 | * to interpose on a packet's traversal of the OpenFlow tables. When | |
314 | * the translation process encounters a "controller" action with the | |
315 | * "pause" flag, it freezes translation, serializes the frozen data, | |
316 | * and sends it to an OpenFlow controller. The controller then | |
317 | * examines and possibly modifies the frozen data and eventually sends | |
318 | * it back to the switch, which thaws it and continues translation. | |
e672ff9b | 319 | * |
1d361a81 BP |
320 | * The main problem of freezing translation is preserving state, so that |
321 | * when the translation is thawed later it resumes from where it left off, | |
322 | * without disruption. In particular, actions must be preserved as follows: | |
323 | * | |
324 | * - If we're freezing because an action needed more information, the | |
325 | * action that prompted it. | |
326 | * | |
327 | * - Any actions remaining to be translated within the current flow. | |
328 | * | |
329 | * - If translation was frozen within a NXAST_RESUBMIT, then any actions | |
330 | * following the resubmit action. Resubmit actions can be nested, so | |
331 | * this has to go all the way up the control stack. | |
e672ff9b JR |
332 | * |
333 | * - The OpenFlow 1.1+ action set. | |
334 | * | |
335 | * State that actions and flow table lookups can depend on, such as the | |
336 | * following, must also be preserved: | |
337 | * | |
338 | * - Metadata fields (input port, registers, OF1.1+ metadata, ...). | |
339 | * | |
1d361a81 | 340 | * - The stack used by NXAST_STACK_PUSH and NXAST_STACK_POP actions. |
e672ff9b JR |
341 | * |
342 | * - The table ID and cookie of the flow being translated at each level | |
1d361a81 BP |
343 | * of the control stack, because these can become visible through |
344 | * OFPAT_CONTROLLER actions (and other ways). | |
e672ff9b JR |
345 | * |
346 | * Translation allows for the control of this state preservation via these | |
1d361a81 BP |
347 | * members. When a need to freeze translation is identified, the |
348 | * translation process: | |
e672ff9b | 349 | * |
1d361a81 | 350 | * 1. Sets 'freezing' to true. |
e672ff9b JR |
351 | * |
352 | * 2. Sets 'exit' to true to tell later steps that we're exiting from the | |
353 | * translation process. | |
354 | * | |
1d361a81 BP |
355 | * 3. Adds an OFPACT_UNROLL_XLATE action to 'frozen_actions', and points |
356 | * frozen_actions.header to the action to make it easy to find it later. | |
357 | * This action holds the current table ID and cookie so that they can be | |
358 | * restored during a post-recirculation upcall translation. | |
e672ff9b JR |
359 | * |
360 | * 4. Adds the action that prompted recirculation and any actions following | |
1d361a81 | 361 | * it within the same flow to 'frozen_actions', so that they can be |
8a5fb3b4 | 362 | * executed during a post-recirculation upcall translation. |
e672ff9b JR |
363 | * |
364 | * 5. Returns. | |
365 | * | |
366 | * 6. The action that prompted recirculation might be nested in a stack of | |
367 | * nested "resubmit"s that have actions remaining. Each of these notices | |
1d361a81 BP |
368 | * that we're exiting and freezing and responds by adding more |
369 | * OFPACT_UNROLL_XLATE actions to 'frozen_actions', as necessary, | |
370 | * followed by any actions that were yet unprocessed. | |
e672ff9b | 371 | * |
1d361a81 BP |
372 | * If we're freezing because of recirculation, the caller generates a |
373 | * recirculation ID and associates all the state produced by this process | |
374 | * with it. For post-recirculation upcall translation, the caller passes it | |
375 | * back in for the new translation to execute. The process yielded a set of | |
376 | * ofpacts that can be translated directly, so it is not much of a special | |
377 | * case at that point. | |
e672ff9b | 378 | */ |
1d361a81 | 379 | bool freezing; |
53cc166a JR |
380 | bool recirc_update_dp_hash; /* Generated recirculation will be preceded |
381 | * by datapath HASH action to get an updated | |
382 | * dp_hash after recirculation. */ | |
383 | uint32_t dp_hash_alg; | |
384 | uint32_t dp_hash_basis; | |
1d361a81 | 385 | struct ofpbuf frozen_actions; |
77ab5fd2 | 386 | const struct ofpact_controller *pause; |
e672ff9b | 387 | |
e12ec36b SH |
388 | /* True if a packet was but is no longer MPLS (due to an MPLS pop action). |
389 | * This is a trigger for recirculation in cases where translating an action | |
390 | * or looking up a flow requires access to the fields of the packet after | |
391 | * the MPLS label stack that was originally present. */ | |
392 | bool was_mpls; | |
393 | ||
07659514 JS |
394 | /* True if conntrack has been performed on this packet during processing |
395 | * on the current bridge. This is used to determine whether conntrack | |
1d361a81 | 396 | * state from the datapath should be honored after thawing. */ |
07659514 JS |
397 | bool conntracked; |
398 | ||
9ac0aada JR |
399 | /* Pointer to an embedded NAT action in a conntrack action, or NULL. */ |
400 | struct ofpact_nat *ct_nat_action; | |
401 | ||
7fdb60a7 SH |
402 | /* OpenFlow 1.1+ action set. |
403 | * | |
404 | * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS. | |
405 | * When translation is otherwise complete, ofpacts_execute_action_set() | |
406 | * converts it to a set of "struct ofpact"s that can be translated into | |
ed9c9e3e | 407 | * datapath actions. */ |
c61f3870 | 408 | bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ |
7fdb60a7 | 409 | struct ofpbuf action_set; /* Action set. */ |
fff1b9c0 JR |
410 | |
411 | enum xlate_error error; /* Translation failed. */ | |
4d0acc70 EJ |
412 | }; |
413 | ||
f0fb825a EG |
414 | /* Structure to track VLAN manipulation */ |
415 | struct xvlan_single { | |
416 | uint16_t tpid; | |
417 | uint16_t vid; | |
418 | uint16_t pcp; | |
419 | }; | |
420 | ||
421 | struct xvlan { | |
422 | struct xvlan_single v[FLOW_MAX_VLAN_HEADERS]; | |
423 | }; | |
424 | ||
fff1b9c0 JR |
425 | const char *xlate_strerror(enum xlate_error error) |
426 | { | |
427 | switch (error) { | |
428 | case XLATE_OK: | |
429 | return "OK"; | |
430 | case XLATE_BRIDGE_NOT_FOUND: | |
431 | return "Bridge not found"; | |
432 | case XLATE_RECURSION_TOO_DEEP: | |
433 | return "Recursion too deep"; | |
434 | case XLATE_TOO_MANY_RESUBMITS: | |
435 | return "Too many resubmits"; | |
436 | case XLATE_STACK_TOO_DEEP: | |
437 | return "Stack too deep"; | |
438 | case XLATE_NO_RECIRCULATION_CONTEXT: | |
439 | return "No recirculation context"; | |
440 | case XLATE_RECIRCULATION_CONFLICT: | |
441 | return "Recirculation conflict"; | |
442 | case XLATE_TOO_MANY_MPLS_LABELS: | |
443 | return "Too many MPLS labels"; | |
8d8ab6c2 JG |
444 | case XLATE_INVALID_TUNNEL_METADATA: |
445 | return "Invalid tunnel metadata"; | |
7873e106 ZB |
446 | case XLATE_UNSUPPORTED_PACKET_TYPE: |
447 | return "Unsupported packet type"; | |
a13a0209 AT |
448 | case XLATE_CONGESTION_DROP: |
449 | return "Congestion Drop"; | |
450 | case XLATE_FORWARDING_DISABLED: | |
451 | return "Forwarding is disabled"; | |
452 | case XLATE_MAX: | |
453 | break; | |
fff1b9c0 JR |
454 | } |
455 | return "Unknown error"; | |
456 | } | |
457 | ||
ed9c9e3e | 458 | static void xlate_action_set(struct xlate_ctx *ctx); |
704bb0bf | 459 | static void xlate_commit_actions(struct xlate_ctx *ctx); |
ed9c9e3e | 460 | |
8bdb2bdb | 461 | static void |
48f704f4 AZ |
462 | patch_port_output(struct xlate_ctx *ctx, const struct xport *in_dev, |
463 | struct xport *out_dev); | |
8bdb2bdb | 464 | |
1d741d6d | 465 | static void |
1d361a81 | 466 | ctx_trigger_freeze(struct xlate_ctx *ctx) |
1d741d6d JR |
467 | { |
468 | ctx->exit = true; | |
1d361a81 | 469 | ctx->freezing = true; |
1d741d6d JR |
470 | } |
471 | ||
53cc166a JR |
472 | static void |
473 | ctx_trigger_recirculate_with_hash(struct xlate_ctx *ctx, uint32_t type, | |
474 | uint32_t basis) | |
475 | { | |
476 | ctx->exit = true; | |
477 | ctx->freezing = true; | |
478 | ctx->recirc_update_dp_hash = true; | |
479 | ctx->dp_hash_alg = type; | |
480 | ctx->dp_hash_basis = basis; | |
481 | } | |
482 | ||
1d741d6d | 483 | static bool |
1d361a81 | 484 | ctx_first_frozen_action(const struct xlate_ctx *ctx) |
1d741d6d | 485 | { |
1d361a81 | 486 | return !ctx->frozen_actions.size; |
e672ff9b JR |
487 | } |
488 | ||
3293cb85 | 489 | static void |
1d361a81 | 490 | ctx_cancel_freeze(struct xlate_ctx *ctx) |
3293cb85 | 491 | { |
1d361a81 BP |
492 | if (ctx->freezing) { |
493 | ctx->freezing = false; | |
53cc166a | 494 | ctx->recirc_update_dp_hash = false; |
1d361a81 BP |
495 | ofpbuf_clear(&ctx->frozen_actions); |
496 | ctx->frozen_actions.header = NULL; | |
5b34f8fc | 497 | ctx->pause = NULL; |
3293cb85 BP |
498 | } |
499 | } | |
500 | ||
77ab5fd2 | 501 | static void finish_freezing(struct xlate_ctx *ctx); |
e672ff9b | 502 | |
9583bc14 EJ |
503 | /* A controller may use OFPP_NONE as the ingress port to indicate that |
504 | * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for | |
505 | * when an input bundle is needed for validation (e.g., mirroring or | |
506 | * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have | |
3548d242 BP |
507 | * any 'port' structs, so care must be taken when dealing with it. */ |
508 | static struct xbundle ofpp_none_bundle = { | |
509 | .name = "OFPP_NONE", | |
510 | .vlan_mode = PORT_VLAN_TRUNK | |
511 | }; | |
9583bc14 | 512 | |
55954f6e EJ |
513 | /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from |
514 | * 'priority' (the datapath's term for QoS queue) to the dscp bits which all | |
515 | * traffic egressing the 'ofport' with that priority should be marked with. */ | |
516 | struct skb_priority_to_dscp { | |
517 | struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */ | |
518 | uint32_t skb_priority; /* Priority of this queue (see struct flow). */ | |
519 | ||
520 | uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */ | |
521 | }; | |
522 | ||
84f0f298 RW |
523 | /* Xlate config contains hash maps of all bridges, bundles and ports. |
524 | * Xcfgp contains the pointer to the current xlate configuration. | |
525 | * When the main thread needs to change the configuration, it copies xcfgp to | |
526 | * new_xcfg and edits new_xcfg. This enables the use of RCU locking which | |
527 | * does not block handler and revalidator threads. */ | |
528 | struct xlate_cfg { | |
529 | struct hmap xbridges; | |
530 | struct hmap xbundles; | |
531 | struct hmap xports; | |
c3594cc3 | 532 | struct hmap xports_uuid; |
84f0f298 | 533 | }; |
b1b72f2d | 534 | static OVSRCU_TYPE(struct xlate_cfg *) xcfgp = OVSRCU_INITIALIZER(NULL); |
f439f23b | 535 | static struct xlate_cfg *new_xcfg = NULL; |
46c88433 | 536 | |
96c3a6e5 | 537 | typedef void xlate_actions_handler(const struct ofpact *, size_t ofpacts_len, |
f5634764 | 538 | struct xlate_ctx *, bool, bool); |
46c88433 | 539 | static bool may_receive(const struct xport *, struct xlate_ctx *); |
9583bc14 | 540 | static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
f5634764 | 541 | struct xlate_ctx *, bool, bool); |
96c3a6e5 | 542 | static void clone_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
f5634764 | 543 | struct xlate_ctx *, bool, bool); |
adcf00ba | 544 | static void xlate_normal(struct xlate_ctx *); |
efc93e66 BP |
545 | static void xlate_normal_flood(struct xlate_ctx *ct, |
546 | struct xbundle *in_xbundle, struct xvlan *); | |
6d328fa2 SH |
547 | static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port, |
548 | uint8_t table_id, bool may_packet_in, | |
feee58b9 | 549 | bool honor_table_miss, bool with_ct_orig, |
96c3a6e5 AZ |
550 | bool is_last_action, xlate_actions_handler *); |
551 | ||
2d9b49dd BP |
552 | static bool input_vid_is_valid(const struct xlate_ctx *, |
553 | uint16_t vid, struct xbundle *); | |
f0fb825a EG |
554 | static void xvlan_copy(struct xvlan *dst, const struct xvlan *src); |
555 | static void xvlan_pop(struct xvlan *src); | |
fed8962a | 556 | static void xvlan_push_uninit(struct xvlan *src); |
f0fb825a | 557 | static void xvlan_extract(const struct flow *, struct xvlan *); |
ffbe41db EB |
558 | static void xvlan_put(struct flow *, const struct xvlan *, |
559 | enum port_priority_tags_mode); | |
f0fb825a EG |
560 | static void xvlan_input_translate(const struct xbundle *, |
561 | const struct xvlan *in, | |
562 | struct xvlan *xvlan); | |
563 | static void xvlan_output_translate(const struct xbundle *, | |
564 | const struct xvlan *xvlan, | |
565 | struct xvlan *out); | |
46c88433 | 566 | static void output_normal(struct xlate_ctx *, const struct xbundle *, |
f0fb825a | 567 | const struct xvlan *); |
e93ef1c7 JR |
568 | |
569 | /* Optional bond recirculation parameter to compose_output_action(). */ | |
570 | struct xlate_bond_recirc { | |
571 | uint32_t recirc_id; /* !0 Use recirculation instead of output. */ | |
572 | uint8_t hash_alg; /* !0 Compute hash for recirc before. */ | |
573 | uint32_t hash_basis; /* Compute hash for recirc before. */ | |
574 | }; | |
575 | ||
576 | static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port, | |
feee58b9 | 577 | const struct xlate_bond_recirc *xr, |
11938578 | 578 | bool is_last_action, bool truncate); |
9583bc14 | 579 | |
84f0f298 RW |
580 | static struct xbridge *xbridge_lookup(struct xlate_cfg *, |
581 | const struct ofproto_dpif *); | |
290835f9 BP |
582 | static struct xbridge *xbridge_lookup_by_uuid(struct xlate_cfg *, |
583 | const struct uuid *); | |
84f0f298 RW |
584 | static struct xbundle *xbundle_lookup(struct xlate_cfg *, |
585 | const struct ofbundle *); | |
586 | static struct xport *xport_lookup(struct xlate_cfg *, | |
587 | const struct ofport_dpif *); | |
00135b86 ZB |
588 | static struct xport *xport_lookup_by_uuid(struct xlate_cfg *, |
589 | const struct uuid *); | |
46c88433 | 590 | static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port); |
55954f6e EJ |
591 | static struct skb_priority_to_dscp *get_skb_priority(const struct xport *, |
592 | uint32_t skb_priority); | |
593 | static void clear_skb_priorities(struct xport *); | |
16194afd | 594 | static size_t count_skb_priorities(const struct xport *); |
55954f6e EJ |
595 | static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, |
596 | uint8_t *dscp); | |
46c88433 | 597 | |
84f0f298 RW |
598 | static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *); |
599 | static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *); | |
600 | static void xlate_xport_init(struct xlate_cfg *, struct xport *); | |
9efd308e | 601 | static void xlate_xbridge_set(struct xbridge *, struct dpif *, |
9efd308e DV |
602 | const struct mac_learning *, struct stp *, |
603 | struct rstp *, const struct mcast_snooping *, | |
604 | const struct mbridge *, | |
605 | const struct dpif_sflow *, | |
606 | const struct dpif_ipfix *, | |
2f47cdf4 | 607 | const struct netflow *, |
84f0f298 | 608 | bool forward_bpdu, bool has_in_band, |
83c2757b ZB |
609 | const struct dpif_backer_support *, |
610 | const struct xbridge_addr *); | |
84f0f298 | 611 | static void xlate_xbundle_set(struct xbundle *xbundle, |
f0fb825a | 612 | enum port_vlan_mode vlan_mode, |
fed8962a EG |
613 | uint16_t qinq_ethtype, int vlan, |
614 | unsigned long *trunks, unsigned long *cvlans, | |
88f52d7f | 615 | enum port_priority_tags_mode, |
84f0f298 | 616 | const struct bond *bond, const struct lacp *lacp, |
c005f976 | 617 | bool floodable, bool protected); |
84f0f298 RW |
618 | static void xlate_xport_set(struct xport *xport, odp_port_t odp_port, |
619 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 DF |
620 | const struct bfd *bfd, const struct lldp *lldp, |
621 | int stp_port_no, const struct rstp_port *rstp_port, | |
84f0f298 RW |
622 | enum ofputil_port_config config, |
623 | enum ofputil_port_state state, bool is_tunnel, | |
624 | bool may_enable); | |
625 | static void xlate_xbridge_remove(struct xlate_cfg *, struct xbridge *); | |
626 | static void xlate_xbundle_remove(struct xlate_cfg *, struct xbundle *); | |
627 | static void xlate_xport_remove(struct xlate_cfg *, struct xport *); | |
628 | static void xlate_xbridge_copy(struct xbridge *); | |
629 | static void xlate_xbundle_copy(struct xbridge *, struct xbundle *); | |
630 | static void xlate_xport_copy(struct xbridge *, struct xbundle *, | |
631 | struct xport *); | |
632 | static void xlate_xcfg_free(struct xlate_cfg *); | |
2d9b49dd BP |
633 | \f |
634 | /* Tracing helpers. */ | |
635 | ||
636 | /* If tracing is enabled in 'ctx', creates a new trace node and appends it to | |
637 | * the list of nodes maintained in ctx->xin. The new node has type 'type' and | |
638 | * its text is created from 'format' by treating it as a printf format string. | |
639 | * Returns the list of nodes embedded within the new trace node; ordinarily, | |
640 | * the calleer can ignore this, but it is useful if the caller needs to nest | |
641 | * more trace nodes within the new node. | |
642 | * | |
643 | * If tracing is not enabled, does nothing and returns NULL. */ | |
644 | static struct ovs_list * OVS_PRINTF_FORMAT(3, 4) | |
645 | xlate_report(const struct xlate_ctx *ctx, enum oftrace_node_type type, | |
646 | const char *format, ...) | |
34dd0d78 | 647 | { |
2d9b49dd BP |
648 | struct ovs_list *subtrace = NULL; |
649 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
c1b3756c | 650 | va_list args; |
c1b3756c | 651 | va_start(args, format); |
2d9b49dd BP |
652 | char *text = xvasprintf(format, args); |
653 | subtrace = &oftrace_report(ctx->xin->trace, type, text)->subs; | |
c1b3756c | 654 | va_end(args); |
2d9b49dd | 655 | free(text); |
34dd0d78 | 656 | } |
2d9b49dd | 657 | return subtrace; |
34dd0d78 | 658 | } |
84f0f298 | 659 | |
2d9b49dd BP |
660 | /* This is like xlate_report() for errors that are serious enough that we |
661 | * should log them even if we are not tracing. */ | |
662 | static void OVS_PRINTF_FORMAT(2, 3) | |
663 | xlate_report_error(const struct xlate_ctx *ctx, const char *format, ...) | |
664 | { | |
665 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
666 | if (!OVS_UNLIKELY(ctx->xin->trace) | |
667 | && (!ctx->xin->packet || VLOG_DROP_WARN(&rl))) { | |
668 | return; | |
669 | } | |
670 | ||
671 | struct ds s = DS_EMPTY_INITIALIZER; | |
672 | va_list args; | |
673 | va_start(args, format); | |
674 | ds_put_format_valist(&s, format, args); | |
675 | va_end(args); | |
676 | ||
677 | if (ctx->xin->trace) { | |
678 | oftrace_report(ctx->xin->trace, OFT_ERROR, ds_cstr(&s)); | |
679 | } else { | |
a4c3463d BP |
680 | ds_put_format(&s, " on bridge %s while processing ", |
681 | ctx->xbridge->name); | |
50f96b10 | 682 | flow_format(&s, &ctx->base_flow, NULL); |
2d9b49dd BP |
683 | VLOG_WARN("%s", ds_cstr(&s)); |
684 | } | |
685 | ds_destroy(&s); | |
686 | } | |
687 | ||
f89547e2 ZW |
688 | /* This is like xlate_report() for messages that should be logged |
689 | at the info level (even when not tracing). */ | |
690 | static void OVS_PRINTF_FORMAT(2, 3) | |
691 | xlate_report_info(const struct xlate_ctx *ctx, const char *format, ...) | |
692 | { | |
693 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
694 | if (!OVS_UNLIKELY(ctx->xin->trace) | |
695 | && (!ctx->xin->packet || VLOG_DROP_INFO(&rl))) { | |
696 | return; | |
697 | } | |
698 | ||
699 | struct ds s = DS_EMPTY_INITIALIZER; | |
700 | va_list args; | |
701 | va_start(args, format); | |
702 | ds_put_format_valist(&s, format, args); | |
703 | va_end(args); | |
704 | ||
705 | if (ctx->xin->trace) { | |
706 | oftrace_report(ctx->xin->trace, OFT_WARN, ds_cstr(&s)); | |
707 | } else { | |
708 | ds_put_format(&s, " on bridge %s while processing ", | |
709 | ctx->xbridge->name); | |
710 | flow_format(&s, &ctx->base_flow, NULL); | |
711 | VLOG_INFO("%s", ds_cstr(&s)); | |
712 | } | |
713 | ds_destroy(&s); | |
714 | } | |
715 | ||
2d9b49dd BP |
716 | /* This is like xlate_report() for messages that should be logged at debug |
717 | * level (even if we are not tracing) because they can be valuable for | |
718 | * debugging. */ | |
719 | static void OVS_PRINTF_FORMAT(3, 4) | |
720 | xlate_report_debug(const struct xlate_ctx *ctx, enum oftrace_node_type type, | |
721 | const char *format, ...) | |
722 | { | |
723 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
724 | if (!OVS_UNLIKELY(ctx->xin->trace) | |
725 | && (!ctx->xin->packet || VLOG_DROP_DBG(&rl))) { | |
726 | return; | |
727 | } | |
728 | ||
729 | struct ds s = DS_EMPTY_INITIALIZER; | |
730 | va_list args; | |
731 | va_start(args, format); | |
732 | ds_put_format_valist(&s, format, args); | |
733 | va_end(args); | |
fff1b9c0 | 734 | |
2d9b49dd BP |
735 | if (ctx->xin->trace) { |
736 | oftrace_report(ctx->xin->trace, type, ds_cstr(&s)); | |
737 | } else { | |
738 | VLOG_DBG("bridge %s: %s", ctx->xbridge->name, ds_cstr(&s)); | |
739 | } | |
740 | ds_destroy(&s); | |
741 | } | |
fff1b9c0 | 742 | |
2d9b49dd BP |
743 | /* If tracing is enabled in 'ctx', appends a node of the given 'type' to the |
744 | * trace, whose text is 'title' followed by a formatted version of the | |
745 | * 'ofpacts_len' OpenFlow actions in 'ofpacts'. | |
746 | * | |
747 | * If tracing is not enabled, does nothing. */ | |
748 | static void | |
749 | xlate_report_actions(const struct xlate_ctx *ctx, enum oftrace_node_type type, | |
750 | const char *title, | |
d6bef3cc BP |
751 | const struct ofpact *ofpacts, size_t ofpacts_len) |
752 | { | |
2d9b49dd | 753 | if (OVS_UNLIKELY(ctx->xin->trace)) { |
d6bef3cc | 754 | struct ds s = DS_EMPTY_INITIALIZER; |
2d9b49dd | 755 | ds_put_format(&s, "%s: ", title); |
efefbcae BP |
756 | struct ofpact_format_params fp = { .s = &s }; |
757 | ofpacts_format(ofpacts, ofpacts_len, &fp); | |
2d9b49dd | 758 | oftrace_report(ctx->xin->trace, type, ds_cstr(&s)); |
d6bef3cc BP |
759 | ds_destroy(&s); |
760 | } | |
761 | } | |
762 | ||
2d9b49dd BP |
763 | /* If tracing is enabled in 'ctx', appends a node of type OFT_DETAIL to the |
764 | * trace, whose the message is a formatted version of the OpenFlow action set. | |
765 | * 'verb' should be "was" or "is", depending on whether the action set reported | |
766 | * is the new action set or the old one. | |
767 | * | |
768 | * If tracing is not enabled, does nothing. */ | |
769 | static void | |
770 | xlate_report_action_set(const struct xlate_ctx *ctx, const char *verb) | |
771 | { | |
772 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
773 | struct ofpbuf action_list; | |
774 | ofpbuf_init(&action_list, 0); | |
775 | ofpacts_execute_action_set(&action_list, &ctx->action_set); | |
776 | if (action_list.size) { | |
777 | struct ds s = DS_EMPTY_INITIALIZER; | |
efefbcae BP |
778 | struct ofpact_format_params fp = { .s = &s }; |
779 | ofpacts_format(action_list.data, action_list.size, &fp); | |
2d9b49dd BP |
780 | xlate_report(ctx, OFT_DETAIL, "action set %s: %s", |
781 | verb, ds_cstr(&s)); | |
782 | ds_destroy(&s); | |
783 | } else { | |
784 | xlate_report(ctx, OFT_DETAIL, "action set %s empty", verb); | |
785 | } | |
786 | ofpbuf_uninit(&action_list); | |
787 | } | |
788 | } | |
789 | ||
790 | ||
791 | /* If tracing is enabled in 'ctx', appends a node representing 'rule' (in | |
792 | * OpenFlow table 'table_id') to the trace and makes this node the parent for | |
793 | * future trace nodes. The caller should save ctx->xin->trace before calling | |
794 | * this function, then after tracing all of the activities under the table, | |
795 | * restore its previous value. | |
796 | * | |
797 | * If tracing is not enabled, does nothing. */ | |
798 | static void | |
799 | xlate_report_table(const struct xlate_ctx *ctx, struct rule_dpif *rule, | |
800 | uint8_t table_id) | |
801 | { | |
802 | if (OVS_LIKELY(!ctx->xin->trace)) { | |
803 | return; | |
804 | } | |
805 | ||
806 | struct ds s = DS_EMPTY_INITIALIZER; | |
807 | ds_put_format(&s, "%2d. ", table_id); | |
808 | if (rule == ctx->xin->ofproto->miss_rule) { | |
809 | ds_put_cstr(&s, "No match, and a \"packet-in\" is called for."); | |
810 | } else if (rule == ctx->xin->ofproto->no_packet_in_rule) { | |
811 | ds_put_cstr(&s, "No match."); | |
812 | } else if (rule == ctx->xin->ofproto->drop_frags_rule) { | |
813 | ds_put_cstr(&s, "Packets are IP fragments and " | |
814 | "the fragment handling mode is \"drop\"."); | |
815 | } else { | |
816 | minimatch_format(&rule->up.cr.match, | |
817 | ofproto_get_tun_tab(&ctx->xin->ofproto->up), | |
50f96b10 | 818 | NULL, &s, OFP_DEFAULT_PRIORITY); |
2d9b49dd BP |
819 | if (ds_last(&s) != ' ') { |
820 | ds_put_cstr(&s, ", "); | |
821 | } | |
822 | ds_put_format(&s, "priority %d", rule->up.cr.priority); | |
823 | if (rule->up.flow_cookie) { | |
824 | ds_put_format(&s, ", cookie %#"PRIx64, | |
825 | ntohll(rule->up.flow_cookie)); | |
826 | } | |
827 | } | |
828 | ctx->xin->trace = &oftrace_report(ctx->xin->trace, OFT_TABLE, | |
829 | ds_cstr(&s))->subs; | |
830 | ds_destroy(&s); | |
831 | } | |
832 | ||
833 | /* If tracing is enabled in 'ctx', adds an OFT_DETAIL trace node to 'ctx' | |
834 | * reporting the value of subfield 'sf'. | |
835 | * | |
836 | * If tracing is not enabled, does nothing. */ | |
837 | static void | |
838 | xlate_report_subfield(const struct xlate_ctx *ctx, | |
839 | const struct mf_subfield *sf) | |
840 | { | |
841 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
842 | struct ds s = DS_EMPTY_INITIALIZER; | |
843 | mf_format_subfield(sf, &s); | |
844 | ds_put_cstr(&s, " is now "); | |
845 | ||
846 | if (sf->ofs == 0 && sf->n_bits >= sf->field->n_bits) { | |
847 | union mf_value value; | |
848 | mf_get_value(sf->field, &ctx->xin->flow, &value); | |
50f96b10 | 849 | mf_format(sf->field, &value, NULL, NULL, &s); |
2d9b49dd BP |
850 | } else { |
851 | union mf_subvalue cst; | |
852 | mf_read_subfield(sf, &ctx->xin->flow, &cst); | |
853 | ds_put_hex(&s, &cst, sizeof cst); | |
854 | } | |
855 | ||
856 | xlate_report(ctx, OFT_DETAIL, "%s", ds_cstr(&s)); | |
857 | ||
858 | ds_destroy(&s); | |
859 | } | |
860 | } | |
861 | \f | |
84f0f298 RW |
862 | static void |
863 | xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
864 | { | |
417e7e66 | 865 | ovs_list_init(&xbridge->xbundles); |
84f0f298 RW |
866 | hmap_init(&xbridge->xports); |
867 | hmap_insert(&xcfg->xbridges, &xbridge->hmap_node, | |
868 | hash_pointer(xbridge->ofproto, 0)); | |
869 | } | |
870 | ||
871 | static void | |
872 | xlate_xbundle_init(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
873 | { | |
417e7e66 BW |
874 | ovs_list_init(&xbundle->xports); |
875 | ovs_list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); | |
84f0f298 RW |
876 | hmap_insert(&xcfg->xbundles, &xbundle->hmap_node, |
877 | hash_pointer(xbundle->ofbundle, 0)); | |
878 | } | |
879 | ||
880 | static void | |
881 | xlate_xport_init(struct xlate_cfg *xcfg, struct xport *xport) | |
882 | { | |
883 | hmap_init(&xport->skb_priorities); | |
884 | hmap_insert(&xcfg->xports, &xport->hmap_node, | |
885 | hash_pointer(xport->ofport, 0)); | |
886 | hmap_insert(&xport->xbridge->xports, &xport->ofp_node, | |
887 | hash_ofp_port(xport->ofp_port)); | |
c3594cc3 ZB |
888 | hmap_insert(&xcfg->xports_uuid, &xport->uuid_node, |
889 | uuid_hash(&xport->uuid)); | |
84f0f298 RW |
890 | } |
891 | ||
83c2757b ZB |
892 | static struct xbridge_addr * |
893 | xbridge_addr_create(struct xbridge *xbridge) | |
894 | { | |
895 | struct xbridge_addr *xbridge_addr = xbridge->addr; | |
896 | struct in6_addr *addr = NULL, *mask = NULL; | |
897 | struct netdev *dev; | |
898 | int err, n_addr = 0; | |
899 | ||
900 | err = netdev_open(xbridge->name, NULL, &dev); | |
901 | if (!err) { | |
902 | err = netdev_get_addr_list(dev, &addr, &mask, &n_addr); | |
903 | if (!err) { | |
904 | if (!xbridge->addr || | |
905 | n_addr != xbridge->addr->n_addr || | |
906 | (xbridge->addr->addr && memcmp(addr, xbridge->addr->addr, | |
907 | sizeof(*addr) * n_addr))) { | |
908 | xbridge_addr = xzalloc(sizeof *xbridge_addr); | |
909 | xbridge_addr->addr = addr; | |
910 | xbridge_addr->n_addr = n_addr; | |
911 | ovs_refcount_init(&xbridge_addr->ref_cnt); | |
912 | } else { | |
913 | free(addr); | |
914 | } | |
915 | free(mask); | |
916 | } | |
917 | netdev_close(dev); | |
918 | } | |
919 | ||
920 | return xbridge_addr; | |
921 | } | |
922 | ||
923 | static struct xbridge_addr * | |
924 | xbridge_addr_ref(const struct xbridge_addr *addr_) | |
925 | { | |
926 | struct xbridge_addr *addr = CONST_CAST(struct xbridge_addr *, addr_); | |
927 | if (addr) { | |
928 | ovs_refcount_ref(&addr->ref_cnt); | |
929 | } | |
930 | return addr; | |
931 | } | |
932 | ||
933 | static void | |
934 | xbridge_addr_unref(struct xbridge_addr *addr) | |
935 | { | |
936 | if (addr && ovs_refcount_unref_relaxed(&addr->ref_cnt) == 1) { | |
937 | free(addr->addr); | |
938 | free(addr); | |
939 | } | |
940 | } | |
941 | ||
84f0f298 RW |
942 | static void |
943 | xlate_xbridge_set(struct xbridge *xbridge, | |
944 | struct dpif *dpif, | |
ec89fc6f | 945 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 946 | struct rstp *rstp, const struct mcast_snooping *ms, |
ec89fc6f | 947 | const struct mbridge *mbridge, |
46c88433 | 948 | const struct dpif_sflow *sflow, |
ce3955be | 949 | const struct dpif_ipfix *ipfix, |
2f47cdf4 | 950 | const struct netflow *netflow, |
4b97b70d | 951 | bool forward_bpdu, bool has_in_band, |
83c2757b ZB |
952 | const struct dpif_backer_support *support, |
953 | const struct xbridge_addr *addr) | |
46c88433 | 954 | { |
46c88433 EJ |
955 | if (xbridge->ml != ml) { |
956 | mac_learning_unref(xbridge->ml); | |
957 | xbridge->ml = mac_learning_ref(ml); | |
958 | } | |
959 | ||
6d95c4e8 FL |
960 | if (xbridge->ms != ms) { |
961 | mcast_snooping_unref(xbridge->ms); | |
962 | xbridge->ms = mcast_snooping_ref(ms); | |
963 | } | |
964 | ||
46c88433 EJ |
965 | if (xbridge->mbridge != mbridge) { |
966 | mbridge_unref(xbridge->mbridge); | |
967 | xbridge->mbridge = mbridge_ref(mbridge); | |
968 | } | |
969 | ||
970 | if (xbridge->sflow != sflow) { | |
971 | dpif_sflow_unref(xbridge->sflow); | |
972 | xbridge->sflow = dpif_sflow_ref(sflow); | |
973 | } | |
974 | ||
975 | if (xbridge->ipfix != ipfix) { | |
976 | dpif_ipfix_unref(xbridge->ipfix); | |
977 | xbridge->ipfix = dpif_ipfix_ref(ipfix); | |
978 | } | |
979 | ||
9d189a50 EJ |
980 | if (xbridge->stp != stp) { |
981 | stp_unref(xbridge->stp); | |
982 | xbridge->stp = stp_ref(stp); | |
983 | } | |
984 | ||
9efd308e DV |
985 | if (xbridge->rstp != rstp) { |
986 | rstp_unref(xbridge->rstp); | |
987 | xbridge->rstp = rstp_ref(rstp); | |
988 | } | |
989 | ||
ce3955be EJ |
990 | if (xbridge->netflow != netflow) { |
991 | netflow_unref(xbridge->netflow); | |
992 | xbridge->netflow = netflow_ref(netflow); | |
993 | } | |
994 | ||
83c2757b ZB |
995 | if (xbridge->addr != addr) { |
996 | xbridge_addr_unref(xbridge->addr); | |
997 | xbridge->addr = xbridge_addr_ref(addr); | |
998 | } | |
999 | ||
89a8a7f0 | 1000 | xbridge->dpif = dpif; |
46c88433 EJ |
1001 | xbridge->forward_bpdu = forward_bpdu; |
1002 | xbridge->has_in_band = has_in_band; | |
b440dd8c | 1003 | xbridge->support = *support; |
46c88433 EJ |
1004 | } |
1005 | ||
84f0f298 RW |
1006 | static void |
1007 | xlate_xbundle_set(struct xbundle *xbundle, | |
fed8962a EG |
1008 | enum port_vlan_mode vlan_mode, uint16_t qinq_ethtype, |
1009 | int vlan, unsigned long *trunks, unsigned long *cvlans, | |
88f52d7f | 1010 | enum port_priority_tags_mode use_priority_tags, |
84f0f298 | 1011 | const struct bond *bond, const struct lacp *lacp, |
c005f976 | 1012 | bool floodable, bool protected) |
84f0f298 RW |
1013 | { |
1014 | ovs_assert(xbundle->xbridge); | |
1015 | ||
1016 | xbundle->vlan_mode = vlan_mode; | |
fed8962a | 1017 | xbundle->qinq_ethtype = qinq_ethtype; |
84f0f298 RW |
1018 | xbundle->vlan = vlan; |
1019 | xbundle->trunks = trunks; | |
fed8962a | 1020 | xbundle->cvlans = cvlans; |
84f0f298 RW |
1021 | xbundle->use_priority_tags = use_priority_tags; |
1022 | xbundle->floodable = floodable; | |
c005f976 | 1023 | xbundle->protected = protected; |
84f0f298 RW |
1024 | |
1025 | if (xbundle->bond != bond) { | |
1026 | bond_unref(xbundle->bond); | |
1027 | xbundle->bond = bond_ref(bond); | |
1028 | } | |
1029 | ||
1030 | if (xbundle->lacp != lacp) { | |
1031 | lacp_unref(xbundle->lacp); | |
1032 | xbundle->lacp = lacp_ref(lacp); | |
1033 | } | |
1034 | } | |
1035 | ||
1036 | static void | |
1037 | xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
1038 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 | 1039 | const struct bfd *bfd, const struct lldp *lldp, int stp_port_no, |
f025bcb7 | 1040 | const struct rstp_port* rstp_port, |
84f0f298 RW |
1041 | enum ofputil_port_config config, enum ofputil_port_state state, |
1042 | bool is_tunnel, bool may_enable) | |
1043 | { | |
1044 | xport->config = config; | |
1045 | xport->state = state; | |
1046 | xport->stp_port_no = stp_port_no; | |
1047 | xport->is_tunnel = is_tunnel; | |
875ab130 | 1048 | xport->pt_mode = netdev_get_pt_mode(netdev); |
84f0f298 RW |
1049 | xport->may_enable = may_enable; |
1050 | xport->odp_port = odp_port; | |
1051 | ||
f025bcb7 JR |
1052 | if (xport->rstp_port != rstp_port) { |
1053 | rstp_port_unref(xport->rstp_port); | |
1054 | xport->rstp_port = rstp_port_ref(rstp_port); | |
1055 | } | |
1056 | ||
84f0f298 RW |
1057 | if (xport->cfm != cfm) { |
1058 | cfm_unref(xport->cfm); | |
1059 | xport->cfm = cfm_ref(cfm); | |
1060 | } | |
1061 | ||
1062 | if (xport->bfd != bfd) { | |
1063 | bfd_unref(xport->bfd); | |
1064 | xport->bfd = bfd_ref(bfd); | |
1065 | } | |
1066 | ||
0477baa9 DF |
1067 | if (xport->lldp != lldp) { |
1068 | lldp_unref(xport->lldp); | |
1069 | xport->lldp = lldp_ref(lldp); | |
1070 | } | |
1071 | ||
84f0f298 RW |
1072 | if (xport->netdev != netdev) { |
1073 | netdev_close(xport->netdev); | |
1074 | xport->netdev = netdev_ref(netdev); | |
1075 | } | |
1076 | } | |
1077 | ||
1078 | static void | |
1079 | xlate_xbridge_copy(struct xbridge *xbridge) | |
1080 | { | |
1081 | struct xbundle *xbundle; | |
1082 | struct xport *xport; | |
1083 | struct xbridge *new_xbridge = xzalloc(sizeof *xbridge); | |
1084 | new_xbridge->ofproto = xbridge->ofproto; | |
1085 | new_xbridge->name = xstrdup(xbridge->name); | |
1086 | xlate_xbridge_init(new_xcfg, new_xbridge); | |
1087 | ||
1088 | xlate_xbridge_set(new_xbridge, | |
34dd0d78 | 1089 | xbridge->dpif, xbridge->ml, xbridge->stp, |
9efd308e DV |
1090 | xbridge->rstp, xbridge->ms, xbridge->mbridge, |
1091 | xbridge->sflow, xbridge->ipfix, xbridge->netflow, | |
b440dd8c | 1092 | xbridge->forward_bpdu, xbridge->has_in_band, |
83c2757b | 1093 | &xbridge->support, xbridge->addr); |
84f0f298 RW |
1094 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
1095 | xlate_xbundle_copy(new_xbridge, xbundle); | |
1096 | } | |
1097 | ||
1098 | /* Copy xports which are not part of a xbundle */ | |
1099 | HMAP_FOR_EACH (xport, ofp_node, &xbridge->xports) { | |
1100 | if (!xport->xbundle) { | |
1101 | xlate_xport_copy(new_xbridge, NULL, xport); | |
1102 | } | |
1103 | } | |
1104 | } | |
1105 | ||
1106 | static void | |
1107 | xlate_xbundle_copy(struct xbridge *xbridge, struct xbundle *xbundle) | |
1108 | { | |
1109 | struct xport *xport; | |
1110 | struct xbundle *new_xbundle = xzalloc(sizeof *xbundle); | |
1111 | new_xbundle->ofbundle = xbundle->ofbundle; | |
1112 | new_xbundle->xbridge = xbridge; | |
1113 | new_xbundle->name = xstrdup(xbundle->name); | |
1114 | xlate_xbundle_init(new_xcfg, new_xbundle); | |
1115 | ||
fed8962a EG |
1116 | xlate_xbundle_set(new_xbundle, xbundle->vlan_mode, xbundle->qinq_ethtype, |
1117 | xbundle->vlan, xbundle->trunks, xbundle->cvlans, | |
84f0f298 | 1118 | xbundle->use_priority_tags, xbundle->bond, xbundle->lacp, |
c005f976 | 1119 | xbundle->floodable, xbundle->protected); |
84f0f298 RW |
1120 | LIST_FOR_EACH (xport, bundle_node, &xbundle->xports) { |
1121 | xlate_xport_copy(xbridge, new_xbundle, xport); | |
1122 | } | |
1123 | } | |
1124 | ||
1125 | static void | |
1126 | xlate_xport_copy(struct xbridge *xbridge, struct xbundle *xbundle, | |
1127 | struct xport *xport) | |
1128 | { | |
1129 | struct skb_priority_to_dscp *pdscp, *new_pdscp; | |
1130 | struct xport *new_xport = xzalloc(sizeof *xport); | |
1131 | new_xport->ofport = xport->ofport; | |
1132 | new_xport->ofp_port = xport->ofp_port; | |
1133 | new_xport->xbridge = xbridge; | |
c3594cc3 | 1134 | new_xport->uuid = xport->uuid; |
84f0f298 RW |
1135 | xlate_xport_init(new_xcfg, new_xport); |
1136 | ||
1137 | xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm, | |
0477baa9 DF |
1138 | xport->bfd, xport->lldp, xport->stp_port_no, |
1139 | xport->rstp_port, xport->config, xport->state, | |
1140 | xport->is_tunnel, xport->may_enable); | |
84f0f298 RW |
1141 | |
1142 | if (xport->peer) { | |
1143 | struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport); | |
1144 | if (peer) { | |
1145 | new_xport->peer = peer; | |
1146 | new_xport->peer->peer = new_xport; | |
1147 | } | |
1148 | } | |
1149 | ||
1150 | if (xbundle) { | |
1151 | new_xport->xbundle = xbundle; | |
417e7e66 | 1152 | ovs_list_insert(&new_xport->xbundle->xports, &new_xport->bundle_node); |
84f0f298 RW |
1153 | } |
1154 | ||
1155 | HMAP_FOR_EACH (pdscp, hmap_node, &xport->skb_priorities) { | |
1156 | new_pdscp = xmalloc(sizeof *pdscp); | |
1157 | new_pdscp->skb_priority = pdscp->skb_priority; | |
1158 | new_pdscp->dscp = pdscp->dscp; | |
1159 | hmap_insert(&new_xport->skb_priorities, &new_pdscp->hmap_node, | |
1160 | hash_int(new_pdscp->skb_priority, 0)); | |
1161 | } | |
1162 | } | |
1163 | ||
1164 | /* Sets the current xlate configuration to new_xcfg and frees the old xlate | |
1165 | * configuration in xcfgp. | |
1166 | * | |
1167 | * This needs to be called after editing the xlate configuration. | |
1168 | * | |
1169 | * Functions that edit the new xlate configuration are | |
6cd20a22 | 1170 | * xlate_<ofproto/bundle/ofport>_set and xlate_<ofproto/bundle/ofport>_remove. |
84f0f298 RW |
1171 | * |
1172 | * A sample workflow: | |
1173 | * | |
586cd310 BP |
1174 | * xlate_txn_start(); |
1175 | * ... | |
1176 | * edit_xlate_configuration(); | |
1177 | * ... | |
1178 | * xlate_txn_commit(); | |
1179 | * | |
1180 | * The ovsrcu_synchronize() call here also ensures that the upcall threads | |
1181 | * retain no references to anything in the previous configuration. | |
1182 | */ | |
46c88433 | 1183 | void |
84f0f298 RW |
1184 | xlate_txn_commit(void) |
1185 | { | |
1186 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
1187 | ||
1188 | ovsrcu_set(&xcfgp, new_xcfg); | |
40a9c4c2 AW |
1189 | ovsrcu_synchronize(); |
1190 | xlate_xcfg_free(xcfg); | |
84f0f298 RW |
1191 | new_xcfg = NULL; |
1192 | } | |
1193 | ||
1194 | /* Copies the current xlate configuration in xcfgp to new_xcfg. | |
1195 | * | |
1196 | * This needs to be called prior to editing the xlate configuration. */ | |
1197 | void | |
1198 | xlate_txn_start(void) | |
1199 | { | |
1200 | struct xbridge *xbridge; | |
1201 | struct xlate_cfg *xcfg; | |
1202 | ||
1203 | ovs_assert(!new_xcfg); | |
1204 | ||
1205 | new_xcfg = xmalloc(sizeof *new_xcfg); | |
1206 | hmap_init(&new_xcfg->xbridges); | |
1207 | hmap_init(&new_xcfg->xbundles); | |
1208 | hmap_init(&new_xcfg->xports); | |
c3594cc3 | 1209 | hmap_init(&new_xcfg->xports_uuid); |
84f0f298 RW |
1210 | |
1211 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
1212 | if (!xcfg) { | |
1213 | return; | |
1214 | } | |
1215 | ||
1216 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
1217 | xlate_xbridge_copy(xbridge); | |
1218 | } | |
1219 | } | |
1220 | ||
1221 | ||
1222 | static void | |
1223 | xlate_xcfg_free(struct xlate_cfg *xcfg) | |
1224 | { | |
1225 | struct xbridge *xbridge, *next_xbridge; | |
1226 | ||
1227 | if (!xcfg) { | |
1228 | return; | |
1229 | } | |
1230 | ||
1231 | HMAP_FOR_EACH_SAFE (xbridge, next_xbridge, hmap_node, &xcfg->xbridges) { | |
1232 | xlate_xbridge_remove(xcfg, xbridge); | |
1233 | } | |
1234 | ||
1235 | hmap_destroy(&xcfg->xbridges); | |
1236 | hmap_destroy(&xcfg->xbundles); | |
1237 | hmap_destroy(&xcfg->xports); | |
c3594cc3 | 1238 | hmap_destroy(&xcfg->xports_uuid); |
84f0f298 RW |
1239 | free(xcfg); |
1240 | } | |
1241 | ||
1242 | void | |
1243 | xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, | |
34dd0d78 | 1244 | struct dpif *dpif, |
84f0f298 | 1245 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 1246 | struct rstp *rstp, const struct mcast_snooping *ms, |
84f0f298 RW |
1247 | const struct mbridge *mbridge, |
1248 | const struct dpif_sflow *sflow, | |
1249 | const struct dpif_ipfix *ipfix, | |
2f47cdf4 | 1250 | const struct netflow *netflow, |
b440dd8c JS |
1251 | bool forward_bpdu, bool has_in_band, |
1252 | const struct dpif_backer_support *support) | |
84f0f298 RW |
1253 | { |
1254 | struct xbridge *xbridge; | |
83c2757b | 1255 | struct xbridge_addr *xbridge_addr, *old_addr; |
84f0f298 RW |
1256 | |
1257 | ovs_assert(new_xcfg); | |
1258 | ||
1259 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
1260 | if (!xbridge) { | |
1261 | xbridge = xzalloc(sizeof *xbridge); | |
1262 | xbridge->ofproto = ofproto; | |
1263 | ||
1264 | xlate_xbridge_init(new_xcfg, xbridge); | |
1265 | } | |
1266 | ||
1267 | free(xbridge->name); | |
1268 | xbridge->name = xstrdup(name); | |
1269 | ||
83c2757b ZB |
1270 | xbridge_addr = xbridge_addr_create(xbridge); |
1271 | old_addr = xbridge->addr; | |
1272 | ||
34dd0d78 | 1273 | xlate_xbridge_set(xbridge, dpif, ml, stp, rstp, ms, mbridge, sflow, ipfix, |
83c2757b ZB |
1274 | netflow, forward_bpdu, has_in_band, support, |
1275 | xbridge_addr); | |
1276 | ||
1277 | if (xbridge_addr != old_addr) { | |
1278 | xbridge_addr_unref(xbridge_addr); | |
1279 | } | |
84f0f298 RW |
1280 | } |
1281 | ||
1282 | static void | |
1283 | xlate_xbridge_remove(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
46c88433 | 1284 | { |
46c88433 EJ |
1285 | struct xbundle *xbundle, *next_xbundle; |
1286 | struct xport *xport, *next_xport; | |
1287 | ||
1288 | if (!xbridge) { | |
1289 | return; | |
1290 | } | |
1291 | ||
1292 | HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) { | |
84f0f298 | 1293 | xlate_xport_remove(xcfg, xport); |
46c88433 EJ |
1294 | } |
1295 | ||
1296 | LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) { | |
84f0f298 | 1297 | xlate_xbundle_remove(xcfg, xbundle); |
46c88433 EJ |
1298 | } |
1299 | ||
84f0f298 | 1300 | hmap_remove(&xcfg->xbridges, &xbridge->hmap_node); |
795cc5c1 | 1301 | mac_learning_unref(xbridge->ml); |
6d95c4e8 | 1302 | mcast_snooping_unref(xbridge->ms); |
795cc5c1 EJ |
1303 | mbridge_unref(xbridge->mbridge); |
1304 | dpif_sflow_unref(xbridge->sflow); | |
1305 | dpif_ipfix_unref(xbridge->ipfix); | |
3570f7e4 | 1306 | netflow_unref(xbridge->netflow); |
795cc5c1 | 1307 | stp_unref(xbridge->stp); |
9efd308e | 1308 | rstp_unref(xbridge->rstp); |
83c2757b | 1309 | xbridge_addr_unref(xbridge->addr); |
795cc5c1 | 1310 | hmap_destroy(&xbridge->xports); |
46c88433 EJ |
1311 | free(xbridge->name); |
1312 | free(xbridge); | |
1313 | } | |
1314 | ||
84f0f298 RW |
1315 | void |
1316 | xlate_remove_ofproto(struct ofproto_dpif *ofproto) | |
1317 | { | |
1318 | struct xbridge *xbridge; | |
1319 | ||
1320 | ovs_assert(new_xcfg); | |
1321 | ||
1322 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
1323 | xlate_xbridge_remove(new_xcfg, xbridge); | |
1324 | } | |
1325 | ||
46c88433 EJ |
1326 | void |
1327 | xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
f0fb825a | 1328 | const char *name, enum port_vlan_mode vlan_mode, |
fed8962a EG |
1329 | uint16_t qinq_ethtype, int vlan, |
1330 | unsigned long *trunks, unsigned long *cvlans, | |
88f52d7f | 1331 | enum port_priority_tags_mode use_priority_tags, |
46c88433 | 1332 | const struct bond *bond, const struct lacp *lacp, |
c005f976 | 1333 | bool floodable, bool protected) |
46c88433 | 1334 | { |
84f0f298 | 1335 | struct xbundle *xbundle; |
46c88433 | 1336 | |
84f0f298 RW |
1337 | ovs_assert(new_xcfg); |
1338 | ||
1339 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
46c88433 EJ |
1340 | if (!xbundle) { |
1341 | xbundle = xzalloc(sizeof *xbundle); | |
1342 | xbundle->ofbundle = ofbundle; | |
84f0f298 | 1343 | xbundle->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 1344 | |
84f0f298 | 1345 | xlate_xbundle_init(new_xcfg, xbundle); |
46c88433 EJ |
1346 | } |
1347 | ||
46c88433 EJ |
1348 | free(xbundle->name); |
1349 | xbundle->name = xstrdup(name); | |
1350 | ||
fed8962a | 1351 | xlate_xbundle_set(xbundle, vlan_mode, qinq_ethtype, vlan, trunks, cvlans, |
c005f976 | 1352 | use_priority_tags, bond, lacp, floodable, protected); |
46c88433 EJ |
1353 | } |
1354 | ||
84f0f298 RW |
1355 | static void |
1356 | xlate_xbundle_remove(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
46c88433 | 1357 | { |
5f03c983 | 1358 | struct xport *xport; |
46c88433 EJ |
1359 | |
1360 | if (!xbundle) { | |
1361 | return; | |
1362 | } | |
1363 | ||
5f03c983 | 1364 | LIST_FOR_EACH_POP (xport, bundle_node, &xbundle->xports) { |
46c88433 EJ |
1365 | xport->xbundle = NULL; |
1366 | } | |
1367 | ||
84f0f298 | 1368 | hmap_remove(&xcfg->xbundles, &xbundle->hmap_node); |
417e7e66 | 1369 | ovs_list_remove(&xbundle->list_node); |
46c88433 EJ |
1370 | bond_unref(xbundle->bond); |
1371 | lacp_unref(xbundle->lacp); | |
1372 | free(xbundle->name); | |
1373 | free(xbundle); | |
1374 | } | |
1375 | ||
84f0f298 RW |
1376 | void |
1377 | xlate_bundle_remove(struct ofbundle *ofbundle) | |
1378 | { | |
1379 | struct xbundle *xbundle; | |
1380 | ||
1381 | ovs_assert(new_xcfg); | |
1382 | ||
1383 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
1384 | xlate_xbundle_remove(new_xcfg, xbundle); | |
1385 | } | |
1386 | ||
46c88433 EJ |
1387 | void |
1388 | xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
1389 | struct ofport_dpif *ofport, ofp_port_t ofp_port, | |
1390 | odp_port_t odp_port, const struct netdev *netdev, | |
1391 | const struct cfm *cfm, const struct bfd *bfd, | |
0477baa9 DF |
1392 | const struct lldp *lldp, struct ofport_dpif *peer, |
1393 | int stp_port_no, const struct rstp_port *rstp_port, | |
55954f6e | 1394 | const struct ofproto_port_queue *qdscp_list, size_t n_qdscp, |
dd8cd4b4 SH |
1395 | enum ofputil_port_config config, |
1396 | enum ofputil_port_state state, bool is_tunnel, | |
9d189a50 | 1397 | bool may_enable) |
46c88433 | 1398 | { |
55954f6e | 1399 | size_t i; |
84f0f298 RW |
1400 | struct xport *xport; |
1401 | ||
1402 | ovs_assert(new_xcfg); | |
46c88433 | 1403 | |
84f0f298 | 1404 | xport = xport_lookup(new_xcfg, ofport); |
46c88433 EJ |
1405 | if (!xport) { |
1406 | xport = xzalloc(sizeof *xport); | |
1407 | xport->ofport = ofport; | |
84f0f298 | 1408 | xport->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 1409 | xport->ofp_port = ofp_port; |
c3594cc3 | 1410 | uuid_generate(&xport->uuid); |
46c88433 | 1411 | |
84f0f298 | 1412 | xlate_xport_init(new_xcfg, xport); |
46c88433 EJ |
1413 | } |
1414 | ||
1415 | ovs_assert(xport->ofp_port == ofp_port); | |
1416 | ||
0477baa9 DF |
1417 | xlate_xport_set(xport, odp_port, netdev, cfm, bfd, lldp, |
1418 | stp_port_no, rstp_port, config, state, is_tunnel, | |
1419 | may_enable); | |
46c88433 EJ |
1420 | |
1421 | if (xport->peer) { | |
1422 | xport->peer->peer = NULL; | |
1423 | } | |
84f0f298 | 1424 | xport->peer = xport_lookup(new_xcfg, peer); |
46c88433 EJ |
1425 | if (xport->peer) { |
1426 | xport->peer->peer = xport; | |
1427 | } | |
1428 | ||
1429 | if (xport->xbundle) { | |
417e7e66 | 1430 | ovs_list_remove(&xport->bundle_node); |
46c88433 | 1431 | } |
84f0f298 | 1432 | xport->xbundle = xbundle_lookup(new_xcfg, ofbundle); |
46c88433 | 1433 | if (xport->xbundle) { |
417e7e66 | 1434 | ovs_list_insert(&xport->xbundle->xports, &xport->bundle_node); |
46c88433 | 1435 | } |
55954f6e EJ |
1436 | |
1437 | clear_skb_priorities(xport); | |
1438 | for (i = 0; i < n_qdscp; i++) { | |
1439 | struct skb_priority_to_dscp *pdscp; | |
1440 | uint32_t skb_priority; | |
1441 | ||
89a8a7f0 EJ |
1442 | if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue, |
1443 | &skb_priority)) { | |
55954f6e EJ |
1444 | continue; |
1445 | } | |
1446 | ||
1447 | pdscp = xmalloc(sizeof *pdscp); | |
1448 | pdscp->skb_priority = skb_priority; | |
1449 | pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK; | |
1450 | hmap_insert(&xport->skb_priorities, &pdscp->hmap_node, | |
1451 | hash_int(pdscp->skb_priority, 0)); | |
1452 | } | |
46c88433 EJ |
1453 | } |
1454 | ||
84f0f298 RW |
1455 | static void |
1456 | xlate_xport_remove(struct xlate_cfg *xcfg, struct xport *xport) | |
46c88433 | 1457 | { |
46c88433 EJ |
1458 | if (!xport) { |
1459 | return; | |
1460 | } | |
1461 | ||
1462 | if (xport->peer) { | |
1463 | xport->peer->peer = NULL; | |
1464 | xport->peer = NULL; | |
1465 | } | |
1466 | ||
e621a12d | 1467 | if (xport->xbundle) { |
417e7e66 | 1468 | ovs_list_remove(&xport->bundle_node); |
e621a12d EJ |
1469 | } |
1470 | ||
55954f6e EJ |
1471 | clear_skb_priorities(xport); |
1472 | hmap_destroy(&xport->skb_priorities); | |
1473 | ||
84f0f298 | 1474 | hmap_remove(&xcfg->xports, &xport->hmap_node); |
c3594cc3 | 1475 | hmap_remove(&xcfg->xports_uuid, &xport->uuid_node); |
46c88433 EJ |
1476 | hmap_remove(&xport->xbridge->xports, &xport->ofp_node); |
1477 | ||
1478 | netdev_close(xport->netdev); | |
f025bcb7 | 1479 | rstp_port_unref(xport->rstp_port); |
46c88433 EJ |
1480 | cfm_unref(xport->cfm); |
1481 | bfd_unref(xport->bfd); | |
0477baa9 | 1482 | lldp_unref(xport->lldp); |
46c88433 EJ |
1483 | free(xport); |
1484 | } | |
1485 | ||
84f0f298 RW |
1486 | void |
1487 | xlate_ofport_remove(struct ofport_dpif *ofport) | |
1488 | { | |
1489 | struct xport *xport; | |
1490 | ||
1491 | ovs_assert(new_xcfg); | |
1492 | ||
1493 | xport = xport_lookup(new_xcfg, ofport); | |
8416e50f VD |
1494 | if (xport) { |
1495 | tnl_neigh_flush(netdev_get_name(xport->netdev)); | |
1496 | } | |
84f0f298 RW |
1497 | xlate_xport_remove(new_xcfg, xport); |
1498 | } | |
1499 | ||
ef377a58 | 1500 | static struct ofproto_dpif * |
d40533fc BP |
1501 | xlate_lookup_ofproto_(const struct dpif_backer *backer, |
1502 | const struct flow *flow, | |
1503 | ofp_port_t *ofp_in_port, const struct xport **xportp, | |
1504 | char **errorp) | |
ef377a58 | 1505 | { |
e672ff9b | 1506 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
ef377a58 | 1507 | const struct xport *xport; |
f9038ef6 | 1508 | |
00135b86 ZB |
1509 | /* If packet is recirculated, xport can be retrieved from frozen state. */ |
1510 | if (flow->recirc_id) { | |
1511 | const struct recirc_id_node *recirc_id_node; | |
1512 | ||
1513 | recirc_id_node = recirc_id_node_find(flow->recirc_id); | |
1514 | ||
1515 | if (OVS_UNLIKELY(!recirc_id_node)) { | |
d40533fc BP |
1516 | if (errorp) { |
1517 | *errorp = xasprintf("no recirculation data for recirc_id " | |
1518 | "%"PRIu32, flow->recirc_id); | |
1519 | } | |
00135b86 ZB |
1520 | return NULL; |
1521 | } | |
1522 | ||
1523 | /* If recirculation was initiated due to bond (in_port = OFPP_NONE) | |
1524 | * then frozen state is static and xport_uuid is not defined, so xport | |
1525 | * cannot be restored from frozen state. */ | |
1526 | if (recirc_id_node->state.metadata.in_port != OFPP_NONE) { | |
1527 | struct uuid xport_uuid = recirc_id_node->state.xport_uuid; | |
1528 | xport = xport_lookup_by_uuid(xcfg, &xport_uuid); | |
1529 | if (xport && xport->xbridge && xport->xbridge->ofproto) { | |
1530 | goto out; | |
1531 | } | |
1532 | } | |
1533 | } | |
1534 | ||
e672ff9b JR |
1535 | xport = xport_lookup(xcfg, tnl_port_should_receive(flow) |
1536 | ? tnl_port_receive(flow) | |
1537 | : odp_port_to_ofport(backer, flow->in_port.odp_port)); | |
1538 | if (OVS_UNLIKELY(!xport)) { | |
d40533fc BP |
1539 | if (errorp) { |
1540 | *errorp = (tnl_port_should_receive(flow) | |
1541 | ? xstrdup("no OpenFlow tunnel port for this packet") | |
1542 | : xasprintf("no OpenFlow tunnel port for datapath " | |
1543 | "port %"PRIu32, flow->in_port.odp_port)); | |
1544 | } | |
e672ff9b | 1545 | return NULL; |
ef377a58 | 1546 | } |
00135b86 ZB |
1547 | |
1548 | out: | |
d40533fc BP |
1549 | if (errorp) { |
1550 | *errorp = NULL; | |
1551 | } | |
e672ff9b | 1552 | *xportp = xport; |
f9038ef6 | 1553 | if (ofp_in_port) { |
e672ff9b | 1554 | *ofp_in_port = xport->ofp_port; |
f9038ef6 | 1555 | } |
e672ff9b | 1556 | return xport->xbridge->ofproto; |
ef377a58 JR |
1557 | } |
1558 | ||
1559 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively) | |
1560 | * returns the corresponding struct ofproto_dpif and OpenFlow port number. */ | |
1561 | struct ofproto_dpif * | |
1562 | xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow, | |
d40533fc | 1563 | ofp_port_t *ofp_in_port, char **errorp) |
ef377a58 JR |
1564 | { |
1565 | const struct xport *xport; | |
1566 | ||
d40533fc | 1567 | return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport, errorp); |
ef377a58 JR |
1568 | } |
1569 | ||
cc377352 | 1570 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively), |
bfc691bb | 1571 | * optionally populates 'ofprotop' with the ofproto_dpif, 'ofp_in_port' with the |
cc377352 | 1572 | * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate |
dcc2c6cd JR |
1573 | * handles for those protocols if they're enabled. Caller may use the returned |
1574 | * pointers until quiescing, for longer term use additional references must | |
1575 | * be taken. | |
8449c4d6 | 1576 | * |
f9038ef6 | 1577 | * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto. |
ef377a58 | 1578 | */ |
8449c4d6 | 1579 | int |
5c476ea3 JR |
1580 | xlate_lookup(const struct dpif_backer *backer, const struct flow *flow, |
1581 | struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix, | |
1582 | struct dpif_sflow **sflow, struct netflow **netflow, | |
1583 | ofp_port_t *ofp_in_port) | |
8449c4d6 | 1584 | { |
ef377a58 | 1585 | struct ofproto_dpif *ofproto; |
84f0f298 | 1586 | const struct xport *xport; |
8449c4d6 | 1587 | |
d40533fc | 1588 | ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport, NULL); |
8449c4d6 | 1589 | |
f9038ef6 | 1590 | if (!ofproto) { |
cc377352 | 1591 | return ENODEV; |
8449c4d6 | 1592 | } |
8449c4d6 | 1593 | |
ef377a58 JR |
1594 | if (ofprotop) { |
1595 | *ofprotop = ofproto; | |
8449c4d6 EJ |
1596 | } |
1597 | ||
1dfdb9b3 | 1598 | if (ipfix) { |
f9038ef6 | 1599 | *ipfix = xport ? xport->xbridge->ipfix : NULL; |
1dfdb9b3 EJ |
1600 | } |
1601 | ||
1602 | if (sflow) { | |
f9038ef6 | 1603 | *sflow = xport ? xport->xbridge->sflow : NULL; |
1dfdb9b3 EJ |
1604 | } |
1605 | ||
1606 | if (netflow) { | |
f9038ef6 | 1607 | *netflow = xport ? xport->xbridge->netflow : NULL; |
1dfdb9b3 | 1608 | } |
f9038ef6 | 1609 | |
cc377352 | 1610 | return 0; |
8449c4d6 EJ |
1611 | } |
1612 | ||
46c88433 | 1613 | static struct xbridge * |
84f0f298 | 1614 | xbridge_lookup(struct xlate_cfg *xcfg, const struct ofproto_dpif *ofproto) |
46c88433 | 1615 | { |
84f0f298 | 1616 | struct hmap *xbridges; |
46c88433 EJ |
1617 | struct xbridge *xbridge; |
1618 | ||
84f0f298 | 1619 | if (!ofproto || !xcfg) { |
5e6af486 EJ |
1620 | return NULL; |
1621 | } | |
1622 | ||
84f0f298 RW |
1623 | xbridges = &xcfg->xbridges; |
1624 | ||
46c88433 | 1625 | HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0), |
84f0f298 | 1626 | xbridges) { |
46c88433 EJ |
1627 | if (xbridge->ofproto == ofproto) { |
1628 | return xbridge; | |
1629 | } | |
1630 | } | |
1631 | return NULL; | |
1632 | } | |
1633 | ||
290835f9 BP |
1634 | static struct xbridge * |
1635 | xbridge_lookup_by_uuid(struct xlate_cfg *xcfg, const struct uuid *uuid) | |
1636 | { | |
1637 | struct xbridge *xbridge; | |
1638 | ||
1639 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
07a3cd5c | 1640 | if (uuid_equals(&xbridge->ofproto->uuid, uuid)) { |
290835f9 BP |
1641 | return xbridge; |
1642 | } | |
1643 | } | |
1644 | return NULL; | |
1645 | } | |
1646 | ||
46c88433 | 1647 | static struct xbundle * |
84f0f298 | 1648 | xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle) |
46c88433 | 1649 | { |
84f0f298 | 1650 | struct hmap *xbundles; |
46c88433 EJ |
1651 | struct xbundle *xbundle; |
1652 | ||
84f0f298 | 1653 | if (!ofbundle || !xcfg) { |
5e6af486 EJ |
1654 | return NULL; |
1655 | } | |
1656 | ||
84f0f298 RW |
1657 | xbundles = &xcfg->xbundles; |
1658 | ||
46c88433 | 1659 | HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), |
84f0f298 | 1660 | xbundles) { |
46c88433 EJ |
1661 | if (xbundle->ofbundle == ofbundle) { |
1662 | return xbundle; | |
1663 | } | |
1664 | } | |
1665 | return NULL; | |
1666 | } | |
1667 | ||
1668 | static struct xport * | |
84f0f298 | 1669 | xport_lookup(struct xlate_cfg *xcfg, const struct ofport_dpif *ofport) |
46c88433 | 1670 | { |
84f0f298 | 1671 | struct hmap *xports; |
46c88433 EJ |
1672 | struct xport *xport; |
1673 | ||
84f0f298 | 1674 | if (!ofport || !xcfg) { |
5e6af486 EJ |
1675 | return NULL; |
1676 | } | |
1677 | ||
84f0f298 RW |
1678 | xports = &xcfg->xports; |
1679 | ||
46c88433 | 1680 | HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0), |
84f0f298 | 1681 | xports) { |
46c88433 EJ |
1682 | if (xport->ofport == ofport) { |
1683 | return xport; | |
1684 | } | |
1685 | } | |
1686 | return NULL; | |
1687 | } | |
1688 | ||
00135b86 ZB |
1689 | static struct xport * |
1690 | xport_lookup_by_uuid(struct xlate_cfg *xcfg, const struct uuid *uuid) | |
1691 | { | |
1692 | struct hmap *xports; | |
1693 | struct xport *xport; | |
1694 | ||
1695 | if (uuid_is_zero(uuid) || !xcfg) { | |
1696 | return NULL; | |
1697 | } | |
1698 | ||
1699 | xports = &xcfg->xports_uuid; | |
1700 | ||
1701 | HMAP_FOR_EACH_IN_BUCKET (xport, uuid_node, uuid_hash(uuid), xports) { | |
1702 | if (uuid_equals(&xport->uuid, uuid)) { | |
1703 | return xport; | |
1704 | } | |
1705 | } | |
1706 | return NULL; | |
1707 | } | |
1708 | ||
40085e56 EJ |
1709 | static struct stp_port * |
1710 | xport_get_stp_port(const struct xport *xport) | |
1711 | { | |
92cf817b | 1712 | return xport->xbridge->stp && xport->stp_port_no != -1 |
40085e56 EJ |
1713 | ? stp_get_port(xport->xbridge->stp, xport->stp_port_no) |
1714 | : NULL; | |
1715 | } | |
9d189a50 | 1716 | |
0d1cee12 | 1717 | static bool |
9d189a50 EJ |
1718 | xport_stp_learn_state(const struct xport *xport) |
1719 | { | |
40085e56 | 1720 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1721 | return sp |
1722 | ? stp_learn_in_state(stp_port_get_state(sp)) | |
1723 | : true; | |
9d189a50 EJ |
1724 | } |
1725 | ||
1726 | static bool | |
1727 | xport_stp_forward_state(const struct xport *xport) | |
1728 | { | |
40085e56 | 1729 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1730 | return sp |
1731 | ? stp_forward_in_state(stp_port_get_state(sp)) | |
1732 | : true; | |
9d189a50 EJ |
1733 | } |
1734 | ||
0d1cee12 | 1735 | static bool |
bacdb85a | 1736 | xport_stp_should_forward_bpdu(const struct xport *xport) |
0d1cee12 K |
1737 | { |
1738 | struct stp_port *sp = xport_get_stp_port(xport); | |
bacdb85a | 1739 | return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED); |
0d1cee12 K |
1740 | } |
1741 | ||
9d189a50 EJ |
1742 | /* Returns true if STP should process 'flow'. Sets fields in 'wc' that |
1743 | * were used to make the determination.*/ | |
1744 | static bool | |
1745 | stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc) | |
1746 | { | |
bbbca389 | 1747 | /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */ |
9d189a50 | 1748 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
bbbca389 | 1749 | return is_stp(flow); |
9d189a50 EJ |
1750 | } |
1751 | ||
1752 | static void | |
cf62fa4c | 1753 | stp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9d189a50 | 1754 | { |
40085e56 | 1755 | struct stp_port *sp = xport_get_stp_port(xport); |
cf62fa4c PS |
1756 | struct dp_packet payload = *packet; |
1757 | struct eth_header *eth = dp_packet_data(&payload); | |
9d189a50 EJ |
1758 | |
1759 | /* Sink packets on ports that have STP disabled when the bridge has | |
1760 | * STP enabled. */ | |
1761 | if (!sp || stp_port_get_state(sp) == STP_DISABLED) { | |
1762 | return; | |
1763 | } | |
1764 | ||
1765 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1766 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1767 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9d189a50 EJ |
1768 | } |
1769 | ||
cf62fa4c PS |
1770 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1771 | stp_received_bpdu(sp, dp_packet_data(&payload), dp_packet_size(&payload)); | |
9d189a50 EJ |
1772 | } |
1773 | } | |
1774 | ||
f025bcb7 JR |
1775 | static enum rstp_state |
1776 | xport_get_rstp_port_state(const struct xport *xport) | |
9efd308e | 1777 | { |
f025bcb7 JR |
1778 | return xport->rstp_port |
1779 | ? rstp_port_get_state(xport->rstp_port) | |
1780 | : RSTP_DISABLED; | |
9efd308e DV |
1781 | } |
1782 | ||
1783 | static bool | |
1784 | xport_rstp_learn_state(const struct xport *xport) | |
1785 | { | |
4b5f1996 DV |
1786 | return xport->xbridge->rstp && xport->rstp_port |
1787 | ? rstp_learn_in_state(xport_get_rstp_port_state(xport)) | |
1788 | : true; | |
9efd308e DV |
1789 | } |
1790 | ||
1791 | static bool | |
1792 | xport_rstp_forward_state(const struct xport *xport) | |
1793 | { | |
4b5f1996 DV |
1794 | return xport->xbridge->rstp && xport->rstp_port |
1795 | ? rstp_forward_in_state(xport_get_rstp_port_state(xport)) | |
1796 | : true; | |
9efd308e DV |
1797 | } |
1798 | ||
1799 | static bool | |
1800 | xport_rstp_should_manage_bpdu(const struct xport *xport) | |
1801 | { | |
f025bcb7 | 1802 | return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport)); |
9efd308e DV |
1803 | } |
1804 | ||
1805 | static void | |
cf62fa4c | 1806 | rstp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9efd308e | 1807 | { |
cf62fa4c PS |
1808 | struct dp_packet payload = *packet; |
1809 | struct eth_header *eth = dp_packet_data(&payload); | |
9efd308e | 1810 | |
f025bcb7 JR |
1811 | /* Sink packets on ports that have no RSTP. */ |
1812 | if (!xport->rstp_port) { | |
9efd308e DV |
1813 | return; |
1814 | } | |
1815 | ||
1816 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1817 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1818 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9efd308e DV |
1819 | } |
1820 | ||
f2f78d5c MM |
1821 | int len = ETH_HEADER_LEN + LLC_HEADER_LEN; |
1822 | if (eth->eth_type == htons(ETH_TYPE_VLAN)) { | |
1823 | len += VLAN_HEADER_LEN; | |
1824 | } | |
1825 | if (dp_packet_try_pull(&payload, len)) { | |
cf62fa4c PS |
1826 | rstp_port_received_bpdu(xport->rstp_port, dp_packet_data(&payload), |
1827 | dp_packet_size(&payload)); | |
9efd308e DV |
1828 | } |
1829 | } | |
1830 | ||
46c88433 EJ |
1831 | static struct xport * |
1832 | get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1833 | { | |
1834 | struct xport *xport; | |
1835 | ||
1836 | HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port), | |
1837 | &xbridge->xports) { | |
1838 | if (xport->ofp_port == ofp_port) { | |
1839 | return xport; | |
1840 | } | |
1841 | } | |
1842 | return NULL; | |
1843 | } | |
1844 | ||
1845 | static odp_port_t | |
1846 | ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1847 | { | |
1848 | const struct xport *xport = get_ofp_port(xbridge, ofp_port); | |
1849 | return xport ? xport->odp_port : ODPP_NONE; | |
1850 | } | |
1851 | ||
dd8cd4b4 SH |
1852 | static bool |
1853 | odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port) | |
1854 | { | |
086fa873 BP |
1855 | struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1856 | return xport && xport->may_enable; | |
dd8cd4b4 SH |
1857 | } |
1858 | ||
1e684d7d | 1859 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1860 | group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *, |
1861 | int depth); | |
1862 | ||
1863 | static bool | |
1864 | group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) | |
1865 | { | |
1866 | struct group_dpif *group; | |
dd8cd4b4 | 1867 | |
5d08a275 | 1868 | group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, |
1f4a8933 | 1869 | ctx->xin->tables_version, false); |
db88b35c | 1870 | if (group) { |
76973237 | 1871 | return group_first_live_bucket(ctx, group, depth) != NULL; |
dc25893e | 1872 | } |
dd8cd4b4 | 1873 | |
dc25893e | 1874 | return false; |
dd8cd4b4 SH |
1875 | } |
1876 | ||
1877 | #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ | |
1878 | ||
1879 | static bool | |
1880 | bucket_is_alive(const struct xlate_ctx *ctx, | |
1e684d7d | 1881 | struct ofputil_bucket *bucket, int depth) |
dd8cd4b4 SH |
1882 | { |
1883 | if (depth >= MAX_LIVENESS_RECURSION) { | |
2d9b49dd BP |
1884 | xlate_report_error(ctx, "bucket chaining exceeded %d links", |
1885 | MAX_LIVENESS_RECURSION); | |
dd8cd4b4 SH |
1886 | return false; |
1887 | } | |
1888 | ||
fdb1999b AZ |
1889 | return (!ofputil_bucket_has_liveness(bucket) |
1890 | || (bucket->watch_port != OFPP_ANY | |
1891 | && odp_port_is_alive(ctx, bucket->watch_port)) | |
1892 | || (bucket->watch_group != OFPG_ANY | |
1893 | && group_is_alive(ctx, bucket->watch_group, depth + 1))); | |
dd8cd4b4 SH |
1894 | } |
1895 | ||
fac4786a BP |
1896 | static void |
1897 | xlate_report_bucket_not_live(const struct xlate_ctx *ctx, | |
1898 | const struct ofputil_bucket *bucket) | |
1899 | { | |
1900 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
1901 | struct ds s = DS_EMPTY_INITIALIZER; | |
1902 | if (bucket->watch_port != OFPP_ANY) { | |
1903 | ds_put_cstr(&s, "port "); | |
1904 | ofputil_format_port(bucket->watch_port, NULL, &s); | |
1905 | } | |
1906 | if (bucket->watch_group != OFPG_ANY) { | |
1907 | if (s.length) { | |
1908 | ds_put_cstr(&s, " and "); | |
1909 | } | |
1910 | ds_put_format(&s, "port %"PRIu32, bucket->watch_group); | |
1911 | } | |
1912 | ||
1913 | xlate_report(ctx, OFT_DETAIL, "bucket %"PRIu32": not live due to %s", | |
1914 | bucket->bucket_id, ds_cstr(&s)); | |
1915 | ||
1916 | ds_destroy(&s); | |
1917 | } | |
1918 | } | |
1919 | ||
1e684d7d | 1920 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1921 | group_first_live_bucket(const struct xlate_ctx *ctx, |
1922 | const struct group_dpif *group, int depth) | |
1923 | { | |
1924 | struct ofputil_bucket *bucket; | |
07a3cd5c | 1925 | LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { |
dd8cd4b4 SH |
1926 | if (bucket_is_alive(ctx, bucket, depth)) { |
1927 | return bucket; | |
1928 | } | |
fac4786a | 1929 | xlate_report_bucket_not_live(ctx, bucket); |
dd8cd4b4 SH |
1930 | } |
1931 | ||
1932 | return NULL; | |
1933 | } | |
1934 | ||
1e684d7d | 1935 | static struct ofputil_bucket * |
fe7e5749 SH |
1936 | group_best_live_bucket(const struct xlate_ctx *ctx, |
1937 | const struct group_dpif *group, | |
1938 | uint32_t basis) | |
1939 | { | |
1e684d7d | 1940 | struct ofputil_bucket *best_bucket = NULL; |
fe7e5749 | 1941 | uint32_t best_score = 0; |
fe7e5749 | 1942 | |
1e684d7d | 1943 | struct ofputil_bucket *bucket; |
07a3cd5c | 1944 | LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { |
fe7e5749 | 1945 | if (bucket_is_alive(ctx, bucket, 0)) { |
c09cb861 LS |
1946 | uint32_t score = |
1947 | (hash_int(bucket->bucket_id, basis) & 0xffff) * bucket->weight; | |
fe7e5749 SH |
1948 | if (score >= best_score) { |
1949 | best_bucket = bucket; | |
1950 | best_score = score; | |
1951 | } | |
fac4786a BP |
1952 | xlate_report(ctx, OFT_DETAIL, "bucket %"PRIu32": score %"PRIu32, |
1953 | bucket->bucket_id, score); | |
1954 | } else { | |
1955 | xlate_report_bucket_not_live(ctx, bucket); | |
fe7e5749 | 1956 | } |
fe7e5749 SH |
1957 | } |
1958 | ||
1959 | return best_bucket; | |
1960 | } | |
1961 | ||
9583bc14 | 1962 | static bool |
46c88433 | 1963 | xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan) |
9583bc14 EJ |
1964 | { |
1965 | return (bundle->vlan_mode != PORT_VLAN_ACCESS | |
1966 | && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan))); | |
1967 | } | |
1968 | ||
fed8962a EG |
1969 | static bool |
1970 | xbundle_allows_cvlan(const struct xbundle *bundle, uint16_t vlan) | |
1971 | { | |
1972 | return (!bundle->cvlans || bitmap_is_set(bundle->cvlans, vlan)); | |
1973 | } | |
1974 | ||
9583bc14 | 1975 | static bool |
f0fb825a | 1976 | xbundle_includes_vlan(const struct xbundle *xbundle, const struct xvlan *xvlan) |
46c88433 | 1977 | { |
f0fb825a EG |
1978 | switch (xbundle->vlan_mode) { |
1979 | case PORT_VLAN_ACCESS: | |
1980 | return xvlan->v[0].vid == xbundle->vlan && xvlan->v[1].vid == 0; | |
1981 | ||
1982 | case PORT_VLAN_TRUNK: | |
1983 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1984 | case PORT_VLAN_NATIVE_TAGGED: | |
1985 | return xbundle_trunks_vlan(xbundle, xvlan->v[0].vid); | |
1986 | ||
fed8962a EG |
1987 | case PORT_VLAN_DOT1Q_TUNNEL: |
1988 | return xvlan->v[0].vid == xbundle->vlan && | |
1989 | xbundle_allows_cvlan(xbundle, xvlan->v[1].vid); | |
1990 | ||
f0fb825a EG |
1991 | default: |
1992 | OVS_NOT_REACHED(); | |
1993 | } | |
46c88433 EJ |
1994 | } |
1995 | ||
1996 | static mirror_mask_t | |
1997 | xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle) | |
1998 | { | |
1999 | return xbundle != &ofpp_none_bundle | |
2000 | ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle) | |
2001 | : 0; | |
2002 | } | |
2003 | ||
2004 | static mirror_mask_t | |
2005 | xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 2006 | { |
46c88433 EJ |
2007 | return xbundle != &ofpp_none_bundle |
2008 | ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle) | |
2009 | : 0; | |
9583bc14 EJ |
2010 | } |
2011 | ||
46c88433 EJ |
2012 | static mirror_mask_t |
2013 | xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 2014 | { |
46c88433 EJ |
2015 | return xbundle != &ofpp_none_bundle |
2016 | ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle) | |
2017 | : 0; | |
2018 | } | |
2019 | ||
2020 | static struct xbundle * | |
2d9b49dd BP |
2021 | lookup_input_bundle__(const struct xbridge *xbridge, |
2022 | ofp_port_t in_port, struct xport **in_xportp) | |
46c88433 EJ |
2023 | { |
2024 | struct xport *xport; | |
9583bc14 EJ |
2025 | |
2026 | /* Find the port and bundle for the received packet. */ | |
46c88433 EJ |
2027 | xport = get_ofp_port(xbridge, in_port); |
2028 | if (in_xportp) { | |
2029 | *in_xportp = xport; | |
9583bc14 | 2030 | } |
46c88433 EJ |
2031 | if (xport && xport->xbundle) { |
2032 | return xport->xbundle; | |
9583bc14 EJ |
2033 | } |
2034 | ||
6362203b YT |
2035 | /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+), |
2036 | * which a controller may use as the ingress port for traffic that | |
2037 | * it is sourcing. */ | |
2038 | if (in_port == OFPP_CONTROLLER || in_port == OFPP_NONE) { | |
9583bc14 EJ |
2039 | return &ofpp_none_bundle; |
2040 | } | |
2d9b49dd BP |
2041 | return NULL; |
2042 | } | |
9583bc14 | 2043 | |
2d9b49dd BP |
2044 | static struct xbundle * |
2045 | lookup_input_bundle(const struct xlate_ctx *ctx, | |
2046 | ofp_port_t in_port, struct xport **in_xportp) | |
2047 | { | |
2048 | struct xbundle *xbundle = lookup_input_bundle__(ctx->xbridge, | |
2049 | in_port, in_xportp); | |
2050 | if (!xbundle) { | |
2051 | /* Odd. A few possible reasons here: | |
2052 | * | |
2053 | * - We deleted a port but there are still a few packets queued up | |
2054 | * from it. | |
2055 | * | |
2056 | * - Someone externally added a port (e.g. "ovs-dpctl add-if") that | |
2057 | * we don't know about. | |
2058 | * | |
2059 | * - The ofproto client didn't configure the port as part of a bundle. | |
2060 | * This is particularly likely to happen if a packet was received on | |
2061 | * the port after it was created, but before the client had a chance | |
2062 | * to configure its bundle. | |
2063 | */ | |
94783c7c | 2064 | xlate_report_error(ctx, "received packet on unknown port %"PRIu32, |
2d9b49dd | 2065 | in_port); |
9583bc14 | 2066 | } |
2d9b49dd | 2067 | return xbundle; |
9583bc14 EJ |
2068 | } |
2069 | ||
faa624b4 BP |
2070 | /* Mirrors the packet represented by 'ctx' to appropriate mirror destinations, |
2071 | * given the packet is ingressing or egressing on 'xbundle', which has ingress | |
2072 | * or egress (as appropriate) mirrors 'mirrors'. */ | |
9583bc14 | 2073 | static void |
7efbc3b7 BP |
2074 | mirror_packet(struct xlate_ctx *ctx, struct xbundle *xbundle, |
2075 | mirror_mask_t mirrors) | |
9583bc14 | 2076 | { |
f0fb825a EG |
2077 | struct xvlan in_xvlan; |
2078 | struct xvlan xvlan; | |
2079 | ||
faa624b4 BP |
2080 | /* Figure out what VLAN the packet is in (because mirrors can select |
2081 | * packets on basis of VLAN). */ | |
f0fb825a EG |
2082 | xvlan_extract(&ctx->xin->flow, &in_xvlan); |
2083 | if (!input_vid_is_valid(ctx, in_xvlan.v[0].vid, xbundle)) { | |
9583bc14 EJ |
2084 | return; |
2085 | } | |
f0fb825a | 2086 | xvlan_input_translate(xbundle, &in_xvlan, &xvlan); |
9583bc14 | 2087 | |
7efbc3b7 | 2088 | const struct xbridge *xbridge = ctx->xbridge; |
9583bc14 | 2089 | |
7efbc3b7 BP |
2090 | /* Don't mirror to destinations that we've already mirrored to. */ |
2091 | mirrors &= ~ctx->mirrors; | |
9583bc14 EJ |
2092 | if (!mirrors) { |
2093 | return; | |
2094 | } | |
2095 | ||
a3954fd8 BP |
2096 | /* 'mirrors' is a bit-mask of candidates for mirroring. Iterate through |
2097 | * the candidates, adding the ones that really should be mirrored to | |
2098 | * 'used_mirrors', as long as some candidates remain. */ | |
2099 | mirror_mask_t used_mirrors = 0; | |
9583bc14 | 2100 | while (mirrors) { |
7efbc3b7 | 2101 | const unsigned long *vlans; |
ec7ceaed EJ |
2102 | mirror_mask_t dup_mirrors; |
2103 | struct ofbundle *out; | |
ec7ceaed | 2104 | int out_vlan; |
1356dbd1 | 2105 | int snaplen; |
ec7ceaed | 2106 | |
faa624b4 | 2107 | /* Get the details of the mirror represented by the rightmost 1-bit. */ |
500db308 BP |
2108 | ovs_assert(mirror_get(xbridge->mbridge, raw_ctz(mirrors), |
2109 | &vlans, &dup_mirrors, | |
2110 | &out, &snaplen, &out_vlan)); | |
ec7ceaed | 2111 | |
1356dbd1 | 2112 | |
faa624b4 BP |
2113 | /* If this mirror selects on the basis of VLAN, and it does not select |
2114 | * 'vlan', then discard this mirror and go on to the next one. */ | |
ec7ceaed | 2115 | if (vlans) { |
f0fb825a | 2116 | ctx->wc->masks.vlans[0].tci |= htons(VLAN_CFI | VLAN_VID_MASK); |
9583bc14 | 2117 | } |
f0fb825a | 2118 | if (vlans && !bitmap_is_set(vlans, xvlan.v[0].vid)) { |
9583bc14 EJ |
2119 | mirrors = zero_rightmost_1bit(mirrors); |
2120 | continue; | |
2121 | } | |
2122 | ||
a3954fd8 BP |
2123 | /* We sent a packet to this mirror. */ |
2124 | used_mirrors |= rightmost_1bit(mirrors); | |
2125 | ||
faa624b4 BP |
2126 | /* Record the mirror, and the mirrors that output to the same |
2127 | * destination, so that we don't mirror to them again. This must be | |
2128 | * done now to ensure that output_normal(), below, doesn't recursively | |
2129 | * output to the same mirrors. */ | |
3d6151f3 | 2130 | ctx->mirrors |= dup_mirrors; |
1356dbd1 | 2131 | ctx->mirror_snaplen = snaplen; |
faa624b4 BP |
2132 | |
2133 | /* Send the packet to the mirror. */ | |
ec7ceaed | 2134 | if (out) { |
0506f184 | 2135 | struct xbundle *out_xbundle = xbundle_lookup(ctx->xcfg, out); |
46c88433 | 2136 | if (out_xbundle) { |
f0fb825a | 2137 | output_normal(ctx, out_xbundle, &xvlan); |
46c88433 | 2138 | } |
f0fb825a | 2139 | } else if (xvlan.v[0].vid != out_vlan |
7efbc3b7 | 2140 | && !eth_addr_is_reserved(ctx->xin->flow.dl_dst)) { |
71f21279 | 2141 | struct xbundle *xb; |
f0fb825a | 2142 | uint16_t old_vid = xvlan.v[0].vid; |
9583bc14 | 2143 | |
f0fb825a | 2144 | xvlan.v[0].vid = out_vlan; |
71f21279 BP |
2145 | LIST_FOR_EACH (xb, list_node, &xbridge->xbundles) { |
2146 | if (xbundle_includes_vlan(xb, &xvlan) | |
2147 | && !xbundle_mirror_out(xbridge, xb)) { | |
2148 | output_normal(ctx, xb, &xvlan); | |
9583bc14 EJ |
2149 | } |
2150 | } | |
f0fb825a | 2151 | xvlan.v[0].vid = old_vid; |
9583bc14 | 2152 | } |
faa624b4 BP |
2153 | |
2154 | /* output_normal() could have recursively output (to different | |
2155 | * mirrors), so make sure that we don't send duplicates. */ | |
2156 | mirrors &= ~ctx->mirrors; | |
1356dbd1 | 2157 | ctx->mirror_snaplen = 0; |
9583bc14 | 2158 | } |
a3954fd8 BP |
2159 | |
2160 | if (used_mirrors) { | |
2161 | if (ctx->xin->resubmit_stats) { | |
2162 | mirror_update_stats(xbridge->mbridge, used_mirrors, | |
2163 | ctx->xin->resubmit_stats->n_packets, | |
2164 | ctx->xin->resubmit_stats->n_bytes); | |
2165 | } | |
2166 | if (ctx->xin->xcache) { | |
2167 | struct xc_entry *entry; | |
2168 | ||
2169 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_MIRROR); | |
2170 | entry->mirror.mbridge = mbridge_ref(xbridge->mbridge); | |
2171 | entry->mirror.mirrors = used_mirrors; | |
2172 | } | |
2173 | } | |
9583bc14 EJ |
2174 | } |
2175 | ||
7efbc3b7 BP |
2176 | static void |
2177 | mirror_ingress_packet(struct xlate_ctx *ctx) | |
2178 | { | |
2179 | if (mbridge_has_mirrors(ctx->xbridge->mbridge)) { | |
7efbc3b7 | 2180 | struct xbundle *xbundle = lookup_input_bundle( |
2d9b49dd | 2181 | ctx, ctx->xin->flow.in_port.ofp_port, NULL); |
7efbc3b7 BP |
2182 | if (xbundle) { |
2183 | mirror_packet(ctx, xbundle, | |
2184 | xbundle_mirror_src(ctx->xbridge, xbundle)); | |
2185 | } | |
2186 | } | |
2187 | } | |
2188 | ||
46c88433 | 2189 | /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'. |
2d9b49dd | 2190 | * If so, returns true. Otherwise, returns false. |
9583bc14 EJ |
2191 | * |
2192 | * 'vid' should be the VID obtained from the 802.1Q header that was received as | |
2193 | * part of a packet (specify 0 if there was no 802.1Q header), in the range | |
2194 | * 0...4095. */ | |
2195 | static bool | |
2d9b49dd BP |
2196 | input_vid_is_valid(const struct xlate_ctx *ctx, |
2197 | uint16_t vid, struct xbundle *in_xbundle) | |
9583bc14 EJ |
2198 | { |
2199 | /* Allow any VID on the OFPP_NONE port. */ | |
46c88433 | 2200 | if (in_xbundle == &ofpp_none_bundle) { |
9583bc14 EJ |
2201 | return true; |
2202 | } | |
2203 | ||
46c88433 | 2204 | switch (in_xbundle->vlan_mode) { |
9583bc14 EJ |
2205 | case PORT_VLAN_ACCESS: |
2206 | if (vid) { | |
2d9b49dd BP |
2207 | xlate_report_error(ctx, "dropping VLAN %"PRIu16" tagged " |
2208 | "packet received on port %s configured as VLAN " | |
fd13c6b5 | 2209 | "%d access port", vid, in_xbundle->name, |
2d9b49dd | 2210 | in_xbundle->vlan); |
9583bc14 EJ |
2211 | return false; |
2212 | } | |
2213 | return true; | |
2214 | ||
2215 | case PORT_VLAN_NATIVE_UNTAGGED: | |
2216 | case PORT_VLAN_NATIVE_TAGGED: | |
2217 | if (!vid) { | |
2218 | /* Port must always carry its native VLAN. */ | |
2219 | return true; | |
2220 | } | |
2221 | /* Fall through. */ | |
2222 | case PORT_VLAN_TRUNK: | |
f0fb825a | 2223 | if (!xbundle_trunks_vlan(in_xbundle, vid)) { |
2d9b49dd BP |
2224 | xlate_report_error(ctx, "dropping VLAN %"PRIu16" packet " |
2225 | "received on port %s not configured for " | |
2226 | "trunking VLAN %"PRIu16, | |
2227 | vid, in_xbundle->name, vid); | |
9583bc14 EJ |
2228 | return false; |
2229 | } | |
2230 | return true; | |
2231 | ||
fed8962a EG |
2232 | case PORT_VLAN_DOT1Q_TUNNEL: |
2233 | if (!xbundle_allows_cvlan(in_xbundle, vid)) { | |
2234 | xlate_report_error(ctx, "dropping VLAN %"PRIu16" packet received " | |
2235 | "on dot1q-tunnel port %s that excludes this " | |
2236 | "VLAN", vid, in_xbundle->name); | |
2237 | return false; | |
2238 | } | |
2239 | return true; | |
2240 | ||
9583bc14 | 2241 | default: |
428b2edd | 2242 | OVS_NOT_REACHED(); |
9583bc14 EJ |
2243 | } |
2244 | ||
2245 | } | |
2246 | ||
f0fb825a EG |
2247 | static void |
2248 | xvlan_copy(struct xvlan *dst, const struct xvlan *src) | |
2249 | { | |
2250 | *dst = *src; | |
2251 | } | |
2252 | ||
2253 | static void | |
2254 | xvlan_pop(struct xvlan *src) | |
2255 | { | |
2256 | memmove(&src->v[0], &src->v[1], sizeof(src->v) - sizeof(src->v[0])); | |
2257 | memset(&src->v[FLOW_MAX_VLAN_HEADERS - 1], 0, | |
2258 | sizeof(src->v[FLOW_MAX_VLAN_HEADERS - 1])); | |
2259 | } | |
2260 | ||
fed8962a EG |
2261 | static void |
2262 | xvlan_push_uninit(struct xvlan *src) | |
2263 | { | |
2264 | memmove(&src->v[1], &src->v[0], sizeof(src->v) - sizeof(src->v[0])); | |
2265 | memset(&src->v[0], 0, sizeof(src->v[0])); | |
2266 | } | |
2267 | ||
f0fb825a EG |
2268 | /* Extract VLAN information (headers) from flow */ |
2269 | static void | |
2270 | xvlan_extract(const struct flow *flow, struct xvlan *xvlan) | |
2271 | { | |
2272 | int i; | |
2273 | memset(xvlan, 0, sizeof(*xvlan)); | |
2274 | for (i = 0; i < FLOW_MAX_VLAN_HEADERS; i++) { | |
2275 | if (!eth_type_vlan(flow->vlans[i].tpid) || | |
2276 | !(flow->vlans[i].tci & htons(VLAN_CFI))) { | |
2277 | break; | |
2278 | } | |
2279 | xvlan->v[i].tpid = ntohs(flow->vlans[i].tpid); | |
2280 | xvlan->v[i].vid = vlan_tci_to_vid(flow->vlans[i].tci); | |
2281 | xvlan->v[i].pcp = ntohs(flow->vlans[i].tci) & VLAN_PCP_MASK; | |
2282 | } | |
2283 | } | |
2284 | ||
2285 | /* Put VLAN information (headers) to flow */ | |
2286 | static void | |
ffbe41db EB |
2287 | xvlan_put(struct flow *flow, const struct xvlan *xvlan, |
2288 | enum port_priority_tags_mode use_priority_tags) | |
f0fb825a EG |
2289 | { |
2290 | ovs_be16 tci; | |
2291 | int i; | |
2292 | for (i = 0; i < FLOW_MAX_VLAN_HEADERS; i++) { | |
2293 | tci = htons(xvlan->v[i].vid | (xvlan->v[i].pcp & VLAN_PCP_MASK)); | |
ffbe41db EB |
2294 | if (tci || ((use_priority_tags == PORT_PRIORITY_TAGS_ALWAYS) && |
2295 | xvlan->v[i].tpid)) { | |
f0fb825a EG |
2296 | tci |= htons(VLAN_CFI); |
2297 | flow->vlans[i].tpid = xvlan->v[i].tpid ? | |
2298 | htons(xvlan->v[i].tpid) : | |
2299 | htons(ETH_TYPE_VLAN_8021Q); | |
2300 | } | |
2301 | flow->vlans[i].tci = tci; | |
2302 | } | |
2303 | } | |
2304 | ||
2305 | /* Given 'in_xvlan', extracted from the input 802.1Q headers received as part | |
2306 | * of a packet, and 'in_xbundle', the bundle on which the packet was received, | |
2307 | * returns the VLANs of the packet during bridge internal processing. */ | |
2308 | static void | |
2309 | xvlan_input_translate(const struct xbundle *in_xbundle, | |
2310 | const struct xvlan *in_xvlan, struct xvlan *xvlan) | |
2311 | { | |
2312 | ||
2313 | switch (in_xbundle->vlan_mode) { | |
2314 | case PORT_VLAN_ACCESS: | |
2315 | memset(xvlan, 0, sizeof(*xvlan)); | |
2316 | xvlan->v[0].tpid = in_xvlan->v[0].tpid ? in_xvlan->v[0].tpid : | |
2317 | ETH_TYPE_VLAN_8021Q; | |
2318 | xvlan->v[0].vid = in_xbundle->vlan; | |
2319 | xvlan->v[0].pcp = in_xvlan->v[0].pcp; | |
2320 | break; | |
2321 | ||
2322 | case PORT_VLAN_TRUNK: | |
2323 | xvlan_copy(xvlan, in_xvlan); | |
2324 | break; | |
2325 | ||
2326 | case PORT_VLAN_NATIVE_UNTAGGED: | |
2327 | case PORT_VLAN_NATIVE_TAGGED: | |
2328 | xvlan_copy(xvlan, in_xvlan); | |
2329 | if (!in_xvlan->v[0].vid) { | |
2330 | xvlan->v[0].tpid = in_xvlan->v[0].tpid ? in_xvlan->v[0].tpid : | |
2331 | ETH_TYPE_VLAN_8021Q; | |
2332 | xvlan->v[0].vid = in_xbundle->vlan; | |
2333 | xvlan->v[0].pcp = in_xvlan->v[0].pcp; | |
2334 | } | |
2335 | break; | |
2336 | ||
fed8962a EG |
2337 | case PORT_VLAN_DOT1Q_TUNNEL: |
2338 | xvlan_copy(xvlan, in_xvlan); | |
2339 | xvlan_push_uninit(xvlan); | |
2340 | xvlan->v[0].tpid = in_xbundle->qinq_ethtype; | |
2341 | xvlan->v[0].vid = in_xbundle->vlan; | |
2342 | xvlan->v[0].pcp = 0; | |
2343 | break; | |
2344 | ||
f0fb825a EG |
2345 | default: |
2346 | OVS_NOT_REACHED(); | |
2347 | } | |
2348 | } | |
2349 | ||
2350 | /* Given 'xvlan', the VLANs of a packet during internal processing, and | |
2351 | * 'out_xbundle', a bundle on which the packet is to be output, returns the | |
2352 | * VLANs that should be included in output packet. */ | |
2353 | static void | |
2354 | xvlan_output_translate(const struct xbundle *out_xbundle, | |
2355 | const struct xvlan *xvlan, struct xvlan *out_xvlan) | |
9583bc14 | 2356 | { |
46c88433 | 2357 | switch (out_xbundle->vlan_mode) { |
9583bc14 | 2358 | case PORT_VLAN_ACCESS: |
f0fb825a EG |
2359 | memset(out_xvlan, 0, sizeof(*out_xvlan)); |
2360 | break; | |
9583bc14 EJ |
2361 | |
2362 | case PORT_VLAN_TRUNK: | |
2363 | case PORT_VLAN_NATIVE_TAGGED: | |
f0fb825a EG |
2364 | xvlan_copy(out_xvlan, xvlan); |
2365 | break; | |
9583bc14 EJ |
2366 | |
2367 | case PORT_VLAN_NATIVE_UNTAGGED: | |
f0fb825a EG |
2368 | xvlan_copy(out_xvlan, xvlan); |
2369 | if (xvlan->v[0].vid == out_xbundle->vlan) { | |
2370 | xvlan_pop(out_xvlan); | |
2371 | } | |
2372 | break; | |
9583bc14 | 2373 | |
fed8962a EG |
2374 | case PORT_VLAN_DOT1Q_TUNNEL: |
2375 | xvlan_copy(out_xvlan, xvlan); | |
2376 | xvlan_pop(out_xvlan); | |
2377 | break; | |
2378 | ||
9583bc14 | 2379 | default: |
428b2edd | 2380 | OVS_NOT_REACHED(); |
9583bc14 EJ |
2381 | } |
2382 | } | |
2383 | ||
fed8962a EG |
2384 | /* If output xbundle is dot1q-tunnel, set mask bits of cvlan */ |
2385 | static void | |
2386 | check_and_set_cvlan_mask(struct flow_wildcards *wc, | |
2387 | const struct xbundle *xbundle) | |
2388 | { | |
2389 | if (xbundle->vlan_mode == PORT_VLAN_DOT1Q_TUNNEL && xbundle->cvlans) { | |
2390 | wc->masks.vlans[1].tci = htons(0xffff); | |
2391 | } | |
2392 | } | |
2393 | ||
9583bc14 | 2394 | static void |
46c88433 | 2395 | output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, |
f0fb825a | 2396 | const struct xvlan *xvlan) |
9583bc14 | 2397 | { |
9583bc14 | 2398 | uint16_t vid; |
f0fb825a | 2399 | union flow_vlan_hdr old_vlans[FLOW_MAX_VLAN_HEADERS]; |
46c88433 | 2400 | struct xport *xport; |
e93ef1c7 JR |
2401 | struct xlate_bond_recirc xr; |
2402 | bool use_recirc = false; | |
f0fb825a | 2403 | struct xvlan out_xvlan; |
9583bc14 | 2404 | |
fed8962a EG |
2405 | check_and_set_cvlan_mask(ctx->wc, out_xbundle); |
2406 | ||
f0fb825a EG |
2407 | xvlan_output_translate(out_xbundle, xvlan, &out_xvlan); |
2408 | if (out_xbundle->use_priority_tags) { | |
2409 | out_xvlan.v[0].pcp = ntohs(ctx->xin->flow.vlans[0].tci) & | |
2410 | VLAN_PCP_MASK; | |
2411 | } | |
2412 | vid = out_xvlan.v[0].vid; | |
417e7e66 | 2413 | if (ovs_list_is_empty(&out_xbundle->xports)) { |
46c88433 EJ |
2414 | /* Partially configured bundle with no slaves. Drop the packet. */ |
2415 | return; | |
2416 | } else if (!out_xbundle->bond) { | |
417e7e66 | 2417 | xport = CONTAINER_OF(ovs_list_front(&out_xbundle->xports), struct xport, |
46c88433 | 2418 | bundle_node); |
9583bc14 | 2419 | } else { |
49a73e0c | 2420 | struct flow_wildcards *wc = ctx->wc; |
84f0f298 | 2421 | struct ofport_dpif *ofport; |
adcf00ba | 2422 | |
a80aba3a AZ |
2423 | if (ctx->xbridge->support.odp.recirc) { |
2424 | /* In case recirculation is not actually in use, 'xr.recirc_id' | |
2425 | * will be set to '0', since a valid 'recirc_id' can | |
82f9f1f5 AZ |
2426 | * not be zero. */ |
2427 | bond_update_post_recirc_rules(out_xbundle->bond, | |
2428 | &xr.recirc_id, | |
2429 | &xr.hash_basis); | |
2430 | if (xr.recirc_id) { | |
2431 | /* Use recirculation instead of output. */ | |
2432 | use_recirc = true; | |
e93ef1c7 | 2433 | xr.hash_alg = OVS_HASH_ALG_L4; |
54ecb5a2 AZ |
2434 | /* Recirculation does not require unmasking hash fields. */ |
2435 | wc = NULL; | |
adcf00ba AZ |
2436 | } |
2437 | } | |
46c88433 | 2438 | |
54ecb5a2 AZ |
2439 | ofport = bond_choose_output_slave(out_xbundle->bond, |
2440 | &ctx->xin->flow, wc, vid); | |
0506f184 | 2441 | xport = xport_lookup(ctx->xcfg, ofport); |
46c88433 EJ |
2442 | |
2443 | if (!xport) { | |
9583bc14 EJ |
2444 | /* No slaves enabled, so drop packet. */ |
2445 | return; | |
2446 | } | |
d6fc5f57 | 2447 | |
e93ef1c7 | 2448 | /* If use_recirc is set, the main thread will handle stats |
b256dc52 | 2449 | * accounting for this bond. */ |
e93ef1c7 | 2450 | if (!use_recirc) { |
b256dc52 JS |
2451 | if (ctx->xin->resubmit_stats) { |
2452 | bond_account(out_xbundle->bond, &ctx->xin->flow, vid, | |
2453 | ctx->xin->resubmit_stats->n_bytes); | |
2454 | } | |
2455 | if (ctx->xin->xcache) { | |
2456 | struct xc_entry *entry; | |
2457 | struct flow *flow; | |
2458 | ||
2459 | flow = &ctx->xin->flow; | |
2460 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND); | |
901a517e JR |
2461 | entry->bond.bond = bond_ref(out_xbundle->bond); |
2462 | entry->bond.flow = xmemdup(flow, sizeof *flow); | |
2463 | entry->bond.vid = vid; | |
b256dc52 | 2464 | } |
d6fc5f57 | 2465 | } |
9583bc14 EJ |
2466 | } |
2467 | ||
f0fb825a | 2468 | memcpy(&old_vlans, &ctx->xin->flow.vlans, sizeof(old_vlans)); |
ffbe41db | 2469 | xvlan_put(&ctx->xin->flow, &out_xvlan, out_xbundle->use_priority_tags); |
9583bc14 | 2470 | |
feee58b9 | 2471 | compose_output_action(ctx, xport->ofp_port, use_recirc ? &xr : NULL, |
11938578 | 2472 | false, false); |
f0fb825a | 2473 | memcpy(&ctx->xin->flow.vlans, &old_vlans, sizeof(old_vlans)); |
9583bc14 EJ |
2474 | } |
2475 | ||
2476 | /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after | |
2477 | * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to | |
2478 | * indicate this; newer upstream kernels use gratuitous ARP requests. */ | |
2479 | static bool | |
2480 | is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) | |
2481 | { | |
2482 | if (flow->dl_type != htons(ETH_TYPE_ARP)) { | |
2483 | return false; | |
2484 | } | |
2485 | ||
2486 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
2487 | if (!eth_addr_is_broadcast(flow->dl_dst)) { | |
2488 | return false; | |
2489 | } | |
2490 | ||
2491 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
2492 | if (flow->nw_proto == ARP_OP_REPLY) { | |
2493 | return true; | |
2494 | } else if (flow->nw_proto == ARP_OP_REQUEST) { | |
2495 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
2496 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
2497 | ||
2498 | return flow->nw_src == flow->nw_dst; | |
2499 | } else { | |
2500 | return false; | |
2501 | } | |
2502 | } | |
2503 | ||
ff69c24a FL |
2504 | /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or |
2505 | * dropped. Returns true if they may be forwarded, false if they should be | |
2506 | * dropped. | |
2507 | * | |
2508 | * 'in_port' must be the xport that corresponds to flow->in_port. | |
2509 | * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). | |
2510 | * | |
2511 | * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as | |
2512 | * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as | |
2513 | * checked by input_vid_is_valid(). | |
2514 | * | |
2515 | * May also add tags to '*tags', although the current implementation only does | |
2516 | * so in one special case. | |
2517 | */ | |
2518 | static bool | |
2519 | is_admissible(struct xlate_ctx *ctx, struct xport *in_port, | |
2520 | uint16_t vlan) | |
2521 | { | |
2522 | struct xbundle *in_xbundle = in_port->xbundle; | |
2523 | const struct xbridge *xbridge = ctx->xbridge; | |
2524 | struct flow *flow = &ctx->xin->flow; | |
2525 | ||
2526 | /* Drop frames for reserved multicast addresses | |
2527 | * only if forward_bpdu option is absent. */ | |
2528 | if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { | |
2d9b49dd BP |
2529 | xlate_report(ctx, OFT_DETAIL, |
2530 | "packet has reserved destination MAC, dropping"); | |
ff69c24a FL |
2531 | return false; |
2532 | } | |
2533 | ||
2534 | if (in_xbundle->bond) { | |
2535 | struct mac_entry *mac; | |
2536 | ||
2537 | switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, | |
2538 | flow->dl_dst)) { | |
2539 | case BV_ACCEPT: | |
2540 | break; | |
2541 | ||
2542 | case BV_DROP: | |
2d9b49dd BP |
2543 | xlate_report(ctx, OFT_DETAIL, |
2544 | "bonding refused admissibility, dropping"); | |
ff69c24a FL |
2545 | return false; |
2546 | ||
2547 | case BV_DROP_IF_MOVED: | |
2548 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
2549 | mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan); | |
9d078ec2 BP |
2550 | if (mac |
2551 | && mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle | |
49a73e0c | 2552 | && (!is_gratuitous_arp(flow, ctx->wc) |
9d078ec2 | 2553 | || mac_entry_is_grat_arp_locked(mac))) { |
ff69c24a | 2554 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
2d9b49dd BP |
2555 | xlate_report(ctx, OFT_DETAIL, |
2556 | "SLB bond thinks this packet looped back, " | |
ff69c24a FL |
2557 | "dropping"); |
2558 | return false; | |
2559 | } | |
2560 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
2561 | break; | |
2562 | } | |
2563 | } | |
2564 | ||
2565 | return true; | |
2566 | } | |
2567 | ||
2d9b49dd BP |
2568 | static bool |
2569 | update_learning_table__(const struct xbridge *xbridge, | |
2570 | struct xbundle *in_xbundle, struct eth_addr dl_src, | |
2571 | int vlan, bool is_grat_arp) | |
2572 | { | |
2573 | return (in_xbundle == &ofpp_none_bundle | |
2574 | || !mac_learning_update(xbridge->ml, dl_src, vlan, | |
2575 | is_grat_arp, | |
2576 | in_xbundle->bond != NULL, | |
2577 | in_xbundle->ofbundle)); | |
2578 | } | |
2579 | ||
ee047520 | 2580 | static void |
2d9b49dd | 2581 | update_learning_table(const struct xlate_ctx *ctx, |
064799a1 JR |
2582 | struct xbundle *in_xbundle, struct eth_addr dl_src, |
2583 | int vlan, bool is_grat_arp) | |
ee047520 | 2584 | { |
2d9b49dd BP |
2585 | if (!update_learning_table__(ctx->xbridge, in_xbundle, dl_src, vlan, |
2586 | is_grat_arp)) { | |
2587 | xlate_report_debug(ctx, OFT_DETAIL, "learned that "ETH_ADDR_FMT" is " | |
2588 | "on port %s in VLAN %d", | |
2589 | ETH_ADDR_ARGS(dl_src), in_xbundle->name, vlan); | |
ee047520 | 2590 | } |
9583bc14 EJ |
2591 | } |
2592 | ||
86e2dcdd FL |
2593 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2594 | * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */ | |
2595 | static void | |
2d9b49dd | 2596 | update_mcast_snooping_table4__(const struct xlate_ctx *ctx, |
06994f87 TLSC |
2597 | const struct flow *flow, |
2598 | struct mcast_snooping *ms, int vlan, | |
2599 | struct xbundle *in_xbundle, | |
2600 | const struct dp_packet *packet) | |
86e2dcdd FL |
2601 | OVS_REQ_WRLOCK(ms->rwlock) |
2602 | { | |
46445c63 | 2603 | const struct igmp_header *igmp; |
e3102e42 | 2604 | int count; |
46445c63 | 2605 | size_t offset; |
06994f87 | 2606 | ovs_be32 ip4 = flow->igmp_group_ip4; |
86e2dcdd | 2607 | |
46445c63 EC |
2608 | offset = (char *) dp_packet_l4(packet) - (char *) dp_packet_data(packet); |
2609 | igmp = dp_packet_at(packet, offset, IGMP_HEADER_LEN); | |
2610 | if (!igmp || csum(igmp, dp_packet_l4_size(packet)) != 0) { | |
2d9b49dd BP |
2611 | xlate_report_debug(ctx, OFT_DETAIL, |
2612 | "multicast snooping received bad IGMP " | |
2613 | "checksum on port %s in VLAN %d", | |
2614 | in_xbundle->name, vlan); | |
46445c63 EC |
2615 | return; |
2616 | } | |
2617 | ||
86e2dcdd FL |
2618 | switch (ntohs(flow->tp_src)) { |
2619 | case IGMP_HOST_MEMBERSHIP_REPORT: | |
2620 | case IGMPV2_HOST_MEMBERSHIP_REPORT: | |
964a4d5f | 2621 | if (mcast_snooping_add_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
2d9b49dd BP |
2622 | xlate_report_debug(ctx, OFT_DETAIL, |
2623 | "multicast snooping learned that " | |
2624 | IP_FMT" is on port %s in VLAN %d", | |
2625 | IP_ARGS(ip4), in_xbundle->name, vlan); | |
86e2dcdd FL |
2626 | } |
2627 | break; | |
2628 | case IGMP_HOST_LEAVE_MESSAGE: | |
964a4d5f | 2629 | if (mcast_snooping_leave_group4(ms, ip4, vlan, in_xbundle->ofbundle)) { |
2d9b49dd BP |
2630 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping leaving " |
2631 | IP_FMT" is on port %s in VLAN %d", | |
2632 | IP_ARGS(ip4), in_xbundle->name, vlan); | |
86e2dcdd FL |
2633 | } |
2634 | break; | |
2635 | case IGMP_HOST_MEMBERSHIP_QUERY: | |
2636 | if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan, | |
2d9b49dd BP |
2637 | in_xbundle->ofbundle)) { |
2638 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping query " | |
2639 | "from "IP_FMT" is on port %s in VLAN %d", | |
2640 | IP_ARGS(flow->nw_src), in_xbundle->name, vlan); | |
86e2dcdd FL |
2641 | } |
2642 | break; | |
e3102e42 | 2643 | case IGMPV3_HOST_MEMBERSHIP_REPORT: |
2d9b49dd BP |
2644 | count = mcast_snooping_add_report(ms, packet, vlan, |
2645 | in_xbundle->ofbundle); | |
2646 | if (count) { | |
2647 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping processed " | |
2648 | "%d addresses on port %s in VLAN %d", | |
2649 | count, in_xbundle->name, vlan); | |
e3102e42 TLSC |
2650 | } |
2651 | break; | |
86e2dcdd FL |
2652 | } |
2653 | } | |
2654 | ||
06994f87 | 2655 | static void |
2d9b49dd | 2656 | update_mcast_snooping_table6__(const struct xlate_ctx *ctx, |
06994f87 TLSC |
2657 | const struct flow *flow, |
2658 | struct mcast_snooping *ms, int vlan, | |
2659 | struct xbundle *in_xbundle, | |
2660 | const struct dp_packet *packet) | |
2661 | OVS_REQ_WRLOCK(ms->rwlock) | |
2662 | { | |
46445c63 | 2663 | const struct mld_header *mld; |
06994f87 | 2664 | int count; |
46445c63 EC |
2665 | size_t offset; |
2666 | ||
2667 | offset = (char *) dp_packet_l4(packet) - (char *) dp_packet_data(packet); | |
2668 | mld = dp_packet_at(packet, offset, MLD_HEADER_LEN); | |
2669 | ||
2670 | if (!mld || | |
2671 | packet_csum_upperlayer6(dp_packet_l3(packet), | |
2672 | mld, IPPROTO_ICMPV6, | |
2673 | dp_packet_l4_size(packet)) != 0) { | |
2d9b49dd BP |
2674 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping received " |
2675 | "bad MLD checksum on port %s in VLAN %d", | |
2676 | in_xbundle->name, vlan); | |
46445c63 EC |
2677 | return; |
2678 | } | |
06994f87 TLSC |
2679 | |
2680 | switch (ntohs(flow->tp_src)) { | |
2681 | case MLD_QUERY: | |
2682 | if (!ipv6_addr_equals(&flow->ipv6_src, &in6addr_any) | |
2683 | && mcast_snooping_add_mrouter(ms, vlan, in_xbundle->ofbundle)) { | |
2d9b49dd BP |
2684 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping query on " |
2685 | "port %s in VLAN %d", in_xbundle->name, vlan); | |
06994f87 TLSC |
2686 | } |
2687 | break; | |
2688 | case MLD_REPORT: | |
2689 | case MLD_DONE: | |
2690 | case MLD2_REPORT: | |
2691 | count = mcast_snooping_add_mld(ms, packet, vlan, in_xbundle->ofbundle); | |
2692 | if (count) { | |
2d9b49dd BP |
2693 | xlate_report_debug(ctx, OFT_DETAIL, "multicast snooping processed " |
2694 | "%d addresses on port %s in VLAN %d", | |
2695 | count, in_xbundle->name, vlan); | |
06994f87 TLSC |
2696 | } |
2697 | break; | |
2698 | } | |
2699 | } | |
2700 | ||
86e2dcdd FL |
2701 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
2702 | * was received on 'in_xbundle' in 'vlan'. */ | |
2703 | static void | |
2d9b49dd | 2704 | update_mcast_snooping_table(const struct xlate_ctx *ctx, |
86e2dcdd | 2705 | const struct flow *flow, int vlan, |
e3102e42 TLSC |
2706 | struct xbundle *in_xbundle, |
2707 | const struct dp_packet *packet) | |
86e2dcdd | 2708 | { |
2d9b49dd | 2709 | struct mcast_snooping *ms = ctx->xbridge->ms; |
86e2dcdd | 2710 | struct xbundle *mcast_xbundle; |
f4ae6e23 | 2711 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2712 | |
2713 | /* Don't learn the OFPP_NONE port. */ | |
2714 | if (in_xbundle == &ofpp_none_bundle) { | |
2715 | return; | |
2716 | } | |
2717 | ||
2718 | /* Don't learn from flood ports */ | |
2719 | mcast_xbundle = NULL; | |
2720 | ovs_rwlock_wrlock(&ms->rwlock); | |
f4ae6e23 | 2721 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
0506f184 | 2722 | mcast_xbundle = xbundle_lookup(ctx->xcfg, fport->port); |
86e2dcdd FL |
2723 | if (mcast_xbundle == in_xbundle) { |
2724 | break; | |
2725 | } | |
2726 | } | |
2727 | ||
2728 | if (!mcast_xbundle || mcast_xbundle != in_xbundle) { | |
06994f87 | 2729 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
2d9b49dd | 2730 | update_mcast_snooping_table4__(ctx, flow, ms, vlan, |
06994f87 TLSC |
2731 | in_xbundle, packet); |
2732 | } else { | |
2d9b49dd | 2733 | update_mcast_snooping_table6__(ctx, flow, ms, vlan, |
06994f87 TLSC |
2734 | in_xbundle, packet); |
2735 | } | |
86e2dcdd FL |
2736 | } |
2737 | ovs_rwlock_unlock(&ms->rwlock); | |
2738 | } | |
efc93e66 BP |
2739 | \f |
2740 | /* A list of multicast output ports. | |
2741 | * | |
2742 | * We accumulate output ports and then do all the outputs afterward. It would | |
2743 | * be more natural to do the outputs one at a time as we discover the need for | |
2744 | * each one, but this can cause a deadlock because we need to take the | |
2745 | * mcast_snooping's rwlock for reading to iterate through the port lists and | |
2746 | * doing an output, if it goes to a patch port, can eventually come back to the | |
2747 | * same mcast_snooping and attempt to take the write lock (see | |
2748 | * https://github.com/openvswitch/ovs-issues/issues/153). */ | |
2749 | struct mcast_output { | |
2750 | /* Discrete ports. */ | |
2751 | struct xbundle **xbundles; | |
2752 | size_t n, allocated; | |
2753 | ||
2754 | /* If set, flood to all ports. */ | |
2755 | bool flood; | |
2756 | }; | |
2757 | #define MCAST_OUTPUT_INIT { NULL, 0, 0, false } | |
2758 | ||
2759 | /* Add 'mcast_bundle' to 'out'. */ | |
2760 | static void | |
2761 | mcast_output_add(struct mcast_output *out, struct xbundle *mcast_xbundle) | |
2762 | { | |
2763 | if (out->n >= out->allocated) { | |
2764 | out->xbundles = x2nrealloc(out->xbundles, &out->allocated, | |
2765 | sizeof *out->xbundles); | |
2766 | } | |
2767 | out->xbundles[out->n++] = mcast_xbundle; | |
2768 | } | |
2769 | ||
2770 | /* Outputs the packet in 'ctx' to all of the output ports in 'out', given input | |
2771 | * bundle 'in_xbundle' and the current 'xvlan'. */ | |
2772 | static void | |
2773 | mcast_output_finish(struct xlate_ctx *ctx, struct mcast_output *out, | |
2774 | struct xbundle *in_xbundle, struct xvlan *xvlan) | |
2775 | { | |
2776 | if (out->flood) { | |
2777 | xlate_normal_flood(ctx, in_xbundle, xvlan); | |
2778 | } else { | |
2779 | for (size_t i = 0; i < out->n; i++) { | |
2780 | output_normal(ctx, out->xbundles[i], xvlan); | |
2781 | } | |
2782 | } | |
2783 | ||
2784 | free(out->xbundles); | |
2785 | } | |
86e2dcdd FL |
2786 | |
2787 | /* send the packet to ports having the multicast group learned */ | |
2788 | static void | |
2789 | xlate_normal_mcast_send_group(struct xlate_ctx *ctx, | |
2790 | struct mcast_snooping *ms OVS_UNUSED, | |
2791 | struct mcast_group *grp, | |
f0fb825a | 2792 | struct xbundle *in_xbundle, |
efc93e66 | 2793 | struct mcast_output *out) |
86e2dcdd FL |
2794 | OVS_REQ_RDLOCK(ms->rwlock) |
2795 | { | |
86e2dcdd FL |
2796 | struct mcast_group_bundle *b; |
2797 | struct xbundle *mcast_xbundle; | |
2798 | ||
86e2dcdd | 2799 | LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) { |
0506f184 | 2800 | mcast_xbundle = xbundle_lookup(ctx->xcfg, b->port); |
86e2dcdd | 2801 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { |
2d9b49dd | 2802 | xlate_report(ctx, OFT_DETAIL, "forwarding to mcast group port"); |
efc93e66 | 2803 | mcast_output_add(out, mcast_xbundle); |
86e2dcdd | 2804 | } else if (!mcast_xbundle) { |
2d9b49dd BP |
2805 | xlate_report(ctx, OFT_WARN, |
2806 | "mcast group port is unknown, dropping"); | |
86e2dcdd | 2807 | } else { |
2d9b49dd BP |
2808 | xlate_report(ctx, OFT_DETAIL, |
2809 | "mcast group port is input port, dropping"); | |
86e2dcdd FL |
2810 | } |
2811 | } | |
2812 | } | |
2813 | ||
2814 | /* send the packet to ports connected to multicast routers */ | |
2815 | static void | |
2816 | xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx, | |
2817 | struct mcast_snooping *ms, | |
f0fb825a | 2818 | struct xbundle *in_xbundle, |
efc93e66 BP |
2819 | const struct xvlan *xvlan, |
2820 | struct mcast_output *out) | |
86e2dcdd FL |
2821 | OVS_REQ_RDLOCK(ms->rwlock) |
2822 | { | |
86e2dcdd FL |
2823 | struct mcast_mrouter_bundle *mrouter; |
2824 | struct xbundle *mcast_xbundle; | |
2825 | ||
86e2dcdd | 2826 | LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) { |
0506f184 | 2827 | mcast_xbundle = xbundle_lookup(ctx->xcfg, mrouter->port); |
94a881c1 | 2828 | if (mcast_xbundle && mcast_xbundle != in_xbundle |
f0fb825a | 2829 | && mrouter->vlan == xvlan->v[0].vid) { |
2d9b49dd | 2830 | xlate_report(ctx, OFT_DETAIL, "forwarding to mcast router port"); |
efc93e66 | 2831 | mcast_output_add(out, mcast_xbundle); |
86e2dcdd | 2832 | } else if (!mcast_xbundle) { |
2d9b49dd BP |
2833 | xlate_report(ctx, OFT_WARN, |
2834 | "mcast router port is unknown, dropping"); | |
f0fb825a | 2835 | } else if (mrouter->vlan != xvlan->v[0].vid) { |
2d9b49dd BP |
2836 | xlate_report(ctx, OFT_DETAIL, |
2837 | "mcast router is on another vlan, dropping"); | |
86e2dcdd | 2838 | } else { |
2d9b49dd BP |
2839 | xlate_report(ctx, OFT_DETAIL, |
2840 | "mcast router port is input port, dropping"); | |
86e2dcdd FL |
2841 | } |
2842 | } | |
2843 | } | |
2844 | ||
2845 | /* send the packet to ports flagged to be flooded */ | |
2846 | static void | |
2847 | xlate_normal_mcast_send_fports(struct xlate_ctx *ctx, | |
2848 | struct mcast_snooping *ms, | |
f0fb825a | 2849 | struct xbundle *in_xbundle, |
efc93e66 | 2850 | struct mcast_output *out) |
86e2dcdd FL |
2851 | OVS_REQ_RDLOCK(ms->rwlock) |
2852 | { | |
f4ae6e23 | 2853 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2854 | struct xbundle *mcast_xbundle; |
2855 | ||
f4ae6e23 | 2856 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
0506f184 | 2857 | mcast_xbundle = xbundle_lookup(ctx->xcfg, fport->port); |
86e2dcdd | 2858 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { |
2d9b49dd | 2859 | xlate_report(ctx, OFT_DETAIL, "forwarding to mcast flood port"); |
efc93e66 | 2860 | mcast_output_add(out, mcast_xbundle); |
86e2dcdd | 2861 | } else if (!mcast_xbundle) { |
2d9b49dd BP |
2862 | xlate_report(ctx, OFT_WARN, |
2863 | "mcast flood port is unknown, dropping"); | |
86e2dcdd | 2864 | } else { |
2d9b49dd BP |
2865 | xlate_report(ctx, OFT_DETAIL, |
2866 | "mcast flood port is input port, dropping"); | |
86e2dcdd FL |
2867 | } |
2868 | } | |
2869 | } | |
2870 | ||
8e04a33f FL |
2871 | /* forward the Reports to configured ports */ |
2872 | static void | |
2873 | xlate_normal_mcast_send_rports(struct xlate_ctx *ctx, | |
2874 | struct mcast_snooping *ms, | |
f0fb825a | 2875 | struct xbundle *in_xbundle, |
efc93e66 | 2876 | struct mcast_output *out) |
8e04a33f FL |
2877 | OVS_REQ_RDLOCK(ms->rwlock) |
2878 | { | |
8e04a33f FL |
2879 | struct mcast_port_bundle *rport; |
2880 | struct xbundle *mcast_xbundle; | |
2881 | ||
8e04a33f | 2882 | LIST_FOR_EACH(rport, node, &ms->rport_list) { |
0506f184 | 2883 | mcast_xbundle = xbundle_lookup(ctx->xcfg, rport->port); |
c781bd52 LJ |
2884 | if (mcast_xbundle |
2885 | && mcast_xbundle != in_xbundle | |
2886 | && mcast_xbundle->ofbundle != in_xbundle->ofbundle) { | |
2d9b49dd BP |
2887 | xlate_report(ctx, OFT_DETAIL, |
2888 | "forwarding report to mcast flagged port"); | |
efc93e66 | 2889 | mcast_output_add(out, mcast_xbundle); |
8e04a33f | 2890 | } else if (!mcast_xbundle) { |
2d9b49dd BP |
2891 | xlate_report(ctx, OFT_WARN, |
2892 | "mcast port is unknown, dropping the report"); | |
8e04a33f | 2893 | } else { |
2d9b49dd BP |
2894 | xlate_report(ctx, OFT_DETAIL, |
2895 | "mcast port is input port, dropping the Report"); | |
8e04a33f FL |
2896 | } |
2897 | } | |
2898 | } | |
2899 | ||
682800a4 FL |
2900 | static void |
2901 | xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, | |
f0fb825a | 2902 | struct xvlan *xvlan) |
682800a4 FL |
2903 | { |
2904 | struct xbundle *xbundle; | |
2905 | ||
2906 | LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { | |
2907 | if (xbundle != in_xbundle | |
c781bd52 | 2908 | && xbundle->ofbundle != in_xbundle->ofbundle |
f0fb825a | 2909 | && xbundle_includes_vlan(xbundle, xvlan) |
682800a4 FL |
2910 | && xbundle->floodable |
2911 | && !xbundle_mirror_out(ctx->xbridge, xbundle)) { | |
f0fb825a | 2912 | output_normal(ctx, xbundle, xvlan); |
682800a4 FL |
2913 | } |
2914 | } | |
2031ef97 | 2915 | ctx->nf_output_iface = NF_OUT_FLOOD; |
682800a4 FL |
2916 | } |
2917 | ||
a75636c8 BP |
2918 | static bool |
2919 | is_ip_local_multicast(const struct flow *flow, struct flow_wildcards *wc) | |
2920 | { | |
2921 | if (flow->dl_type == htons(ETH_TYPE_IP)) { | |
2922 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
2923 | return ip_is_local_multicast(flow->nw_dst); | |
2924 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
2925 | memset(&wc->masks.ipv6_dst, 0xff, sizeof wc->masks.ipv6_dst); | |
2926 | return ipv6_is_all_hosts(&flow->ipv6_dst); | |
2927 | } else { | |
2928 | return false; | |
2929 | } | |
2930 | } | |
2931 | ||
9583bc14 EJ |
2932 | static void |
2933 | xlate_normal(struct xlate_ctx *ctx) | |
2934 | { | |
49a73e0c | 2935 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 2936 | struct flow *flow = &ctx->xin->flow; |
46c88433 EJ |
2937 | struct xbundle *in_xbundle; |
2938 | struct xport *in_port; | |
9583bc14 | 2939 | struct mac_entry *mac; |
d6d5bbc9 | 2940 | void *mac_port; |
f0fb825a EG |
2941 | struct xvlan in_xvlan; |
2942 | struct xvlan xvlan; | |
9583bc14 | 2943 | uint16_t vlan; |
9583bc14 | 2944 | |
33bf9176 BP |
2945 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
2946 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
f0fb825a | 2947 | wc->masks.vlans[0].tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
9583bc14 | 2948 | |
2d9b49dd | 2949 | in_xbundle = lookup_input_bundle(ctx, flow->in_port.ofp_port, &in_port); |
46c88433 | 2950 | if (!in_xbundle) { |
2d9b49dd | 2951 | xlate_report(ctx, OFT_WARN, "no input bundle, dropping"); |
9583bc14 EJ |
2952 | return; |
2953 | } | |
2954 | ||
2955 | /* Drop malformed frames. */ | |
f0fb825a EG |
2956 | if (eth_type_vlan(flow->dl_type) && |
2957 | !(flow->vlans[0].tci & htons(VLAN_CFI))) { | |
9583bc14 | 2958 | if (ctx->xin->packet != NULL) { |
2d9b49dd BP |
2959 | xlate_report_error(ctx, "dropping packet with partial " |
2960 | "VLAN tag received on port %s", | |
2961 | in_xbundle->name); | |
9583bc14 | 2962 | } |
2d9b49dd | 2963 | xlate_report(ctx, OFT_WARN, "partial VLAN tag, dropping"); |
9583bc14 EJ |
2964 | return; |
2965 | } | |
2966 | ||
2967 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 2968 | if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) { |
9583bc14 | 2969 | if (ctx->xin->packet != NULL) { |
2d9b49dd BP |
2970 | xlate_report_error(ctx, "dropping packet received on port %s, " |
2971 | "which is reserved exclusively for mirroring", | |
2972 | in_xbundle->name); | |
9583bc14 | 2973 | } |
2d9b49dd BP |
2974 | xlate_report(ctx, OFT_WARN, |
2975 | "input port is mirror output port, dropping"); | |
9583bc14 EJ |
2976 | return; |
2977 | } | |
2978 | ||
2979 | /* Check VLAN. */ | |
f0fb825a EG |
2980 | xvlan_extract(flow, &in_xvlan); |
2981 | if (!input_vid_is_valid(ctx, in_xvlan.v[0].vid, in_xbundle)) { | |
2d9b49dd BP |
2982 | xlate_report(ctx, OFT_WARN, |
2983 | "disallowed VLAN VID for this input port, dropping"); | |
9583bc14 EJ |
2984 | return; |
2985 | } | |
f0fb825a EG |
2986 | xvlan_input_translate(in_xbundle, &in_xvlan, &xvlan); |
2987 | vlan = xvlan.v[0].vid; | |
9583bc14 EJ |
2988 | |
2989 | /* Check other admissibility requirements. */ | |
2990 | if (in_port && !is_admissible(ctx, in_port, vlan)) { | |
2991 | return; | |
2992 | } | |
2993 | ||
2994 | /* Learn source MAC. */ | |
064799a1 | 2995 | bool is_grat_arp = is_gratuitous_arp(flow, wc); |
875ab130 BP |
2996 | if (ctx->xin->allow_side_effects |
2997 | && flow->packet_type == htonl(PT_ETH) | |
2998 | && in_port->pt_mode != NETDEV_PT_LEGACY_L3 | |
2999 | ) { | |
2d9b49dd | 3000 | update_learning_table(ctx, in_xbundle, flow->dl_src, vlan, |
064799a1 | 3001 | is_grat_arp); |
9583bc14 | 3002 | } |
064799a1 | 3003 | if (ctx->xin->xcache && in_xbundle != &ofpp_none_bundle) { |
b256dc52 JS |
3004 | struct xc_entry *entry; |
3005 | ||
064799a1 | 3006 | /* Save just enough info to update mac learning table later. */ |
b256dc52 | 3007 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL); |
901a517e JR |
3008 | entry->normal.ofproto = ctx->xbridge->ofproto; |
3009 | entry->normal.in_port = flow->in_port.ofp_port; | |
3010 | entry->normal.dl_src = flow->dl_src; | |
3011 | entry->normal.vlan = vlan; | |
3012 | entry->normal.is_gratuitous_arp = is_grat_arp; | |
b256dc52 | 3013 | } |
9583bc14 EJ |
3014 | |
3015 | /* Determine output bundle. */ | |
86e2dcdd FL |
3016 | if (mcast_snooping_enabled(ctx->xbridge->ms) |
3017 | && !eth_addr_is_broadcast(flow->dl_dst) | |
3018 | && eth_addr_is_multicast(flow->dl_dst) | |
06994f87 | 3019 | && is_ip_any(flow)) { |
86e2dcdd | 3020 | struct mcast_snooping *ms = ctx->xbridge->ms; |
06994f87 | 3021 | struct mcast_group *grp = NULL; |
86e2dcdd | 3022 | |
a75636c8 | 3023 | if (is_igmp(flow, wc)) { |
1bc24169 BP |
3024 | /* |
3025 | * IGMP packets need to take the slow path, in order to be | |
3026 | * processed for mdb updates. That will prevent expires | |
3027 | * firing off even after hosts have sent reports. | |
3028 | */ | |
3029 | ctx->xout->slow |= SLOW_ACTION; | |
3030 | ||
a75636c8 | 3031 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); |
d29f137b TLSC |
3032 | if (mcast_snooping_is_membership(flow->tp_src) || |
3033 | mcast_snooping_is_query(flow->tp_src)) { | |
df70a773 | 3034 | if (ctx->xin->allow_side_effects && ctx->xin->packet) { |
2d9b49dd | 3035 | update_mcast_snooping_table(ctx, flow, vlan, |
e3102e42 | 3036 | in_xbundle, ctx->xin->packet); |
d29f137b | 3037 | } |
86e2dcdd | 3038 | } |
d6d5bbc9 | 3039 | |
86e2dcdd | 3040 | if (mcast_snooping_is_membership(flow->tp_src)) { |
efc93e66 BP |
3041 | struct mcast_output out = MCAST_OUTPUT_INIT; |
3042 | ||
86e2dcdd | 3043 | ovs_rwlock_rdlock(&ms->rwlock); |
efc93e66 BP |
3044 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, &xvlan, |
3045 | &out); | |
8e04a33f FL |
3046 | /* RFC4541: section 2.1.1, item 1: A snooping switch should |
3047 | * forward IGMP Membership Reports only to those ports where | |
3048 | * multicast routers are attached. Alternatively stated: a | |
3049 | * snooping switch should not forward IGMP Membership Reports | |
3050 | * to ports on which only hosts are attached. | |
3051 | * An administrative control may be provided to override this | |
3052 | * restriction, allowing the report messages to be flooded to | |
3053 | * other ports. */ | |
efc93e66 | 3054 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, &out); |
86e2dcdd | 3055 | ovs_rwlock_unlock(&ms->rwlock); |
efc93e66 BP |
3056 | |
3057 | mcast_output_finish(ctx, &out, in_xbundle, &xvlan); | |
86e2dcdd | 3058 | } else { |
2d9b49dd | 3059 | xlate_report(ctx, OFT_DETAIL, "multicast traffic, flooding"); |
f0fb825a | 3060 | xlate_normal_flood(ctx, in_xbundle, &xvlan); |
86e2dcdd FL |
3061 | } |
3062 | return; | |
a75636c8 | 3063 | } else if (is_mld(flow, wc)) { |
06994f87 | 3064 | ctx->xout->slow |= SLOW_ACTION; |
df70a773 | 3065 | if (ctx->xin->allow_side_effects && ctx->xin->packet) { |
2d9b49dd | 3066 | update_mcast_snooping_table(ctx, flow, vlan, |
06994f87 TLSC |
3067 | in_xbundle, ctx->xin->packet); |
3068 | } | |
a75636c8 | 3069 | if (is_mld_report(flow, wc)) { |
efc93e66 BP |
3070 | struct mcast_output out = MCAST_OUTPUT_INIT; |
3071 | ||
06994f87 | 3072 | ovs_rwlock_rdlock(&ms->rwlock); |
efc93e66 BP |
3073 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, &xvlan, |
3074 | &out); | |
3075 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, &out); | |
06994f87 | 3076 | ovs_rwlock_unlock(&ms->rwlock); |
efc93e66 BP |
3077 | |
3078 | mcast_output_finish(ctx, &out, in_xbundle, &xvlan); | |
06994f87 | 3079 | } else { |
2d9b49dd | 3080 | xlate_report(ctx, OFT_DETAIL, "MLD query, flooding"); |
f0fb825a | 3081 | xlate_normal_flood(ctx, in_xbundle, &xvlan); |
06994f87 | 3082 | } |
86e2dcdd | 3083 | } else { |
a75636c8 | 3084 | if (is_ip_local_multicast(flow, wc)) { |
86e2dcdd FL |
3085 | /* RFC4541: section 2.1.2, item 2: Packets with a dst IP |
3086 | * address in the 224.0.0.x range which are not IGMP must | |
3087 | * be forwarded on all ports */ | |
2d9b49dd BP |
3088 | xlate_report(ctx, OFT_DETAIL, |
3089 | "RFC4541: section 2.1.2, item 2, flooding"); | |
f0fb825a | 3090 | xlate_normal_flood(ctx, in_xbundle, &xvlan); |
86e2dcdd FL |
3091 | return; |
3092 | } | |
3093 | } | |
3094 | ||
3095 | /* forwarding to group base ports */ | |
efc93e66 BP |
3096 | struct mcast_output out = MCAST_OUTPUT_INIT; |
3097 | ||
86e2dcdd | 3098 | ovs_rwlock_rdlock(&ms->rwlock); |
06994f87 TLSC |
3099 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
3100 | grp = mcast_snooping_lookup4(ms, flow->nw_dst, vlan); | |
3101 | } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { | |
3102 | grp = mcast_snooping_lookup(ms, &flow->ipv6_dst, vlan); | |
3103 | } | |
86e2dcdd | 3104 | if (grp) { |
efc93e66 BP |
3105 | xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, &out); |
3106 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, &out); | |
3107 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, &xvlan, | |
3108 | &out); | |
9583bc14 | 3109 | } else { |
86e2dcdd | 3110 | if (mcast_snooping_flood_unreg(ms)) { |
2d9b49dd BP |
3111 | xlate_report(ctx, OFT_DETAIL, |
3112 | "unregistered multicast, flooding"); | |
efc93e66 | 3113 | out.flood = true; |
86e2dcdd | 3114 | } else { |
efc93e66 BP |
3115 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, &xvlan, |
3116 | &out); | |
3117 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, &out); | |
86e2dcdd | 3118 | } |
9583bc14 | 3119 | } |
86e2dcdd | 3120 | ovs_rwlock_unlock(&ms->rwlock); |
efc93e66 BP |
3121 | |
3122 | mcast_output_finish(ctx, &out, in_xbundle, &xvlan); | |
9583bc14 | 3123 | } else { |
86e2dcdd FL |
3124 | ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); |
3125 | mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); | |
9d078ec2 | 3126 | mac_port = mac ? mac_entry_get_port(ctx->xbridge->ml, mac) : NULL; |
86e2dcdd FL |
3127 | ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); |
3128 | ||
3129 | if (mac_port) { | |
0506f184 | 3130 | struct xbundle *mac_xbundle = xbundle_lookup(ctx->xcfg, mac_port); |
84dd881b DL |
3131 | |
3132 | if (mac_xbundle && xbundle_mirror_out(ctx->xbridge, mac_xbundle)) { | |
3133 | xlate_report(ctx, OFT_WARN, | |
3134 | "learned port is a mirror port, dropping"); | |
3135 | return; | |
3136 | } | |
3137 | ||
c781bd52 LJ |
3138 | if (mac_xbundle |
3139 | && mac_xbundle != in_xbundle | |
3140 | && mac_xbundle->ofbundle != in_xbundle->ofbundle) { | |
2d9b49dd | 3141 | xlate_report(ctx, OFT_DETAIL, "forwarding to learned port"); |
f0fb825a | 3142 | output_normal(ctx, mac_xbundle, &xvlan); |
86e2dcdd | 3143 | } else if (!mac_xbundle) { |
2d9b49dd BP |
3144 | xlate_report(ctx, OFT_WARN, |
3145 | "learned port is unknown, dropping"); | |
86e2dcdd | 3146 | } else { |
2d9b49dd BP |
3147 | xlate_report(ctx, OFT_DETAIL, |
3148 | "learned port is input port, dropping"); | |
86e2dcdd FL |
3149 | } |
3150 | } else { | |
2d9b49dd BP |
3151 | xlate_report(ctx, OFT_DETAIL, |
3152 | "no learned MAC for destination, flooding"); | |
f0fb825a | 3153 | xlate_normal_flood(ctx, in_xbundle, &xvlan); |
86e2dcdd | 3154 | } |
9583bc14 EJ |
3155 | } |
3156 | } | |
3157 | ||
a6092018 BP |
3158 | /* Appends a "sample" action for sFlow or IPFIX to 'ctx->odp_actions'. The |
3159 | * 'probability' is the number of packets out of UINT32_MAX to sample. The | |
8de6ff3e JP |
3160 | * 'cookie' is passed back in the callback for each sampled packet. |
3161 | * 'tunnel_out_port', if not ODPP_NONE, is added as the | |
3162 | * OVS_USERSPACE_ATTR_EGRESS_TUN_PORT attribute. If 'include_actions', | |
3163 | * an OVS_USERSPACE_ATTR_ACTIONS attribute is added. If | |
3164 | * 'emit_set_tunnel', sample(sampling_port=1) would translate into | |
3165 | * datapath sample action set(tunnel(...)), sample(...) and it is used | |
3166 | * for sampling egress tunnel information. | |
9583bc14 EJ |
3167 | */ |
3168 | static size_t | |
a6092018 | 3169 | compose_sample_action(struct xlate_ctx *ctx, |
9583bc14 | 3170 | const uint32_t probability, |
8de6ff3e | 3171 | const struct user_action_cookie *cookie, |
7321bda3 NM |
3172 | const odp_port_t tunnel_out_port, |
3173 | bool include_actions) | |
9583bc14 | 3174 | { |
b97f2c3a BY |
3175 | if (probability == 0) { |
3176 | /* No need to generate sampling or the inner action. */ | |
3177 | return 0; | |
3178 | } | |
3179 | ||
31b29c2e AZ |
3180 | /* If the slow path meter is configured by the controller, |
3181 | * insert a meter action before the user space action. */ | |
3182 | struct ofproto *ofproto = &ctx->xin->ofproto->up; | |
3183 | uint32_t meter_id = ofproto->slowpath_meter_id; | |
3184 | ||
3185 | /* When meter action is not required, avoid generate sample action | |
3186 | * for 100% sampling rate. */ | |
3187 | bool is_sample = probability < UINT32_MAX || meter_id != UINT32_MAX; | |
dc2c9ce3 | 3188 | size_t sample_offset = 0, actions_offset = 0; |
72471622 BY |
3189 | if (is_sample) { |
3190 | sample_offset = nl_msg_start_nested(ctx->odp_actions, | |
3191 | OVS_ACTION_ATTR_SAMPLE); | |
3192 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, | |
3193 | probability); | |
3194 | actions_offset = nl_msg_start_nested(ctx->odp_actions, | |
3195 | OVS_SAMPLE_ATTR_ACTIONS); | |
3196 | } | |
9583bc14 | 3197 | |
31b29c2e AZ |
3198 | if (meter_id != UINT32_MAX) { |
3199 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_METER, meter_id); | |
3200 | } | |
3201 | ||
a6092018 BP |
3202 | odp_port_t odp_port = ofp_port_to_odp_port( |
3203 | ctx->xbridge, ctx->xin->flow.in_port.ofp_port); | |
769b5034 | 3204 | uint32_t pid = dpif_port_get_pid(ctx->xbridge->dpif, odp_port); |
3276f2f2 LH |
3205 | size_t cookie_offset = odp_put_userspace_action(pid, cookie, |
3206 | sizeof *cookie, | |
3207 | tunnel_out_port, | |
3208 | include_actions, | |
3209 | ctx->odp_actions); | |
89a8a7f0 | 3210 | |
72471622 BY |
3211 | if (is_sample) { |
3212 | nl_msg_end_nested(ctx->odp_actions, actions_offset); | |
3213 | nl_msg_end_nested(ctx->odp_actions, sample_offset); | |
3214 | } | |
9583bc14 | 3215 | |
9583bc14 EJ |
3216 | return cookie_offset; |
3217 | } | |
3218 | ||
a6092018 BP |
3219 | /* If sFLow is not enabled, returns 0 without doing anything. |
3220 | * | |
3221 | * If sFlow is enabled, appends a template "sample" action to the ODP actions | |
3222 | * in 'ctx'. This action is a template because some of the information needed | |
3223 | * to fill it out is not available until flow translation is complete. In this | |
3224 | * case, this functions returns an offset, which is always nonzero, to pass | |
3225 | * later to fix_sflow_action() to fill in the rest of the template. */ | |
9583bc14 | 3226 | static size_t |
a6092018 | 3227 | compose_sflow_action(struct xlate_ctx *ctx) |
9583bc14 | 3228 | { |
a6092018 BP |
3229 | struct dpif_sflow *sflow = ctx->xbridge->sflow; |
3230 | if (!sflow || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { | |
9583bc14 EJ |
3231 | return 0; |
3232 | } | |
3233 | ||
24a4bbe1 IM |
3234 | struct user_action_cookie cookie; |
3235 | ||
3236 | memset(&cookie, 0, sizeof cookie); | |
3237 | cookie.type = USER_ACTION_COOKIE_SFLOW; | |
3238 | cookie.ofp_in_port = ctx->xin->flow.in_port.ofp_port; | |
3239 | cookie.ofproto_uuid = ctx->xbridge->ofproto->uuid; | |
3240 | ||
a6092018 | 3241 | return compose_sample_action(ctx, dpif_sflow_get_probability(sflow), |
8de6ff3e | 3242 | &cookie, ODPP_NONE, true); |
9583bc14 EJ |
3243 | } |
3244 | ||
f69f713b BY |
3245 | /* If flow IPFIX is enabled, make sure IPFIX flow sample action |
3246 | * at egress point of tunnel port is just in front of corresponding | |
3247 | * output action. If bridge IPFIX is enabled, this appends an IPFIX | |
3248 | * sample action to 'ctx->odp_actions'. */ | |
9583bc14 | 3249 | static void |
a6092018 | 3250 | compose_ipfix_action(struct xlate_ctx *ctx, odp_port_t output_odp_port) |
9583bc14 | 3251 | { |
a6092018 | 3252 | struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; |
8b7ea2d4 | 3253 | odp_port_t tunnel_out_port = ODPP_NONE; |
9583bc14 | 3254 | |
a6092018 | 3255 | if (!ipfix || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
3256 | return; |
3257 | } | |
3258 | ||
8b7ea2d4 WZ |
3259 | /* For input case, output_odp_port is ODPP_NONE, which is an invalid port |
3260 | * number. */ | |
3261 | if (output_odp_port == ODPP_NONE && | |
a6092018 | 3262 | !dpif_ipfix_get_bridge_exporter_input_sampling(ipfix)) { |
8b7ea2d4 WZ |
3263 | return; |
3264 | } | |
3265 | ||
f69f713b | 3266 | /* For output case, output_odp_port is valid. */ |
8b7ea2d4 | 3267 | if (output_odp_port != ODPP_NONE) { |
a6092018 | 3268 | if (!dpif_ipfix_get_bridge_exporter_output_sampling(ipfix)) { |
8b7ea2d4 WZ |
3269 | return; |
3270 | } | |
3271 | /* If tunnel sampling is enabled, put an additional option attribute: | |
3272 | * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT | |
3273 | */ | |
a6092018 | 3274 | if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(ipfix) && |
cd32509e | 3275 | dpif_ipfix_is_tunnel_port(ipfix, output_odp_port) ) { |
8b7ea2d4 WZ |
3276 | tunnel_out_port = output_odp_port; |
3277 | } | |
3278 | } | |
3279 | ||
24a4bbe1 IM |
3280 | struct user_action_cookie cookie; |
3281 | ||
3282 | memset(&cookie, 0, sizeof cookie); | |
3283 | cookie.type = USER_ACTION_COOKIE_IPFIX; | |
3284 | cookie.ofp_in_port = ctx->xin->flow.in_port.ofp_port; | |
3285 | cookie.ofproto_uuid = ctx->xbridge->ofproto->uuid; | |
3286 | cookie.ipfix.output_odp_port = output_odp_port; | |
3287 | ||
a6092018 BP |
3288 | compose_sample_action(ctx, |
3289 | dpif_ipfix_get_bridge_exporter_probability(ipfix), | |
8de6ff3e | 3290 | &cookie, tunnel_out_port, false); |
9583bc14 EJ |
3291 | } |
3292 | ||
a6092018 BP |
3293 | /* Fix "sample" action according to data collected while composing ODP actions, |
3294 | * as described in compose_sflow_action(). | |
3295 | * | |
8de6ff3e JP |
3296 | * 'user_cookie_offset' must be the offset returned by |
3297 | * compose_sflow_action(). */ | |
9583bc14 | 3298 | static void |
a6092018 | 3299 | fix_sflow_action(struct xlate_ctx *ctx, unsigned int user_cookie_offset) |
9583bc14 EJ |
3300 | { |
3301 | const struct flow *base = &ctx->base_flow; | |
8de6ff3e | 3302 | struct user_action_cookie *cookie; |
9583bc14 | 3303 | |
8de6ff3e | 3304 | cookie = ofpbuf_at(ctx->odp_actions, user_cookie_offset, sizeof *cookie); |
9583bc14 EJ |
3305 | ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); |
3306 | ||
f0fb825a | 3307 | cookie->sflow.vlan_tci = base->vlans[0].tci; |
a6092018 BP |
3308 | |
3309 | /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output | |
3310 | * port information") for the interpretation of cookie->output. */ | |
3311 | switch (ctx->sflow_n_outputs) { | |
3312 | case 0: | |
3313 | /* 0x40000000 | 256 means "packet dropped for unknown reason". */ | |
3314 | cookie->sflow.output = 0x40000000 | 256; | |
3315 | break; | |
3316 | ||
3317 | case 1: | |
3318 | cookie->sflow.output = dpif_sflow_odp_port_to_ifindex( | |
3319 | ctx->xbridge->sflow, ctx->sflow_odp_port); | |
3320 | if (cookie->sflow.output) { | |
3321 | break; | |
3322 | } | |
3323 | /* Fall through. */ | |
3324 | default: | |
3325 | /* 0x80000000 means "multiple output ports. */ | |
3326 | cookie->sflow.output = 0x80000000 | ctx->sflow_n_outputs; | |
3327 | break; | |
3328 | } | |
9583bc14 EJ |
3329 | } |
3330 | ||
515793d5 BP |
3331 | static bool |
3332 | process_special(struct xlate_ctx *ctx, const struct xport *xport) | |
db7d4e46 | 3333 | { |
515793d5 | 3334 | const struct flow *flow = &ctx->xin->flow; |
49a73e0c | 3335 | struct flow_wildcards *wc = ctx->wc; |
46c88433 | 3336 | const struct xbridge *xbridge = ctx->xbridge; |
515793d5 BP |
3337 | const struct dp_packet *packet = ctx->xin->packet; |
3338 | enum slow_path_reason slow; | |
a8448cb1 | 3339 | bool lacp_may_enable; |
642dc74d | 3340 | |
46c88433 | 3341 | if (!xport) { |
515793d5 | 3342 | slow = 0; |
46c88433 | 3343 | } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) { |
db7d4e46 | 3344 | if (packet) { |
46c88433 | 3345 | cfm_process_heartbeat(xport->cfm, packet); |
db7d4e46 | 3346 | } |
515793d5 | 3347 | slow = SLOW_CFM; |
fab52e16 | 3348 | } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) { |
db7d4e46 | 3349 | if (packet) { |
46c88433 | 3350 | bfd_process_packet(xport->bfd, flow, packet); |
60d02c72 AW |
3351 | /* If POLL received, immediately sends FINAL back. */ |
3352 | if (bfd_should_send_packet(xport->bfd)) { | |
6d308b28 | 3353 | ofproto_dpif_monitor_port_send_soon(xport->ofport); |
60d02c72 | 3354 | } |
db7d4e46 | 3355 | } |
515793d5 | 3356 | slow = SLOW_BFD; |
46c88433 | 3357 | } else if (xport->xbundle && xport->xbundle->lacp |
db7d4e46 JP |
3358 | && flow->dl_type == htons(ETH_TYPE_LACP)) { |
3359 | if (packet) { | |
a8448cb1 NK |
3360 | lacp_may_enable = lacp_process_packet(xport->xbundle->lacp, |
3361 | xport->ofport, packet); | |
3362 | /* Update LACP status in bond-slave to avoid packet-drops until | |
3363 | * LACP state machine is run by the main thread. */ | |
3364 | if (xport->xbundle->bond && lacp_may_enable) { | |
3365 | bond_slave_set_may_enable(xport->xbundle->bond, xport->ofport, | |
3366 | lacp_may_enable); | |
3367 | } | |
db7d4e46 | 3368 | } |
515793d5 | 3369 | slow = SLOW_LACP; |
9efd308e DV |
3370 | } else if ((xbridge->stp || xbridge->rstp) && |
3371 | stp_should_process_flow(flow, wc)) { | |
db7d4e46 | 3372 | if (packet) { |
f025bcb7 JR |
3373 | xbridge->stp |
3374 | ? stp_process_packet(xport, packet) | |
3375 | : rstp_process_packet(xport, packet); | |
db7d4e46 | 3376 | } |
515793d5 | 3377 | slow = SLOW_STP; |
19aef6ef | 3378 | } else if (xport->lldp && lldp_should_process_flow(xport->lldp, flow)) { |
0477baa9 DF |
3379 | if (packet) { |
3380 | lldp_process_packet(xport->lldp, packet); | |
3381 | } | |
515793d5 | 3382 | slow = SLOW_LLDP; |
db7d4e46 | 3383 | } else { |
515793d5 BP |
3384 | slow = 0; |
3385 | } | |
3386 | ||
3387 | if (slow) { | |
3388 | ctx->xout->slow |= slow; | |
3389 | return true; | |
3390 | } else { | |
3391 | return false; | |
db7d4e46 JP |
3392 | } |
3393 | } | |
3394 | ||
a36de779 | 3395 | static int |
0506f184 HH |
3396 | tnl_route_lookup_flow(const struct xlate_ctx *ctx, |
3397 | const struct flow *oflow, | |
a8704b50 PS |
3398 | struct in6_addr *ip, struct in6_addr *src, |
3399 | struct xport **out_port) | |
a36de779 PS |
3400 | { |
3401 | char out_dev[IFNAMSIZ]; | |
3402 | struct xbridge *xbridge; | |
c2b878e0 TLSC |
3403 | struct in6_addr gw; |
3404 | struct in6_addr dst; | |
a36de779 | 3405 | |
c2b878e0 | 3406 | dst = flow_tnl_dst(&oflow->tunnel); |
ed52ca57 | 3407 | if (!ovs_router_lookup(oflow->pkt_mark, &dst, out_dev, src, &gw)) { |
a36de779 PS |
3408 | return -ENOENT; |
3409 | } | |
3410 | ||
c2b878e0 TLSC |
3411 | if (ipv6_addr_is_set(&gw) && |
3412 | (!IN6_IS_ADDR_V4MAPPED(&gw) || in6_addr_get_mapped_ipv4(&gw))) { | |
a36de779 PS |
3413 | *ip = gw; |
3414 | } else { | |
c2b878e0 | 3415 | *ip = dst; |
a36de779 PS |
3416 | } |
3417 | ||
0506f184 | 3418 | HMAP_FOR_EACH (xbridge, hmap_node, &ctx->xcfg->xbridges) { |
a36de779 PS |
3419 | if (!strncmp(xbridge->name, out_dev, IFNAMSIZ)) { |
3420 | struct xport *port; | |
3421 | ||
3422 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
3423 | if (!strncmp(netdev_get_name(port->netdev), out_dev, IFNAMSIZ)) { | |
3424 | *out_port = port; | |
3425 | return 0; | |
3426 | } | |
3427 | } | |
3428 | } | |
3429 | } | |
dc0bd12f YS |
3430 | |
3431 | /* If tunnel IP isn't configured on bridges, then we search all ports. */ | |
3432 | HMAP_FOR_EACH (xbridge, hmap_node, &ctx->xcfg->xbridges) { | |
3433 | struct xport *port; | |
3434 | ||
3435 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
3436 | if (!strncmp(netdev_get_name(port->netdev), | |
3437 | out_dev, IFNAMSIZ)) { | |
3438 | *out_port = port; | |
3439 | return 0; | |
3440 | } | |
3441 | } | |
3442 | } | |
a36de779 PS |
3443 | return -ENOENT; |
3444 | } | |
3445 | ||
3446 | static int | |
cdd42eda JG |
3447 | compose_table_xlate(struct xlate_ctx *ctx, const struct xport *out_dev, |
3448 | struct dp_packet *packet) | |
a36de779 | 3449 | { |
cdd42eda | 3450 | struct xbridge *xbridge = out_dev->xbridge; |
ea41c034 | 3451 | ovs_version_t version = ofproto_dpif_get_tables_version(xbridge->ofproto); |
a36de779 PS |
3452 | struct ofpact_output output; |
3453 | struct flow flow; | |
3454 | ||
3455 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); | |
cf62fa4c | 3456 | flow_extract(packet, &flow); |
cdd42eda JG |
3457 | flow.in_port.ofp_port = out_dev->ofp_port; |
3458 | output.port = OFPP_TABLE; | |
a36de779 PS |
3459 | output.max_len = 0; |
3460 | ||
ea41c034 | 3461 | return ofproto_dpif_execute_actions__(xbridge->ofproto, version, &flow, |
1f4a8933 | 3462 | NULL, &output.ofpact, sizeof output, |
2d9b49dd | 3463 | ctx->depth, ctx->resubmits, packet); |
a36de779 PS |
3464 | } |
3465 | ||
c2b878e0 TLSC |
3466 | static void |
3467 | tnl_send_nd_request(struct xlate_ctx *ctx, const struct xport *out_dev, | |
3468 | const struct eth_addr eth_src, | |
3469 | struct in6_addr * ipv6_src, struct in6_addr * ipv6_dst) | |
3470 | { | |
3471 | struct dp_packet packet; | |
3472 | ||
3473 | dp_packet_init(&packet, 0); | |
16187903 | 3474 | compose_nd_ns(&packet, eth_src, ipv6_src, ipv6_dst); |
c2b878e0 TLSC |
3475 | compose_table_xlate(ctx, out_dev, &packet); |
3476 | dp_packet_uninit(&packet); | |
3477 | } | |
3478 | ||
a36de779 | 3479 | static void |
cdd42eda | 3480 | tnl_send_arp_request(struct xlate_ctx *ctx, const struct xport *out_dev, |
74ff3298 | 3481 | const struct eth_addr eth_src, |
a36de779 PS |
3482 | ovs_be32 ip_src, ovs_be32 ip_dst) |
3483 | { | |
cf62fa4c | 3484 | struct dp_packet packet; |
a36de779 | 3485 | |
cf62fa4c | 3486 | dp_packet_init(&packet, 0); |
eb0b295e BP |
3487 | compose_arp(&packet, ARP_OP_REQUEST, |
3488 | eth_src, eth_addr_zero, true, ip_src, ip_dst); | |
a36de779 | 3489 | |
cdd42eda | 3490 | compose_table_xlate(ctx, out_dev, &packet); |
cf62fa4c | 3491 | dp_packet_uninit(&packet); |
a36de779 PS |
3492 | } |
3493 | ||
7c12dfc5 SC |
3494 | static void |
3495 | propagate_tunnel_data_to_flow__(struct flow *dst_flow, | |
3496 | const struct flow *src_flow, | |
3497 | struct eth_addr dmac, struct eth_addr smac, | |
3498 | struct in6_addr s_ip6, ovs_be32 s_ip, | |
3499 | bool is_tnl_ipv6, uint8_t nw_proto) | |
3500 | { | |
3501 | dst_flow->dl_dst = dmac; | |
3502 | dst_flow->dl_src = smac; | |
3503 | ||
3504 | dst_flow->packet_type = htonl(PT_ETH); | |
3505 | dst_flow->nw_dst = src_flow->tunnel.ip_dst; | |
3506 | dst_flow->nw_src = src_flow->tunnel.ip_src; | |
3507 | dst_flow->ipv6_dst = src_flow->tunnel.ipv6_dst; | |
3508 | dst_flow->ipv6_src = src_flow->tunnel.ipv6_src; | |
3509 | ||
9cce7e71 | 3510 | dst_flow->nw_frag = 0; /* Tunnel packets are unfragmented. */ |
7c12dfc5 SC |
3511 | dst_flow->nw_tos = src_flow->tunnel.ip_tos; |
3512 | dst_flow->nw_ttl = src_flow->tunnel.ip_ttl; | |
3513 | dst_flow->tp_dst = src_flow->tunnel.tp_dst; | |
3514 | dst_flow->tp_src = src_flow->tunnel.tp_src; | |
3515 | ||
3516 | if (is_tnl_ipv6) { | |
3517 | dst_flow->dl_type = htons(ETH_TYPE_IPV6); | |
3518 | if (ipv6_mask_is_any(&dst_flow->ipv6_src) | |
3519 | && !ipv6_mask_is_any(&s_ip6)) { | |
3520 | dst_flow->ipv6_src = s_ip6; | |
3521 | } | |
3522 | } else { | |
3523 | dst_flow->dl_type = htons(ETH_TYPE_IP); | |
3524 | if (dst_flow->nw_src == 0 && s_ip) { | |
3525 | dst_flow->nw_src = s_ip; | |
3526 | } | |
3527 | } | |
3528 | dst_flow->nw_proto = nw_proto; | |
3529 | } | |
3530 | ||
3531 | /* | |
3532 | * Populate the 'flow' and 'base_flow' L3 fields to do the post tunnel push | |
3533 | * translations. | |
3534 | */ | |
3535 | static void | |
3536 | propagate_tunnel_data_to_flow(struct xlate_ctx *ctx, struct eth_addr dmac, | |
3537 | struct eth_addr smac, struct in6_addr s_ip6, | |
3538 | ovs_be32 s_ip, bool is_tnl_ipv6, | |
3539 | enum ovs_vport_type tnl_type) | |
3540 | { | |
3541 | struct flow *base_flow, *flow; | |
3542 | flow = &ctx->xin->flow; | |
3543 | base_flow = &ctx->base_flow; | |
3544 | uint8_t nw_proto = 0; | |
3545 | ||
3546 | switch (tnl_type) { | |
3547 | case OVS_VPORT_TYPE_GRE: | |
c387d817 GR |
3548 | case OVS_VPORT_TYPE_ERSPAN: |
3549 | case OVS_VPORT_TYPE_IP6ERSPAN: | |
3550 | case OVS_VPORT_TYPE_IP6GRE: | |
7c12dfc5 SC |
3551 | nw_proto = IPPROTO_GRE; |
3552 | break; | |
3553 | case OVS_VPORT_TYPE_VXLAN: | |
3554 | case OVS_VPORT_TYPE_GENEVE: | |
3555 | nw_proto = IPPROTO_UDP; | |
3556 | break; | |
3557 | case OVS_VPORT_TYPE_LISP: | |
3558 | case OVS_VPORT_TYPE_STT: | |
3559 | case OVS_VPORT_TYPE_UNSPEC: | |
3560 | case OVS_VPORT_TYPE_NETDEV: | |
3561 | case OVS_VPORT_TYPE_INTERNAL: | |
3562 | case __OVS_VPORT_TYPE_MAX: | |
3563 | default: | |
3564 | OVS_NOT_REACHED(); | |
7c12dfc5 SC |
3565 | } |
3566 | /* | |
3567 | * Update base_flow first followed by flow as the dst_flow gets modified | |
3568 | * in the function. | |
3569 | */ | |
3570 | propagate_tunnel_data_to_flow__(base_flow, flow, dmac, smac, s_ip6, s_ip, | |
3571 | is_tnl_ipv6, nw_proto); | |
3572 | propagate_tunnel_data_to_flow__(flow, flow, dmac, smac, s_ip6, s_ip, | |
3573 | is_tnl_ipv6, nw_proto); | |
3574 | } | |
3575 | ||
a36de779 | 3576 | static int |
11938578 AZ |
3577 | native_tunnel_output(struct xlate_ctx *ctx, const struct xport *xport, |
3578 | const struct flow *flow, odp_port_t tunnel_odp_port, | |
3579 | bool truncate) | |
a36de779 | 3580 | { |
4975aa3e | 3581 | struct netdev_tnl_build_header_params tnl_params; |
a36de779 PS |
3582 | struct ovs_action_push_tnl tnl_push_data; |
3583 | struct xport *out_dev = NULL; | |
c2b878e0 TLSC |
3584 | ovs_be32 s_ip = 0, d_ip = 0; |
3585 | struct in6_addr s_ip6 = in6addr_any; | |
3586 | struct in6_addr d_ip6 = in6addr_any; | |
74ff3298 JR |
3587 | struct eth_addr smac; |
3588 | struct eth_addr dmac; | |
a36de779 | 3589 | int err; |
c2b878e0 TLSC |
3590 | char buf_sip6[INET6_ADDRSTRLEN]; |
3591 | char buf_dip6[INET6_ADDRSTRLEN]; | |
a36de779 | 3592 | |
283d8662 ZB |
3593 | /* Store sFlow data. */ |
3594 | uint32_t sflow_n_outputs = ctx->sflow_n_outputs; | |
3595 | ||
7c12dfc5 SC |
3596 | /* Structures to backup Ethernet and IP of base_flow. */ |
3597 | struct flow old_base_flow; | |
3598 | struct flow old_flow; | |
3599 | ||
3600 | /* Backup flow & base_flow data. */ | |
3601 | memcpy(&old_base_flow, &ctx->base_flow, sizeof old_base_flow); | |
3602 | memcpy(&old_flow, &ctx->xin->flow, sizeof old_flow); | |
5a0e4aec | 3603 | |
8e4e4588 | 3604 | if (flow->tunnel.ip_src) { |
3605 | in6_addr_set_mapped_ipv4(&s_ip6, flow->tunnel.ip_src); | |
3606 | } | |
7c12dfc5 | 3607 | |
0506f184 | 3608 | err = tnl_route_lookup_flow(ctx, flow, &d_ip6, &s_ip6, &out_dev); |
a36de779 | 3609 | if (err) { |
2d9b49dd | 3610 | xlate_report(ctx, OFT_WARN, "native tunnel routing failed"); |
a36de779 PS |
3611 | return err; |
3612 | } | |
c2b878e0 | 3613 | |
2d9b49dd | 3614 | xlate_report(ctx, OFT_DETAIL, "tunneling to %s via %s", |
c2b878e0 TLSC |
3615 | ipv6_string_mapped(buf_dip6, &d_ip6), |
3616 | netdev_get_name(out_dev->netdev)); | |
a36de779 PS |
3617 | |
3618 | /* Use mac addr of bridge port of the peer. */ | |
74ff3298 | 3619 | err = netdev_get_etheraddr(out_dev->netdev, &smac); |
a36de779 | 3620 | if (err) { |
2d9b49dd BP |
3621 | xlate_report(ctx, OFT_WARN, |
3622 | "tunnel output device lacks Ethernet address"); | |
a36de779 PS |
3623 | return err; |
3624 | } | |
3625 | ||
c2b878e0 TLSC |
3626 | d_ip = in6_addr_get_mapped_ipv4(&d_ip6); |
3627 | if (d_ip) { | |
a8704b50 | 3628 | s_ip = in6_addr_get_mapped_ipv4(&s_ip6); |
a36de779 PS |
3629 | } |
3630 | ||
c2b878e0 | 3631 | err = tnl_neigh_lookup(out_dev->xbridge->name, &d_ip6, &dmac); |
a36de779 | 3632 | if (err) { |
2d9b49dd BP |
3633 | xlate_report(ctx, OFT_DETAIL, |
3634 | "neighbor cache miss for %s on bridge %s, " | |
c2b878e0 TLSC |
3635 | "sending %s request", |
3636 | buf_dip6, out_dev->xbridge->name, d_ip ? "ARP" : "ND"); | |
3637 | if (d_ip) { | |
3638 | tnl_send_arp_request(ctx, out_dev, smac, s_ip, d_ip); | |
3639 | } else { | |
3640 | tnl_send_nd_request(ctx, out_dev, smac, &s_ip6, &d_ip6); | |
3641 | } | |
a36de779 PS |
3642 | return err; |
3643 | } | |
c2b878e0 | 3644 | |
a36de779 PS |
3645 | if (ctx->xin->xcache) { |
3646 | struct xc_entry *entry; | |
3647 | ||
53902038 | 3648 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_NEIGH); |
901a517e JR |
3649 | ovs_strlcpy(entry->tnl_neigh_cache.br_name, out_dev->xbridge->name, |
3650 | sizeof entry->tnl_neigh_cache.br_name); | |
3651 | entry->tnl_neigh_cache.d_ipv6 = d_ip6; | |
a36de779 | 3652 | } |
81de18ec | 3653 | |
2d9b49dd | 3654 | xlate_report(ctx, OFT_DETAIL, "tunneling from "ETH_ADDR_FMT" %s" |
c2b878e0 TLSC |
3655 | " to "ETH_ADDR_FMT" %s", |
3656 | ETH_ADDR_ARGS(smac), ipv6_string_mapped(buf_sip6, &s_ip6), | |
3657 | ETH_ADDR_ARGS(dmac), buf_dip6); | |
3658 | ||
4975aa3e PS |
3659 | netdev_init_tnl_build_header_params(&tnl_params, flow, &s_ip6, dmac, smac); |
3660 | err = tnl_port_build_header(xport->ofport, &tnl_push_data, &tnl_params); | |
a36de779 PS |
3661 | if (err) { |
3662 | return err; | |
3663 | } | |
81765c00 BP |
3664 | tnl_push_data.tnl_port = tunnel_odp_port; |
3665 | tnl_push_data.out_port = out_dev->odp_port; | |
beb75a40 | 3666 | |
7c12dfc5 SC |
3667 | /* After tunnel header has been added, MAC and IP data of flow and |
3668 | * base_flow need to be set properly, since there is not recirculation | |
3669 | * any more when sending packet to tunnel. */ | |
beb75a40 | 3670 | |
11938578 AZ |
3671 | propagate_tunnel_data_to_flow(ctx, dmac, smac, s_ip6, |
3672 | s_ip, tnl_params.is_ipv6, | |
3673 | tnl_push_data.tnl_type); | |
7c12dfc5 | 3674 | |
11938578 AZ |
3675 | size_t clone_ofs = 0; |
3676 | size_t push_action_size; | |
7c12dfc5 | 3677 | |
11938578 AZ |
3678 | clone_ofs = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CLONE); |
3679 | odp_put_tnl_push_action(ctx->odp_actions, &tnl_push_data); | |
3680 | push_action_size = ctx->odp_actions->size; | |
3681 | ||
3682 | if (!truncate) { | |
3683 | const struct dpif_flow_stats *backup_resubmit_stats; | |
3684 | struct xlate_cache *backup_xcache; | |
3685 | struct flow_wildcards *backup_wc, wc; | |
3686 | bool backup_side_effects; | |
3687 | const struct dp_packet *backup_packet; | |
3688 | ||
3689 | memset(&wc, 0 , sizeof wc); | |
3690 | backup_wc = ctx->wc; | |
3691 | ctx->wc = &wc; | |
3692 | ctx->xin->wc = NULL; | |
3693 | backup_resubmit_stats = ctx->xin->resubmit_stats; | |
3694 | backup_xcache = ctx->xin->xcache; | |
3695 | backup_side_effects = ctx->xin->allow_side_effects; | |
3696 | backup_packet = ctx->xin->packet; | |
3697 | ||
3698 | ctx->xin->resubmit_stats = NULL; | |
3699 | ctx->xin->xcache = xlate_cache_new(); /* Use new temporary cache. */ | |
3700 | ctx->xin->allow_side_effects = false; | |
3701 | ctx->xin->packet = NULL; | |
3702 | ||
3703 | /* Push the cache entry for the tunnel first. */ | |
3704 | struct xc_entry *entry; | |
3705 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TUNNEL_HEADER); | |
3706 | entry->tunnel_hdr.hdr_size = tnl_push_data.header_len; | |
3707 | entry->tunnel_hdr.operation = ADD; | |
3708 | ||
3709 | patch_port_output(ctx, xport, out_dev); | |
3710 | ||
3711 | /* Similar to the stats update in revalidation, the x_cache entries | |
3712 | * are populated by the previous translation are used to update the | |
3713 | * stats correctly. | |
7c12dfc5 | 3714 | */ |
11938578 AZ |
3715 | if (backup_resubmit_stats) { |
3716 | struct dpif_flow_stats stats = *backup_resubmit_stats; | |
16441315 | 3717 | xlate_push_stats(ctx->xin->xcache, &stats, false); |
11938578 AZ |
3718 | } |
3719 | xlate_cache_steal_entries(backup_xcache, ctx->xin->xcache); | |
3720 | ||
3721 | if (ctx->odp_actions->size > push_action_size) { | |
3722 | nl_msg_end_non_empty_nested(ctx->odp_actions, clone_ofs); | |
3723 | } else { | |
3724 | nl_msg_cancel_nested(ctx->odp_actions, clone_ofs); | |
11938578 AZ |
3725 | } |
3726 | ||
3727 | /* Restore context status. */ | |
3728 | ctx->xin->resubmit_stats = backup_resubmit_stats; | |
3729 | xlate_cache_delete(ctx->xin->xcache); | |
3730 | ctx->xin->xcache = backup_xcache; | |
3731 | ctx->xin->allow_side_effects = backup_side_effects; | |
3732 | ctx->xin->packet = backup_packet; | |
3733 | ctx->wc = backup_wc; | |
3734 | } else { | |
3735 | /* In order to maintain accurate stats, use recirc for | |
3736 | * natvie tunneling. */ | |
7c12dfc5 SC |
3737 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, 0); |
3738 | nl_msg_end_nested(ctx->odp_actions, clone_ofs); | |
3739 | } | |
11938578 | 3740 | |
7c12dfc5 SC |
3741 | /* Restore the flows after the translation. */ |
3742 | memcpy(&ctx->xin->flow, &old_flow, sizeof ctx->xin->flow); | |
3743 | memcpy(&ctx->base_flow, &old_base_flow, sizeof ctx->base_flow); | |
283d8662 ZB |
3744 | |
3745 | /* Restore sFlow data. */ | |
3746 | ctx->sflow_n_outputs = sflow_n_outputs; | |
3747 | ||
a36de779 PS |
3748 | return 0; |
3749 | } | |
3750 | ||
704bb0bf JS |
3751 | static void |
3752 | xlate_commit_actions(struct xlate_ctx *ctx) | |
3753 | { | |
3754 | bool use_masked = ctx->xbridge->support.masked_set_action; | |
3755 | ||
3756 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, | |
3757 | ctx->odp_actions, ctx->wc, | |
1fc11c59 | 3758 | use_masked, ctx->pending_encap, |
88ec1e0a | 3759 | ctx->pending_decap, ctx->encap_data); |
f839892a | 3760 | ctx->pending_encap = false; |
88ec1e0a | 3761 | ctx->pending_decap = false; |
1fc11c59 JS |
3762 | ofpbuf_delete(ctx->encap_data); |
3763 | ctx->encap_data = NULL; | |
704bb0bf JS |
3764 | } |
3765 | ||
07659514 | 3766 | static void |
72fe7578 | 3767 | clear_conntrack(struct xlate_ctx *ctx) |
07659514 | 3768 | { |
72fe7578 | 3769 | ctx->conntracked = false; |
6846e91e | 3770 | flow_clear_conntrack(&ctx->xin->flow); |
07659514 JS |
3771 | } |
3772 | ||
58d636ee BK |
3773 | static bool |
3774 | xlate_flow_is_protected(const struct xlate_ctx *ctx, const struct flow *flow, const struct xport *xport_out) | |
3775 | { | |
3776 | const struct xport *xport_in; | |
3777 | ||
3778 | if (!xport_out) { | |
3779 | return false; | |
3780 | } | |
3781 | ||
3782 | xport_in = get_ofp_port(ctx->xbridge, flow->in_port.ofp_port); | |
3783 | ||
3784 | return (xport_in && xport_in->xbundle && xport_out->xbundle && | |
3785 | xport_in->xbundle->protected && xport_out->xbundle->protected); | |
3786 | } | |
3787 | ||
48f704f4 | 3788 | /* Function handles when a packet is sent from one bridge to another bridge. |
8bdb2bdb | 3789 | * |
48f704f4 AZ |
3790 | * The bridges are internally connected, either with patch ports or with |
3791 | * tunnel ports. | |
3792 | * | |
3793 | * The output action to another bridge causes translation to continue within | |
3794 | * the next bridge. This process can be recursive; the next bridge can | |
3795 | * output yet to another bridge. | |
3796 | * | |
3797 | * The translated actions from the second bridge onwards are enclosed within | |
3798 | * the clone action, so that any modification to the packet will not be visible | |
3799 | * to the remaining actions of the originating bridge. | |
8bdb2bdb SC |
3800 | */ |
3801 | static void | |
48f704f4 AZ |
3802 | patch_port_output(struct xlate_ctx *ctx, const struct xport *in_dev, |
3803 | struct xport *out_dev) | |
8bdb2bdb SC |
3804 | { |
3805 | struct flow *flow = &ctx->xin->flow; | |
3806 | struct flow old_flow = ctx->xin->flow; | |
3807 | struct flow_tnl old_flow_tnl_wc = ctx->wc->masks.tunnel; | |
3808 | bool old_conntrack = ctx->conntracked; | |
3809 | bool old_was_mpls = ctx->was_mpls; | |
3810 | ovs_version_t old_version = ctx->xin->tables_version; | |
3811 | struct ofpbuf old_stack = ctx->stack; | |
3812 | uint8_t new_stack[1024]; | |
3813 | struct ofpbuf old_action_set = ctx->action_set; | |
3814 | struct ovs_list *old_trace = ctx->xin->trace; | |
3815 | uint64_t actset_stub[1024 / 8]; | |
3816 | ||
3817 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); | |
3818 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); | |
3819 | flow->in_port.ofp_port = out_dev->ofp_port; | |
3820 | flow->metadata = htonll(0); | |
3821 | memset(&flow->tunnel, 0, sizeof flow->tunnel); | |
86bfb29a | 3822 | memset(&ctx->wc->masks.tunnel, 0, sizeof ctx->wc->masks.tunnel); |
8bdb2bdb SC |
3823 | flow->tunnel.metadata.tab = |
3824 | ofproto_get_tun_tab(&out_dev->xbridge->ofproto->up); | |
3825 | ctx->wc->masks.tunnel.metadata.tab = flow->tunnel.metadata.tab; | |
3826 | memset(flow->regs, 0, sizeof flow->regs); | |
3827 | flow->actset_output = OFPP_UNSET; | |
3828 | clear_conntrack(ctx); | |
3829 | ctx->xin->trace = xlate_report(ctx, OFT_BRIDGE, "bridge(\"%s\")", | |
3830 | out_dev->xbridge->name); | |
3831 | mirror_mask_t old_mirrors = ctx->mirrors; | |
3832 | bool independent_mirrors = out_dev->xbridge != ctx->xbridge; | |
3833 | if (independent_mirrors) { | |
3834 | ctx->mirrors = 0; | |
3835 | } | |
3836 | ctx->xbridge = out_dev->xbridge; | |
3837 | ||
3838 | /* The bridge is now known so obtain its table version. */ | |
3839 | ctx->xin->tables_version | |
3840 | = ofproto_dpif_get_tables_version(ctx->xbridge->ofproto); | |
3841 | ||
3842 | if (!process_special(ctx, out_dev) && may_receive(out_dev, ctx)) { | |
3843 | if (xport_stp_forward_state(out_dev) && | |
3844 | xport_rstp_forward_state(out_dev)) { | |
3845 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true, | |
96c3a6e5 | 3846 | false, true, clone_xlate_actions); |
8bdb2bdb SC |
3847 | if (!ctx->freezing) { |
3848 | xlate_action_set(ctx); | |
3849 | } | |
3850 | if (ctx->freezing) { | |
3851 | finish_freezing(ctx); | |
3852 | } | |
3853 | } else { | |
3854 | /* Forwarding is disabled by STP and RSTP. Let OFPP_NORMAL and | |
3855 | * the learning action look at the packet, then drop it. */ | |
3856 | struct flow old_base_flow = ctx->base_flow; | |
3857 | size_t old_size = ctx->odp_actions->size; | |
3858 | mirror_mask_t old_mirrors2 = ctx->mirrors; | |
3859 | ||
3860 | xlate_table_action(ctx, flow->in_port.ofp_port, 0, true, true, | |
96c3a6e5 | 3861 | false, true, clone_xlate_actions); |
8bdb2bdb SC |
3862 | ctx->mirrors = old_mirrors2; |
3863 | ctx->base_flow = old_base_flow; | |
3864 | ctx->odp_actions->size = old_size; | |
3865 | ||
3866 | /* Undo changes that may have been done for freezing. */ | |
3867 | ctx_cancel_freeze(ctx); | |
3868 | } | |
3869 | } | |
3870 | ||
3871 | ctx->xin->trace = old_trace; | |
3872 | if (independent_mirrors) { | |
3873 | ctx->mirrors = old_mirrors; | |
3874 | } | |
3875 | ctx->xin->flow = old_flow; | |
3876 | ctx->xbridge = in_dev->xbridge; | |
3877 | ofpbuf_uninit(&ctx->action_set); | |
3878 | ctx->action_set = old_action_set; | |
3879 | ofpbuf_uninit(&ctx->stack); | |
3880 | ctx->stack = old_stack; | |
3881 | ||
3882 | /* Restore calling bridge's lookup version. */ | |
3883 | ctx->xin->tables_version = old_version; | |
3884 | ||
3885 | /* Restore to calling bridge tunneling information */ | |
3886 | ctx->wc->masks.tunnel = old_flow_tnl_wc; | |
3887 | ||
3888 | /* The out bridge popping MPLS should have no effect on the original | |
3889 | * bridge. */ | |
3890 | ctx->was_mpls = old_was_mpls; | |
3891 | ||
3892 | /* The out bridge's conntrack execution should have no effect on the | |
3893 | * original bridge. */ | |
3894 | ctx->conntracked = old_conntrack; | |
3895 | ||
3896 | /* The fact that the out bridge exits (for any reason) does not mean | |
3897 | * that the original bridge should exit. Specifically, if the out | |
3898 | * bridge freezes translation, the original bridge must continue | |
3899 | * processing with the original, not the frozen packet! */ | |
3900 | ctx->exit = false; | |
3901 | ||
3902 | /* Out bridge errors do not propagate back. */ | |
3903 | ctx->error = XLATE_OK; | |
3904 | ||
3905 | if (ctx->xin->resubmit_stats) { | |
3906 | netdev_vport_inc_tx(in_dev->netdev, ctx->xin->resubmit_stats); | |
3907 | netdev_vport_inc_rx(out_dev->netdev, ctx->xin->resubmit_stats); | |
3908 | if (out_dev->bfd) { | |
3909 | bfd_account_rx(out_dev->bfd, ctx->xin->resubmit_stats); | |
3910 | } | |
3911 | } | |
3912 | if (ctx->xin->xcache) { | |
3913 | struct xc_entry *entry; | |
3914 | ||
3915 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
3916 | entry->dev.tx = netdev_ref(in_dev->netdev); | |
3917 | entry->dev.rx = netdev_ref(out_dev->netdev); | |
3918 | entry->dev.bfd = bfd_ref(out_dev->bfd); | |
3919 | } | |
3920 | } | |
3921 | ||
081617f0 JS |
3922 | static bool |
3923 | check_output_prerequisites(struct xlate_ctx *ctx, | |
3924 | const struct xport *xport, | |
3925 | struct flow *flow, | |
3926 | bool check_stp) | |
9583bc14 | 3927 | { |
49a73e0c | 3928 | struct flow_wildcards *wc = ctx->wc; |
9583bc14 | 3929 | |
46c88433 | 3930 | if (!xport) { |
2d9b49dd | 3931 | xlate_report(ctx, OFT_WARN, "Nonexistent output port"); |
081617f0 | 3932 | return false; |
46c88433 | 3933 | } else if (xport->config & OFPUTIL_PC_NO_FWD) { |
2d9b49dd | 3934 | xlate_report(ctx, OFT_DETAIL, "OFPPC_NO_FWD set, skipping output"); |
081617f0 | 3935 | return false; |
1356dbd1 | 3936 | } else if (ctx->mirror_snaplen != 0 && xport->odp_port == ODPP_NONE) { |
2d9b49dd BP |
3937 | xlate_report(ctx, OFT_WARN, |
3938 | "Mirror truncate to ODPP_NONE, skipping output"); | |
081617f0 | 3939 | return false; |
58d636ee | 3940 | } else if (xlate_flow_is_protected(ctx, flow, xport)) { |
2d9b49dd BP |
3941 | xlate_report(ctx, OFT_WARN, |
3942 | "Flow is between protected ports, skipping output."); | |
081617f0 | 3943 | return false; |
0d1cee12 | 3944 | } else if (check_stp) { |
bbbca389 | 3945 | if (is_stp(&ctx->base_flow)) { |
9efd308e DV |
3946 | if (!xport_stp_should_forward_bpdu(xport) && |
3947 | !xport_rstp_should_manage_bpdu(xport)) { | |
3948 | if (ctx->xbridge->stp != NULL) { | |
2d9b49dd BP |
3949 | xlate_report(ctx, OFT_WARN, |
3950 | "STP not in listening state, " | |
3951 | "skipping bpdu output"); | |
9efd308e | 3952 | } else if (ctx->xbridge->rstp != NULL) { |
2d9b49dd BP |
3953 | xlate_report(ctx, OFT_WARN, |
3954 | "RSTP not managing BPDU in this state, " | |
3955 | "skipping bpdu output"); | |
9efd308e | 3956 | } |
081617f0 | 3957 | return false; |
0d1cee12 | 3958 | } |
67818616 MV |
3959 | } else if ((xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) |
3960 | || (xport->bfd && bfd_should_process_flow(xport->bfd, flow, | |
3961 | wc))) { | |
3962 | /* Pass; STP should not block link health detection. */ | |
9efd308e DV |
3963 | } else if (!xport_stp_forward_state(xport) || |
3964 | !xport_rstp_forward_state(xport)) { | |
3965 | if (ctx->xbridge->stp != NULL) { | |
2d9b49dd BP |
3966 | xlate_report(ctx, OFT_WARN, |
3967 | "STP not in forwarding state, skipping output"); | |
9efd308e | 3968 | } else if (ctx->xbridge->rstp != NULL) { |
2d9b49dd BP |
3969 | xlate_report(ctx, OFT_WARN, |
3970 | "RSTP not in forwarding state, skipping output"); | |
9efd308e | 3971 | } |
081617f0 | 3972 | return false; |
0d1cee12 | 3973 | } |
9583bc14 | 3974 | } |
5dbfe239 ZB |
3975 | |
3976 | if (xport->pt_mode == NETDEV_PT_LEGACY_L2 && | |
3977 | flow->packet_type != htonl(PT_ETH)) { | |
3978 | xlate_report(ctx, OFT_WARN, "Trying to send non-Ethernet packet " | |
3979 | "through legacy L2 port. Dropping packet."); | |
3980 | return false; | |
3981 | } | |
3982 | ||
081617f0 JS |
3983 | return true; |
3984 | } | |
3985 | ||
83c2757b ZB |
3986 | /* Function verifies if destination address of received Neighbor Advertisement |
3987 | * message stored in 'flow' is correct. It should be either FF02::1:FFXX:XXXX | |
3988 | * where XX:XXXX stands for the last 24 bits of 'ipv6_addr' or it should match | |
3989 | * 'ipv6_addr'. */ | |
3990 | static bool | |
3991 | is_nd_dst_correct(const struct flow *flow, const struct in6_addr *ipv6_addr) | |
3992 | { | |
3993 | const uint8_t *flow_ipv6_addr = (uint8_t *) &flow->ipv6_dst; | |
3994 | const uint8_t *addr = (uint8_t *) ipv6_addr; | |
3995 | ||
68fd9251 | 3996 | return (IN6_IS_ADDR_MC_LINKLOCAL(&flow->ipv6_dst) && |
83c2757b ZB |
3997 | flow_ipv6_addr[11] == 0x01 && |
3998 | flow_ipv6_addr[12] == 0xff && | |
3999 | flow_ipv6_addr[13] == addr[13] && | |
4000 | flow_ipv6_addr[14] == addr[14] && | |
4001 | flow_ipv6_addr[15] == addr[15]) || | |
4002 | IN6_ARE_ADDR_EQUAL(&flow->ipv6_dst, ipv6_addr); | |
4003 | } | |
4004 | ||
dc0bd12f YS |
4005 | static bool |
4006 | is_neighbor_reply_matched(const struct flow *flow, struct in6_addr *ip_addr) | |
4007 | { | |
4008 | return ((IN6_IS_ADDR_V4MAPPED(ip_addr) && | |
4009 | flow->dl_type == htons(ETH_TYPE_ARP) && | |
4010 | in6_addr_get_mapped_ipv4(ip_addr) == flow->nw_dst) || | |
4011 | (!IN6_IS_ADDR_V4MAPPED(ip_addr) && | |
4012 | is_nd_dst_correct(flow, ip_addr))); | |
4013 | } | |
4014 | ||
83c2757b ZB |
4015 | /* Function verifies if the ARP reply or Neighbor Advertisement represented by |
4016 | * 'flow' addresses the 'xbridge' of 'ctx'. Returns true if the ARP TA or | |
4017 | * neighbor discovery destination is in the list of configured IP addresses of | |
4018 | * the bridge. Otherwise, it returns false. */ | |
4019 | static bool | |
4020 | is_neighbor_reply_correct(const struct xlate_ctx *ctx, const struct flow *flow) | |
4021 | { | |
4022 | bool ret = false; | |
4023 | int i; | |
4024 | struct xbridge_addr *xbridge_addr = xbridge_addr_ref(ctx->xbridge->addr); | |
4025 | ||
4026 | /* Verify if 'nw_dst' of ARP or 'ipv6_dst' of ICMPV6 is in the list. */ | |
4027 | for (i = 0; xbridge_addr && i < xbridge_addr->n_addr; i++) { | |
4028 | struct in6_addr *ip_addr = &xbridge_addr->addr[i]; | |
dc0bd12f | 4029 | if (is_neighbor_reply_matched(flow, ip_addr)) { |
83c2757b ZB |
4030 | /* Found a match. */ |
4031 | ret = true; | |
4032 | break; | |
4033 | } | |
4034 | } | |
4035 | ||
4036 | xbridge_addr_unref(xbridge_addr); | |
dc0bd12f YS |
4037 | |
4038 | /* If not found in bridge's IPs, search in its ports. */ | |
4039 | if (!ret) { | |
4040 | struct in6_addr *ip_addr, *mask; | |
4041 | struct xport *port; | |
4042 | int error, n_in6; | |
4043 | ||
4044 | HMAP_FOR_EACH (port, ofp_node, &ctx->xbridge->xports) { | |
4045 | error = netdev_get_addr_list(port->netdev, &ip_addr, | |
4046 | &mask, &n_in6); | |
26228f44 DS |
4047 | if (!error) { |
4048 | ret = is_neighbor_reply_matched(flow, ip_addr); | |
4049 | free(ip_addr); | |
4050 | free(mask); | |
4051 | if (ret) { | |
4052 | /* Found a match. */ | |
4053 | break; | |
4054 | } | |
dc0bd12f YS |
4055 | } |
4056 | } | |
4057 | } | |
83c2757b ZB |
4058 | return ret; |
4059 | } | |
4060 | ||
081617f0 | 4061 | static bool |
dc0bd12f YS |
4062 | terminate_native_tunnel(struct xlate_ctx *ctx, struct flow *flow, |
4063 | struct flow_wildcards *wc, odp_port_t *tnl_port) | |
081617f0 JS |
4064 | { |
4065 | *tnl_port = ODPP_NONE; | |
4066 | ||
4067 | /* XXX: Write better Filter for tunnel port. We can use in_port | |
4068 | * in tunnel-port flow to avoid these checks completely. */ | |
dc0bd12f | 4069 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
081617f0 | 4070 | *tnl_port = tnl_port_map_lookup(flow, wc); |
83c2757b ZB |
4071 | |
4072 | /* If no tunnel port was found and it's about an ARP or ICMPv6 packet, | |
4073 | * do tunnel neighbor snooping. */ | |
4074 | if (*tnl_port == ODPP_NONE && | |
4075 | (flow->dl_type == htons(ETH_TYPE_ARP) || | |
4076 | flow->nw_proto == IPPROTO_ICMPV6) && | |
4077 | is_neighbor_reply_correct(ctx, flow)) { | |
4078 | tnl_neigh_snoop(flow, wc, ctx->xbridge->name); | |
4079 | } | |
081617f0 JS |
4080 | } |
4081 | ||
4082 | return *tnl_port != ODPP_NONE; | |
4083 | } | |
4084 | ||
4085 | static void | |
4086 | compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, | |
feee58b9 | 4087 | const struct xlate_bond_recirc *xr, bool check_stp, |
11938578 | 4088 | bool is_last_action OVS_UNUSED, bool truncate) |
081617f0 JS |
4089 | { |
4090 | const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); | |
4091 | struct flow_wildcards *wc = ctx->wc; | |
4092 | struct flow *flow = &ctx->xin->flow; | |
4093 | struct flow_tnl flow_tnl; | |
4094 | union flow_vlan_hdr flow_vlans[FLOW_MAX_VLAN_HEADERS]; | |
4095 | uint8_t flow_nw_tos; | |
4096 | odp_port_t out_port, odp_port, odp_tnl_port; | |
4097 | bool is_native_tunnel = false; | |
4098 | uint8_t dscp; | |
5dbfe239 ZB |
4099 | struct eth_addr flow_dl_dst = flow->dl_dst; |
4100 | struct eth_addr flow_dl_src = flow->dl_src; | |
4101 | ovs_be32 flow_packet_type = flow->packet_type; | |
4102 | ovs_be16 flow_dl_type = flow->dl_type; | |
081617f0 JS |
4103 | |
4104 | /* If 'struct flow' gets additional metadata, we'll need to zero it out | |
4105 | * before traversing a patch port. */ | |
7dc18ae9 | 4106 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 41); |
081617f0 JS |
4107 | memset(&flow_tnl, 0, sizeof flow_tnl); |
4108 | ||
4109 | if (!check_output_prerequisites(ctx, xport, flow, check_stp)) { | |
4110 | return; | |
4111 | } | |
9583bc14 | 4112 | |
875ab130 BP |
4113 | if (flow->packet_type == htonl(PT_ETH)) { |
4114 | /* Strip Ethernet header for legacy L3 port. */ | |
4115 | if (xport->pt_mode == NETDEV_PT_LEGACY_L3) { | |
4116 | flow->packet_type = PACKET_TYPE_BE(OFPHTN_ETHERTYPE, | |
4117 | ntohs(flow->dl_type)); | |
4118 | } | |
beb75a40 JS |
4119 | } |
4120 | ||
46c88433 | 4121 | if (xport->peer) { |
5cb92182 BP |
4122 | if (truncate) { |
4123 | xlate_report_error(ctx, "Cannot truncate output to patch port"); | |
4124 | } | |
48f704f4 | 4125 | patch_port_output(ctx, xport, xport->peer); |
8bdb2bdb | 4126 | return; |
9583bc14 EJ |
4127 | } |
4128 | ||
f0fb825a | 4129 | memcpy(flow_vlans, flow->vlans, sizeof flow_vlans); |
33bf9176 | 4130 | flow_nw_tos = flow->nw_tos; |
9583bc14 | 4131 | |
16194afd DDP |
4132 | if (count_skb_priorities(xport)) { |
4133 | memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority); | |
4134 | if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { | |
4135 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
4136 | flow->nw_tos &= ~IP_DSCP_MASK; | |
4137 | flow->nw_tos |= dscp; | |
4138 | } | |
9583bc14 EJ |
4139 | } |
4140 | ||
46c88433 | 4141 | if (xport->is_tunnel) { |
c2b878e0 | 4142 | struct in6_addr dst; |
9583bc14 EJ |
4143 | /* Save tunnel metadata so that changes made due to |
4144 | * the Logical (tunnel) Port are not visible for any further | |
4145 | * matches, while explicit set actions on tunnel metadata are. | |
4146 | */ | |
a36de779 | 4147 | flow_tnl = flow->tunnel; |
49a73e0c | 4148 | odp_port = tnl_port_send(xport->ofport, flow, ctx->wc); |
4e022ec0 | 4149 | if (odp_port == ODPP_NONE) { |
2d9b49dd | 4150 | xlate_report(ctx, OFT_WARN, "Tunneling decided against output"); |
9583bc14 EJ |
4151 | goto out; /* restore flow_nw_tos */ |
4152 | } | |
c2b878e0 TLSC |
4153 | dst = flow_tnl_dst(&flow->tunnel); |
4154 | if (ipv6_addr_equals(&dst, &ctx->orig_tunnel_ipv6_dst)) { | |
2d9b49dd | 4155 | xlate_report(ctx, OFT_WARN, "Not tunneling to our own address"); |
9583bc14 EJ |
4156 | goto out; /* restore flow_nw_tos */ |
4157 | } | |
4158 | if (ctx->xin->resubmit_stats) { | |
46c88433 | 4159 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
9583bc14 | 4160 | } |
b256dc52 JS |
4161 | if (ctx->xin->xcache) { |
4162 | struct xc_entry *entry; | |
4163 | ||
4164 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
901a517e | 4165 | entry->dev.tx = netdev_ref(xport->netdev); |
b256dc52 | 4166 | } |
9583bc14 | 4167 | out_port = odp_port; |
a36de779 | 4168 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
2d9b49dd | 4169 | xlate_report(ctx, OFT_DETAIL, "output to native tunnel"); |
081617f0 | 4170 | is_native_tunnel = true; |
a36de779 | 4171 | } else { |
c6d87201 WT |
4172 | const char *tnl_type; |
4173 | ||
2d9b49dd | 4174 | xlate_report(ctx, OFT_DETAIL, "output to kernel tunnel"); |
c6d87201 WT |
4175 | tnl_type = tnl_port_get_type(xport->ofport); |
4176 | commit_odp_tunnel_action(flow, &ctx->base_flow, | |
4177 | ctx->odp_actions, tnl_type); | |
a36de779 PS |
4178 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ |
4179 | } | |
9583bc14 | 4180 | } else { |
46c88433 | 4181 | odp_port = xport->odp_port; |
7614e5d0 | 4182 | out_port = odp_port; |
9583bc14 | 4183 | } |
9583bc14 | 4184 | |
4e022ec0 | 4185 | if (out_port != ODPP_NONE) { |
081617f0 | 4186 | /* Commit accumulated flow updates before output. */ |
704bb0bf | 4187 | xlate_commit_actions(ctx); |
adcf00ba | 4188 | |
e93ef1c7 | 4189 | if (xr) { |
081617f0 | 4190 | /* Recirculate the packet. */ |
347bf289 | 4191 | struct ovs_action_hash *act_hash; |
adcf00ba | 4192 | |
347bf289 | 4193 | /* Hash action. */ |
6a0b0d3b JS |
4194 | enum ovs_hash_alg hash_alg = xr->hash_alg; |
4195 | if (hash_alg > ctx->xbridge->support.max_hash_alg) { | |
4196 | /* Algorithm supported by all datapaths. */ | |
4197 | hash_alg = OVS_HASH_ALG_L4; | |
4198 | } | |
1520ef4f | 4199 | act_hash = nl_msg_put_unspec_uninit(ctx->odp_actions, |
347bf289 AZ |
4200 | OVS_ACTION_ATTR_HASH, |
4201 | sizeof *act_hash); | |
6a0b0d3b | 4202 | act_hash->hash_alg = hash_alg; |
62ac1f20 | 4203 | act_hash->hash_basis = xr->hash_basis; |
347bf289 AZ |
4204 | |
4205 | /* Recirc action. */ | |
1520ef4f | 4206 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, |
347bf289 | 4207 | xr->recirc_id); |
081617f0 JS |
4208 | } else if (is_native_tunnel) { |
4209 | /* Output to native tunnel port. */ | |
11938578 | 4210 | native_tunnel_output(ctx, xport, flow, odp_port, truncate); |
081617f0 | 4211 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ |
a36de779 | 4212 | |
dc0bd12f | 4213 | } else if (terminate_native_tunnel(ctx, flow, wc, |
081617f0 JS |
4214 | &odp_tnl_port)) { |
4215 | /* Intercept packet to be received on native tunnel port. */ | |
4216 | nl_msg_put_odp_port(ctx->odp_actions, OVS_ACTION_ATTR_TUNNEL_POP, | |
4217 | odp_tnl_port); | |
a36de779 | 4218 | |
081617f0 JS |
4219 | } else { |
4220 | /* Tunnel push-pop action is not compatible with | |
4221 | * IPFIX action. */ | |
4222 | compose_ipfix_action(ctx, out_port); | |
4223 | ||
4224 | /* Handle truncation of the mirrored packet. */ | |
4225 | if (ctx->mirror_snaplen > 0 && | |
4226 | ctx->mirror_snaplen < UINT16_MAX) { | |
4227 | struct ovs_action_trunc *trunc; | |
4228 | ||
4229 | trunc = nl_msg_put_unspec_uninit(ctx->odp_actions, | |
4230 | OVS_ACTION_ATTR_TRUNC, | |
4231 | sizeof *trunc); | |
4232 | trunc->max_len = ctx->mirror_snaplen; | |
4233 | if (!ctx->xbridge->support.trunc) { | |
4234 | ctx->xout->slow |= SLOW_ACTION; | |
1356dbd1 WT |
4235 | } |
4236 | } | |
081617f0 JS |
4237 | |
4238 | nl_msg_put_odp_port(ctx->odp_actions, | |
4239 | OVS_ACTION_ATTR_OUTPUT, | |
4240 | out_port); | |
adcf00ba | 4241 | } |
9583bc14 | 4242 | |
6cbbf4fa EJ |
4243 | ctx->sflow_odp_port = odp_port; |
4244 | ctx->sflow_n_outputs++; | |
2031ef97 | 4245 | ctx->nf_output_iface = ofp_port; |
6cbbf4fa EJ |
4246 | } |
4247 | ||
7efbc3b7 BP |
4248 | if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { |
4249 | mirror_packet(ctx, xport->xbundle, | |
4250 | xbundle_mirror_dst(xport->xbundle->xbridge, | |
4251 | xport->xbundle)); | |
4252 | } | |
4253 | ||
6cbbf4fa | 4254 | out: |
9583bc14 | 4255 | /* Restore flow */ |
f0fb825a | 4256 | memcpy(flow->vlans, flow_vlans, sizeof flow->vlans); |
33bf9176 | 4257 | flow->nw_tos = flow_nw_tos; |
5dbfe239 ZB |
4258 | flow->dl_dst = flow_dl_dst; |
4259 | flow->dl_src = flow_dl_src; | |
4260 | flow->packet_type = flow_packet_type; | |
4261 | flow->dl_type = flow_dl_type; | |
9583bc14 EJ |
4262 | } |
4263 | ||
4264 | static void | |
e93ef1c7 | 4265 | compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
feee58b9 | 4266 | const struct xlate_bond_recirc *xr, |
11938578 | 4267 | bool is_last_action, bool truncate) |
9583bc14 | 4268 | { |
11938578 AZ |
4269 | compose_output_action__(ctx, ofp_port, xr, true, |
4270 | is_last_action, truncate); | |
9583bc14 EJ |
4271 | } |
4272 | ||
bb61b33d | 4273 | static void |
feee58b9 | 4274 | xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule, |
96c3a6e5 AZ |
4275 | bool deepens, bool is_last_action, |
4276 | xlate_actions_handler *actions_xlator) | |
bb61b33d BP |
4277 | { |
4278 | struct rule_dpif *old_rule = ctx->rule; | |
8b1e5560 | 4279 | ovs_be64 old_cookie = ctx->rule_cookie; |
dc723c44 | 4280 | const struct rule_actions *actions; |
bb61b33d BP |
4281 | |
4282 | if (ctx->xin->resubmit_stats) { | |
16441315 | 4283 | rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats, false); |
bb61b33d BP |
4284 | } |
4285 | ||
98b07853 | 4286 | ctx->resubmits++; |
790c5d26 | 4287 | |
790c5d26 | 4288 | ctx->depth += deepens; |
bb61b33d | 4289 | ctx->rule = rule; |
07a3cd5c BP |
4290 | ctx->rule_cookie = rule->up.flow_cookie; |
4291 | actions = rule_get_actions(&rule->up); | |
96c3a6e5 | 4292 | actions_xlator(actions->ofpacts, actions->ofpacts_len, ctx, |
f5634764 | 4293 | is_last_action, false); |
8b1e5560 | 4294 | ctx->rule_cookie = old_cookie; |
bb61b33d | 4295 | ctx->rule = old_rule; |
790c5d26 | 4296 | ctx->depth -= deepens; |
bb61b33d BP |
4297 | } |
4298 | ||
bd3240ba SH |
4299 | static bool |
4300 | xlate_resubmit_resource_check(struct xlate_ctx *ctx) | |
9583bc14 | 4301 | { |
790c5d26 | 4302 | if (ctx->depth >= MAX_DEPTH) { |
2d9b49dd | 4303 | xlate_report_error(ctx, "over max translation depth %d", MAX_DEPTH); |
fff1b9c0 | 4304 | ctx->error = XLATE_RECURSION_TOO_DEEP; |
790c5d26 | 4305 | } else if (ctx->resubmits >= MAX_RESUBMITS) { |
2d9b49dd | 4306 | xlate_report_error(ctx, "over %d resubmit actions", MAX_RESUBMITS); |
fff1b9c0 | 4307 | ctx->error = XLATE_TOO_MANY_RESUBMITS; |
1520ef4f | 4308 | } else if (ctx->odp_actions->size > UINT16_MAX) { |
2d9b49dd | 4309 | xlate_report_error(ctx, "resubmits yielded over 64 kB of actions"); |
fff1b9c0 JR |
4310 | /* NOT an error, as we'll be slow-pathing the flow in this case? */ |
4311 | ctx->exit = true; /* XXX: translation still terminated! */ | |
6fd6ed71 | 4312 | } else if (ctx->stack.size >= 65536) { |
2d9b49dd | 4313 | xlate_report_error(ctx, "resubmits yielded over 64 kB of stack"); |
fff1b9c0 | 4314 | ctx->error = XLATE_STACK_TOO_DEEP; |
98b07853 | 4315 | } else { |
bd3240ba SH |
4316 | return true; |
4317 | } | |
4318 | ||
4319 | return false; | |
4320 | } | |
4321 | ||
2cd20955 JR |
4322 | static void |
4323 | tuple_swap_flow(struct flow *flow, bool ipv4) | |
4324 | { | |
4325 | uint8_t nw_proto = flow->nw_proto; | |
4326 | flow->nw_proto = flow->ct_nw_proto; | |
4327 | flow->ct_nw_proto = nw_proto; | |
4328 | ||
4329 | if (ipv4) { | |
4330 | ovs_be32 nw_src = flow->nw_src; | |
4331 | flow->nw_src = flow->ct_nw_src; | |
4332 | flow->ct_nw_src = nw_src; | |
4333 | ||
4334 | ovs_be32 nw_dst = flow->nw_dst; | |
4335 | flow->nw_dst = flow->ct_nw_dst; | |
4336 | flow->ct_nw_dst = nw_dst; | |
4337 | } else { | |
4338 | struct in6_addr ipv6_src = flow->ipv6_src; | |
4339 | flow->ipv6_src = flow->ct_ipv6_src; | |
4340 | flow->ct_ipv6_src = ipv6_src; | |
4341 | ||
4342 | struct in6_addr ipv6_dst = flow->ipv6_dst; | |
4343 | flow->ipv6_dst = flow->ct_ipv6_dst; | |
4344 | flow->ct_ipv6_dst = ipv6_dst; | |
4345 | } | |
4346 | ||
4347 | ovs_be16 tp_src = flow->tp_src; | |
4348 | flow->tp_src = flow->ct_tp_src; | |
4349 | flow->ct_tp_src = tp_src; | |
4350 | ||
4351 | ovs_be16 tp_dst = flow->tp_dst; | |
4352 | flow->tp_dst = flow->ct_tp_dst; | |
4353 | flow->ct_tp_dst = tp_dst; | |
4354 | } | |
4355 | ||
4356 | static void | |
4357 | tuple_swap(struct flow *flow, struct flow_wildcards *wc) | |
4358 | { | |
4359 | bool ipv4 = (flow->dl_type == htons(ETH_TYPE_IP)); | |
4360 | ||
4361 | tuple_swap_flow(flow, ipv4); | |
4362 | tuple_swap_flow(&wc->masks, ipv4); | |
4363 | } | |
4364 | ||
bd3240ba | 4365 | static void |
6d328fa2 | 4366 | xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, |
2cd20955 | 4367 | bool may_packet_in, bool honor_table_miss, |
96c3a6e5 AZ |
4368 | bool with_ct_orig, bool is_last_action, |
4369 | xlate_actions_handler *xlator) | |
bd3240ba | 4370 | { |
e12ec36b SH |
4371 | /* Check if we need to recirculate before matching in a table. */ |
4372 | if (ctx->was_mpls) { | |
4373 | ctx_trigger_freeze(ctx); | |
4374 | return; | |
4375 | } | |
bd3240ba | 4376 | if (xlate_resubmit_resource_check(ctx)) { |
9583bc14 | 4377 | uint8_t old_table_id = ctx->table_id; |
3f207910 | 4378 | struct rule_dpif *rule; |
9583bc14 EJ |
4379 | |
4380 | ctx->table_id = table_id; | |
4381 | ||
2cd20955 JR |
4382 | /* Swap packet fields with CT 5-tuple if requested. */ |
4383 | if (with_ct_orig) { | |
4384 | /* Do not swap if there is no CT tuple, or if key is not IP. */ | |
4385 | if (ctx->xin->flow.ct_nw_proto == 0 || | |
4386 | !is_ip_any(&ctx->xin->flow)) { | |
4387 | xlate_report_error(ctx, | |
4388 | "resubmit(ct) with non-tracked or non-IP packet!"); | |
7b950521 | 4389 | ctx->table_id = old_table_id; |
2cd20955 JR |
4390 | return; |
4391 | } | |
4392 | tuple_swap(&ctx->xin->flow, ctx->wc); | |
4393 | } | |
34dd0d78 | 4394 | rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, |
1f4a8933 | 4395 | ctx->xin->tables_version, |
c0e638aa | 4396 | &ctx->xin->flow, ctx->wc, |
34dd0d78 JR |
4397 | ctx->xin->resubmit_stats, |
4398 | &ctx->table_id, in_port, | |
a027899e JR |
4399 | may_packet_in, honor_table_miss, |
4400 | ctx->xin->xcache); | |
2cd20955 JR |
4401 | /* Swap back. */ |
4402 | if (with_ct_orig) { | |
4403 | tuple_swap(&ctx->xin->flow, ctx->wc); | |
4404 | } | |
ad3efdcb | 4405 | |
a2143702 | 4406 | if (rule) { |
83709dfa JR |
4407 | /* Fill in the cache entry here instead of xlate_recursively |
4408 | * to make the reference counting more explicit. We take a | |
4409 | * reference in the lookups above if we are going to cache the | |
4410 | * rule. */ | |
4411 | if (ctx->xin->xcache) { | |
4412 | struct xc_entry *entry; | |
4413 | ||
4414 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE); | |
901a517e | 4415 | entry->rule = rule; |
07a3cd5c | 4416 | ofproto_rule_ref(&rule->up); |
83709dfa | 4417 | } |
2d9b49dd BP |
4418 | |
4419 | struct ovs_list *old_trace = ctx->xin->trace; | |
4420 | xlate_report_table(ctx, rule, table_id); | |
feee58b9 | 4421 | xlate_recursively(ctx, rule, table_id <= old_table_id, |
96c3a6e5 | 4422 | is_last_action, xlator); |
2d9b49dd | 4423 | ctx->xin->trace = old_trace; |
ad3efdcb EJ |
4424 | } |
4425 | ||
9583bc14 | 4426 | ctx->table_id = old_table_id; |
98b07853 | 4427 | return; |
9583bc14 EJ |
4428 | } |
4429 | } | |
4430 | ||
76973237 | 4431 | /* Consumes the group reference, which is only taken if xcache exists. */ |
f4fb341b | 4432 | static void |
1e684d7d RW |
4433 | xlate_group_stats(struct xlate_ctx *ctx, struct group_dpif *group, |
4434 | struct ofputil_bucket *bucket) | |
4435 | { | |
4436 | if (ctx->xin->resubmit_stats) { | |
4437 | group_dpif_credit_stats(group, bucket, ctx->xin->resubmit_stats); | |
4438 | } | |
4439 | if (ctx->xin->xcache) { | |
4440 | struct xc_entry *entry; | |
4441 | ||
4442 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_GROUP); | |
901a517e JR |
4443 | entry->group.group = group; |
4444 | entry->group.bucket = bucket; | |
1e684d7d RW |
4445 | } |
4446 | } | |
4447 | ||
4448 | static void | |
feee58b9 AZ |
4449 | xlate_group_bucket(struct xlate_ctx *ctx, struct ofputil_bucket *bucket, |
4450 | bool is_last_action) | |
f4fb341b | 4451 | { |
fac4786a BP |
4452 | struct ovs_list *old_trace = ctx->xin->trace; |
4453 | if (OVS_UNLIKELY(ctx->xin->trace)) { | |
4454 | char *s = xasprintf("bucket %"PRIu32, bucket->bucket_id); | |
4455 | ctx->xin->trace = &oftrace_report(ctx->xin->trace, OFT_BUCKET, | |
4456 | s)->subs; | |
4457 | free(s); | |
4458 | } | |
4459 | ||
f4fb341b | 4460 | uint64_t action_list_stub[1024 / 8]; |
0a2869d5 BP |
4461 | struct ofpbuf action_list = OFPBUF_STUB_INITIALIZER(action_list_stub); |
4462 | struct ofpbuf action_set = ofpbuf_const_initializer(bucket->ofpacts, | |
4463 | bucket->ofpacts_len); | |
5b09e569 | 4464 | struct flow old_flow = ctx->xin->flow; |
e12ec36b | 4465 | bool old_was_mpls = ctx->was_mpls; |
f4fb341b | 4466 | |
f4fb341b | 4467 | ofpacts_execute_action_set(&action_list, &action_set); |
790c5d26 | 4468 | ctx->depth++; |
f5634764 KG |
4469 | do_xlate_actions(action_list.data, action_list.size, ctx, is_last_action, |
4470 | true); | |
790c5d26 | 4471 | ctx->depth--; |
f4fb341b | 4472 | |
f4fb341b | 4473 | ofpbuf_uninit(&action_list); |
5b09e569 | 4474 | |
77ab5fd2 | 4475 | /* Check if need to freeze. */ |
1d361a81 | 4476 | if (ctx->freezing) { |
77ab5fd2 | 4477 | finish_freezing(ctx); |
e672ff9b JR |
4478 | } |
4479 | ||
5b09e569 JR |
4480 | /* Roll back flow to previous state. |
4481 | * This is equivalent to cloning the packet for each bucket. | |
4482 | * | |
4483 | * As a side effect any subsequently applied actions will | |
4484 | * also effectively be applied to a clone of the packet taken | |
4485 | * just before applying the all or indirect group. | |
4486 | * | |
4487 | * Note that group buckets are action sets, hence they cannot modify the | |
4488 | * main action set. Also any stack actions are ignored when executing an | |
afc0c379 | 4489 | * action set, so group buckets cannot directly change the stack either. |
5b09e569 | 4490 | * However, we do allow resubmit actions in group buckets, which could |
afc0c379 BP |
4491 | * recursively execute actions that do modify the action set or change the |
4492 | * stack. The controller must be careful about what it does to the | |
4493 | * action_set and stack in the tables resubmitted to from group buckets. */ | |
5b09e569 JR |
4494 | ctx->xin->flow = old_flow; |
4495 | ||
e12ec36b SH |
4496 | /* The group bucket popping MPLS should have no effect after bucket |
4497 | * execution. */ | |
4498 | ctx->was_mpls = old_was_mpls; | |
4499 | ||
5b09e569 JR |
4500 | /* The fact that the group bucket exits (for any reason) does not mean that |
4501 | * the translation after the group action should exit. Specifically, if | |
1d361a81 BP |
4502 | * the group bucket freezes translation, the actions after the group action |
4503 | * must continue processing with the original, not the frozen packet! */ | |
5b09e569 | 4504 | ctx->exit = false; |
b99654b0 VDA |
4505 | |
4506 | /* Context error in a bucket should not impact processing of other buckets | |
4507 | * or actions. This is similar to cloning a packet for group buckets. | |
4508 | * There is no need to restore the error back to old value due to the fact | |
4509 | * that we actually processed group action which can happen only when there | |
4510 | * is no previous context error. | |
4511 | * | |
4512 | * Exception to above is errors which are system limits to protect | |
4513 | * translation from running too long or occupy too much space. These errors | |
4514 | * should not be masked. XLATE_RECURSION_TOO_DEEP, XLATE_TOO_MANY_RESUBMITS | |
4515 | * and XLATE_STACK_TOO_DEEP fall in this category. */ | |
4516 | if (ctx->error == XLATE_TOO_MANY_MPLS_LABELS || | |
4517 | ctx->error == XLATE_UNSUPPORTED_PACKET_TYPE) { | |
4518 | /* reset the error and continue processing other buckets */ | |
4519 | ctx->error = XLATE_OK; | |
4520 | } | |
f4fb341b | 4521 | |
fac4786a | 4522 | ctx->xin->trace = old_trace; |
f4fb341b SH |
4523 | } |
4524 | ||
a04e5888 BP |
4525 | static struct ofputil_bucket * |
4526 | pick_ff_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
dd8cd4b4 | 4527 | { |
a04e5888 | 4528 | return group_first_live_bucket(ctx, group, 0); |
dd8cd4b4 SH |
4529 | } |
4530 | ||
a04e5888 BP |
4531 | static struct ofputil_bucket * |
4532 | pick_default_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
fe7e5749 | 4533 | { |
a04e5888 BP |
4534 | flow_mask_hash_fields(&ctx->xin->flow, ctx->wc, |
4535 | NX_HASH_FIELDS_SYMMETRIC_L4); | |
4536 | return group_best_live_bucket(ctx, group, | |
4537 | flow_hash_symmetric_l4(&ctx->xin->flow, 0)); | |
fe7e5749 SH |
4538 | } |
4539 | ||
a04e5888 BP |
4540 | static struct ofputil_bucket * |
4541 | pick_hash_fields_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
0c4b9393 | 4542 | { |
07a3cd5c BP |
4543 | const struct field_array *fields = &group->up.props.fields; |
4544 | const uint8_t *mask_values = fields->values; | |
4545 | uint32_t basis = hash_uint64(group->up.props.selection_method_param); | |
0c4b9393 | 4546 | |
07a3cd5c | 4547 | size_t i; |
e8dba719 JR |
4548 | BITMAP_FOR_EACH_1 (i, MFF_N_IDS, fields->used.bm) { |
4549 | const struct mf_field *mf = mf_from_id(i); | |
0c4b9393 | 4550 | |
5bcd4754 | 4551 | /* Skip fields for which prerequisites are not met. */ |
e8dba719 JR |
4552 | if (!mf_are_prereqs_ok(mf, &ctx->xin->flow, ctx->wc)) { |
4553 | /* Skip the mask bytes for this field. */ | |
4554 | mask_values += mf->n_bytes; | |
4555 | continue; | |
4556 | } | |
0c4b9393 | 4557 | |
e8dba719 JR |
4558 | union mf_value value; |
4559 | union mf_value mask; | |
0c4b9393 | 4560 | |
e8dba719 JR |
4561 | mf_get_value(mf, &ctx->xin->flow, &value); |
4562 | /* Mask the value. */ | |
4563 | for (int j = 0; j < mf->n_bytes; j++) { | |
4564 | mask.b[j] = *mask_values++; | |
4565 | value.b[j] &= mask.b[j]; | |
4566 | } | |
4567 | basis = hash_bytes(&value, mf->n_bytes, basis); | |
1cb20095 | 4568 | |
e8dba719 JR |
4569 | /* For tunnels, hash in whether the field is present. */ |
4570 | if (mf_is_tun_metadata(mf)) { | |
4571 | basis = hash_boolean(mf_is_set(mf, &ctx->xin->flow), basis); | |
0c4b9393 | 4572 | } |
e8dba719 JR |
4573 | |
4574 | mf_mask_field_masked(mf, &mask, ctx->wc); | |
0c4b9393 SH |
4575 | } |
4576 | ||
a04e5888 | 4577 | return group_best_live_bucket(ctx, group, basis); |
0c4b9393 SH |
4578 | } |
4579 | ||
a04e5888 BP |
4580 | static struct ofputil_bucket * |
4581 | pick_dp_hash_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
53cc166a | 4582 | { |
2e3fd24c JS |
4583 | uint32_t dp_hash = ctx->xin->flow.dp_hash; |
4584 | ||
53cc166a JR |
4585 | /* dp_hash value 0 is special since it means that the dp_hash has not been |
4586 | * computed, as all computed dp_hash values are non-zero. Therefore | |
4587 | * compare to zero can be used to decide if the dp_hash value is valid | |
4588 | * without masking the dp_hash field. */ | |
2e3fd24c JS |
4589 | if (!dp_hash) { |
4590 | enum ovs_hash_alg hash_alg = group->hash_alg; | |
4591 | if (hash_alg > ctx->xbridge->support.max_hash_alg) { | |
4592 | /* Algorithm supported by all datapaths. */ | |
4593 | hash_alg = OVS_HASH_ALG_L4; | |
4594 | } | |
4595 | ctx_trigger_recirculate_with_hash(ctx, hash_alg, group->hash_basis); | |
a04e5888 | 4596 | return NULL; |
53cc166a | 4597 | } else { |
2e3fd24c JS |
4598 | uint32_t hash_mask = group->hash_mask; |
4599 | ctx->wc->masks.dp_hash |= hash_mask; | |
4600 | ||
4601 | /* Starting from the original masked dp_hash value iterate over the | |
4602 | * hash mapping table to find the first live bucket. As the buckets | |
4603 | * are quasi-randomly spread over the hash values, this maintains | |
4604 | * a distribution according to bucket weights even when some buckets | |
4605 | * are non-live. */ | |
4606 | for (int i = 0; i <= hash_mask; i++) { | |
4607 | struct ofputil_bucket *b = | |
4608 | group->hash_map[(dp_hash + i) & hash_mask]; | |
4609 | if (bucket_is_alive(ctx, b, 0)) { | |
4610 | return b; | |
4611 | } | |
53cc166a | 4612 | } |
2e3fd24c | 4613 | |
a04e5888 | 4614 | return NULL; |
53cc166a JR |
4615 | } |
4616 | } | |
4617 | ||
a04e5888 BP |
4618 | static struct ofputil_bucket * |
4619 | pick_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
7565c3e4 | 4620 | { |
e12ec36b SH |
4621 | /* Select groups may access flow keys beyond L2 in order to |
4622 | * select a bucket. Recirculate as appropriate to make this possible. | |
4623 | */ | |
4624 | if (ctx->was_mpls) { | |
4625 | ctx_trigger_freeze(ctx); | |
991f1f63 | 4626 | return NULL; |
e12ec36b SH |
4627 | } |
4628 | ||
2e3fd24c JS |
4629 | switch (group->selection_method) { |
4630 | case SEL_METHOD_DEFAULT: | |
a04e5888 | 4631 | return pick_default_select_group(ctx, group); |
2e3fd24c JS |
4632 | break; |
4633 | case SEL_METHOD_HASH: | |
a04e5888 | 4634 | return pick_hash_fields_select_group(ctx, group); |
2e3fd24c JS |
4635 | break; |
4636 | case SEL_METHOD_DP_HASH: | |
a04e5888 | 4637 | return pick_dp_hash_select_group(ctx, group); |
2e3fd24c JS |
4638 | break; |
4639 | default: | |
4640 | /* Parsing of groups ensures this never happens */ | |
7565c3e4 SH |
4641 | OVS_NOT_REACHED(); |
4642 | } | |
2e3fd24c JS |
4643 | |
4644 | return NULL; | |
7565c3e4 SH |
4645 | } |
4646 | ||
f4fb341b | 4647 | static void |
feee58b9 AZ |
4648 | xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group, |
4649 | bool is_last_action) | |
f4fb341b | 4650 | { |
a04e5888 | 4651 | if (group->up.type == OFPGT11_ALL || group->up.type == OFPGT11_INDIRECT) { |
03121ac4 | 4652 | struct ovs_list *last_bucket = group->up.buckets.prev; |
a04e5888 BP |
4653 | struct ofputil_bucket *bucket; |
4654 | LIST_FOR_EACH (bucket, list_node, &group->up.buckets) { | |
4655 | bool is_last_bucket = &bucket->list_node == last_bucket; | |
4656 | xlate_group_bucket(ctx, bucket, is_last_action && is_last_bucket); | |
4657 | } | |
4658 | xlate_group_stats(ctx, group, NULL); | |
4659 | } else { | |
4660 | struct ofputil_bucket *bucket; | |
4661 | if (group->up.type == OFPGT11_SELECT) { | |
4662 | bucket = pick_select_group(ctx, group); | |
4663 | } else if (group->up.type == OFPGT11_FF) { | |
4664 | bucket = pick_ff_group(ctx, group); | |
4665 | } else { | |
4666 | OVS_NOT_REACHED(); | |
4667 | } | |
fac4786a | 4668 | |
a04e5888 | 4669 | if (bucket) { |
fac4786a BP |
4670 | xlate_report(ctx, OFT_DETAIL, "using bucket %"PRIu32, |
4671 | bucket->bucket_id); | |
a04e5888 BP |
4672 | xlate_group_bucket(ctx, bucket, is_last_action); |
4673 | xlate_group_stats(ctx, group, bucket); | |
fac4786a BP |
4674 | } else { |
4675 | xlate_report(ctx, OFT_DETAIL, "no live bucket"); | |
4676 | if (ctx->xin->xcache) { | |
4677 | ofproto_group_unref(&group->up); | |
4678 | } | |
a04e5888 | 4679 | } |
f4fb341b | 4680 | } |
f4fb341b SH |
4681 | } |
4682 | ||
4683 | static bool | |
feee58b9 AZ |
4684 | xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id, |
4685 | bool is_last_action) | |
f4fb341b | 4686 | { |
0eb48fe1 | 4687 | if (xlate_resubmit_resource_check(ctx)) { |
f4fb341b | 4688 | struct group_dpif *group; |
f4fb341b | 4689 | |
76973237 JR |
4690 | /* Take ref only if xcache exists. */ |
4691 | group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, | |
1f4a8933 | 4692 | ctx->xin->tables_version, ctx->xin->xcache); |
db88b35c JR |
4693 | if (!group) { |
4694 | /* XXX: Should set ctx->error ? */ | |
2d9b49dd BP |
4695 | xlate_report(ctx, OFT_WARN, "output to nonexistent group %"PRIu32, |
4696 | group_id); | |
f4fb341b SH |
4697 | return true; |
4698 | } | |
feee58b9 | 4699 | xlate_group_action__(ctx, group, is_last_action); |
f4fb341b SH |
4700 | } |
4701 | ||
4702 | return false; | |
4703 | } | |
4704 | ||
9583bc14 EJ |
4705 | static void |
4706 | xlate_ofpact_resubmit(struct xlate_ctx *ctx, | |
feee58b9 AZ |
4707 | const struct ofpact_resubmit *resubmit, |
4708 | bool is_last_action) | |
9583bc14 | 4709 | { |
4e022ec0 | 4710 | ofp_port_t in_port; |
9583bc14 | 4711 | uint8_t table_id; |
adcf00ba AZ |
4712 | bool may_packet_in = false; |
4713 | bool honor_table_miss = false; | |
4714 | ||
4715 | if (ctx->rule && rule_dpif_is_internal(ctx->rule)) { | |
4716 | /* Still allow missed packets to be sent to the controller | |
4717 | * if resubmitting from an internal table. */ | |
4718 | may_packet_in = true; | |
4719 | honor_table_miss = true; | |
4720 | } | |
9583bc14 EJ |
4721 | |
4722 | in_port = resubmit->in_port; | |
4723 | if (in_port == OFPP_IN_PORT) { | |
4e022ec0 | 4724 | in_port = ctx->xin->flow.in_port.ofp_port; |
9583bc14 EJ |
4725 | } |
4726 | ||
4727 | table_id = resubmit->table_id; | |
4728 | if (table_id == 255) { | |
4729 | table_id = ctx->table_id; | |
4730 | } | |
4731 | ||
adcf00ba | 4732 | xlate_table_action(ctx, in_port, table_id, may_packet_in, |
feee58b9 | 4733 | honor_table_miss, resubmit->with_ct_orig, |
96c3a6e5 | 4734 | is_last_action, do_xlate_actions); |
9583bc14 EJ |
4735 | } |
4736 | ||
4737 | static void | |
feee58b9 AZ |
4738 | flood_packet_to_port(struct xlate_ctx *ctx, const struct xport *xport, |
4739 | bool all, bool is_last_action) | |
9583bc14 | 4740 | { |
feee58b9 AZ |
4741 | if (!xport) { |
4742 | return; | |
4743 | } | |
4744 | ||
4745 | if (all) { | |
4746 | compose_output_action__(ctx, xport->ofp_port, NULL, false, | |
11938578 | 4747 | is_last_action, false); |
feee58b9 | 4748 | } else { |
11938578 AZ |
4749 | compose_output_action(ctx, xport->ofp_port, NULL, is_last_action, |
4750 | false); | |
feee58b9 AZ |
4751 | } |
4752 | } | |
4753 | ||
4754 | static void | |
4755 | flood_packets(struct xlate_ctx *ctx, bool all, bool is_last_action) | |
4756 | { | |
4757 | const struct xport *xport, *last = NULL; | |
9583bc14 | 4758 | |
feee58b9 | 4759 | /* Use 'last' the keep track of the last output port. */ |
46c88433 EJ |
4760 | HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) { |
4761 | if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
4762 | continue; |
4763 | } | |
4764 | ||
feee58b9 AZ |
4765 | if (all || !(xport->config & OFPUTIL_PC_NO_FLOOD)) { |
4766 | /* 'last' is not the last port, send a packet out, and | |
4767 | * update 'last'. */ | |
4768 | flood_packet_to_port(ctx, last, all, false); | |
4769 | last = xport; | |
9583bc14 EJ |
4770 | } |
4771 | } | |
4772 | ||
feee58b9 AZ |
4773 | /* Send the packet to the 'last' port. */ |
4774 | flood_packet_to_port(ctx, last, all, is_last_action); | |
2031ef97 | 4775 | ctx->nf_output_iface = NF_OUT_FLOOD; |
9583bc14 EJ |
4776 | } |
4777 | ||
74c4530d JP |
4778 | static void |
4779 | put_controller_user_action(struct xlate_ctx *ctx, | |
4780 | bool dont_send, bool continuation, | |
4781 | uint32_t recirc_id, int len, | |
4782 | enum ofp_packet_in_reason reason, | |
4783 | uint16_t controller_id) | |
4784 | { | |
4785 | struct user_action_cookie cookie; | |
4786 | ||
4787 | memset(&cookie, 0, sizeof cookie); | |
4788 | cookie.type = USER_ACTION_COOKIE_CONTROLLER; | |
4789 | cookie.ofp_in_port = OFPP_NONE, | |
4790 | cookie.ofproto_uuid = ctx->xbridge->ofproto->uuid; | |
4791 | cookie.controller.dont_send = dont_send; | |
4792 | cookie.controller.continuation = continuation; | |
4793 | cookie.controller.reason = reason; | |
4794 | cookie.controller.recirc_id = recirc_id; | |
4795 | put_32aligned_be64(&cookie.controller.rule_cookie, ctx->rule_cookie); | |
4796 | cookie.controller.controller_id = controller_id; | |
4797 | cookie.controller.max_len = len; | |
4798 | ||
4799 | odp_port_t odp_port = ofp_port_to_odp_port(ctx->xbridge, | |
4800 | ctx->xin->flow.in_port.ofp_port); | |
769b5034 | 4801 | uint32_t pid = dpif_port_get_pid(ctx->xbridge->dpif, odp_port); |
74c4530d JP |
4802 | odp_put_userspace_action(pid, &cookie, sizeof cookie, ODPP_NONE, |
4803 | false, ctx->odp_actions); | |
4804 | } | |
4805 | ||
9583bc14 | 4806 | static void |
d39ec23d JP |
4807 | xlate_controller_action(struct xlate_ctx *ctx, int len, |
4808 | enum ofp_packet_in_reason reason, | |
4809 | uint16_t controller_id, | |
206ddb9a | 4810 | uint32_t provider_meter_id, |
d39ec23d | 4811 | const uint8_t *userdata, size_t userdata_len) |
9583bc14 | 4812 | { |
b476e2f2 | 4813 | xlate_commit_actions(ctx); |
beb75a40 | 4814 | |
9bfe9334 BP |
4815 | /* A packet sent by an action in a table-miss rule is considered an |
4816 | * explicit table miss. OpenFlow before 1.3 doesn't have that concept so | |
4817 | * it will get translated back to OFPR_ACTION for those versions. */ | |
4818 | if (reason == OFPR_ACTION | |
07a3cd5c | 4819 | && ctx->rule && rule_is_table_miss(&ctx->rule->up)) { |
9bfe9334 BP |
4820 | reason = OFPR_EXPLICIT_MISS; |
4821 | } | |
4822 | ||
d39ec23d JP |
4823 | struct frozen_state state = { |
4824 | .table_id = ctx->table_id, | |
4825 | .ofproto_uuid = ctx->xbridge->ofproto->uuid, | |
4826 | .stack = ctx->stack.data, | |
4827 | .stack_size = ctx->stack.size, | |
4828 | .mirrors = ctx->mirrors, | |
4829 | .conntracked = ctx->conntracked, | |
aeb6566d | 4830 | .was_mpls = ctx->was_mpls, |
d39ec23d JP |
4831 | .ofpacts = NULL, |
4832 | .ofpacts_len = 0, | |
4833 | .action_set = NULL, | |
4834 | .action_set_len = 0, | |
4835 | .userdata = CONST_CAST(uint8_t *, userdata), | |
4836 | .userdata_len = userdata_len, | |
9bfe9334 | 4837 | }; |
d39ec23d | 4838 | frozen_metadata_from_flow(&state.metadata, &ctx->xin->flow); |
9583bc14 | 4839 | |
d39ec23d JP |
4840 | uint32_t recirc_id = recirc_alloc_id_ctx(&state); |
4841 | if (!recirc_id) { | |
4842 | xlate_report_error(ctx, "Failed to allocate recirculation id"); | |
4843 | ctx->error = XLATE_NO_RECIRCULATION_CONTEXT; | |
4844 | return; | |
4845 | } | |
4846 | recirc_refs_add(&ctx->xout->recircs, recirc_id); | |
df70a773 | 4847 | |
206ddb9a JP |
4848 | /* If the controller action didn't request a meter (indicated by a |
4849 | * 'meter_id' argument other than NX_CTLR_NO_METER), see if one was | |
4850 | * configured through the "controller" virtual meter. | |
4851 | * | |
4852 | * Internally, ovs-vswitchd uses UINT32_MAX to indicate no meter is | |
4853 | * configured. */ | |
4854 | uint32_t meter_id; | |
4855 | if (provider_meter_id == UINT32_MAX) { | |
4856 | meter_id = ctx->xbridge->ofproto->up.controller_meter_id; | |
4857 | } else { | |
4858 | meter_id = provider_meter_id; | |
4859 | } | |
4860 | ||
d39ec23d JP |
4861 | size_t offset; |
4862 | size_t ac_offset; | |
d39ec23d JP |
4863 | if (meter_id != UINT32_MAX) { |
4864 | /* If controller meter is configured, generate clone(meter, userspace) | |
4865 | * action. */ | |
4866 | offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_SAMPLE); | |
4867 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, | |
4868 | UINT32_MAX); | |
4869 | ac_offset = nl_msg_start_nested(ctx->odp_actions, | |
4870 | OVS_SAMPLE_ATTR_ACTIONS); | |
4871 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_METER, meter_id); | |
df70a773 | 4872 | } |
3b4fff43 | 4873 | |
d39ec23d JP |
4874 | /* Generate the datapath flows even if we don't send the packet-in |
4875 | * so that debugging more closely represents normal state. */ | |
74c4530d | 4876 | bool dont_send = false; |
d39ec23d | 4877 | if (!ctx->xin->allow_side_effects && !ctx->xin->xcache) { |
74c4530d | 4878 | dont_send = true; |
d39ec23d | 4879 | } |
74c4530d JP |
4880 | put_controller_user_action(ctx, dont_send, false, recirc_id, len, |
4881 | reason, controller_id); | |
d39ec23d JP |
4882 | |
4883 | if (meter_id != UINT32_MAX) { | |
4884 | nl_msg_end_nested(ctx->odp_actions, ac_offset); | |
4885 | nl_msg_end_nested(ctx->odp_actions, offset); | |
4886 | } | |
9583bc14 EJ |
4887 | } |
4888 | ||
e6bc8e74 YHW |
4889 | /* Creates a frozen state, and allocates a unique recirc id for the given |
4890 | * state. Returns a non-zero recirc id if it is allocated successfully. | |
4891 | * Returns 0 otherwise. | |
4892 | **/ | |
4893 | static uint32_t | |
77ab5fd2 BP |
4894 | finish_freezing__(struct xlate_ctx *ctx, uint8_t table) |
4895 | { | |
1d361a81 | 4896 | ovs_assert(ctx->freezing); |
7bbdd84f | 4897 | |
1d361a81 | 4898 | struct frozen_state state = { |
07659514 | 4899 | .table_id = table, |
07a3cd5c | 4900 | .ofproto_uuid = ctx->xbridge->ofproto->uuid, |
5c1b2314 | 4901 | .stack = ctx->stack.data, |
84cf3c1f | 4902 | .stack_size = ctx->stack.size, |
29bae541 | 4903 | .mirrors = ctx->mirrors, |
07659514 | 4904 | .conntracked = ctx->conntracked, |
aeb6566d | 4905 | .was_mpls = ctx->was_mpls, |
00135b86 | 4906 | .xport_uuid = ctx->xin->xport_uuid, |
1d361a81 BP |
4907 | .ofpacts = ctx->frozen_actions.data, |
4908 | .ofpacts_len = ctx->frozen_actions.size, | |
417509fa | 4909 | .action_set = ctx->action_set.data, |
8a5fb3b4 | 4910 | .action_set_len = ctx->action_set.size, |
74c4530d JP |
4911 | .userdata = ctx->pause ? CONST_CAST(uint8_t *,ctx->pause->userdata) |
4912 | : NULL, | |
4913 | .userdata_len = ctx->pause ? ctx->pause->userdata_len : 0, | |
2082425c | 4914 | }; |
77ab5fd2 | 4915 | frozen_metadata_from_flow(&state.metadata, &ctx->xin->flow); |
2082425c | 4916 | |
74c4530d JP |
4917 | /* Allocate a unique recirc id for the given metadata state in the |
4918 | * flow. An existing id, with a new reference to the corresponding | |
4919 | * recirculation context, will be returned if possible. | |
4920 | * The life-cycle of this recirc id is managed by associating it | |
4921 | * with the udpif key ('ukey') created for each new datapath flow. */ | |
4922 | uint32_t recirc_id = recirc_alloc_id_ctx(&state); | |
4923 | if (!recirc_id) { | |
4924 | xlate_report_error(ctx, "Failed to allocate recirculation id"); | |
4925 | ctx->error = XLATE_NO_RECIRCULATION_CONTEXT; | |
4926 | return 0; | |
4927 | } | |
4928 | recirc_refs_add(&ctx->xout->recircs, recirc_id); | |
4929 | ||
77ab5fd2 | 4930 | if (ctx->pause) { |
74c4530d | 4931 | if (!ctx->xin->allow_side_effects && !ctx->xin->xcache) { |
e6bc8e74 | 4932 | return 0; |
77ab5fd2 | 4933 | } |
7bbdd84f | 4934 | |
74c4530d JP |
4935 | put_controller_user_action(ctx, false, true, recirc_id, |
4936 | ctx->pause->max_len, | |
4937 | ctx->pause->reason, | |
4938 | ctx->pause->controller_id); | |
4939 | } else { | |
53cc166a JR |
4940 | if (ctx->recirc_update_dp_hash) { |
4941 | struct ovs_action_hash *act_hash; | |
4942 | ||
4943 | /* Hash action. */ | |
4944 | act_hash = nl_msg_put_unspec_uninit(ctx->odp_actions, | |
4945 | OVS_ACTION_ATTR_HASH, | |
4946 | sizeof *act_hash); | |
2e3fd24c JS |
4947 | act_hash->hash_alg = ctx->dp_hash_alg; |
4948 | act_hash->hash_basis = ctx->dp_hash_basis; | |
53cc166a | 4949 | } |
74c4530d | 4950 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_RECIRC, recirc_id); |
77ab5fd2 | 4951 | } |
e672ff9b | 4952 | |
1d361a81 BP |
4953 | /* Undo changes done by freezing. */ |
4954 | ctx_cancel_freeze(ctx); | |
74c4530d | 4955 | return recirc_id; |
7bbdd84f SH |
4956 | } |
4957 | ||
1d361a81 | 4958 | /* Called only when we're freezing. */ |
07659514 | 4959 | static void |
77ab5fd2 | 4960 | finish_freezing(struct xlate_ctx *ctx) |
07659514 JS |
4961 | { |
4962 | xlate_commit_actions(ctx); | |
77ab5fd2 | 4963 | finish_freezing__(ctx, 0); |
07659514 JS |
4964 | } |
4965 | ||
e37b8437 JS |
4966 | /* Fork the pipeline here. The current packet will continue processing the |
4967 | * current action list. A clone of the current packet will recirculate, skip | |
4968 | * the remainder of the current action list and asynchronously resume pipeline | |
4969 | * processing in 'table' with the current metadata and action set. */ | |
4970 | static void | |
5fdd80cc YHW |
4971 | compose_recirculate_and_fork(struct xlate_ctx *ctx, uint8_t table, |
4972 | const uint16_t zone) | |
e37b8437 | 4973 | { |
e6bc8e74 | 4974 | uint32_t recirc_id; |
1d361a81 | 4975 | ctx->freezing = true; |
e6bc8e74 YHW |
4976 | recirc_id = finish_freezing__(ctx, table); |
4977 | ||
4978 | if (OVS_UNLIKELY(ctx->xin->trace) && recirc_id) { | |
4979 | if (oftrace_add_recirc_node(ctx->xin->recirc_queue, | |
4980 | OFT_RECIRC_CONNTRACK, &ctx->xin->flow, | |
5fdd80cc | 4981 | ctx->xin->packet, recirc_id, zone)) { |
e6bc8e74 YHW |
4982 | xlate_report(ctx, OFT_DETAIL, "A clone of the packet is forked to " |
4983 | "recirculate. The forked pipeline will be resumed at " | |
4984 | "table %u.", table); | |
4985 | } else { | |
4986 | xlate_report(ctx, OFT_DETAIL, "Failed to trace the conntrack " | |
4987 | "forked pipeline with recirc_id = %d.", recirc_id); | |
4988 | } | |
4989 | } | |
e37b8437 JS |
4990 | } |
4991 | ||
8bfd0fda BP |
4992 | static void |
4993 | compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) | |
9583bc14 | 4994 | { |
33bf9176 | 4995 | struct flow *flow = &ctx->xin->flow; |
8bfd0fda | 4996 | int n; |
33bf9176 | 4997 | |
8bfd0fda | 4998 | ovs_assert(eth_type_mpls(mpls->ethertype)); |
b0a17866 | 4999 | |
49a73e0c | 5000 | n = flow_count_mpls_labels(flow, ctx->wc); |
8bfd0fda | 5001 | if (!n) { |
704bb0bf | 5002 | xlate_commit_actions(ctx); |
8bfd0fda BP |
5003 | } else if (n >= FLOW_MAX_MPLS_LABELS) { |
5004 | if (ctx->xin->packet != NULL) { | |
2d9b49dd BP |
5005 | xlate_report_error(ctx, "dropping packet on which an MPLS push " |
5006 | "action can't be performed as it would have " | |
5007 | "more MPLS LSEs than the %d supported.", | |
5008 | FLOW_MAX_MPLS_LABELS); | |
9583bc14 | 5009 | } |
fff1b9c0 | 5010 | ctx->error = XLATE_TOO_MANY_MPLS_LABELS; |
8bfd0fda | 5011 | return; |
9583bc14 | 5012 | } |
b0a17866 | 5013 | |
742c0ac3 JR |
5014 | /* Update flow's MPLS stack, and clear L3/4 fields to mark them invalid. */ |
5015 | flow_push_mpls(flow, n, mpls->ethertype, ctx->wc, true); | |
9583bc14 EJ |
5016 | } |
5017 | ||
8bfd0fda | 5018 | static void |
9cfef3d0 | 5019 | compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) |
9583bc14 | 5020 | { |
8bfd0fda | 5021 | struct flow *flow = &ctx->xin->flow; |
49a73e0c | 5022 | int n = flow_count_mpls_labels(flow, ctx->wc); |
33bf9176 | 5023 | |
49a73e0c | 5024 | if (flow_pop_mpls(flow, n, eth_type, ctx->wc)) { |
8bf009bf | 5025 | if (!eth_type_mpls(eth_type) && ctx->xbridge->support.odp.recirc) { |
e12ec36b | 5026 | ctx->was_mpls = true; |
7bbdd84f SH |
5027 | } |
5028 | } else if (n >= FLOW_MAX_MPLS_LABELS) { | |
8bfd0fda | 5029 | if (ctx->xin->packet != NULL) { |
2d9b49dd BP |
5030 | xlate_report_error(ctx, "dropping packet on which an " |
5031 | "MPLS pop action can't be performed as it has " | |
5032 | "more MPLS LSEs than the %d supported.", | |
5033 | FLOW_MAX_MPLS_LABELS); | |
8bfd0fda | 5034 | } |
fff1b9c0 | 5035 | ctx->error = XLATE_TOO_MANY_MPLS_LABELS; |
1520ef4f | 5036 | ofpbuf_clear(ctx->odp_actions); |
9583bc14 EJ |
5037 | } |
5038 | } | |
5039 | ||
5040 | static bool | |
5041 | compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids) | |
5042 | { | |
33bf9176 BP |
5043 | struct flow *flow = &ctx->xin->flow; |
5044 | ||
5045 | if (!is_ip_any(flow)) { | |
9583bc14 EJ |
5046 | return false; |
5047 | } | |
5048 | ||
49a73e0c | 5049 | ctx->wc->masks.nw_ttl = 0xff; |
33bf9176 BP |
5050 | if (flow->nw_ttl > 1) { |
5051 | flow->nw_ttl--; | |
9583bc14 EJ |
5052 | return false; |
5053 | } else { | |
5054 | size_t i; | |
5055 | ||
5056 | for (i = 0; i < ids->n_controllers; i++) { | |
d39ec23d | 5057 | xlate_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, |
206ddb9a | 5058 | ids->cnt_ids[i], UINT32_MAX, NULL, 0); |
9583bc14 EJ |
5059 | } |
5060 | ||
5061 | /* Stop processing for current table. */ | |
2d9b49dd BP |
5062 | xlate_report(ctx, OFT_WARN, "IPv%d decrement TTL exception", |
5063 | flow->dl_type == htons(ETH_TYPE_IP) ? 4 : 6); | |
9583bc14 EJ |
5064 | return true; |
5065 | } | |
5066 | } | |
5067 | ||
8bfd0fda | 5068 | static void |
097d4939 JR |
5069 | compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label) |
5070 | { | |
8bfd0fda | 5071 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 5072 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK); |
8bfd0fda | 5073 | set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label); |
097d4939 | 5074 | } |
097d4939 JR |
5075 | } |
5076 | ||
8bfd0fda | 5077 | static void |
097d4939 JR |
5078 | compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc) |
5079 | { | |
8bfd0fda | 5080 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 5081 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TC_MASK); |
8bfd0fda | 5082 | set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc); |
097d4939 | 5083 | } |
097d4939 JR |
5084 | } |
5085 | ||
491e05c2 YY |
5086 | static bool |
5087 | compose_dec_nsh_ttl_action(struct xlate_ctx *ctx) | |
5088 | { | |
5089 | struct flow *flow = &ctx->xin->flow; | |
5090 | ||
5091 | if ((flow->packet_type == htonl(PT_NSH)) || | |
5092 | (flow->dl_type == htons(ETH_TYPE_NSH))) { | |
5093 | ctx->wc->masks.nsh.ttl = 0xff; | |
5094 | if (flow->nsh.ttl > 1) { | |
5095 | flow->nsh.ttl--; | |
5096 | return false; | |
5097 | } else { | |
5098 | xlate_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, | |
206ddb9a | 5099 | 0, UINT32_MAX, NULL, 0); |
491e05c2 YY |
5100 | } |
5101 | } | |
5102 | ||
5103 | /* Stop processing for current table. */ | |
5104 | xlate_report(ctx, OFT_WARN, "NSH decrement TTL exception"); | |
5105 | return true; | |
5106 | } | |
5107 | ||
8bfd0fda | 5108 | static void |
9cfef3d0 | 5109 | compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl) |
9583bc14 | 5110 | { |
8bfd0fda | 5111 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
49a73e0c | 5112 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda | 5113 | set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl); |
b0a17866 | 5114 | } |
9583bc14 EJ |
5115 | } |
5116 | ||
5117 | static bool | |
9cfef3d0 | 5118 | compose_dec_mpls_ttl_action(struct xlate_ctx *ctx) |
9583bc14 | 5119 | { |
33bf9176 | 5120 | struct flow *flow = &ctx->xin->flow; |
1dd35f8a | 5121 | |
8bfd0fda | 5122 | if (eth_type_mpls(flow->dl_type)) { |
22d38fca JR |
5123 | uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]); |
5124 | ||
49a73e0c | 5125 | ctx->wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); |
8bfd0fda BP |
5126 | if (ttl > 1) { |
5127 | ttl--; | |
5128 | set_mpls_lse_ttl(&flow->mpls_lse[0], ttl); | |
5129 | return false; | |
5130 | } else { | |
d39ec23d | 5131 | xlate_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0, |
206ddb9a | 5132 | UINT32_MAX, NULL, 0); |
8bfd0fda | 5133 | } |
9583bc14 | 5134 | } |
22d38fca JR |
5135 | |
5136 | /* Stop processing for current table. */ | |
2d9b49dd | 5137 | xlate_report(ctx, OFT_WARN, "MPLS decrement TTL exception"); |
22d38fca | 5138 | return true; |
9583bc14 EJ |
5139 | } |
5140 | ||
8b496c72 BP |
5141 | /* Emits an action that outputs to 'port', within 'ctx'. |
5142 | * | |
5143 | * 'controller_len' affects only packets sent to an OpenFlow controller. It | |
5144 | * is the maximum number of bytes of the packet to send. UINT16_MAX means to | |
5145 | * send the whole packet (and 0 means to omit the packet entirely). | |
5146 | * | |
5147 | * 'may_packet_in' determines whether the packet may be sent to an OpenFlow | |
5148 | * controller. If it is false, then the packet is never sent to the OpenFlow | |
5149 | * controller. | |
5150 | * | |
5151 | * 'is_last_action' should be true if this output is the last OpenFlow action | |
5152 | * to be processed, which enables certain optimizations. | |
5153 | * | |
5154 | * 'truncate' should be true if the packet to be output is being truncated, | |
5155 | * which suppresses certain optimizations. */ | |
9583bc14 | 5156 | static void |
8b496c72 BP |
5157 | xlate_output_action(struct xlate_ctx *ctx, ofp_port_t port, |
5158 | uint16_t controller_len, bool may_packet_in, | |
f5634764 KG |
5159 | bool is_last_action, bool truncate, |
5160 | bool group_bucket_action) | |
9583bc14 | 5161 | { |
2031ef97 | 5162 | ofp_port_t prev_nf_output_iface = ctx->nf_output_iface; |
9583bc14 | 5163 | |
2031ef97 | 5164 | ctx->nf_output_iface = NF_OUT_DROP; |
9583bc14 EJ |
5165 | |
5166 | switch (port) { | |
5167 | case OFPP_IN_PORT: | |
feee58b9 | 5168 | compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port, NULL, |
8bbbda3c | 5169 | is_last_action, truncate); |
9583bc14 EJ |
5170 | break; |
5171 | case OFPP_TABLE: | |
4e022ec0 | 5172 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
11938578 | 5173 | 0, may_packet_in, true, false, false, |
96c3a6e5 | 5174 | do_xlate_actions); |
9583bc14 EJ |
5175 | break; |
5176 | case OFPP_NORMAL: | |
5177 | xlate_normal(ctx); | |
5178 | break; | |
5179 | case OFPP_FLOOD: | |
feee58b9 | 5180 | flood_packets(ctx, false, is_last_action); |
9583bc14 EJ |
5181 | break; |
5182 | case OFPP_ALL: | |
feee58b9 | 5183 | flood_packets(ctx, true, is_last_action); |
9583bc14 EJ |
5184 | break; |
5185 | case OFPP_CONTROLLER: | |
d39ec23d JP |
5186 | xlate_controller_action(ctx, controller_len, |
5187 | (ctx->in_packet_out ? OFPR_PACKET_OUT | |
f5634764 | 5188 | : group_bucket_action ? OFPR_GROUP |
d39ec23d JP |
5189 | : ctx->in_action_set ? OFPR_ACTION_SET |
5190 | : OFPR_ACTION), | |
206ddb9a | 5191 | 0, UINT32_MAX, NULL, 0); |
9583bc14 EJ |
5192 | break; |
5193 | case OFPP_NONE: | |
5194 | break; | |
5195 | case OFPP_LOCAL: | |
5196 | default: | |
4e022ec0 | 5197 | if (port != ctx->xin->flow.in_port.ofp_port) { |
8bbbda3c | 5198 | compose_output_action(ctx, port, NULL, is_last_action, truncate); |
9583bc14 | 5199 | } else { |
f89547e2 | 5200 | xlate_report_info(ctx, "skipping output to input port"); |
9583bc14 EJ |
5201 | } |
5202 | break; | |
5203 | } | |
5204 | ||
5205 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
2031ef97 BP |
5206 | ctx->nf_output_iface = NF_OUT_FLOOD; |
5207 | } else if (ctx->nf_output_iface == NF_OUT_DROP) { | |
5208 | ctx->nf_output_iface = prev_nf_output_iface; | |
9583bc14 | 5209 | } else if (prev_nf_output_iface != NF_OUT_DROP && |
2031ef97 BP |
5210 | ctx->nf_output_iface != NF_OUT_FLOOD) { |
5211 | ctx->nf_output_iface = NF_OUT_MULTI; | |
9583bc14 EJ |
5212 | } |
5213 | } | |
5214 | ||
5215 | static void | |
5216 | xlate_output_reg_action(struct xlate_ctx *ctx, | |
feee58b9 | 5217 | const struct ofpact_output_reg *or, |
f5634764 KG |
5218 | bool is_last_action, |
5219 | bool group_bucket_action) | |
9583bc14 EJ |
5220 | { |
5221 | uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow); | |
5222 | if (port <= UINT16_MAX) { | |
2d9b49dd BP |
5223 | xlate_report(ctx, OFT_DETAIL, "output port is %"PRIu64, port); |
5224 | ||
9583bc14 EJ |
5225 | union mf_subvalue value; |
5226 | ||
5227 | memset(&value, 0xff, sizeof value); | |
49a73e0c | 5228 | mf_write_subfield_flow(&or->src, &value, &ctx->wc->masks); |
8b496c72 | 5229 | xlate_output_action(ctx, u16_to_ofp(port), or->max_len, |
f5634764 KG |
5230 | false, is_last_action, false, |
5231 | group_bucket_action); | |
2d9b49dd BP |
5232 | } else { |
5233 | xlate_report(ctx, OFT_WARN, "output port %"PRIu64" is out of range", | |
5234 | port); | |
9583bc14 EJ |
5235 | } |
5236 | } | |
5237 | ||
aaca4fe0 WT |
5238 | static void |
5239 | xlate_output_trunc_action(struct xlate_ctx *ctx, | |
feee58b9 | 5240 | ofp_port_t port, uint32_t max_len, |
f5634764 KG |
5241 | bool is_last_action, |
5242 | bool group_bucket_action) | |
aaca4fe0 WT |
5243 | { |
5244 | bool support_trunc = ctx->xbridge->support.trunc; | |
5245 | struct ovs_action_trunc *trunc; | |
29718ad4 | 5246 | char name[OFP_MAX_PORT_NAME_LEN]; |
aaca4fe0 WT |
5247 | |
5248 | switch (port) { | |
5249 | case OFPP_TABLE: | |
5250 | case OFPP_NORMAL: | |
5251 | case OFPP_FLOOD: | |
5252 | case OFPP_ALL: | |
5253 | case OFPP_CONTROLLER: | |
5254 | case OFPP_NONE: | |
50f96b10 | 5255 | ofputil_port_to_string(port, NULL, name, sizeof name); |
2d9b49dd BP |
5256 | xlate_report(ctx, OFT_WARN, |
5257 | "output_trunc does not support port: %s", name); | |
aaca4fe0 WT |
5258 | break; |
5259 | case OFPP_LOCAL: | |
5260 | case OFPP_IN_PORT: | |
5261 | default: | |
5262 | if (port != ctx->xin->flow.in_port.ofp_port) { | |
5263 | const struct xport *xport = get_ofp_port(ctx->xbridge, port); | |
5264 | ||
5265 | if (xport == NULL || xport->odp_port == ODPP_NONE) { | |
5266 | /* Since truncate happens at its following output action, if | |
5267 | * the output port is a patch port, the behavior is somehow | |
49f17344 | 5268 | * unpredictable. For simplicity, disallow this case. */ |
50f96b10 | 5269 | ofputil_port_to_string(port, NULL, name, sizeof name); |
2d9b49dd BP |
5270 | xlate_report_error(ctx, "output_trunc does not support " |
5271 | "patch port %s", name); | |
aaca4fe0 WT |
5272 | break; |
5273 | } | |
5274 | ||
5275 | trunc = nl_msg_put_unspec_uninit(ctx->odp_actions, | |
5276 | OVS_ACTION_ATTR_TRUNC, | |
5277 | sizeof *trunc); | |
5278 | trunc->max_len = max_len; | |
f5634764 KG |
5279 | xlate_output_action(ctx, port, 0, false, is_last_action, true, |
5280 | group_bucket_action); | |
aaca4fe0 WT |
5281 | if (!support_trunc) { |
5282 | ctx->xout->slow |= SLOW_ACTION; | |
5283 | } | |
5284 | } else { | |
f89547e2 | 5285 | xlate_report_info(ctx, "skipping output to input port"); |
aaca4fe0 WT |
5286 | } |
5287 | break; | |
5288 | } | |
5289 | } | |
5290 | ||
9583bc14 EJ |
5291 | static void |
5292 | xlate_enqueue_action(struct xlate_ctx *ctx, | |
feee58b9 | 5293 | const struct ofpact_enqueue *enqueue, |
f5634764 KG |
5294 | bool is_last_action, |
5295 | bool group_bucket_action) | |
9583bc14 | 5296 | { |
4e022ec0 | 5297 | ofp_port_t ofp_port = enqueue->port; |
9583bc14 EJ |
5298 | uint32_t queue_id = enqueue->queue; |
5299 | uint32_t flow_priority, priority; | |
5300 | int error; | |
5301 | ||
5302 | /* Translate queue to priority. */ | |
89a8a7f0 | 5303 | error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority); |
9583bc14 EJ |
5304 | if (error) { |
5305 | /* Fall back to ordinary output action. */ | |
8b496c72 | 5306 | xlate_output_action(ctx, enqueue->port, 0, false, |
f5634764 KG |
5307 | is_last_action, false, |
5308 | group_bucket_action); | |
9583bc14 EJ |
5309 | return; |
5310 | } | |
5311 | ||
5312 | /* Check output port. */ | |
5313 | if (ofp_port == OFPP_IN_PORT) { | |
4e022ec0 AW |
5314 | ofp_port = ctx->xin->flow.in_port.ofp_port; |
5315 | } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
5316 | return; |
5317 | } | |
5318 | ||
5319 | /* Add datapath actions. */ | |
5320 | flow_priority = ctx->xin->flow.skb_priority; | |
5321 | ctx->xin->flow.skb_priority = priority; | |
11938578 | 5322 | compose_output_action(ctx, ofp_port, NULL, is_last_action, false); |
9583bc14 EJ |
5323 | ctx->xin->flow.skb_priority = flow_priority; |
5324 | ||
5325 | /* Update NetFlow output port. */ | |
2031ef97 BP |
5326 | if (ctx->nf_output_iface == NF_OUT_DROP) { |
5327 | ctx->nf_output_iface = ofp_port; | |
5328 | } else if (ctx->nf_output_iface != NF_OUT_FLOOD) { | |
5329 | ctx->nf_output_iface = NF_OUT_MULTI; | |
9583bc14 EJ |
5330 | } |
5331 | } | |
5332 | ||
5333 | static void | |
5334 | xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id) | |
5335 | { | |
5336 | uint32_t skb_priority; | |
5337 | ||
89a8a7f0 | 5338 | if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) { |
9583bc14 EJ |
5339 | ctx->xin->flow.skb_priority = skb_priority; |
5340 | } else { | |
5341 | /* Couldn't translate queue to a priority. Nothing to do. A warning | |
5342 | * has already been logged. */ | |
5343 | } | |
5344 | } | |
5345 | ||
5346 | static bool | |
46c88433 | 5347 | slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_) |
9583bc14 | 5348 | { |
46c88433 EJ |
5349 | const struct xbridge *xbridge = xbridge_; |
5350 | struct xport *port; | |
9583bc14 EJ |
5351 | |
5352 | switch (ofp_port) { | |
5353 | case OFPP_IN_PORT: | |
5354 | case OFPP_TABLE: | |
5355 | case OFPP_NORMAL: | |
5356 | case OFPP_FLOOD: | |
5357 | case OFPP_ALL: | |
5358 | case OFPP_NONE: | |
5359 | return true; | |
5360 | case OFPP_CONTROLLER: /* Not supported by the bundle action. */ | |
5361 | return false; | |
5362 | default: | |
46c88433 | 5363 | port = get_ofp_port(xbridge, ofp_port); |
9583bc14 EJ |
5364 | return port ? port->may_enable : false; |
5365 | } | |
5366 | } | |
5367 | ||
5368 | static void | |
5369 | xlate_bundle_action(struct xlate_ctx *ctx, | |
feee58b9 | 5370 | const struct ofpact_bundle *bundle, |
f5634764 KG |
5371 | bool is_last_action, |
5372 | bool group_bucket_action) | |
9583bc14 | 5373 | { |
4e022ec0 | 5374 | ofp_port_t port; |
9583bc14 | 5375 | |
49a73e0c | 5376 | port = bundle_execute(bundle, &ctx->xin->flow, ctx->wc, slave_enabled_cb, |
46c88433 | 5377 | CONST_CAST(struct xbridge *, ctx->xbridge)); |
9583bc14 | 5378 | if (bundle->dst.field) { |
49a73e0c | 5379 | nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow, ctx->wc); |
2d9b49dd | 5380 | xlate_report_subfield(ctx, &bundle->dst); |
9583bc14 | 5381 | } else { |
f5634764 KG |
5382 | xlate_output_action(ctx, port, 0, false, is_last_action, false, |
5383 | group_bucket_action); | |
9583bc14 EJ |
5384 | } |
5385 | } | |
5386 | ||
4165b5e0 JS |
5387 | static void |
5388 | xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) | |
5389 | { | |
49a73e0c | 5390 | learn_mask(learn, ctx->wc); |
9583bc14 | 5391 | |
df70a773 | 5392 | if (ctx->xin->xcache || ctx->xin->allow_side_effects) { |
4165b5e0 JS |
5393 | uint64_t ofpacts_stub[1024 / 8]; |
5394 | struct ofputil_flow_mod fm; | |
2c7ee524 | 5395 | struct ofproto_flow_mod ofm__, *ofm; |
4165b5e0 | 5396 | struct ofpbuf ofpacts; |
2c7ee524 JR |
5397 | enum ofperr error; |
5398 | ||
5399 | if (ctx->xin->xcache) { | |
3f3b97b0 | 5400 | ofm = xmalloc(sizeof *ofm); |
2c7ee524 JR |
5401 | } else { |
5402 | ofm = &ofm__; | |
5403 | } | |
4165b5e0 JS |
5404 | |
5405 | ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); | |
2c7ee524 | 5406 | learn_execute(learn, &ctx->xin->flow, &fm, &ofpacts); |
2d9b49dd BP |
5407 | if (OVS_UNLIKELY(ctx->xin->trace)) { |
5408 | struct ds s = DS_EMPTY_INITIALIZER; | |
5409 | ds_put_format(&s, "table=%"PRIu8" ", fm.table_id); | |
6a6b7060 BP |
5410 | minimatch_format(&fm.match, |
5411 | ofproto_get_tun_tab(&ctx->xin->ofproto->up), | |
5412 | NULL, &s, OFP_DEFAULT_PRIORITY); | |
2d9b49dd BP |
5413 | ds_chomp(&s, ' '); |
5414 | ds_put_format(&s, " priority=%d", fm.priority); | |
5415 | if (fm.new_cookie) { | |
5416 | ds_put_format(&s, " cookie=%#"PRIx64, ntohll(fm.new_cookie)); | |
5417 | } | |
5418 | if (fm.idle_timeout != OFP_FLOW_PERMANENT) { | |
5419 | ds_put_format(&s, " idle=%"PRIu16, fm.idle_timeout); | |
5420 | } | |
5421 | if (fm.hard_timeout != OFP_FLOW_PERMANENT) { | |
5422 | ds_put_format(&s, " hard=%"PRIu16, fm.hard_timeout); | |
5423 | } | |
5424 | if (fm.flags & NX_LEARN_F_SEND_FLOW_REM) { | |
5425 | ds_put_cstr(&s, " send_flow_rem"); | |
5426 | } | |
5427 | ds_put_cstr(&s, " actions="); | |
efefbcae BP |
5428 | struct ofpact_format_params fp = { .s = &s }; |
5429 | ofpacts_format(fm.ofpacts, fm.ofpacts_len, &fp); | |
2d9b49dd BP |
5430 | xlate_report(ctx, OFT_DETAIL, "%s", ds_cstr(&s)); |
5431 | ds_destroy(&s); | |
5432 | } | |
2c7ee524 JR |
5433 | error = ofproto_dpif_flow_mod_init_for_learn(ctx->xbridge->ofproto, |
5434 | &fm, ofm); | |
4165b5e0 | 5435 | ofpbuf_uninit(&ofpacts); |
2c7ee524 | 5436 | |
3f3b97b0 | 5437 | if (!error) { |
4c71600d | 5438 | bool success = true; |
3f3b97b0 | 5439 | if (ctx->xin->allow_side_effects) { |
4c71600d DDP |
5440 | error = ofproto_flow_mod_learn(ofm, ctx->xin->xcache != NULL, |
5441 | learn->limit, &success); | |
5442 | } else if (learn->limit) { | |
5443 | if (!ofm->temp_rule | |
5444 | || ofm->temp_rule->state != RULE_INSERTED) { | |
5445 | /* The learned rule expired and there are no packets, so | |
5446 | * we cannot learn again. Since the translated actions | |
5447 | * depend on the result of learning, we tell the caller | |
5448 | * that there's no point in caching this result. */ | |
5449 | ctx->xout->avoid_caching = true; | |
5450 | } | |
3f3b97b0 DDP |
5451 | } |
5452 | ||
4c71600d DDP |
5453 | if (learn->flags & NX_LEARN_F_WRITE_RESULT) { |
5454 | nxm_reg_load(&learn->result_dst, success ? 1 : 0, | |
5455 | &ctx->xin->flow, ctx->wc); | |
5456 | xlate_report_subfield(ctx, &learn->result_dst); | |
5457 | } | |
5458 | ||
5459 | if (success && ctx->xin->xcache) { | |
3f3b97b0 DDP |
5460 | struct xc_entry *entry; |
5461 | ||
5462 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN); | |
5463 | entry->learn.ofm = ofm; | |
4c71600d | 5464 | entry->learn.limit = learn->limit; |
3f3b97b0 | 5465 | ofm = NULL; |
1bddcb5d YS |
5466 | } else { |
5467 | ofproto_flow_mod_uninit(ofm); | |
3f3b97b0 | 5468 | } |
4c71600d DDP |
5469 | |
5470 | if (OVS_UNLIKELY(ctx->xin->trace && !success)) { | |
5471 | xlate_report(ctx, OFT_DETAIL, "Limit exceeded, learn failed"); | |
5472 | } | |
3f3b97b0 DDP |
5473 | } |
5474 | ||
3f76c123 | 5475 | if (ofm != &ofm__) { |
3f3b97b0 | 5476 | free(ofm); |
2c7ee524 JR |
5477 | } |
5478 | ||
5479 | if (error) { | |
2d9b49dd BP |
5480 | xlate_report_error(ctx, "LEARN action execution failed (%s).", |
5481 | ofperr_to_string(error)); | |
2c7ee524 | 5482 | } |
6a6b7060 BP |
5483 | |
5484 | minimatch_destroy(&fm.match); | |
2d9b49dd BP |
5485 | } else { |
5486 | xlate_report(ctx, OFT_WARN, | |
5487 | "suppressing side effects, so learn action ignored"); | |
b256dc52 JS |
5488 | } |
5489 | } | |
5490 | ||
5491 | static void | |
5492 | xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags, | |
5493 | uint16_t idle_timeout, uint16_t hard_timeout) | |
5494 | { | |
5495 | if (tcp_flags & (TCP_FIN | TCP_RST)) { | |
07a3cd5c | 5496 | ofproto_rule_reduce_timeouts(&rule->up, idle_timeout, hard_timeout); |
b256dc52 | 5497 | } |
9583bc14 EJ |
5498 | } |
5499 | ||
9583bc14 EJ |
5500 | static void |
5501 | xlate_fin_timeout(struct xlate_ctx *ctx, | |
5502 | const struct ofpact_fin_timeout *oft) | |
5503 | { | |
b256dc52 | 5504 | if (ctx->rule) { |
df70a773 JR |
5505 | if (ctx->xin->allow_side_effects) { |
5506 | xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags, | |
5507 | oft->fin_idle_timeout, oft->fin_hard_timeout); | |
5508 | } | |
b256dc52 JS |
5509 | if (ctx->xin->xcache) { |
5510 | struct xc_entry *entry; | |
5511 | ||
5512 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT); | |
83709dfa JR |
5513 | /* XC_RULE already holds a reference on the rule, none is taken |
5514 | * here. */ | |
901a517e JR |
5515 | entry->fin.rule = ctx->rule; |
5516 | entry->fin.idle = oft->fin_idle_timeout; | |
5517 | entry->fin.hard = oft->fin_hard_timeout; | |
b256dc52 | 5518 | } |
9583bc14 EJ |
5519 | } |
5520 | } | |
5521 | ||
5522 | static void | |
5523 | xlate_sample_action(struct xlate_ctx *ctx, | |
5524 | const struct ofpact_sample *os) | |
5525 | { | |
f69f713b BY |
5526 | odp_port_t output_odp_port = ODPP_NONE; |
5527 | odp_port_t tunnel_out_port = ODPP_NONE; | |
5528 | struct dpif_ipfix *ipfix = ctx->xbridge->ipfix; | |
5529 | bool emit_set_tunnel = false; | |
5530 | ||
5531 | if (!ipfix || ctx->xin->flow.in_port.ofp_port == OFPP_NONE) { | |
5532 | return; | |
5533 | } | |
5534 | ||
e824d78d JR |
5535 | /* Scale the probability from 16-bit to 32-bit while representing |
5536 | * the same percentage. */ | |
5537 | uint32_t probability = (os->probability << 16) | os->probability; | |
5538 | ||
f69f713b BY |
5539 | /* If ofp_port in flow sample action is equel to ofp_port, |
5540 | * this sample action is a input port action. */ | |
5541 | if (os->sampling_port != OFPP_NONE && | |
5542 | os->sampling_port != ctx->xin->flow.in_port.ofp_port) { | |
5543 | output_odp_port = ofp_port_to_odp_port(ctx->xbridge, | |
5544 | os->sampling_port); | |
5545 | if (output_odp_port == ODPP_NONE) { | |
2d9b49dd BP |
5546 | xlate_report_error(ctx, "can't use unknown port %d in flow sample " |
5547 | "action", os->sampling_port); | |
f69f713b BY |
5548 | return; |
5549 | } | |
5550 | ||
5551 | if (dpif_ipfix_get_flow_exporter_tunnel_sampling(ipfix, | |
5552 | os->collector_set_id) | |
cd32509e | 5553 | && dpif_ipfix_is_tunnel_port(ipfix, output_odp_port)) { |
f69f713b BY |
5554 | tunnel_out_port = output_odp_port; |
5555 | emit_set_tunnel = true; | |
5556 | } | |
5557 | } | |
5558 | ||
5559 | xlate_commit_actions(ctx); | |
5560 | /* If 'emit_set_tunnel', sample(sampling_port=1) would translate | |
5561 | * into datapath sample action set(tunnel(...)), sample(...) and | |
5562 | * it is used for sampling egress tunnel information. */ | |
5563 | if (emit_set_tunnel) { | |
5564 | const struct xport *xport = get_ofp_port(ctx->xbridge, | |
5565 | os->sampling_port); | |
5566 | ||
5567 | if (xport && xport->is_tunnel) { | |
5568 | struct flow *flow = &ctx->xin->flow; | |
5569 | tnl_port_send(xport->ofport, flow, ctx->wc); | |
5570 | if (!ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { | |
5571 | struct flow_tnl flow_tnl = flow->tunnel; | |
c6d87201 | 5572 | const char *tnl_type; |
f69f713b | 5573 | |
c6d87201 | 5574 | tnl_type = tnl_port_get_type(xport->ofport); |
f69f713b | 5575 | commit_odp_tunnel_action(flow, &ctx->base_flow, |
c6d87201 | 5576 | ctx->odp_actions, tnl_type); |
f69f713b BY |
5577 | flow->tunnel = flow_tnl; |
5578 | } | |
5579 | } else { | |
2d9b49dd BP |
5580 | xlate_report_error(ctx, |
5581 | "sampling_port:%d should be a tunnel port.", | |
5582 | os->sampling_port); | |
f69f713b BY |
5583 | } |
5584 | } | |
e824d78d | 5585 | |
24a4bbe1 IM |
5586 | struct user_action_cookie cookie; |
5587 | ||
5588 | memset(&cookie, 0, sizeof cookie); | |
5589 | cookie.type = USER_ACTION_COOKIE_FLOW_SAMPLE; | |
5590 | cookie.ofp_in_port = ctx->xin->flow.in_port.ofp_port; | |
5591 | cookie.ofproto_uuid = ctx->xbridge->ofproto->uuid; | |
5592 | cookie.flow_sample.probability = os->probability; | |
5593 | cookie.flow_sample.collector_set_id = os->collector_set_id; | |
5594 | cookie.flow_sample.obs_domain_id = os->obs_domain_id; | |
5595 | cookie.flow_sample.obs_point_id = os->obs_point_id; | |
5596 | cookie.flow_sample.output_odp_port = output_odp_port; | |
5597 | cookie.flow_sample.direction = os->direction; | |
5598 | ||
8de6ff3e | 5599 | compose_sample_action(ctx, probability, &cookie, tunnel_out_port, false); |
9583bc14 EJ |
5600 | } |
5601 | ||
eee69393 AZ |
5602 | /* Determine if an datapath action translated from the openflow action |
5603 | * can be reversed by another datapath action. | |
5604 | * | |
5605 | * Openflow actions that do not emit datapath actions are trivially | |
5606 | * reversible. Reversiblity of other actions depends on nature of | |
5607 | * action and their translation. */ | |
5608 | static bool | |
5609 | reversible_actions(const struct ofpact *ofpacts, size_t ofpacts_len) | |
bef503e8 | 5610 | { |
eee69393 | 5611 | const struct ofpact *a; |
bef503e8 | 5612 | |
eee69393 AZ |
5613 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
5614 | switch (a->type) { | |
5615 | case OFPACT_BUNDLE: | |
5616 | case OFPACT_CLEAR_ACTIONS: | |
5617 | case OFPACT_CLONE: | |
5618 | case OFPACT_CONJUNCTION: | |
5619 | case OFPACT_CONTROLLER: | |
5620 | case OFPACT_CT_CLEAR: | |
5621 | case OFPACT_DEBUG_RECIRC: | |
a934a3dd | 5622 | case OFPACT_DEBUG_SLOW: |
eee69393 AZ |
5623 | case OFPACT_DEC_MPLS_TTL: |
5624 | case OFPACT_DEC_TTL: | |
5625 | case OFPACT_ENQUEUE: | |
5626 | case OFPACT_EXIT: | |
5627 | case OFPACT_FIN_TIMEOUT: | |
5628 | case OFPACT_GOTO_TABLE: | |
5629 | case OFPACT_GROUP: | |
5630 | case OFPACT_LEARN: | |
5631 | case OFPACT_MULTIPATH: | |
5632 | case OFPACT_NOTE: | |
5633 | case OFPACT_OUTPUT: | |
5634 | case OFPACT_OUTPUT_REG: | |
5635 | case OFPACT_POP_MPLS: | |
5636 | case OFPACT_POP_QUEUE: | |
5637 | case OFPACT_PUSH_MPLS: | |
5638 | case OFPACT_PUSH_VLAN: | |
5639 | case OFPACT_REG_MOVE: | |
5640 | case OFPACT_RESUBMIT: | |
5641 | case OFPACT_SAMPLE: | |
5642 | case OFPACT_SET_ETH_DST: | |
5643 | case OFPACT_SET_ETH_SRC: | |
5644 | case OFPACT_SET_FIELD: | |
5645 | case OFPACT_SET_IP_DSCP: | |
5646 | case OFPACT_SET_IP_ECN: | |
5647 | case OFPACT_SET_IP_TTL: | |
5648 | case OFPACT_SET_IPV4_DST: | |
5649 | case OFPACT_SET_IPV4_SRC: | |
5650 | case OFPACT_SET_L4_DST_PORT: | |
5651 | case OFPACT_SET_L4_SRC_PORT: | |
5652 | case OFPACT_SET_MPLS_LABEL: | |
5653 | case OFPACT_SET_MPLS_TC: | |
5654 | case OFPACT_SET_MPLS_TTL: | |
5655 | case OFPACT_SET_QUEUE: | |
5656 | case OFPACT_SET_TUNNEL: | |
5657 | case OFPACT_SET_VLAN_PCP: | |
5658 | case OFPACT_SET_VLAN_VID: | |
5659 | case OFPACT_STACK_POP: | |
5660 | case OFPACT_STACK_PUSH: | |
5661 | case OFPACT_STRIP_VLAN: | |
5662 | case OFPACT_UNROLL_XLATE: | |
5663 | case OFPACT_WRITE_ACTIONS: | |
5664 | case OFPACT_WRITE_METADATA: | |
5b34f8fc | 5665 | case OFPACT_CHECK_PKT_LARGER: |
eee69393 AZ |
5666 | break; |
5667 | ||
5668 | case OFPACT_CT: | |
5669 | case OFPACT_METER: | |
5670 | case OFPACT_NAT: | |
5671 | case OFPACT_OUTPUT_TRUNC: | |
2142be1f BP |
5672 | case OFPACT_ENCAP: |
5673 | case OFPACT_DECAP: | |
491e05c2 | 5674 | case OFPACT_DEC_NSH_TTL: |
eee69393 | 5675 | return false; |
9c2a44dc | 5676 | } |
456024cb | 5677 | } |
eee69393 | 5678 | return true; |
bef503e8 AZ |
5679 | } |
5680 | ||
5681 | static void | |
c9f0a445 | 5682 | clone_xlate_actions(const struct ofpact *actions, size_t actions_len, |
f5634764 KG |
5683 | struct xlate_ctx *ctx, bool is_last_action, |
5684 | bool group_bucket_action OVS_UNUSED) | |
7ae62a67 | 5685 | { |
b827b231 BP |
5686 | struct ofpbuf old_stack = ctx->stack; |
5687 | union mf_subvalue new_stack[1024 / sizeof(union mf_subvalue)]; | |
5688 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); | |
5689 | ofpbuf_put(&ctx->stack, old_stack.data, old_stack.size); | |
5690 | ||
5691 | struct ofpbuf old_action_set = ctx->action_set; | |
5692 | uint64_t actset_stub[1024 / 8]; | |
5693 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); | |
5694 | ofpbuf_put(&ctx->action_set, old_action_set.data, old_action_set.size); | |
5695 | ||
eee69393 | 5696 | size_t offset, ac_offset; |
eee69393 AZ |
5697 | struct flow old_flow = ctx->xin->flow; |
5698 | ||
feee58b9 | 5699 | if (reversible_actions(actions, actions_len) || is_last_action) { |
eee69393 | 5700 | old_flow = ctx->xin->flow; |
f5634764 | 5701 | do_xlate_actions(actions, actions_len, ctx, is_last_action, false); |
118b21d9 EG |
5702 | if (!ctx->freezing) { |
5703 | xlate_action_set(ctx); | |
5704 | } | |
60eebf12 AZ |
5705 | if (ctx->freezing) { |
5706 | finish_freezing(ctx); | |
5707 | } | |
eee69393 AZ |
5708 | goto xlate_done; |
5709 | } | |
5710 | ||
5711 | /* Commit datapath actions before emitting the clone action to | |
5712 | * avoid emitting those actions twice. Once inside | |
5713 | * the clone, another time for the action after clone. */ | |
5714 | xlate_commit_actions(ctx); | |
9c2a44dc | 5715 | struct flow old_base = ctx->base_flow; |
eee69393 AZ |
5716 | bool old_was_mpls = ctx->was_mpls; |
5717 | bool old_conntracked = ctx->conntracked; | |
ba653d2a | 5718 | |
eee69393 AZ |
5719 | /* The actions are not reversible, a datapath clone action is |
5720 | * required to encode the translation. Select the clone action | |
5721 | * based on datapath capabilities. */ | |
5722 | if (ctx->xbridge->support.clone) { /* Use clone action */ | |
5723 | /* Use clone action as datapath clone. */ | |
5724 | offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CLONE); | |
f5634764 | 5725 | do_xlate_actions(actions, actions_len, ctx, true, false); |
118b21d9 EG |
5726 | if (!ctx->freezing) { |
5727 | xlate_action_set(ctx); | |
5728 | } | |
60eebf12 AZ |
5729 | if (ctx->freezing) { |
5730 | finish_freezing(ctx); | |
5731 | } | |
eee69393 AZ |
5732 | nl_msg_end_non_empty_nested(ctx->odp_actions, offset); |
5733 | goto dp_clone_done; | |
5734 | } | |
b827b231 | 5735 | |
eee69393 AZ |
5736 | if (ctx->xbridge->support.sample_nesting > 3) { |
5737 | /* Use sample action as datapath clone. */ | |
5738 | offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_SAMPLE); | |
5739 | ac_offset = nl_msg_start_nested(ctx->odp_actions, | |
5740 | OVS_SAMPLE_ATTR_ACTIONS); | |
f5634764 | 5741 | do_xlate_actions(actions, actions_len, ctx, true, false); |
118b21d9 EG |
5742 | if (!ctx->freezing) { |
5743 | xlate_action_set(ctx); | |
5744 | } | |
60eebf12 AZ |
5745 | if (ctx->freezing) { |
5746 | finish_freezing(ctx); | |
5747 | } | |
eee69393 AZ |
5748 | if (nl_msg_end_non_empty_nested(ctx->odp_actions, ac_offset)) { |
5749 | nl_msg_cancel_nested(ctx->odp_actions, offset); | |
5750 | } else { | |
5751 | nl_msg_put_u32(ctx->odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, | |
5752 | UINT32_MAX); /* 100% probability. */ | |
5753 | nl_msg_end_nested(ctx->odp_actions, offset); | |
5754 | } | |
5755 | goto dp_clone_done; | |
5756 | } | |
5757 | ||
5758 | /* Datapath does not support clone, skip xlate 'oc' and | |
5759 | * report an error */ | |
5760 | xlate_report_error(ctx, "Failed to compose clone action"); | |
ba653d2a | 5761 | |
eee69393 | 5762 | dp_clone_done: |
ba653d2a BP |
5763 | /* The clone's conntrack execution should have no effect on the original |
5764 | * packet. */ | |
5765 | ctx->conntracked = old_conntracked; | |
bd3c2df3 BP |
5766 | |
5767 | /* Popping MPLS from the clone should have no effect on the original | |
5768 | * packet. */ | |
5769 | ctx->was_mpls = old_was_mpls; | |
eee69393 AZ |
5770 | |
5771 | /* Restore the 'base_flow' for the next action. */ | |
5772 | ctx->base_flow = old_base; | |
5773 | ||
5774 | xlate_done: | |
5775 | ofpbuf_uninit(&ctx->action_set); | |
5776 | ctx->action_set = old_action_set; | |
5777 | ofpbuf_uninit(&ctx->stack); | |
5778 | ctx->stack = old_stack; | |
5779 | ctx->xin->flow = old_flow; | |
7ae62a67 WT |
5780 | } |
5781 | ||
c9f0a445 | 5782 | static void |
feee58b9 AZ |
5783 | compose_clone(struct xlate_ctx *ctx, const struct ofpact_nest *oc, |
5784 | bool is_last_action) | |
c9f0a445 AZ |
5785 | { |
5786 | size_t oc_actions_len = ofpact_nest_get_action_len(oc); | |
5787 | ||
f5634764 KG |
5788 | clone_xlate_actions(oc->actions, oc_actions_len, ctx, is_last_action, |
5789 | false); | |
c9f0a445 AZ |
5790 | } |
5791 | ||
076caa2f JR |
5792 | static void |
5793 | xlate_meter_action(struct xlate_ctx *ctx, const struct ofpact_meter *meter) | |
5794 | { | |
5795 | if (meter->provider_meter_id != UINT32_MAX) { | |
5796 | nl_msg_put_u32(ctx->odp_actions, OVS_ACTION_ATTR_METER, | |
5797 | meter->provider_meter_id); | |
5798 | } | |
5799 | } | |
5800 | ||
9583bc14 | 5801 | static bool |
46c88433 | 5802 | may_receive(const struct xport *xport, struct xlate_ctx *ctx) |
9583bc14 | 5803 | { |
bbbca389 | 5804 | if (xport->config & (is_stp(&ctx->xin->flow) |
46c88433 EJ |
5805 | ? OFPUTIL_PC_NO_RECV_STP |
5806 | : OFPUTIL_PC_NO_RECV)) { | |
9583bc14 EJ |
5807 | return false; |
5808 | } | |
5809 | ||
5810 | /* Only drop packets here if both forwarding and learning are | |
5811 | * disabled. If just learning is enabled, we need to have | |
5812 | * OFPP_NORMAL and the learning action have a look at the packet | |
5813 | * before we can drop it. */ | |
9efd308e DV |
5814 | if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) || |
5815 | (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) { | |
9583bc14 EJ |
5816 | return false; |
5817 | } | |
5818 | ||
5819 | return true; | |
5820 | } | |
5821 | ||
7fdb60a7 | 5822 | static void |
7e7e8dbb BP |
5823 | xlate_write_actions__(struct xlate_ctx *ctx, |
5824 | const struct ofpact *ofpacts, size_t ofpacts_len) | |
7fdb60a7 | 5825 | { |
c61f3870 BP |
5826 | /* Maintain actset_output depending on the contents of the action set: |
5827 | * | |
5828 | * - OFPP_UNSET, if there is no "output" action. | |
5829 | * | |
5830 | * - The output port, if there is an "output" action and no "group" | |
5831 | * action. | |
5832 | * | |
5833 | * - OFPP_UNSET, if there is a "group" action. | |
5834 | */ | |
5835 | if (!ctx->action_set_has_group) { | |
7e7e8dbb BP |
5836 | const struct ofpact *a; |
5837 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { | |
5838 | if (a->type == OFPACT_OUTPUT) { | |
5839 | ctx->xin->flow.actset_output = ofpact_get_OUTPUT(a)->port; | |
5840 | } else if (a->type == OFPACT_GROUP) { | |
c61f3870 BP |
5841 | ctx->xin->flow.actset_output = OFPP_UNSET; |
5842 | ctx->action_set_has_group = true; | |
9055ca9a | 5843 | break; |
c61f3870 BP |
5844 | } |
5845 | } | |
5846 | } | |
5847 | ||
7e7e8dbb BP |
5848 | ofpbuf_put(&ctx->action_set, ofpacts, ofpacts_len); |
5849 | } | |
5850 | ||
5851 | static void | |
5852 | xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact_nest *a) | |
5853 | { | |
5854 | xlate_write_actions__(ctx, a->actions, ofpact_nest_get_action_len(a)); | |
7fdb60a7 SH |
5855 | } |
5856 | ||
5857 | static void | |
5858 | xlate_action_set(struct xlate_ctx *ctx) | |
5859 | { | |
2d9b49dd BP |
5860 | uint64_t action_list_stub[1024 / 8]; |
5861 | struct ofpbuf action_list = OFPBUF_STUB_INITIALIZER(action_list_stub); | |
7fdb60a7 | 5862 | ofpacts_execute_action_set(&action_list, &ctx->action_set); |
ed9c9e3e JR |
5863 | /* Clear the action set, as it is not needed any more. */ |
5864 | ofpbuf_clear(&ctx->action_set); | |
2d9b49dd BP |
5865 | if (action_list.size) { |
5866 | ctx->in_action_set = true; | |
5867 | ||
5868 | struct ovs_list *old_trace = ctx->xin->trace; | |
5869 | ctx->xin->trace = xlate_report(ctx, OFT_TABLE, | |
5870 | "--. Executing action set:"); | |
f5634764 | 5871 | do_xlate_actions(action_list.data, action_list.size, ctx, true, false); |
2d9b49dd BP |
5872 | ctx->xin->trace = old_trace; |
5873 | ||
5874 | ctx->in_action_set = false; | |
5875 | } | |
7fdb60a7 SH |
5876 | ofpbuf_uninit(&action_list); |
5877 | } | |
5878 | ||
e672ff9b | 5879 | static void |
1d361a81 | 5880 | freeze_put_unroll_xlate(struct xlate_ctx *ctx) |
e672ff9b | 5881 | { |
1d361a81 | 5882 | struct ofpact_unroll_xlate *unroll = ctx->frozen_actions.header; |
e672ff9b JR |
5883 | |
5884 | /* Restore the table_id and rule cookie for a potential PACKET | |
5885 | * IN if needed. */ | |
5886 | if (!unroll || | |
5887 | (ctx->table_id != unroll->rule_table_id | |
5888 | || ctx->rule_cookie != unroll->rule_cookie)) { | |
1d361a81 | 5889 | unroll = ofpact_put_UNROLL_XLATE(&ctx->frozen_actions); |
e672ff9b JR |
5890 | unroll->rule_table_id = ctx->table_id; |
5891 | unroll->rule_cookie = ctx->rule_cookie; | |
1d361a81 | 5892 | ctx->frozen_actions.header = unroll; |
e672ff9b JR |
5893 | } |
5894 | } | |
5895 | ||
5896 | ||
1d361a81 BP |
5897 | /* Copy actions 'a' through 'end' to ctx->frozen_actions, which will be |
5898 | * executed after thawing. Inserts an UNROLL_XLATE action, if none is already | |
5899 | * present, before any action that may depend on the current table ID or flow | |
5900 | * cookie. */ | |
e672ff9b | 5901 | static void |
1d361a81 | 5902 | freeze_unroll_actions(const struct ofpact *a, const struct ofpact *end, |
e672ff9b JR |
5903 | struct xlate_ctx *ctx) |
5904 | { | |
c2b283b7 | 5905 | for (; a < end; a = ofpact_next(a)) { |
e672ff9b | 5906 | switch (a->type) { |
e672ff9b | 5907 | case OFPACT_OUTPUT_REG: |
aaca4fe0 | 5908 | case OFPACT_OUTPUT_TRUNC: |
e672ff9b JR |
5909 | case OFPACT_GROUP: |
5910 | case OFPACT_OUTPUT: | |
5911 | case OFPACT_CONTROLLER: | |
5912 | case OFPACT_DEC_MPLS_TTL: | |
491e05c2 | 5913 | case OFPACT_DEC_NSH_TTL: |
e672ff9b | 5914 | case OFPACT_DEC_TTL: |
83a31283 BP |
5915 | /* These actions may generate asynchronous messages, which include |
5916 | * table ID and flow cookie information. */ | |
1d361a81 | 5917 | freeze_put_unroll_xlate(ctx); |
e672ff9b JR |
5918 | break; |
5919 | ||
83a31283 BP |
5920 | case OFPACT_RESUBMIT: |
5921 | if (ofpact_get_RESUBMIT(a)->table_id == 0xff) { | |
5922 | /* This resubmit action is relative to the current table, so we | |
5923 | * need to track what table that is.*/ | |
1d361a81 | 5924 | freeze_put_unroll_xlate(ctx); |
83a31283 BP |
5925 | } |
5926 | break; | |
5927 | ||
e672ff9b JR |
5928 | case OFPACT_SET_TUNNEL: |
5929 | case OFPACT_REG_MOVE: | |
5930 | case OFPACT_SET_FIELD: | |
5931 | case OFPACT_STACK_PUSH: | |
5932 | case OFPACT_STACK_POP: | |
5933 | case OFPACT_LEARN: | |
5934 | case OFPACT_WRITE_METADATA: | |
83a31283 | 5935 | case OFPACT_GOTO_TABLE: |
e672ff9b JR |
5936 | case OFPACT_ENQUEUE: |
5937 | case OFPACT_SET_VLAN_VID: | |
5938 | case OFPACT_SET_VLAN_PCP: | |
5939 | case OFPACT_STRIP_VLAN: | |
5940 | case OFPACT_PUSH_VLAN: | |
5941 | case OFPACT_SET_ETH_SRC: | |
5942 | case OFPACT_SET_ETH_DST: | |
5943 | case OFPACT_SET_IPV4_SRC: | |
5944 | case OFPACT_SET_IPV4_DST: | |
5945 | case OFPACT_SET_IP_DSCP: | |
5946 | case OFPACT_SET_IP_ECN: | |
5947 | case OFPACT_SET_IP_TTL: | |
5948 | case OFPACT_SET_L4_SRC_PORT: | |
5949 | case OFPACT_SET_L4_DST_PORT: | |
5950 | case OFPACT_SET_QUEUE: | |
5951 | case OFPACT_POP_QUEUE: | |
5952 | case OFPACT_PUSH_MPLS: | |
5953 | case OFPACT_POP_MPLS: | |
5954 | case OFPACT_SET_MPLS_LABEL: | |
5955 | case OFPACT_SET_MPLS_TC: | |
5956 | case OFPACT_SET_MPLS_TTL: | |
5957 | case OFPACT_MULTIPATH: | |
5958 | case OFPACT_BUNDLE: | |
5959 | case OFPACT_EXIT: | |
5960 | case OFPACT_UNROLL_XLATE: | |
5961 | case OFPACT_FIN_TIMEOUT: | |
5962 | case OFPACT_CLEAR_ACTIONS: | |
5963 | case OFPACT_WRITE_ACTIONS: | |
5964 | case OFPACT_METER: | |
5965 | case OFPACT_SAMPLE: | |
7ae62a67 | 5966 | case OFPACT_CLONE: |
f839892a JS |
5967 | case OFPACT_ENCAP: |
5968 | case OFPACT_DECAP: | |
d4abaff5 | 5969 | case OFPACT_DEBUG_RECIRC: |
a934a3dd | 5970 | case OFPACT_DEBUG_SLOW: |
07659514 | 5971 | case OFPACT_CT: |
72fe7578 | 5972 | case OFPACT_CT_CLEAR: |
9ac0aada | 5973 | case OFPACT_NAT: |
5b34f8fc | 5974 | case OFPACT_CHECK_PKT_LARGER: |
83a31283 | 5975 | /* These may not generate PACKET INs. */ |
e672ff9b JR |
5976 | break; |
5977 | ||
e672ff9b JR |
5978 | case OFPACT_NOTE: |
5979 | case OFPACT_CONJUNCTION: | |
83a31283 | 5980 | /* These need not be copied for restoration. */ |
e672ff9b JR |
5981 | continue; |
5982 | } | |
5983 | /* Copy the action over. */ | |
1d361a81 | 5984 | ofpbuf_put(&ctx->frozen_actions, a, OFPACT_ALIGN(a->len)); |
e672ff9b JR |
5985 | } |
5986 | } | |
5987 | ||
8e53fe8c | 5988 | static void |
f2d105b5 JS |
5989 | put_ct_mark(const struct flow *flow, struct ofpbuf *odp_actions, |
5990 | struct flow_wildcards *wc) | |
8e53fe8c | 5991 | { |
2a754f4a JS |
5992 | if (wc->masks.ct_mark) { |
5993 | struct { | |
5994 | uint32_t key; | |
5995 | uint32_t mask; | |
5996 | } *odp_ct_mark; | |
5997 | ||
5998 | odp_ct_mark = nl_msg_put_unspec_uninit(odp_actions, OVS_CT_ATTR_MARK, | |
5999 | sizeof(*odp_ct_mark)); | |
6000 | odp_ct_mark->key = flow->ct_mark & wc->masks.ct_mark; | |
6001 | odp_ct_mark->mask = wc->masks.ct_mark; | |
8e53fe8c JS |
6002 | } |
6003 | } | |
6004 | ||
9daf2348 | 6005 | static void |
f2d105b5 JS |
6006 | put_ct_label(const struct flow *flow, struct ofpbuf *odp_actions, |
6007 | struct flow_wildcards *wc) | |
9daf2348 | 6008 | { |
2ff8484b | 6009 | if (!ovs_u128_is_zero(wc->masks.ct_label)) { |
9daf2348 JS |
6010 | struct { |
6011 | ovs_u128 key; | |
6012 | ovs_u128 mask; | |
89cf41ec | 6013 | } odp_ct_label; |
9daf2348 | 6014 | |
89cf41ec BP |
6015 | odp_ct_label.key = ovs_u128_and(flow->ct_label, wc->masks.ct_label); |
6016 | odp_ct_label.mask = wc->masks.ct_label; | |
6017 | nl_msg_put_unspec(odp_actions, OVS_CT_ATTR_LABELS, | |
6018 | &odp_ct_label, sizeof odp_ct_label); | |
9daf2348 JS |
6019 | } |
6020 | } | |
6021 | ||
a13a0209 AT |
6022 | static void |
6023 | put_drop_action(struct ofpbuf *odp_actions, enum xlate_error error) | |
6024 | { | |
6025 | nl_msg_put_u32(odp_actions, OVS_ACTION_ATTR_DROP, error); | |
6026 | } | |
6027 | ||
d787ad39 | 6028 | static void |
2d9b49dd BP |
6029 | put_ct_helper(struct xlate_ctx *ctx, |
6030 | struct ofpbuf *odp_actions, struct ofpact_conntrack *ofc) | |
d787ad39 JS |
6031 | { |
6032 | if (ofc->alg) { | |
40c7b2fc JS |
6033 | switch(ofc->alg) { |
6034 | case IPPORT_FTP: | |
d787ad39 | 6035 | nl_msg_put_string(odp_actions, OVS_CT_ATTR_HELPER, "ftp"); |
40c7b2fc JS |
6036 | break; |
6037 | case IPPORT_TFTP: | |
6038 | nl_msg_put_string(odp_actions, OVS_CT_ATTR_HELPER, "tftp"); | |
6039 | break; | |
6040 | default: | |
2d9b49dd | 6041 | xlate_report_error(ctx, "cannot serialize ct_helper %d", ofc->alg); |
40c7b2fc | 6042 | break; |
d787ad39 JS |
6043 | } |
6044 | } | |
6045 | } | |
6046 | ||
187bb41f YHW |
6047 | static void |
6048 | put_ct_timeout(struct ofpbuf *odp_actions, const struct dpif_backer *backer, | |
6049 | const struct flow *flow, struct flow_wildcards *wc, | |
6050 | uint16_t zone_id) | |
6051 | { | |
6052 | bool unwildcard; | |
6053 | char *tp_name = NULL; | |
6054 | ||
6055 | if (ofproto_dpif_ct_zone_timeout_policy_get_name(backer, zone_id, | |
6056 | ntohs(flow->dl_type), flow->nw_proto, &tp_name, &unwildcard)) { | |
6057 | nl_msg_put_string(odp_actions, OVS_CT_ATTR_TIMEOUT, tp_name); | |
6058 | ||
6059 | if (unwildcard) { | |
6060 | /* The underlying datapath requires separate timeout | |
6061 | * policies for different Ethertypes and IP protocols. We | |
6062 | * don't need to unwildcard 'wc->masks.dl_type' since that | |
6063 | * field is always unwildcarded in megaflows. */ | |
6064 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
6065 | } | |
6066 | } | |
6067 | free(tp_name); | |
6068 | } | |
6069 | ||
9ac0aada JR |
6070 | static void |
6071 | put_ct_nat(struct xlate_ctx *ctx) | |
6072 | { | |
6073 | struct ofpact_nat *ofn = ctx->ct_nat_action; | |
6074 | size_t nat_offset; | |
6075 | ||
6076 | if (!ofn) { | |
6077 | return; | |
6078 | } | |
6079 | ||
6080 | nat_offset = nl_msg_start_nested(ctx->odp_actions, OVS_CT_ATTR_NAT); | |
6081 | if (ofn->flags & NX_NAT_F_SRC || ofn->flags & NX_NAT_F_DST) { | |
6082 | nl_msg_put_flag(ctx->odp_actions, ofn->flags & NX_NAT_F_SRC | |
6083 | ? OVS_NAT_ATTR_SRC : OVS_NAT_ATTR_DST); | |
6084 | if (ofn->flags & NX_NAT_F_PERSISTENT) { | |
6085 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PERSISTENT); | |
6086 | } | |
6087 | if (ofn->flags & NX_NAT_F_PROTO_HASH) { | |
6088 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_HASH); | |
6089 | } else if (ofn->flags & NX_NAT_F_PROTO_RANDOM) { | |
6090 | nl_msg_put_flag(ctx->odp_actions, OVS_NAT_ATTR_PROTO_RANDOM); | |
6091 | } | |
6092 | if (ofn->range_af == AF_INET) { | |
73e8bc23 | 6093 | nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, |
9ac0aada JR |
6094 | ofn->range.addr.ipv4.min); |
6095 | if (ofn->range.addr.ipv4.max && | |
73e8bc23 BP |
6096 | (ntohl(ofn->range.addr.ipv4.max) |
6097 | > ntohl(ofn->range.addr.ipv4.min))) { | |
6098 | nl_msg_put_be32(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, | |
6099 | ofn->range.addr.ipv4.max); | |
9ac0aada JR |
6100 | } |
6101 | } else if (ofn->range_af == AF_INET6) { | |
6102 | nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MIN, | |
6103 | &ofn->range.addr.ipv6.min, | |
6104 | sizeof ofn->range.addr.ipv6.min); | |
6105 | if (!ipv6_mask_is_any(&ofn->range.addr.ipv6.max) && | |
6106 | memcmp(&ofn->range.addr.ipv6.max, &ofn->range.addr.ipv6.min, | |
6107 | sizeof ofn->range.addr.ipv6.max) > 0) { | |
6108 | nl_msg_put_unspec(ctx->odp_actions, OVS_NAT_ATTR_IP_MAX, | |
6109 | &ofn->range.addr.ipv6.max, | |
6110 | sizeof ofn->range.addr.ipv6.max); | |
6111 | } | |
6112 | } | |
6113 | if (ofn->range_af != AF_UNSPEC && ofn->range.proto.min) { | |
6114 | nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MIN, | |
6115 | ofn->range.proto.min); | |
6116 | if (ofn->range.proto.max && | |
6117 | ofn->range.proto.max > ofn->range.proto.min) { | |
6118 | nl_msg_put_u16(ctx->odp_actions, OVS_NAT_ATTR_PROTO_MAX, | |
6119 | ofn->range.proto.max); | |
6120 | } | |
6121 | } | |
6122 | } | |
6123 | nl_msg_end_nested(ctx->odp_actions, nat_offset); | |
6124 | } | |
6125 | ||
07659514 | 6126 | static void |
feee58b9 AZ |
6127 | compose_conntrack_action(struct xlate_ctx *ctx, struct ofpact_conntrack *ofc, |
6128 | bool is_last_action) | |
07659514 | 6129 | { |
f2d105b5 | 6130 | ovs_u128 old_ct_label_mask = ctx->wc->masks.ct_label; |
f2d105b5 | 6131 | uint32_t old_ct_mark_mask = ctx->wc->masks.ct_mark; |
07659514 JS |
6132 | size_t ct_offset; |
6133 | uint16_t zone; | |
6134 | ||
6135 | /* Ensure that any prior actions are applied before composing the new | |
6136 | * conntrack action. */ | |
6137 | xlate_commit_actions(ctx); | |
6138 | ||
8e53fe8c | 6139 | /* Process nested actions first, to populate the key. */ |
9ac0aada | 6140 | ctx->ct_nat_action = NULL; |
f2d105b5 | 6141 | ctx->wc->masks.ct_mark = 0; |
f6fabcc6 | 6142 | ctx->wc->masks.ct_label = OVS_U128_ZERO; |
feee58b9 | 6143 | do_xlate_actions(ofc->actions, ofpact_ct_get_action_len(ofc), ctx, |
f5634764 | 6144 | is_last_action, false); |
8e53fe8c | 6145 | |
07659514 JS |
6146 | if (ofc->zone_src.field) { |
6147 | zone = mf_get_subfield(&ofc->zone_src, &ctx->xin->flow); | |
6148 | } else { | |
6149 | zone = ofc->zone_imm; | |
6150 | } | |
6151 | ||
6152 | ct_offset = nl_msg_start_nested(ctx->odp_actions, OVS_ACTION_ATTR_CT); | |
6153 | if (ofc->flags & NX_CT_F_COMMIT) { | |
a76a37ef JR |
6154 | nl_msg_put_flag(ctx->odp_actions, ofc->flags & NX_CT_F_FORCE ? |
6155 | OVS_CT_ATTR_FORCE_COMMIT : OVS_CT_ATTR_COMMIT); | |
adfe7a0b JR |
6156 | if (ctx->xbridge->support.ct_eventmask) { |
6157 | nl_msg_put_u32(ctx->odp_actions, OVS_CT_ATTR_EVENTMASK, | |
975954af | 6158 | OVS_CT_EVENTMASK_DEFAULT); |
adfe7a0b | 6159 | } |
187bb41f YHW |
6160 | if (ctx->xbridge->support.ct_timeout) { |
6161 | put_ct_timeout(ctx->odp_actions, ctx->xbridge->ofproto->backer, | |
6162 | &ctx->xin->flow, ctx->wc, zone); | |
6163 | } | |
07659514 JS |
6164 | } |
6165 | nl_msg_put_u16(ctx->odp_actions, OVS_CT_ATTR_ZONE, zone); | |
f2d105b5 JS |
6166 | put_ct_mark(&ctx->xin->flow, ctx->odp_actions, ctx->wc); |
6167 | put_ct_label(&ctx->xin->flow, ctx->odp_actions, ctx->wc); | |
2d9b49dd | 6168 | put_ct_helper(ctx, ctx->odp_actions, ofc); |
9ac0aada JR |
6169 | put_ct_nat(ctx); |
6170 | ctx->ct_nat_action = NULL; | |
07659514 JS |
6171 | nl_msg_end_nested(ctx->odp_actions, ct_offset); |
6172 | ||
f2d105b5 | 6173 | ctx->wc->masks.ct_mark = old_ct_mark_mask; |
f2d105b5 | 6174 | ctx->wc->masks.ct_label = old_ct_label_mask; |
8e53fe8c | 6175 | |
f6fabcc6 | 6176 | if (ofc->recirc_table != NX_CT_RECIRC_NONE) { |
07659514 | 6177 | ctx->conntracked = true; |
5fdd80cc | 6178 | compose_recirculate_and_fork(ctx, ofc->recirc_table, zone); |
07659514 | 6179 | } |
f6fabcc6 JP |
6180 | |
6181 | /* The ct_* fields are only available in the scope of the 'recirc_table' | |
6182 | * call chain. */ | |
6183 | flow_clear_conntrack(&ctx->xin->flow); | |
5a046d3b YHW |
6184 | xlate_report(ctx, OFT_DETAIL, "Sets the packet to an untracked state, " |
6185 | "and clears all the conntrack fields."); | |
f6fabcc6 | 6186 | ctx->conntracked = false; |
07659514 JS |
6187 | } |
6188 | ||
1fe178d2 EG |
6189 | static void |
6190 | compose_ct_clear_action(struct xlate_ctx *ctx) | |
6191 | { | |
6192 | clear_conntrack(ctx); | |
6193 | /* This action originally existed without dpif support. So to preserve | |
6194 | * compatibility, only append it if the dpif supports it. */ | |
6195 | if (ctx->xbridge->support.ct_clear) { | |
6196 | nl_msg_put_flag(ctx->odp_actions, OVS_ACTION_ATTR_CT_CLEAR); | |
6197 | } | |
6198 | } | |
6199 | ||
5b34f8fc NS |
6200 | /* check_pkt_larger action checks the packet length and stores the |
6201 | * result in the register bit. We translate this action to the | |
6202 | * datapath action - 'check_pkt_len' whose format | |
6203 | * is: 'check_pkt_len(pkt_len, ge(actions), le(actions))'. | |
6204 | * | |
6205 | * We first set the destination register bit to 1 and call | |
6206 | * 'do_xlate_actions' for the case - packet len greater than | |
6207 | * the specified packet length. | |
6208 | * | |
6209 | * We then set the destination register bit to 0 and call | |
6210 | * 'do_xlate_actions' for the case - packet length is lesser or | |
6211 | * equal to the specified packet length. | |
6212 | * | |
6213 | * It is possible for freezing to happen for both the cases. | |
6214 | */ | |
6215 | static void | |
6216 | xlate_check_pkt_larger(struct xlate_ctx *ctx, | |
6217 | struct ofpact_check_pkt_larger *check_pkt_larger, | |
6218 | const struct ofpact *remaining_acts, | |
6219 | size_t remaining_acts_len) | |
6220 | { | |
6221 | union mf_subvalue value; | |
6222 | memset(&value, 0, sizeof value); | |
6223 | if (!ctx->xbridge->support.check_pkt_len) { | |
6224 | uint8_t is_pkt_larger = 0; | |
6225 | if (ctx->xin->packet) { | |
6226 | is_pkt_larger = | |
6227 | dp_packet_size(ctx->xin->packet) > check_pkt_larger->pkt_len; | |
6228 | } | |
6229 | value.u8_val = is_pkt_larger; | |
6230 | mf_write_subfield_flow(&check_pkt_larger->dst, &value, | |
6231 | &ctx->xin->flow); | |
6232 | /* If datapath doesn't support check_pkt_len action, then set the | |
6233 | * SLOW_ACTION flag. If we don't set SLOW_ACTION, we | |
6234 | * will push a flow to the datapath based on the packet length | |
6235 | * in ctx->xin->packet. For subsequent patches which match the | |
6236 | * same flow, datapath will apply the actions without considering | |
6237 | * the packet length. This results in wrong actions being applied. | |
6238 | */ | |
6239 | ctx->xout->slow |= SLOW_ACTION; | |
6240 | return; | |
6241 | } | |
6242 | ||
6243 | struct ofpbuf old_stack = ctx->stack; | |
6244 | union mf_subvalue new_stack[1024 / sizeof(union mf_subvalue)]; | |
6245 | ofpbuf_use_stub(&ctx->stack, new_stack, sizeof new_stack); | |
6246 | ofpbuf_put(&ctx->stack, old_stack.data, old_stack.size); | |
6247 | ||
6248 | struct ofpbuf old_action_set = ctx->action_set; | |
6249 | uint64_t actset_stub[1024 / 8]; | |
6250 | ofpbuf_use_stub(&ctx->action_set, actset_stub, sizeof actset_stub); | |
6251 | ofpbuf_put(&ctx->action_set, old_action_set.data, old_action_set.size); | |
6252 | ||
6253 | struct flow old_flow = ctx->xin->flow; | |
6254 | xlate_commit_actions(ctx); | |
6255 | struct flow old_base = ctx->base_flow; | |
6256 | bool old_was_mpls = ctx->was_mpls; | |
6257 | bool old_conntracked = ctx->conntracked; | |
6258 | ||
6259 | size_t offset = nl_msg_start_nested(ctx->odp_actions, | |
6260 | OVS_ACTION_ATTR_CHECK_PKT_LEN); | |
6261 | nl_msg_put_u16(ctx->odp_actions, OVS_CHECK_PKT_LEN_ATTR_PKT_LEN, | |
6262 | check_pkt_larger->pkt_len); | |
6263 | size_t offset_attr = nl_msg_start_nested( | |
6264 | ctx->odp_actions, OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_GREATER); | |
6265 | value.u8_val = 1; | |
6266 | mf_write_subfield_flow(&check_pkt_larger->dst, &value, &ctx->xin->flow); | |
6267 | do_xlate_actions(remaining_acts, remaining_acts_len, ctx, true, false); | |
6268 | if (!ctx->freezing) { | |
6269 | xlate_action_set(ctx); | |
6270 | } | |
6271 | if (ctx->freezing) { | |
6272 | finish_freezing(ctx); | |
6273 | } | |
6274 | nl_msg_end_nested(ctx->odp_actions, offset_attr); | |
6275 | ||
6276 | ctx->base_flow = old_base; | |
6277 | ctx->was_mpls = old_was_mpls; | |
6278 | ctx->conntracked = old_conntracked; | |
6279 | ctx->xin->flow = old_flow; | |
6280 | ||
6281 | /* If the flow translation for the IF_GREATER case requires freezing, | |
6282 | * then ctx->exit would be true. Reset to false so that we can | |
6283 | * do flow translation for 'IF_LESS_EQUAL' case. finish_freezing() | |
6284 | * would have taken care of Undoing the changes done for freeze. */ | |
6285 | ctx->exit = false; | |
6286 | ||
6287 | offset_attr = nl_msg_start_nested( | |
6288 | ctx->odp_actions, OVS_CHECK_PKT_LEN_ATTR_ACTIONS_IF_LESS_EQUAL); | |
6289 | value.u8_val = 0; | |
6290 | mf_write_subfield_flow(&check_pkt_larger->dst, &value, &ctx->xin->flow); | |
6291 | do_xlate_actions(remaining_acts, remaining_acts_len, ctx, true, false); | |
6292 | if (!ctx->freezing) { | |
6293 | xlate_action_set(ctx); | |
6294 | } | |
6295 | if (ctx->freezing) { | |
6296 | finish_freezing(ctx); | |
6297 | } | |
6298 | nl_msg_end_nested(ctx->odp_actions, offset_attr); | |
6299 | nl_msg_end_nested(ctx->odp_actions, offset); | |
6300 | ||
6301 | ofpbuf_uninit(&ctx->action_set); | |
6302 | ctx->action_set = old_action_set; | |
6303 | ofpbuf_uninit(&ctx->stack); | |
6304 | ctx->stack = old_stack; | |
6305 | ctx->base_flow = old_base; | |
6306 | ctx->was_mpls = old_was_mpls; | |
6307 | ctx->conntracked = old_conntracked; | |
6308 | ctx->xin->flow = old_flow; | |
6309 | ctx->exit = true; | |
6310 | } | |
6311 | ||
f839892a JS |
6312 | static void |
6313 | rewrite_flow_encap_ethernet(struct xlate_ctx *ctx, | |
6314 | struct flow *flow, | |
6315 | struct flow_wildcards *wc) | |
6316 | { | |
6317 | wc->masks.packet_type = OVS_BE32_MAX; | |
6318 | if (pt_ns(flow->packet_type) == OFPHTN_ETHERTYPE) { | |
6319 | /* Only adjust the packet_type and zero the dummy Ethernet addresses. */ | |
6320 | ovs_be16 ethertype = pt_ns_type_be(flow->packet_type); | |
6321 | flow->packet_type = htonl(PT_ETH); | |
6322 | flow->dl_src = eth_addr_zero; | |
6323 | flow->dl_dst = eth_addr_zero; | |
6324 | flow->dl_type = ethertype; | |
6325 | } else { | |
1fc11c59 | 6326 | /* Error handling: drop packet. */ |
f839892a | 6327 | xlate_report_debug(ctx, OFT_ACTION, |
1fc11c59 JS |
6328 | "Dropping packet as encap(ethernet) is not " |
6329 | "supported for packet type ethernet."); | |
7873e106 | 6330 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
f839892a JS |
6331 | } |
6332 | } | |
6333 | ||
1fc11c59 JS |
6334 | /* For an MD2 NSH header returns a pointer to an ofpbuf with the encoded |
6335 | * MD2 TLVs provided as encap properties to the encap operation. This | |
f59cb331 | 6336 | * will be stored as encap_data in the ctx and copied into the push_nsh |
1fc11c59 JS |
6337 | * action at the next commit. */ |
6338 | static struct ofpbuf * | |
f59cb331 YY |
6339 | rewrite_flow_push_nsh(struct xlate_ctx *ctx, |
6340 | const struct ofpact_encap *encap, | |
6341 | struct flow *flow, | |
6342 | struct flow_wildcards *wc) | |
1fc11c59 JS |
6343 | { |
6344 | ovs_be32 packet_type = flow->packet_type; | |
6345 | const char *ptr = (char *) encap->props; | |
f59cb331 | 6346 | struct ofpbuf *buf = ofpbuf_new(NSH_CTX_HDRS_MAX_LEN); |
1fc11c59 JS |
6347 | uint8_t md_type = NSH_M_TYPE1; |
6348 | uint8_t np = 0; | |
6349 | int i; | |
6350 | ||
6351 | /* Scan the optional NSH encap TLV properties, if any. */ | |
6352 | for (i = 0; i < encap->n_props; i++) { | |
6353 | struct ofpact_ed_prop *prop_ptr = | |
6354 | ALIGNED_CAST(struct ofpact_ed_prop *, ptr); | |
6355 | if (prop_ptr->prop_class == OFPPPC_NSH) { | |
6356 | switch (prop_ptr->type) { | |
6357 | case OFPPPT_PROP_NSH_MDTYPE: { | |
6358 | struct ofpact_ed_prop_nsh_md_type *prop_md_type = | |
6359 | ALIGNED_CAST(struct ofpact_ed_prop_nsh_md_type *, | |
6360 | prop_ptr); | |
6361 | md_type = prop_md_type->md_type; | |
6362 | break; | |
6363 | } | |
6364 | case OFPPPT_PROP_NSH_TLV: { | |
6365 | struct ofpact_ed_prop_nsh_tlv *tlv_prop = | |
6366 | ALIGNED_CAST(struct ofpact_ed_prop_nsh_tlv *, | |
6367 | prop_ptr); | |
6368 | struct nsh_md2_tlv *md2_ctx = | |
6369 | ofpbuf_put_uninit(buf, sizeof(*md2_ctx)); | |
6370 | md2_ctx->md_class = tlv_prop->tlv_class; | |
6371 | md2_ctx->type = tlv_prop->tlv_type; | |
6372 | md2_ctx->length = tlv_prop->tlv_len; | |
6373 | size_t len = ROUND_UP(md2_ctx->length, 4); | |
6374 | size_t padding = len - md2_ctx->length; | |
6375 | ofpbuf_put(buf, tlv_prop->data, md2_ctx->length); | |
6376 | ofpbuf_put_zeros(buf, padding); | |
6377 | break; | |
6378 | } | |
6379 | default: | |
6380 | /* No other NSH encap properties defined yet. */ | |
6381 | break; | |
6382 | } | |
6383 | } | |
6384 | ptr += ROUND_UP(prop_ptr->len, 8); | |
6385 | } | |
f59cb331 | 6386 | if (buf->size == 0 || buf->size > NSH_CTX_HDRS_MAX_LEN) { |
1fc11c59 JS |
6387 | ofpbuf_delete(buf); |
6388 | buf = NULL; | |
6389 | } | |
6390 | ||
6391 | /* Determine the Next Protocol field for NSH header. */ | |
6392 | switch (ntohl(packet_type)) { | |
6393 | case PT_ETH: | |
6394 | np = NSH_P_ETHERNET; | |
6395 | break; | |
6396 | case PT_IPV4: | |
6397 | np = NSH_P_IPV4; | |
6398 | break; | |
6399 | case PT_IPV6: | |
6400 | np = NSH_P_IPV6; | |
6401 | break; | |
6402 | case PT_NSH: | |
6403 | np = NSH_P_NSH; | |
6404 | break; | |
6405 | default: | |
6406 | /* Error handling: drop packet. */ | |
6407 | xlate_report_debug(ctx, OFT_ACTION, | |
6408 | "Dropping packet as encap(nsh) is not " | |
6409 | "supported for packet type (%d,0x%x)", | |
6410 | pt_ns(packet_type), pt_ns_type(packet_type)); | |
7873e106 | 6411 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
1fc11c59 JS |
6412 | return buf; |
6413 | } | |
6414 | /* Note that we have matched on packet_type! */ | |
6415 | wc->masks.packet_type = OVS_BE32_MAX; | |
6416 | ||
6417 | /* Reset all current flow packet headers. */ | |
6418 | memset(&flow->dl_dst, 0, | |
6419 | sizeof(struct flow) - offsetof(struct flow, dl_dst)); | |
6420 | ||
6421 | /* Populate the flow with the new NSH header. */ | |
6422 | flow->packet_type = htonl(PT_NSH); | |
6423 | flow->dl_type = htons(ETH_TYPE_NSH); | |
17553f27 YY |
6424 | flow->nsh.flags = 0; |
6425 | flow->nsh.ttl = 63; | |
1fc11c59 | 6426 | flow->nsh.np = np; |
17553f27 | 6427 | flow->nsh.path_hdr = htonl(255); |
1fc11c59 JS |
6428 | |
6429 | if (md_type == NSH_M_TYPE1) { | |
6430 | flow->nsh.mdtype = NSH_M_TYPE1; | |
f59cb331 | 6431 | memset(flow->nsh.context, 0, sizeof flow->nsh.context); |
1fc11c59 JS |
6432 | if (buf) { |
6433 | /* Drop any MD2 context TLVs. */ | |
6434 | ofpbuf_delete(buf); | |
6435 | buf = NULL; | |
6436 | } | |
6437 | } else if (md_type == NSH_M_TYPE2) { | |
6438 | flow->nsh.mdtype = NSH_M_TYPE2; | |
6439 | } | |
17553f27 | 6440 | flow->nsh.mdtype &= NSH_MDTYPE_MASK; |
1fc11c59 JS |
6441 | |
6442 | return buf; | |
6443 | } | |
6444 | ||
f839892a JS |
6445 | static void |
6446 | xlate_generic_encap_action(struct xlate_ctx *ctx, | |
6447 | const struct ofpact_encap *encap) | |
6448 | { | |
6449 | struct flow *flow = &ctx->xin->flow; | |
6450 | struct flow_wildcards *wc = ctx->wc; | |
1fc11c59 | 6451 | struct ofpbuf *encap_data = NULL; |
f839892a JS |
6452 | |
6453 | /* Ensure that any pending actions on the inner packet are applied before | |
6454 | * rewriting the flow */ | |
6455 | xlate_commit_actions(ctx); | |
6456 | ||
6457 | /* Rewrite the flow to reflect the effect of pushing the new encap header. */ | |
6458 | switch (ntohl(encap->new_pkt_type)) { | |
6459 | case PT_ETH: | |
6460 | rewrite_flow_encap_ethernet(ctx, flow, wc); | |
6461 | break; | |
1fc11c59 | 6462 | case PT_NSH: |
f59cb331 | 6463 | encap_data = rewrite_flow_push_nsh(ctx, encap, flow, wc); |
1fc11c59 | 6464 | break; |
f839892a | 6465 | default: |
1fc11c59 JS |
6466 | /* New packet type was checked during decoding. */ |
6467 | OVS_NOT_REACHED(); | |
f839892a JS |
6468 | } |
6469 | ||
6470 | if (!ctx->error) { | |
6471 | /* The actual encap datapath action will be generated at next commit. */ | |
6472 | ctx->pending_encap = true; | |
1fc11c59 | 6473 | ctx->encap_data = encap_data; |
f839892a JS |
6474 | } |
6475 | } | |
6476 | ||
6477 | /* Returns true if packet must be recirculated after decapsulation. */ | |
6478 | static bool | |
6479 | xlate_generic_decap_action(struct xlate_ctx *ctx, | |
6480 | const struct ofpact_decap *decap OVS_UNUSED) | |
6481 | { | |
6482 | struct flow *flow = &ctx->xin->flow; | |
6483 | ||
6484 | /* Ensure that any pending actions on the current packet are applied | |
6485 | * before generating the decap action. */ | |
6486 | xlate_commit_actions(ctx); | |
6487 | ||
6488 | /* We assume for now that the new_pkt_type is PT_USE_NEXT_PROTO. */ | |
6489 | switch (ntohl(flow->packet_type)) { | |
6490 | case PT_ETH: | |
6491 | if (flow->vlans[0].tci & htons(VLAN_CFI)) { | |
6492 | /* Error handling: drop packet. */ | |
6493 | xlate_report_debug(ctx, OFT_ACTION, "Dropping packet, cannot " | |
6494 | "decap Ethernet if VLAN is present."); | |
7873e106 | 6495 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
f839892a JS |
6496 | } else { |
6497 | /* Just change the packet_type. | |
6498 | * Delay generating pop_eth to the next commit. */ | |
6499 | flow->packet_type = htonl(PACKET_TYPE(OFPHTN_ETHERTYPE, | |
6500 | ntohs(flow->dl_type))); | |
6501 | ctx->wc->masks.dl_type = OVS_BE16_MAX; | |
6502 | } | |
6503 | return false; | |
1fc11c59 | 6504 | case PT_NSH: |
f59cb331 | 6505 | /* The pop_nsh action is generated at the commit executed as |
1fc11c59 JS |
6506 | * part of freezing the ctx for recirculation. Here we just set |
6507 | * the new packet type based on the NSH next protocol field. */ | |
6508 | switch (flow->nsh.np) { | |
6509 | case NSH_P_ETHERNET: | |
6510 | flow->packet_type = htonl(PT_ETH); | |
6511 | break; | |
6512 | case NSH_P_IPV4: | |
6513 | flow->packet_type = htonl(PT_IPV4); | |
6514 | break; | |
6515 | case NSH_P_IPV6: | |
6516 | flow->packet_type = htonl(PT_IPV6); | |
6517 | break; | |
6518 | case NSH_P_NSH: | |
6519 | flow->packet_type = htonl(PT_NSH); | |
6520 | break; | |
6521 | default: | |
6522 | /* Error handling: drop packet. */ | |
6523 | xlate_report_debug(ctx, OFT_ACTION, | |
6524 | "Dropping packet as NSH next protocol %d " | |
6525 | "is not supported", flow->nsh.np); | |
7873e106 | 6526 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
1fc11c59 JS |
6527 | return false; |
6528 | break; | |
6529 | } | |
6530 | ctx->wc->masks.nsh.np = UINT8_MAX; | |
88ec1e0a | 6531 | ctx->pending_decap = true; |
1fc11c59 JS |
6532 | /* Trigger recirculation. */ |
6533 | return true; | |
f839892a | 6534 | default: |
1fc11c59 JS |
6535 | /* Error handling: drop packet. */ |
6536 | xlate_report_debug( | |
6537 | ctx, OFT_ACTION, | |
6538 | "Dropping packet as the decap() does not support " | |
6539 | "packet type (%d,0x%x)", | |
6540 | pt_ns(flow->packet_type), pt_ns_type(flow->packet_type)); | |
7873e106 | 6541 | ctx->error = XLATE_UNSUPPORTED_PACKET_TYPE; |
f839892a JS |
6542 | return false; |
6543 | } | |
6544 | } | |
6545 | ||
e12ec36b SH |
6546 | static void |
6547 | recirc_for_mpls(const struct ofpact *a, struct xlate_ctx *ctx) | |
6548 | { | |
6549 | /* No need to recirculate if already exiting. */ | |
6550 | if (ctx->exit) { | |
6551 | return; | |
6552 | } | |
6553 | ||
6554 | /* Do not consider recirculating unless the packet was previously MPLS. */ | |
6555 | if (!ctx->was_mpls) { | |
6556 | return; | |
6557 | } | |
6558 | ||
6559 | /* Special case these actions, only recirculating if necessary. | |
6560 | * This avoids the overhead of recirculation in common use-cases. | |
6561 | */ | |
6562 | switch (a->type) { | |
6563 | ||
6564 | /* Output actions do not require recirculation. */ | |
6565 | case OFPACT_OUTPUT: | |
aaca4fe0 | 6566 | case OFPACT_OUTPUT_TRUNC: |
e12ec36b SH |
6567 | case OFPACT_ENQUEUE: |
6568 | case OFPACT_OUTPUT_REG: | |
6569 | /* Set actions that don't touch L3+ fields do not require recirculation. */ | |
6570 | case OFPACT_SET_VLAN_VID: | |
6571 | case OFPACT_SET_VLAN_PCP: | |
6572 | case OFPACT_SET_ETH_SRC: | |
6573 | case OFPACT_SET_ETH_DST: | |
6574 | case OFPACT_SET_TUNNEL: | |
6575 | case OFPACT_SET_QUEUE: | |
6576 | /* If actions of a group require recirculation that can be detected | |
6577 | * when translating them. */ | |
6578 | case OFPACT_GROUP: | |
6579 | return; | |
6580 | ||
6581 | /* Set field that don't touch L3+ fields don't require recirculation. */ | |
6582 | case OFPACT_SET_FIELD: | |
6583 | if (mf_is_l3_or_higher(ofpact_get_SET_FIELD(a)->field)) { | |
6584 | break; | |
6585 | } | |
6586 | return; | |
6587 | ||
6588 | /* For simplicity, recirculate in all other cases. */ | |
6589 | case OFPACT_CONTROLLER: | |
6590 | case OFPACT_BUNDLE: | |
6591 | case OFPACT_STRIP_VLAN: | |
6592 | case OFPACT_PUSH_VLAN: | |
6593 | case OFPACT_SET_IPV4_SRC: | |
6594 | case OFPACT_SET_IPV4_DST: | |
6595 | case OFPACT_SET_IP_DSCP: | |
6596 | case OFPACT_SET_IP_ECN: | |
6597 | case OFPACT_SET_IP_TTL: | |
6598 | case OFPACT_SET_L4_SRC_PORT: | |
6599 | case OFPACT_SET_L4_DST_PORT: | |
6600 | case OFPACT_REG_MOVE: | |
6601 | case OFPACT_STACK_PUSH: | |
6602 | case OFPACT_STACK_POP: | |
6603 | case OFPACT_DEC_TTL: | |
6604 | case OFPACT_SET_MPLS_LABEL: | |
6605 | case OFPACT_SET_MPLS_TC: | |
6606 | case OFPACT_SET_MPLS_TTL: | |
6607 | case OFPACT_DEC_MPLS_TTL: | |
6608 | case OFPACT_PUSH_MPLS: | |
6609 | case OFPACT_POP_MPLS: | |
6610 | case OFPACT_POP_QUEUE: | |
6611 | case OFPACT_FIN_TIMEOUT: | |
6612 | case OFPACT_RESUBMIT: | |
6613 | case OFPACT_LEARN: | |
6614 | case OFPACT_CONJUNCTION: | |
6615 | case OFPACT_MULTIPATH: | |
6616 | case OFPACT_NOTE: | |
6617 | case OFPACT_EXIT: | |
6618 | case OFPACT_SAMPLE: | |
7ae62a67 | 6619 | case OFPACT_CLONE: |
f839892a JS |
6620 | case OFPACT_ENCAP: |
6621 | case OFPACT_DECAP: | |
491e05c2 | 6622 | case OFPACT_DEC_NSH_TTL: |
e12ec36b SH |
6623 | case OFPACT_UNROLL_XLATE: |
6624 | case OFPACT_CT: | |
72fe7578 | 6625 | case OFPACT_CT_CLEAR: |
e12ec36b SH |
6626 | case OFPACT_NAT: |
6627 | case OFPACT_DEBUG_RECIRC: | |
a934a3dd | 6628 | case OFPACT_DEBUG_SLOW: |
e12ec36b SH |
6629 | case OFPACT_METER: |
6630 | case OFPACT_CLEAR_ACTIONS: | |
6631 | case OFPACT_WRITE_ACTIONS: | |
6632 | case OFPACT_WRITE_METADATA: | |
6633 | case OFPACT_GOTO_TABLE: | |
5b34f8fc | 6634 | case OFPACT_CHECK_PKT_LARGER: |
e12ec36b SH |
6635 | default: |
6636 | break; | |
6637 | } | |
6638 | ||
6639 | /* Recirculate */ | |
6640 | ctx_trigger_freeze(ctx); | |
6641 | } | |
6642 | ||
2d9b49dd BP |
6643 | static void |
6644 | xlate_ofpact_reg_move(struct xlate_ctx *ctx, const struct ofpact_reg_move *a) | |
6645 | { | |
6646 | mf_subfield_copy(&a->src, &a->dst, &ctx->xin->flow, ctx->wc); | |
6647 | xlate_report_subfield(ctx, &a->dst); | |
6648 | } | |
6649 | ||
6650 | static void | |
6651 | xlate_ofpact_stack_pop(struct xlate_ctx *ctx, const struct ofpact_stack *a) | |
6652 | { | |
6653 | if (nxm_execute_stack_pop(a, &ctx->xin->flow, ctx->wc, &ctx->stack)) { | |
6654 | xlate_report_subfield(ctx, &a->subfield); | |
6655 | } else { | |
6656 | xlate_report_error(ctx, "stack underflow"); | |
6657 | } | |
6658 | } | |
6659 | ||
6660 | /* Restore translation context data that was stored earlier. */ | |
6661 | static void | |
6662 | xlate_ofpact_unroll_xlate(struct xlate_ctx *ctx, | |
6663 | const struct ofpact_unroll_xlate *a) | |
6664 | { | |
6665 | ctx->table_id = a->rule_table_id; | |
6666 | ctx->rule_cookie = a->rule_cookie; | |
6667 | xlate_report(ctx, OFT_THAW, "restored state: table=%"PRIu8", " | |
6668 | "cookie=%#"PRIx64, a->rule_table_id, a->rule_cookie); | |
6669 | } | |
6670 | ||
9583bc14 EJ |
6671 | static void |
6672 | do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
f5634764 KG |
6673 | struct xlate_ctx *ctx, bool is_last_action, |
6674 | bool group_bucket_action) | |
9583bc14 | 6675 | { |
49a73e0c | 6676 | struct flow_wildcards *wc = ctx->wc; |
33bf9176 | 6677 | struct flow *flow = &ctx->xin->flow; |
9583bc14 EJ |
6678 | const struct ofpact *a; |
6679 | ||
f47ea021 JR |
6680 | /* dl_type already in the mask, not set below. */ |
6681 | ||
2d9b49dd BP |
6682 | if (!ofpacts_len) { |
6683 | xlate_report(ctx, OFT_ACTION, "drop"); | |
6684 | return; | |
6685 | } | |
6686 | ||
9583bc14 EJ |
6687 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
6688 | struct ofpact_controller *controller; | |
6689 | const struct ofpact_metadata *metadata; | |
b2dd70be JR |
6690 | const struct ofpact_set_field *set_field; |
6691 | const struct mf_field *mf; | |
feee58b9 AZ |
6692 | bool last = is_last_action && ofpact_last(a, ofpacts, ofpacts_len) |
6693 | && ctx->action_set.size; | |
9583bc14 | 6694 | |
fff1b9c0 JR |
6695 | if (ctx->error) { |
6696 | break; | |
6697 | } | |
6698 | ||
e12ec36b SH |
6699 | recirc_for_mpls(a, ctx); |
6700 | ||
e672ff9b JR |
6701 | if (ctx->exit) { |
6702 | /* Check if need to store the remaining actions for later | |
6703 | * execution. */ | |
1d361a81 BP |
6704 | if (ctx->freezing) { |
6705 | freeze_unroll_actions(a, ofpact_end(ofpacts, ofpacts_len), | |
e672ff9b JR |
6706 | ctx); |
6707 | } | |
6708 | break; | |
7bbdd84f SH |
6709 | } |
6710 | ||
2d9b49dd BP |
6711 | if (OVS_UNLIKELY(ctx->xin->trace)) { |
6712 | struct ds s = DS_EMPTY_INITIALIZER; | |
efefbcae BP |
6713 | struct ofpact_format_params fp = { .s = &s }; |
6714 | ofpacts_format(a, OFPACT_ALIGN(a->len), &fp); | |
2d9b49dd BP |
6715 | xlate_report(ctx, OFT_ACTION, "%s", ds_cstr(&s)); |
6716 | ds_destroy(&s); | |
6717 | } | |
6718 | ||
9583bc14 EJ |
6719 | switch (a->type) { |
6720 | case OFPACT_OUTPUT: | |
6721 | xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port, | |
8b496c72 | 6722 | ofpact_get_OUTPUT(a)->max_len, true, last, |
f5634764 | 6723 | false, group_bucket_action); |
9583bc14 EJ |
6724 | break; |
6725 | ||
7395c052 | 6726 | case OFPACT_GROUP: |
feee58b9 | 6727 | if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id, last)) { |
1d741d6d | 6728 | /* Group could not be found. */ |
db88b35c JR |
6729 | |
6730 | /* XXX: Terminates action list translation, but does not | |
6731 | * terminate the pipeline. */ | |
f4fb341b SH |
6732 | return; |
6733 | } | |
7395c052 NZ |
6734 | break; |
6735 | ||
9583bc14 EJ |
6736 | case OFPACT_CONTROLLER: |
6737 | controller = ofpact_get_CONTROLLER(a); | |
77ab5fd2 BP |
6738 | if (controller->pause) { |
6739 | ctx->pause = controller; | |
77ab5fd2 BP |
6740 | ctx_trigger_freeze(ctx); |
6741 | a = ofpact_next(a); | |
6742 | } else { | |
d39ec23d JP |
6743 | xlate_controller_action(ctx, controller->max_len, |
6744 | controller->reason, | |
6745 | controller->controller_id, | |
206ddb9a | 6746 | controller->provider_meter_id, |
d39ec23d JP |
6747 | controller->userdata, |
6748 | controller->userdata_len); | |
77ab5fd2 | 6749 | } |
9583bc14 EJ |
6750 | break; |
6751 | ||
6752 | case OFPACT_ENQUEUE: | |
16194afd DDP |
6753 | memset(&wc->masks.skb_priority, 0xff, |
6754 | sizeof wc->masks.skb_priority); | |
f5634764 KG |
6755 | xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a), last, |
6756 | group_bucket_action); | |
9583bc14 EJ |
6757 | break; |
6758 | ||
6759 | case OFPACT_SET_VLAN_VID: | |
f0fb825a EG |
6760 | wc->masks.vlans[0].tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
6761 | if (flow->vlans[0].tci & htons(VLAN_CFI) || | |
ca287d20 | 6762 | ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) { |
f0fb825a EG |
6763 | if (!flow->vlans[0].tpid) { |
6764 | flow->vlans[0].tpid = htons(ETH_TYPE_VLAN); | |
6765 | } | |
6766 | flow->vlans[0].tci &= ~htons(VLAN_VID_MASK); | |
6767 | flow->vlans[0].tci |= | |
6768 | (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid) | | |
6769 | htons(VLAN_CFI)); | |
ca287d20 | 6770 | } |
9583bc14 EJ |
6771 | break; |
6772 | ||
6773 | case OFPACT_SET_VLAN_PCP: | |
f0fb825a EG |
6774 | wc->masks.vlans[0].tci |= htons(VLAN_PCP_MASK | VLAN_CFI); |
6775 | if (flow->vlans[0].tci & htons(VLAN_CFI) || | |
ca287d20 | 6776 | ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) { |
f0fb825a EG |
6777 | if (!flow->vlans[0].tpid) { |
6778 | flow->vlans[0].tpid = htons(ETH_TYPE_VLAN); | |
6779 | } | |
6780 | flow->vlans[0].tci &= ~htons(VLAN_PCP_MASK); | |
6781 | flow->vlans[0].tci |= | |
6782 | htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp | |
6783 | << VLAN_PCP_SHIFT) | VLAN_CFI); | |
ca287d20 | 6784 | } |
9583bc14 EJ |
6785 | break; |
6786 | ||
6787 | case OFPACT_STRIP_VLAN: | |
f0fb825a | 6788 | flow_pop_vlan(flow, wc); |
9583bc14 EJ |
6789 | break; |
6790 | ||
6791 | case OFPACT_PUSH_VLAN: | |
f0fb825a EG |
6792 | flow_push_vlan_uninit(flow, wc); |
6793 | flow->vlans[0].tpid = ofpact_get_PUSH_VLAN(a)->ethertype; | |
6794 | flow->vlans[0].tci = htons(VLAN_CFI); | |
9583bc14 EJ |
6795 | break; |
6796 | ||
6797 | case OFPACT_SET_ETH_SRC: | |
74ff3298 JR |
6798 | WC_MASK_FIELD(wc, dl_src); |
6799 | flow->dl_src = ofpact_get_SET_ETH_SRC(a)->mac; | |
9583bc14 EJ |
6800 | break; |
6801 | ||
6802 | case OFPACT_SET_ETH_DST: | |
74ff3298 JR |
6803 | WC_MASK_FIELD(wc, dl_dst); |
6804 | flow->dl_dst = ofpact_get_SET_ETH_DST(a)->mac; | |
9583bc14 EJ |
6805 | break; |
6806 | ||
6807 | case OFPACT_SET_IPV4_SRC: | |
33bf9176 | 6808 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 6809 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); |
33bf9176 | 6810 | flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4; |
9583bc14 EJ |
6811 | } |
6812 | break; | |
6813 | ||
6814 | case OFPACT_SET_IPV4_DST: | |
33bf9176 | 6815 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 6816 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); |
33bf9176 | 6817 | flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4; |
9583bc14 EJ |
6818 | } |
6819 | break; | |
6820 | ||
04f01c24 BP |
6821 | case OFPACT_SET_IP_DSCP: |
6822 | if (is_ip_any(flow)) { | |
f47ea021 | 6823 | wc->masks.nw_tos |= IP_DSCP_MASK; |
33bf9176 | 6824 | flow->nw_tos &= ~IP_DSCP_MASK; |
04f01c24 | 6825 | flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp; |
9583bc14 EJ |
6826 | } |
6827 | break; | |
6828 | ||
ff14eb7a JR |
6829 | case OFPACT_SET_IP_ECN: |
6830 | if (is_ip_any(flow)) { | |
6831 | wc->masks.nw_tos |= IP_ECN_MASK; | |
6832 | flow->nw_tos &= ~IP_ECN_MASK; | |
6833 | flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn; | |
6834 | } | |
6835 | break; | |
6836 | ||
0c20dbe4 JR |
6837 | case OFPACT_SET_IP_TTL: |
6838 | if (is_ip_any(flow)) { | |
6839 | wc->masks.nw_ttl = 0xff; | |
6840 | flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl; | |
6841 | } | |
6842 | break; | |
6843 | ||
9583bc14 | 6844 | case OFPACT_SET_L4_SRC_PORT: |
b8778a0d | 6845 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
6846 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
6847 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
33bf9176 | 6848 | flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port); |
9583bc14 EJ |
6849 | } |
6850 | break; | |
6851 | ||
6852 | case OFPACT_SET_L4_DST_PORT: | |
b8778a0d | 6853 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
6854 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
6855 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
33bf9176 | 6856 | flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port); |
9583bc14 EJ |
6857 | } |
6858 | break; | |
6859 | ||
6860 | case OFPACT_RESUBMIT: | |
8bf009bf JR |
6861 | /* Freezing complicates resubmit. Some action in the flow |
6862 | * entry found by resubmit might trigger freezing. If that | |
6863 | * happens, then we do not want to execute the resubmit again after | |
6864 | * during thawing, so we want to skip back to the head of the loop | |
6865 | * to avoid that, only adding any actions that follow the resubmit | |
6866 | * to the frozen actions. | |
6b1c5734 | 6867 | */ |
feee58b9 | 6868 | xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a), last); |
6b1c5734 | 6869 | continue; |
9583bc14 EJ |
6870 | |
6871 | case OFPACT_SET_TUNNEL: | |
33bf9176 | 6872 | flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id); |
9583bc14 EJ |
6873 | break; |
6874 | ||
6875 | case OFPACT_SET_QUEUE: | |
16194afd DDP |
6876 | memset(&wc->masks.skb_priority, 0xff, |
6877 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
6878 | xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id); |
6879 | break; | |
6880 | ||
6881 | case OFPACT_POP_QUEUE: | |
16194afd DDP |
6882 | memset(&wc->masks.skb_priority, 0xff, |
6883 | sizeof wc->masks.skb_priority); | |
2d9b49dd BP |
6884 | if (flow->skb_priority != ctx->orig_skb_priority) { |
6885 | flow->skb_priority = ctx->orig_skb_priority; | |
6886 | xlate_report(ctx, OFT_DETAIL, "queue = %#"PRIx32, | |
6887 | flow->skb_priority); | |
6888 | } | |
9583bc14 EJ |
6889 | break; |
6890 | ||
6891 | case OFPACT_REG_MOVE: | |
2d9b49dd | 6892 | xlate_ofpact_reg_move(ctx, ofpact_get_REG_MOVE(a)); |
9583bc14 EJ |
6893 | break; |
6894 | ||
b2dd70be JR |
6895 | case OFPACT_SET_FIELD: |
6896 | set_field = ofpact_get_SET_FIELD(a); | |
6897 | mf = set_field->field; | |
b2dd70be | 6898 | |
aff49b8c JR |
6899 | /* Set the field only if the packet actually has it. */ |
6900 | if (mf_are_prereqs_ok(mf, flow, wc)) { | |
128684a6 JR |
6901 | mf_mask_field_masked(mf, ofpact_set_field_mask(set_field), wc); |
6902 | mf_set_flow_value_masked(mf, set_field->value, | |
6903 | ofpact_set_field_mask(set_field), | |
6904 | flow); | |
2d9b49dd BP |
6905 | } else { |
6906 | xlate_report(ctx, OFT_WARN, | |
6907 | "unmet prerequisites for %s, set_field ignored", | |
6908 | mf->name); | |
6909 | ||
b8778a0d | 6910 | } |
b2dd70be JR |
6911 | break; |
6912 | ||
9583bc14 | 6913 | case OFPACT_STACK_PUSH: |
33bf9176 BP |
6914 | nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc, |
6915 | &ctx->stack); | |
9583bc14 EJ |
6916 | break; |
6917 | ||
6918 | case OFPACT_STACK_POP: | |
2d9b49dd | 6919 | xlate_ofpact_stack_pop(ctx, ofpact_get_STACK_POP(a)); |
9583bc14 EJ |
6920 | break; |
6921 | ||
6922 | case OFPACT_PUSH_MPLS: | |
8bfd0fda | 6923 | compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)); |
9583bc14 EJ |
6924 | break; |
6925 | ||
6926 | case OFPACT_POP_MPLS: | |
8bfd0fda | 6927 | compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype); |
9583bc14 EJ |
6928 | break; |
6929 | ||
097d4939 | 6930 | case OFPACT_SET_MPLS_LABEL: |
8bfd0fda BP |
6931 | compose_set_mpls_label_action( |
6932 | ctx, ofpact_get_SET_MPLS_LABEL(a)->label); | |
1d741d6d | 6933 | break; |
097d4939 JR |
6934 | |
6935 | case OFPACT_SET_MPLS_TC: | |
8bfd0fda | 6936 | compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc); |
097d4939 JR |
6937 | break; |
6938 | ||
9583bc14 | 6939 | case OFPACT_SET_MPLS_TTL: |
8bfd0fda | 6940 | compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl); |
9583bc14 EJ |
6941 | break; |
6942 | ||
6943 | case OFPACT_DEC_MPLS_TTL: | |
9cfef3d0 | 6944 | if (compose_dec_mpls_ttl_action(ctx)) { |
ad3efdcb | 6945 | return; |
9583bc14 EJ |
6946 | } |
6947 | break; | |
6948 | ||
491e05c2 YY |
6949 | case OFPACT_DEC_NSH_TTL: |
6950 | if (compose_dec_nsh_ttl_action(ctx)) { | |
6951 | return; | |
6952 | } | |
6953 | break; | |
6954 | ||
9583bc14 | 6955 | case OFPACT_DEC_TTL: |
f74e7df7 | 6956 | wc->masks.nw_ttl = 0xff; |
9583bc14 | 6957 | if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) { |
ad3efdcb | 6958 | return; |
9583bc14 EJ |
6959 | } |
6960 | break; | |
6961 | ||
6962 | case OFPACT_NOTE: | |
6963 | /* Nothing to do. */ | |
6964 | break; | |
6965 | ||
6966 | case OFPACT_MULTIPATH: | |
33bf9176 | 6967 | multipath_execute(ofpact_get_MULTIPATH(a), flow, wc); |
2d9b49dd | 6968 | xlate_report_subfield(ctx, &ofpact_get_MULTIPATH(a)->dst); |
9583bc14 EJ |
6969 | break; |
6970 | ||
6971 | case OFPACT_BUNDLE: | |
f5634764 KG |
6972 | xlate_bundle_action(ctx, ofpact_get_BUNDLE(a), last, |
6973 | group_bucket_action); | |
9583bc14 EJ |
6974 | break; |
6975 | ||
6976 | case OFPACT_OUTPUT_REG: | |
f5634764 KG |
6977 | xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a), last, |
6978 | group_bucket_action); | |
9583bc14 EJ |
6979 | break; |
6980 | ||
aaca4fe0 WT |
6981 | case OFPACT_OUTPUT_TRUNC: |
6982 | xlate_output_trunc_action(ctx, ofpact_get_OUTPUT_TRUNC(a)->port, | |
f5634764 KG |
6983 | ofpact_get_OUTPUT_TRUNC(a)->max_len, last, |
6984 | group_bucket_action); | |
aaca4fe0 WT |
6985 | break; |
6986 | ||
9583bc14 EJ |
6987 | case OFPACT_LEARN: |
6988 | xlate_learn_action(ctx, ofpact_get_LEARN(a)); | |
6989 | break; | |
6990 | ||
2d9b49dd | 6991 | case OFPACT_CONJUNCTION: |
afc3987b BP |
6992 | /* A flow with a "conjunction" action represents part of a special |
6993 | * kind of "set membership match". Such a flow should not actually | |
6994 | * get executed, but it could via, say, a "packet-out", even though | |
6995 | * that wouldn't be useful. Log it to help debugging. */ | |
2d9b49dd | 6996 | xlate_report_error(ctx, "executing no-op conjunction action"); |
18080541 BP |
6997 | break; |
6998 | ||
9583bc14 EJ |
6999 | case OFPACT_EXIT: |
7000 | ctx->exit = true; | |
7001 | break; | |
7002 | ||
2d9b49dd BP |
7003 | case OFPACT_UNROLL_XLATE: |
7004 | xlate_ofpact_unroll_xlate(ctx, ofpact_get_UNROLL_XLATE(a)); | |
e672ff9b | 7005 | break; |
2d9b49dd | 7006 | |
9583bc14 | 7007 | case OFPACT_FIN_TIMEOUT: |
33bf9176 | 7008 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
9583bc14 EJ |
7009 | xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a)); |
7010 | break; | |
7011 | ||
7012 | case OFPACT_CLEAR_ACTIONS: | |
2d9b49dd | 7013 | xlate_report_action_set(ctx, "was"); |
7fdb60a7 | 7014 | ofpbuf_clear(&ctx->action_set); |
c61f3870 BP |
7015 | ctx->xin->flow.actset_output = OFPP_UNSET; |
7016 | ctx->action_set_has_group = false; | |
7fdb60a7 SH |
7017 | break; |
7018 | ||
7019 | case OFPACT_WRITE_ACTIONS: | |
7e7e8dbb | 7020 | xlate_write_actions(ctx, ofpact_get_WRITE_ACTIONS(a)); |
2d9b49dd | 7021 | xlate_report_action_set(ctx, "is"); |
9583bc14 EJ |
7022 | break; |
7023 | ||
7024 | case OFPACT_WRITE_METADATA: | |
7025 | metadata = ofpact_get_WRITE_METADATA(a); | |
33bf9176 BP |
7026 | flow->metadata &= ~metadata->mask; |
7027 | flow->metadata |= metadata->metadata & metadata->mask; | |
9583bc14 EJ |
7028 | break; |
7029 | ||
638a19b0 | 7030 | case OFPACT_METER: |
076caa2f | 7031 | xlate_meter_action(ctx, ofpact_get_METER(a)); |
638a19b0 JR |
7032 | break; |
7033 | ||
9583bc14 | 7034 | case OFPACT_GOTO_TABLE: { |
9583bc14 | 7035 | struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a); |
9583bc14 | 7036 | |
9167fc1a JR |
7037 | ovs_assert(ctx->table_id < ogt->table_id); |
7038 | ||
4468099e | 7039 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
96c3a6e5 AZ |
7040 | ogt->table_id, true, true, false, last, |
7041 | do_xlate_actions); | |
9583bc14 EJ |
7042 | break; |
7043 | } | |
7044 | ||
7045 | case OFPACT_SAMPLE: | |
7046 | xlate_sample_action(ctx, ofpact_get_SAMPLE(a)); | |
7047 | break; | |
d4abaff5 | 7048 | |
7ae62a67 | 7049 | case OFPACT_CLONE: |
feee58b9 | 7050 | compose_clone(ctx, ofpact_get_CLONE(a), last); |
7ae62a67 WT |
7051 | break; |
7052 | ||
f839892a JS |
7053 | case OFPACT_ENCAP: |
7054 | xlate_generic_encap_action(ctx, ofpact_get_ENCAP(a)); | |
7055 | break; | |
7056 | ||
7057 | case OFPACT_DECAP: { | |
7058 | bool recirc_needed = | |
7059 | xlate_generic_decap_action(ctx, ofpact_get_DECAP(a)); | |
7060 | if (!ctx->error && recirc_needed) { | |
7061 | /* Recirculate for parsing of inner packet. */ | |
7062 | ctx_trigger_freeze(ctx); | |
7063 | /* Then continue with next action. */ | |
7064 | a = ofpact_next(a); | |
7065 | } | |
7066 | break; | |
7067 | } | |
7068 | ||
07659514 | 7069 | case OFPACT_CT: |
feee58b9 | 7070 | compose_conntrack_action(ctx, ofpact_get_CT(a), last); |
07659514 JS |
7071 | break; |
7072 | ||
72fe7578 | 7073 | case OFPACT_CT_CLEAR: |
1fe178d2 | 7074 | compose_ct_clear_action(ctx); |
72fe7578 BP |
7075 | break; |
7076 | ||
9ac0aada JR |
7077 | case OFPACT_NAT: |
7078 | /* This will be processed by compose_conntrack_action(). */ | |
7079 | ctx->ct_nat_action = ofpact_get_NAT(a); | |
7080 | break; | |
7081 | ||
d4abaff5 | 7082 | case OFPACT_DEBUG_RECIRC: |
1d361a81 | 7083 | ctx_trigger_freeze(ctx); |
d4abaff5 BP |
7084 | a = ofpact_next(a); |
7085 | break; | |
a934a3dd JP |
7086 | |
7087 | case OFPACT_DEBUG_SLOW: | |
7088 | ctx->xout->slow |= SLOW_ACTION; | |
7089 | break; | |
5b34f8fc NS |
7090 | |
7091 | case OFPACT_CHECK_PKT_LARGER: { | |
7092 | if (last) { | |
7093 | /* If this is last action, then there is no need to | |
7094 | * translate the action. */ | |
7095 | break; | |
7096 | } | |
7097 | const struct ofpact *remaining_acts = ofpact_next(a); | |
7098 | size_t remaining_acts_len = ofpact_remaining_len(remaining_acts, | |
7099 | ofpacts, | |
7100 | ofpacts_len); | |
7101 | xlate_check_pkt_larger(ctx, ofpact_get_CHECK_PKT_LARGER(a), | |
7102 | remaining_acts, remaining_acts_len); | |
7103 | break; | |
7104 | } | |
9583bc14 | 7105 | } |
1d741d6d JR |
7106 | |
7107 | /* Check if need to store this and the remaining actions for later | |
7108 | * execution. */ | |
1d361a81 BP |
7109 | if (!ctx->error && ctx->exit && ctx_first_frozen_action(ctx)) { |
7110 | freeze_unroll_actions(a, ofpact_end(ofpacts, ofpacts_len), ctx); | |
1d741d6d JR |
7111 | break; |
7112 | } | |
9583bc14 | 7113 | } |
9583bc14 EJ |
7114 | } |
7115 | ||
7116 | void | |
7117 | xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, | |
1f4a8933 JR |
7118 | ovs_version_t version, const struct flow *flow, |
7119 | ofp_port_t in_port, struct rule_dpif *rule, uint16_t tcp_flags, | |
1520ef4f BP |
7120 | const struct dp_packet *packet, struct flow_wildcards *wc, |
7121 | struct ofpbuf *odp_actions) | |
9583bc14 EJ |
7122 | { |
7123 | xin->ofproto = ofproto; | |
1f4a8933 | 7124 | xin->tables_version = version; |
9583bc14 | 7125 | xin->flow = *flow; |
8d8ab6c2 | 7126 | xin->upcall_flow = flow; |
cc377352 | 7127 | xin->flow.in_port.ofp_port = in_port; |
c61f3870 | 7128 | xin->flow.actset_output = OFPP_UNSET; |
9583bc14 | 7129 | xin->packet = packet; |
df70a773 | 7130 | xin->allow_side_effects = packet != NULL; |
9583bc14 | 7131 | xin->rule = rule; |
b256dc52 | 7132 | xin->xcache = NULL; |
9583bc14 EJ |
7133 | xin->ofpacts = NULL; |
7134 | xin->ofpacts_len = 0; | |
7135 | xin->tcp_flags = tcp_flags; | |
2d9b49dd | 7136 | xin->trace = NULL; |
9583bc14 | 7137 | xin->resubmit_stats = NULL; |
790c5d26 | 7138 | xin->depth = 0; |
cdd42eda | 7139 | xin->resubmits = 0; |
49a73e0c | 7140 | xin->wc = wc; |
1520ef4f | 7141 | xin->odp_actions = odp_actions; |
331c07ac | 7142 | xin->in_packet_out = false; |
e6bc8e74 | 7143 | xin->recirc_queue = NULL; |
00135b86 | 7144 | xin->xport_uuid = UUID_ZERO; |
e672ff9b JR |
7145 | |
7146 | /* Do recirc lookup. */ | |
1d361a81 | 7147 | xin->frozen_state = NULL; |
29b1ea3f BP |
7148 | if (flow->recirc_id) { |
7149 | const struct recirc_id_node *node | |
7150 | = recirc_id_node_find(flow->recirc_id); | |
7151 | if (node) { | |
1d361a81 | 7152 | xin->frozen_state = &node->state; |
29b1ea3f BP |
7153 | } |
7154 | } | |
9583bc14 EJ |
7155 | } |
7156 | ||
7157 | void | |
7158 | xlate_out_uninit(struct xlate_out *xout) | |
7159 | { | |
e672ff9b | 7160 | if (xout) { |
fbf5d6ec | 7161 | recirc_refs_unref(&xout->recircs); |
9583bc14 EJ |
7162 | } |
7163 | } | |
9583bc14 | 7164 | \f |
55954f6e EJ |
7165 | static struct skb_priority_to_dscp * |
7166 | get_skb_priority(const struct xport *xport, uint32_t skb_priority) | |
7167 | { | |
7168 | struct skb_priority_to_dscp *pdscp; | |
7169 | uint32_t hash; | |
7170 | ||
7171 | hash = hash_int(skb_priority, 0); | |
7172 | HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) { | |
7173 | if (pdscp->skb_priority == skb_priority) { | |
7174 | return pdscp; | |
7175 | } | |
7176 | } | |
7177 | return NULL; | |
7178 | } | |
7179 | ||
7180 | static bool | |
7181 | dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority, | |
7182 | uint8_t *dscp) | |
7183 | { | |
7184 | struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority); | |
7185 | *dscp = pdscp ? pdscp->dscp : 0; | |
7186 | return pdscp != NULL; | |
7187 | } | |
7188 | ||
16194afd DDP |
7189 | static size_t |
7190 | count_skb_priorities(const struct xport *xport) | |
7191 | { | |
7192 | return hmap_count(&xport->skb_priorities); | |
7193 | } | |
7194 | ||
55954f6e EJ |
7195 | static void |
7196 | clear_skb_priorities(struct xport *xport) | |
7197 | { | |
4ec3d7c7 | 7198 | struct skb_priority_to_dscp *pdscp; |
55954f6e | 7199 | |
4ec3d7c7 | 7200 | HMAP_FOR_EACH_POP (pdscp, hmap_node, &xport->skb_priorities) { |
55954f6e EJ |
7201 | free(pdscp); |
7202 | } | |
7203 | } | |
7204 | ||
ce4a6b76 BP |
7205 | static bool |
7206 | actions_output_to_local_port(const struct xlate_ctx *ctx) | |
7207 | { | |
46c88433 | 7208 | odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL); |
ce4a6b76 BP |
7209 | const struct nlattr *a; |
7210 | unsigned int left; | |
7211 | ||
1520ef4f BP |
7212 | NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->odp_actions->data, |
7213 | ctx->odp_actions->size) { | |
ce4a6b76 BP |
7214 | if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT |
7215 | && nl_attr_get_odp_port(a) == local_odp_port) { | |
7216 | return true; | |
7217 | } | |
7218 | } | |
7219 | return false; | |
7220 | } | |
9583bc14 | 7221 | |
5e2a6702 | 7222 | #if defined(__linux__) |
7d031d7e BP |
7223 | /* Returns the maximum number of packets that the Linux kernel is willing to |
7224 | * queue up internally to certain kinds of software-implemented ports, or the | |
7225 | * default (and rarely modified) value if it cannot be determined. */ | |
7226 | static int | |
7227 | netdev_max_backlog(void) | |
7228 | { | |
7229 | static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; | |
7230 | static int max_backlog = 1000; /* The normal default value. */ | |
7231 | ||
7232 | if (ovsthread_once_start(&once)) { | |
7233 | static const char filename[] = "/proc/sys/net/core/netdev_max_backlog"; | |
7234 | FILE *stream; | |
7235 | int n; | |
7236 | ||
7237 | stream = fopen(filename, "r"); | |
7238 | if (!stream) { | |
120c348f | 7239 | VLOG_INFO("%s: open failed (%s)", filename, ovs_strerror(errno)); |
7d031d7e BP |
7240 | } else { |
7241 | if (fscanf(stream, "%d", &n) != 1) { | |
7242 | VLOG_WARN("%s: read error", filename); | |
7243 | } else if (n <= 100) { | |
7244 | VLOG_WARN("%s: unexpectedly small value %d", filename, n); | |
7245 | } else { | |
7246 | max_backlog = n; | |
7247 | } | |
7248 | fclose(stream); | |
7249 | } | |
7250 | ovsthread_once_done(&once); | |
7251 | ||
7252 | VLOG_DBG("%s: using %d max_backlog", filename, max_backlog); | |
7253 | } | |
7254 | ||
7255 | return max_backlog; | |
7256 | } | |
7257 | ||
7258 | /* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in | |
7259 | * 'odp_actions'. */ | |
7260 | static int | |
7261 | count_output_actions(const struct ofpbuf *odp_actions) | |
7262 | { | |
7263 | const struct nlattr *a; | |
7264 | size_t left; | |
7265 | int n = 0; | |
7266 | ||
6fd6ed71 | 7267 | NL_ATTR_FOR_EACH_UNSAFE (a, left, odp_actions->data, odp_actions->size) { |
7d031d7e BP |
7268 | if (a->nla_type == OVS_ACTION_ATTR_OUTPUT) { |
7269 | n++; | |
7270 | } | |
7271 | } | |
7272 | return n; | |
7273 | } | |
5e2a6702 | 7274 | #endif /* defined(__linux__) */ |
7d031d7e BP |
7275 | |
7276 | /* Returns true if 'odp_actions' contains more output actions than the datapath | |
7277 | * can reliably handle in one go. On Linux, this is the value of the | |
7278 | * net.core.netdev_max_backlog sysctl, which limits the maximum number of | |
7279 | * packets that the kernel is willing to queue up for processing while the | |
7280 | * datapath is processing a set of actions. */ | |
7281 | static bool | |
5e2a6702 | 7282 | too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) |
7d031d7e BP |
7283 | { |
7284 | #ifdef __linux__ | |
6fd6ed71 | 7285 | return (odp_actions->size / NL_A_U32_SIZE > netdev_max_backlog() |
7d031d7e BP |
7286 | && count_output_actions(odp_actions) > netdev_max_backlog()); |
7287 | #else | |
7288 | /* OSes other than Linux might have similar limits, but we don't know how | |
7289 | * to determine them.*/ | |
7290 | return false; | |
7291 | #endif | |
7292 | } | |
7293 | ||
234c3da9 BP |
7294 | static void |
7295 | xlate_wc_init(struct xlate_ctx *ctx) | |
7296 | { | |
7297 | flow_wildcards_init_catchall(ctx->wc); | |
7298 | ||
7299 | /* Some fields we consider to always be examined. */ | |
3d4b2e6e | 7300 | WC_MASK_FIELD(ctx->wc, packet_type); |
5e2e998a | 7301 | WC_MASK_FIELD(ctx->wc, in_port); |
29b5c0c3 | 7302 | WC_MASK_FIELD(ctx->wc, dl_type); |
234c3da9 | 7303 | if (is_ip_any(&ctx->xin->flow)) { |
5e2e998a | 7304 | WC_MASK_FIELD_MASK(ctx->wc, nw_frag, FLOW_NW_FRAG_MASK); |
234c3da9 BP |
7305 | } |
7306 | ||
7307 | if (ctx->xbridge->support.odp.recirc) { | |
7308 | /* Always exactly match recirc_id when datapath supports | |
7309 | * recirculation. */ | |
5e2e998a | 7310 | WC_MASK_FIELD(ctx->wc, recirc_id); |
234c3da9 BP |
7311 | } |
7312 | ||
7313 | if (ctx->xbridge->netflow) { | |
7314 | netflow_mask_wc(&ctx->xin->flow, ctx->wc); | |
7315 | } | |
7316 | ||
7317 | tnl_wc_init(&ctx->xin->flow, ctx->wc); | |
7318 | } | |
7319 | ||
7320 | static void | |
7321 | xlate_wc_finish(struct xlate_ctx *ctx) | |
7322 | { | |
f0fb825a EG |
7323 | int i; |
7324 | ||
234c3da9 BP |
7325 | /* Clear the metadata and register wildcard masks, because we won't |
7326 | * use non-header fields as part of the cache. */ | |
7327 | flow_wildcards_clear_non_packet_fields(ctx->wc); | |
7328 | ||
29b5c0c3 BP |
7329 | /* Wildcard Ethernet address fields if the original packet type was not |
7330 | * Ethernet. | |
7331 | * | |
7332 | * (The Ethertype field is used even when the original packet type is not | |
7333 | * Ethernet.) */ | |
beb75a40 JS |
7334 | if (ctx->xin->upcall_flow->packet_type != htonl(PT_ETH)) { |
7335 | ctx->wc->masks.dl_dst = eth_addr_zero; | |
7336 | ctx->wc->masks.dl_src = eth_addr_zero; | |
7337 | } | |
7338 | ||
234c3da9 BP |
7339 | /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow |
7340 | * uses the low 8 bits of the 16-bit tp_src and tp_dst members to | |
7341 | * represent these fields. The datapath interface, on the other hand, | |
7342 | * represents them with just 8 bits each. This means that if the high | |
7343 | * 8 bits of the masks for these fields somehow become set, then they | |
7344 | * will get chopped off by a round trip through the datapath, and | |
7345 | * revalidation will spot that as an inconsistency and delete the flow. | |
7346 | * Avoid the problem here by making sure that only the low 8 bits of | |
7347 | * either field can be unwildcarded for ICMP. | |
7348 | */ | |
a75636c8 | 7349 | if (is_icmpv4(&ctx->xin->flow, NULL) || is_icmpv6(&ctx->xin->flow, NULL)) { |
234c3da9 BP |
7350 | ctx->wc->masks.tp_src &= htons(UINT8_MAX); |
7351 | ctx->wc->masks.tp_dst &= htons(UINT8_MAX); | |
7352 | } | |
7353 | /* VLAN_TCI CFI bit must be matched if any of the TCI is matched. */ | |
f0fb825a EG |
7354 | for (i = 0; i < FLOW_MAX_VLAN_HEADERS; i++) { |
7355 | if (ctx->wc->masks.vlans[i].tci) { | |
7356 | ctx->wc->masks.vlans[i].tci |= htons(VLAN_CFI); | |
7357 | } | |
234c3da9 | 7358 | } |
4a7ab326 DDP |
7359 | |
7360 | /* The classifier might return masks that match on tp_src and tp_dst even | |
7361 | * for later fragments. This happens because there might be flows that | |
7362 | * match on tp_src or tp_dst without matching on the frag bits, because | |
7363 | * it is not a prerequisite for OpenFlow. Since it is a prerequisite for | |
7364 | * datapath flows and since tp_src and tp_dst are always going to be 0, | |
7365 | * wildcard the fields here. */ | |
7366 | if (ctx->xin->flow.nw_frag & FLOW_NW_FRAG_LATER) { | |
7367 | ctx->wc->masks.tp_src = 0; | |
7368 | ctx->wc->masks.tp_dst = 0; | |
7369 | } | |
736a8111 VDA |
7370 | |
7371 | /* Clear flow wildcard bits for fields which are not present | |
7372 | * in the original packet header. These wildcards may get set | |
7373 | * due to push/set_field actions. This results into frequent | |
7374 | * invalidation of datapath flows by revalidator thread. */ | |
7375 | ||
7376 | /* Clear mpls label wc bits if original packet is non-mpls. */ | |
7377 | if (!eth_type_mpls(ctx->xin->upcall_flow->dl_type)) { | |
7378 | for (i = 0; i < FLOW_MAX_MPLS_LABELS; i++) { | |
7379 | ctx->wc->masks.mpls_lse[i] = 0; | |
7380 | } | |
7381 | } | |
7382 | /* Clear vlan header wc bits if original packet does not have | |
7383 | * vlan header. */ | |
7384 | for (i = 0; i < FLOW_MAX_VLAN_HEADERS; i++) { | |
7385 | if (!eth_type_vlan(ctx->xin->upcall_flow->vlans[i].tpid)) { | |
7386 | ctx->wc->masks.vlans[i].tpid = 0; | |
7387 | ctx->wc->masks.vlans[i].tci = 0; | |
7388 | } | |
7389 | } | |
234c3da9 BP |
7390 | } |
7391 | ||
e672ff9b JR |
7392 | /* Translates the flow, actions, or rule in 'xin' into datapath actions in |
7393 | * 'xout'. | |
56450a41 | 7394 | * The caller must take responsibility for eventually freeing 'xout', with |
fff1b9c0 JR |
7395 | * xlate_out_uninit(). |
7396 | * Returns 'XLATE_OK' if translation was successful. In case of an error an | |
7397 | * empty set of actions will be returned in 'xin->odp_actions' (if non-NULL), | |
7398 | * so that most callers may ignore the return value and transparently install a | |
7399 | * drop flow when the translation fails. */ | |
7400 | enum xlate_error | |
84f0f298 | 7401 | xlate_actions(struct xlate_in *xin, struct xlate_out *xout) |
9583bc14 | 7402 | { |
e467ea42 BP |
7403 | *xout = (struct xlate_out) { |
7404 | .slow = 0, | |
fbf5d6ec | 7405 | .recircs = RECIRC_REFS_EMPTY_INITIALIZER, |
e467ea42 BP |
7406 | }; |
7407 | ||
84f0f298 | 7408 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
bb00fdef BP |
7409 | struct xbridge *xbridge = xbridge_lookup(xcfg, xin->ofproto); |
7410 | if (!xbridge) { | |
fff1b9c0 | 7411 | return XLATE_BRIDGE_NOT_FOUND; |
bb00fdef BP |
7412 | } |
7413 | ||
33bf9176 BP |
7414 | struct flow *flow = &xin->flow; |
7415 | ||
84cf3c1f | 7416 | uint8_t stack_stub[1024]; |
bb00fdef | 7417 | uint64_t action_set_stub[1024 / 8]; |
1d361a81 | 7418 | uint64_t frozen_actions_stub[1024 / 8]; |
1520ef4f BP |
7419 | uint64_t actions_stub[256 / 8]; |
7420 | struct ofpbuf scratch_actions = OFPBUF_STUB_INITIALIZER(actions_stub); | |
bb00fdef BP |
7421 | struct xlate_ctx ctx = { |
7422 | .xin = xin, | |
7423 | .xout = xout, | |
7424 | .base_flow = *flow, | |
c2b878e0 | 7425 | .orig_tunnel_ipv6_dst = flow_tnl_dst(&flow->tunnel), |
0506f184 | 7426 | .xcfg = xcfg, |
bb00fdef BP |
7427 | .xbridge = xbridge, |
7428 | .stack = OFPBUF_STUB_INITIALIZER(stack_stub), | |
7429 | .rule = xin->rule, | |
c0e638aa BP |
7430 | .wc = (xin->wc |
7431 | ? xin->wc | |
f36efd90 | 7432 | : &(struct flow_wildcards) { .masks = { .dl_type = 0 } }), |
1520ef4f | 7433 | .odp_actions = xin->odp_actions ? xin->odp_actions : &scratch_actions, |
bb00fdef | 7434 | |
790c5d26 | 7435 | .depth = xin->depth, |
cdd42eda | 7436 | .resubmits = xin->resubmits, |
bb00fdef | 7437 | .in_action_set = false, |
331c07ac | 7438 | .in_packet_out = xin->in_packet_out, |
f839892a | 7439 | .pending_encap = false, |
88ec1e0a | 7440 | .pending_decap = false, |
1fc11c59 | 7441 | .encap_data = NULL, |
bb00fdef BP |
7442 | |
7443 | .table_id = 0, | |
7444 | .rule_cookie = OVS_BE64_MAX, | |
7445 | .orig_skb_priority = flow->skb_priority, | |
7446 | .sflow_n_outputs = 0, | |
7447 | .sflow_odp_port = 0, | |
2031ef97 | 7448 | .nf_output_iface = NF_OUT_DROP, |
bb00fdef | 7449 | .exit = false, |
fff1b9c0 | 7450 | .error = XLATE_OK, |
3d6151f3 | 7451 | .mirrors = 0, |
bb00fdef | 7452 | |
1d361a81 | 7453 | .freezing = false, |
53cc166a | 7454 | .recirc_update_dp_hash = false, |
1d361a81 | 7455 | .frozen_actions = OFPBUF_STUB_INITIALIZER(frozen_actions_stub), |
77ab5fd2 | 7456 | .pause = NULL, |
bb00fdef | 7457 | |
e12ec36b | 7458 | .was_mpls = false, |
07659514 | 7459 | .conntracked = false, |
bb00fdef | 7460 | |
9ac0aada JR |
7461 | .ct_nat_action = NULL, |
7462 | ||
bb00fdef BP |
7463 | .action_set_has_group = false, |
7464 | .action_set = OFPBUF_STUB_INITIALIZER(action_set_stub), | |
7465 | }; | |
865ca6cf BP |
7466 | |
7467 | /* 'base_flow' reflects the packet as it came in, but we need it to reflect | |
42deb67d PS |
7468 | * the packet as the datapath will treat it for output actions. Our |
7469 | * datapath doesn't retain tunneling information without us re-setting | |
7470 | * it, so clear the tunnel data. | |
865ca6cf | 7471 | */ |
42deb67d | 7472 | |
bb00fdef | 7473 | memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); |
865ca6cf | 7474 | |
1520ef4f | 7475 | ofpbuf_reserve(ctx.odp_actions, NL_A_U32_SIZE); |
c0e638aa | 7476 | xlate_wc_init(&ctx); |
bb00fdef | 7477 | |
46c88433 | 7478 | COVERAGE_INC(xlate_actions); |
9583bc14 | 7479 | |
2d9b49dd BP |
7480 | xin->trace = xlate_report(&ctx, OFT_BRIDGE, "bridge(\"%s\")", |
7481 | xbridge->name); | |
1d361a81 BP |
7482 | if (xin->frozen_state) { |
7483 | const struct frozen_state *state = xin->frozen_state; | |
e672ff9b | 7484 | |
2d9b49dd BP |
7485 | struct ovs_list *old_trace = xin->trace; |
7486 | xin->trace = xlate_report(&ctx, OFT_THAW, "thaw"); | |
d6bef3cc | 7487 | |
e672ff9b | 7488 | if (xin->ofpacts_len > 0 || ctx.rule) { |
2d9b49dd BP |
7489 | xlate_report_error(&ctx, "Recirculation conflict (%s)!", |
7490 | xin->ofpacts_len ? "actions" : "rule"); | |
fff1b9c0 | 7491 | ctx.error = XLATE_RECIRCULATION_CONFLICT; |
1520ef4f | 7492 | goto exit; |
e672ff9b JR |
7493 | } |
7494 | ||
7495 | /* Set the bridge for post-recirculation processing if needed. */ | |
07a3cd5c | 7496 | if (!uuid_equals(&ctx.xbridge->ofproto->uuid, &state->ofproto_uuid)) { |
2082425c | 7497 | const struct xbridge *new_bridge |
290835f9 | 7498 | = xbridge_lookup_by_uuid(xcfg, &state->ofproto_uuid); |
e672ff9b JR |
7499 | |
7500 | if (OVS_UNLIKELY(!new_bridge)) { | |
7501 | /* Drop the packet if the bridge cannot be found. */ | |
2d9b49dd | 7502 | xlate_report_error(&ctx, "Frozen bridge no longer exists."); |
fff1b9c0 | 7503 | ctx.error = XLATE_BRIDGE_NOT_FOUND; |
2d9b49dd | 7504 | xin->trace = old_trace; |
1520ef4f | 7505 | goto exit; |
e672ff9b JR |
7506 | } |
7507 | ctx.xbridge = new_bridge; | |
1f4a8933 JR |
7508 | /* The bridge is now known so obtain its table version. */ |
7509 | ctx.xin->tables_version | |
7510 | = ofproto_dpif_get_tables_version(ctx.xbridge->ofproto); | |
e672ff9b JR |
7511 | } |
7512 | ||
1d361a81 BP |
7513 | /* Set the thawed table id. Note: A table lookup is done only if there |
7514 | * are no frozen actions. */ | |
2082425c | 7515 | ctx.table_id = state->table_id; |
2d9b49dd BP |
7516 | xlate_report(&ctx, OFT_THAW, |
7517 | "Resuming from table %"PRIu8, ctx.table_id); | |
e672ff9b | 7518 | |
40b0fbd3 | 7519 | ctx.conntracked = state->conntracked; |
07659514 | 7520 | if (!state->conntracked) { |
72fe7578 | 7521 | clear_conntrack(&ctx); |
07659514 JS |
7522 | } |
7523 | ||
e672ff9b | 7524 | /* Restore pipeline metadata. May change flow's in_port and other |
1d361a81 BP |
7525 | * metadata to the values that existed when freezing was triggered. */ |
7526 | frozen_metadata_to_flow(&state->metadata, flow); | |
e672ff9b JR |
7527 | |
7528 | /* Restore stack, if any. */ | |
2082425c | 7529 | if (state->stack) { |
84cf3c1f | 7530 | ofpbuf_put(&ctx.stack, state->stack, state->stack_size); |
e672ff9b JR |
7531 | } |
7532 | ||
29bae541 BP |
7533 | /* Restore mirror state. */ |
7534 | ctx.mirrors = state->mirrors; | |
7535 | ||
e672ff9b | 7536 | /* Restore action set, if any. */ |
2082425c | 7537 | if (state->action_set_len) { |
2d9b49dd | 7538 | xlate_report_actions(&ctx, OFT_THAW, "Restoring action set", |
417509fa | 7539 | state->action_set, state->action_set_len); |
d6bef3cc | 7540 | |
7e7e8dbb BP |
7541 | flow->actset_output = OFPP_UNSET; |
7542 | xlate_write_actions__(&ctx, state->action_set, | |
7543 | state->action_set_len); | |
e672ff9b JR |
7544 | } |
7545 | ||
1d361a81 BP |
7546 | /* Restore frozen actions. If there are no actions, processing will |
7547 | * start with a lookup in the table set above. */ | |
417509fa BP |
7548 | xin->ofpacts = state->ofpacts; |
7549 | xin->ofpacts_len = state->ofpacts_len; | |
7550 | if (state->ofpacts_len) { | |
2d9b49dd | 7551 | xlate_report_actions(&ctx, OFT_THAW, "Restoring actions", |
d6bef3cc | 7552 | xin->ofpacts, xin->ofpacts_len); |
e672ff9b | 7553 | } |
e672ff9b | 7554 | |
2d9b49dd BP |
7555 | xin->trace = old_trace; |
7556 | } else if (OVS_UNLIKELY(flow->recirc_id)) { | |
7557 | xlate_report_error(&ctx, | |
7558 | "Recirculation context not found for ID %"PRIx32, | |
7559 | flow->recirc_id); | |
fff1b9c0 | 7560 | ctx.error = XLATE_NO_RECIRCULATION_CONTEXT; |
1520ef4f | 7561 | goto exit; |
e672ff9b | 7562 | } |
9583bc14 | 7563 | |
8d8ab6c2 JG |
7564 | /* Tunnel metadata in udpif format must be normalized before translation. */ |
7565 | if (flow->tunnel.flags & FLOW_TNL_F_UDPIF) { | |
5b09d9f7 MS |
7566 | const struct tun_table *tun_tab = ofproto_get_tun_tab( |
7567 | &ctx.xbridge->ofproto->up); | |
8d8ab6c2 JG |
7568 | int err; |
7569 | ||
7570 | err = tun_metadata_from_geneve_udpif(tun_tab, &xin->upcall_flow->tunnel, | |
7571 | &xin->upcall_flow->tunnel, | |
7572 | &flow->tunnel); | |
7573 | if (err) { | |
2d9b49dd | 7574 | xlate_report_error(&ctx, "Invalid Geneve tunnel metadata"); |
8d8ab6c2 JG |
7575 | ctx.error = XLATE_INVALID_TUNNEL_METADATA; |
7576 | goto exit; | |
7577 | } | |
254878c1 | 7578 | } else if (!flow->tunnel.metadata.tab || xin->frozen_state) { |
8d8ab6c2 JG |
7579 | /* If the original flow did not come in on a tunnel, then it won't have |
7580 | * FLOW_TNL_F_UDPIF set. However, we still need to have a metadata | |
7581 | * table in case we generate tunnel actions. */ | |
254878c1 YHW |
7582 | /* If the translation is from a frozen state, we use the latest |
7583 | * TLV map to avoid segmentation fault in case the old TLV map is | |
7584 | * replaced by a new one. | |
7585 | * XXX: It is better to abort translation if the table is changed. */ | |
5b09d9f7 MS |
7586 | flow->tunnel.metadata.tab = ofproto_get_tun_tab( |
7587 | &ctx.xbridge->ofproto->up); | |
8d8ab6c2 JG |
7588 | } |
7589 | ctx.wc->masks.tunnel.metadata.tab = flow->tunnel.metadata.tab; | |
7590 | ||
beb75a40 JS |
7591 | /* Get the proximate input port of the packet. (If xin->frozen_state, |
7592 | * flow->in_port is the ultimate input port of the packet.) */ | |
7593 | struct xport *in_port = get_ofp_port(xbridge, | |
7594 | ctx.base_flow.in_port.ofp_port); | |
00135b86 ZB |
7595 | if (in_port && !in_port->peer) { |
7596 | ctx.xin->xport_uuid = in_port->uuid; | |
7597 | } | |
beb75a40 | 7598 | |
875ab130 BP |
7599 | if (flow->packet_type != htonl(PT_ETH) && in_port && |
7600 | in_port->pt_mode == NETDEV_PT_LEGACY_L3 && ctx.table_id == 0) { | |
beb75a40 JS |
7601 | /* Add dummy Ethernet header to non-L2 packet if it's coming from a |
7602 | * L3 port. So all packets will be L2 packets for lookup. | |
7603 | * The dl_type has already been set from the packet_type. */ | |
7604 | flow->packet_type = htonl(PT_ETH); | |
7605 | flow->dl_src = eth_addr_zero; | |
7606 | flow->dl_dst = eth_addr_zero; | |
f839892a | 7607 | ctx.pending_encap = true; |
beb75a40 JS |
7608 | } |
7609 | ||
10c44245 | 7610 | if (!xin->ofpacts && !ctx.rule) { |
b2e89cc9 | 7611 | ctx.rule = rule_dpif_lookup_from_table( |
1f4a8933 | 7612 | ctx.xbridge->ofproto, ctx.xin->tables_version, flow, ctx.wc, |
1e1e1d19 | 7613 | ctx.xin->resubmit_stats, &ctx.table_id, |
a027899e | 7614 | flow->in_port.ofp_port, true, true, ctx.xin->xcache); |
10c44245 | 7615 | if (ctx.xin->resubmit_stats) { |
16441315 | 7616 | rule_dpif_credit_stats(ctx.rule, ctx.xin->resubmit_stats, false); |
10c44245 | 7617 | } |
b256dc52 JS |
7618 | if (ctx.xin->xcache) { |
7619 | struct xc_entry *entry; | |
7620 | ||
7621 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE); | |
901a517e | 7622 | entry->rule = ctx.rule; |
07a3cd5c | 7623 | ofproto_rule_ref(&ctx.rule->up); |
b256dc52 | 7624 | } |
a8c31348 | 7625 | |
2d9b49dd | 7626 | xlate_report_table(&ctx, ctx.rule, ctx.table_id); |
10c44245 | 7627 | } |
10c44245 | 7628 | |
1d361a81 BP |
7629 | /* Tunnel stats only for not-thawed packets. */ |
7630 | if (!xin->frozen_state && in_port && in_port->is_tunnel) { | |
b256dc52 JS |
7631 | if (ctx.xin->resubmit_stats) { |
7632 | netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats); | |
7633 | if (in_port->bfd) { | |
7634 | bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats); | |
7635 | } | |
7636 | } | |
7637 | if (ctx.xin->xcache) { | |
7638 | struct xc_entry *entry; | |
7639 | ||
7640 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV); | |
901a517e JR |
7641 | entry->dev.rx = netdev_ref(in_port->netdev); |
7642 | entry->dev.bfd = bfd_ref(in_port->bfd); | |
d6fc5f57 EJ |
7643 | } |
7644 | } | |
7645 | ||
1d361a81 | 7646 | if (!xin->frozen_state && process_special(&ctx, in_port)) { |
bef1403e BP |
7647 | /* process_special() did all the processing for this packet. |
7648 | * | |
1d361a81 BP |
7649 | * We do not perform special processing on thawed packets, since that |
7650 | * was done before they were frozen and should not be redone. */ | |
27a9c0e3 | 7651 | mirror_ingress_packet(&ctx); |
bef1403e BP |
7652 | } else if (in_port && in_port->xbundle |
7653 | && xbundle_mirror_out(xbridge, in_port->xbundle)) { | |
2d9b49dd BP |
7654 | xlate_report_error(&ctx, "dropping packet received on port " |
7655 | "%s, which is reserved exclusively for mirroring", | |
7656 | in_port->xbundle->name); | |
bef1403e | 7657 | } else { |
1d361a81 | 7658 | /* Sampling is done on initial reception; don't redo after thawing. */ |
a6092018 | 7659 | unsigned int user_cookie_offset = 0; |
1d361a81 | 7660 | if (!xin->frozen_state) { |
a6092018 BP |
7661 | user_cookie_offset = compose_sflow_action(&ctx); |
7662 | compose_ipfix_action(&ctx, ODPP_NONE); | |
e672ff9b | 7663 | } |
0731abc5 | 7664 | size_t sample_actions_len = ctx.odp_actions->size; |
a13a0209 | 7665 | bool ecn_drop = !tnl_process_ecn(flow); |
9583bc14 | 7666 | |
a13a0209 | 7667 | if (!ecn_drop |
234c3da9 | 7668 | && (!in_port || may_receive(in_port, &ctx))) { |
1806291d BP |
7669 | const struct ofpact *ofpacts; |
7670 | size_t ofpacts_len; | |
7671 | ||
7672 | if (xin->ofpacts) { | |
7673 | ofpacts = xin->ofpacts; | |
7674 | ofpacts_len = xin->ofpacts_len; | |
7675 | } else if (ctx.rule) { | |
7676 | const struct rule_actions *actions | |
07a3cd5c | 7677 | = rule_get_actions(&ctx.rule->up); |
1806291d BP |
7678 | ofpacts = actions->ofpacts; |
7679 | ofpacts_len = actions->ofpacts_len; | |
07a3cd5c | 7680 | ctx.rule_cookie = ctx.rule->up.flow_cookie; |
1806291d BP |
7681 | } else { |
7682 | OVS_NOT_REACHED(); | |
7683 | } | |
7684 | ||
7efbc3b7 | 7685 | mirror_ingress_packet(&ctx); |
f5634764 | 7686 | do_xlate_actions(ofpacts, ofpacts_len, &ctx, true, false); |
fff1b9c0 JR |
7687 | if (ctx.error) { |
7688 | goto exit; | |
7689 | } | |
9583bc14 EJ |
7690 | |
7691 | /* We've let OFPP_NORMAL and the learning action look at the | |
1d361a81 | 7692 | * packet, so cancel all actions and freezing if forwarding is |
8a5fb3b4 | 7693 | * disabled. */ |
9efd308e DV |
7694 | if (in_port && (!xport_stp_forward_state(in_port) || |
7695 | !xport_rstp_forward_state(in_port))) { | |
1520ef4f | 7696 | ctx.odp_actions->size = sample_actions_len; |
1d361a81 | 7697 | ctx_cancel_freeze(&ctx); |
8a5fb3b4 | 7698 | ofpbuf_clear(&ctx.action_set); |
a13a0209 | 7699 | ctx.error = XLATE_FORWARDING_DISABLED; |
8a5fb3b4 BP |
7700 | } |
7701 | ||
1d361a81 | 7702 | if (!ctx.freezing) { |
8a5fb3b4 | 7703 | xlate_action_set(&ctx); |
e672ff9b | 7704 | } |
1d361a81 | 7705 | if (ctx.freezing) { |
77ab5fd2 | 7706 | finish_freezing(&ctx); |
9583bc14 | 7707 | } |
a13a0209 AT |
7708 | } else if (ecn_drop) { |
7709 | ctx.error = XLATE_CONGESTION_DROP; | |
9583bc14 EJ |
7710 | } |
7711 | ||
e672ff9b | 7712 | /* Output only fully processed packets. */ |
1d361a81 | 7713 | if (!ctx.freezing |
e672ff9b | 7714 | && xbridge->has_in_band |
ce4a6b76 BP |
7715 | && in_band_must_output_to_local_port(flow) |
7716 | && !actions_output_to_local_port(&ctx)) { | |
c9dc050d AT |
7717 | WC_MASK_FIELD(ctx.wc, nw_proto); |
7718 | WC_MASK_FIELD(ctx.wc, tp_src); | |
7719 | WC_MASK_FIELD(ctx.wc, tp_dst); | |
7720 | WC_MASK_FIELD(ctx.wc, dl_type); | |
b1986b08 BP |
7721 | xlate_report(&ctx, OFT_DETAIL, "outputting DHCP packet " |
7722 | "to local port for in-band control"); | |
11938578 | 7723 | compose_output_action(&ctx, OFPP_LOCAL, NULL, false, false); |
9583bc14 | 7724 | } |
aaa0fbae | 7725 | |
a6092018 BP |
7726 | if (user_cookie_offset) { |
7727 | fix_sflow_action(&ctx, user_cookie_offset); | |
e672ff9b | 7728 | } |
9583bc14 EJ |
7729 | } |
7730 | ||
1520ef4f | 7731 | if (nl_attr_oversized(ctx.odp_actions->size)) { |
542024c4 | 7732 | /* These datapath actions are too big for a Netlink attribute, so we |
0f032e95 BP |
7733 | * can't hand them to the kernel directly. dpif_execute() can execute |
7734 | * them one by one with help, so just mark the result as SLOW_ACTION to | |
7735 | * prevent the flow from being installed. */ | |
7736 | COVERAGE_INC(xlate_actions_oversize); | |
7737 | ctx.xout->slow |= SLOW_ACTION; | |
1520ef4f | 7738 | } else if (too_many_output_actions(ctx.odp_actions)) { |
7d031d7e BP |
7739 | COVERAGE_INC(xlate_actions_too_many_output); |
7740 | ctx.xout->slow |= SLOW_ACTION; | |
542024c4 BP |
7741 | } |
7742 | ||
64fb5f82 JP |
7743 | /* Update NetFlow for non-frozen traffic. */ |
7744 | if (xbridge->netflow && !xin->frozen_state) { | |
1806291d BP |
7745 | if (ctx.xin->resubmit_stats) { |
7746 | netflow_flow_update(xbridge->netflow, flow, | |
2031ef97 | 7747 | ctx.nf_output_iface, |
1806291d BP |
7748 | ctx.xin->resubmit_stats); |
7749 | } | |
7750 | if (ctx.xin->xcache) { | |
7751 | struct xc_entry *entry; | |
b256dc52 | 7752 | |
1806291d | 7753 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW); |
901a517e JR |
7754 | entry->nf.netflow = netflow_ref(xbridge->netflow); |
7755 | entry->nf.flow = xmemdup(flow, sizeof *flow); | |
7756 | entry->nf.iface = ctx.nf_output_iface; | |
d6fc5f57 EJ |
7757 | } |
7758 | } | |
7759 | ||
8d8ab6c2 JG |
7760 | /* Translate tunnel metadata masks to udpif format if necessary. */ |
7761 | if (xin->upcall_flow->tunnel.flags & FLOW_TNL_F_UDPIF) { | |
7762 | if (ctx.wc->masks.tunnel.metadata.present.map) { | |
7763 | const struct flow_tnl *upcall_tnl = &xin->upcall_flow->tunnel; | |
7764 | struct geneve_opt opts[TLV_TOT_OPT_SIZE / | |
7765 | sizeof(struct geneve_opt)]; | |
7766 | ||
7767 | tun_metadata_to_geneve_udpif_mask(&flow->tunnel, | |
7768 | &ctx.wc->masks.tunnel, | |
7769 | upcall_tnl->metadata.opts.gnv, | |
7770 | upcall_tnl->metadata.present.len, | |
7771 | opts); | |
7772 | memset(&ctx.wc->masks.tunnel.metadata, 0, | |
7773 | sizeof ctx.wc->masks.tunnel.metadata); | |
7774 | memcpy(&ctx.wc->masks.tunnel.metadata.opts.gnv, opts, | |
7775 | upcall_tnl->metadata.present.len); | |
7776 | } | |
7777 | ctx.wc->masks.tunnel.metadata.present.len = 0xff; | |
7778 | ctx.wc->masks.tunnel.metadata.tab = NULL; | |
7779 | ctx.wc->masks.tunnel.flags |= FLOW_TNL_F_UDPIF; | |
7780 | } else if (!xin->upcall_flow->tunnel.metadata.tab) { | |
7781 | /* If we didn't have options in UDPIF format and didn't have an existing | |
7782 | * metadata table, then it means that there were no options at all when | |
7783 | * we started processing and any wildcards we picked up were from | |
7784 | * action generation. Without options on the incoming packet, wildcards | |
7785 | * aren't meaningful. To avoid them possibly getting misinterpreted, | |
7786 | * just clear everything. */ | |
7787 | if (ctx.wc->masks.tunnel.metadata.present.map) { | |
7788 | memset(&ctx.wc->masks.tunnel.metadata, 0, | |
7789 | sizeof ctx.wc->masks.tunnel.metadata); | |
7790 | } else { | |
7791 | ctx.wc->masks.tunnel.metadata.tab = NULL; | |
7792 | } | |
7793 | } | |
7794 | ||
c0e638aa | 7795 | xlate_wc_finish(&ctx); |
1520ef4f BP |
7796 | |
7797 | exit: | |
8d8ab6c2 JG |
7798 | /* Reset the table to what it was when we came in. If we only fetched |
7799 | * it locally, then it has no meaning outside of flow translation. */ | |
7800 | flow->tunnel.metadata.tab = xin->upcall_flow->tunnel.metadata.tab; | |
7801 | ||
1520ef4f BP |
7802 | ofpbuf_uninit(&ctx.stack); |
7803 | ofpbuf_uninit(&ctx.action_set); | |
1d361a81 | 7804 | ofpbuf_uninit(&ctx.frozen_actions); |
1520ef4f | 7805 | ofpbuf_uninit(&scratch_actions); |
1fc11c59 | 7806 | ofpbuf_delete(ctx.encap_data); |
fff1b9c0 JR |
7807 | |
7808 | /* Make sure we return a "drop flow" in case of an error. */ | |
7809 | if (ctx.error) { | |
7810 | xout->slow = 0; | |
7811 | if (xin->odp_actions) { | |
7812 | ofpbuf_clear(xin->odp_actions); | |
7813 | } | |
7814 | } | |
a13a0209 AT |
7815 | |
7816 | /* Install drop action if datapath supports explicit drop action. */ | |
7817 | if (xin->odp_actions && !xin->odp_actions->size && | |
7818 | ovs_explicit_drop_action_supported(ctx.xbridge->ofproto)) { | |
7819 | put_drop_action(xin->odp_actions, ctx.error); | |
7820 | } | |
7821 | ||
7822 | /* Since congestion drop and forwarding drop are not exactly | |
7823 | * translation error, we are resetting the translation error. | |
7824 | */ | |
7825 | if (ctx.error == XLATE_CONGESTION_DROP || | |
7826 | ctx.error == XLATE_FORWARDING_DISABLED) { | |
7827 | ctx.error = XLATE_OK; | |
7828 | } | |
7829 | ||
fff1b9c0 | 7830 | return ctx.error; |
91d6cd12 AW |
7831 | } |
7832 | ||
77ab5fd2 BP |
7833 | enum ofperr |
7834 | xlate_resume(struct ofproto_dpif *ofproto, | |
7835 | const struct ofputil_packet_in_private *pin, | |
7836 | struct ofpbuf *odp_actions, | |
2f355bff YHW |
7837 | enum slow_path_reason *slow, |
7838 | struct flow *flow, | |
7839 | struct xlate_cache *xcache) | |
77ab5fd2 BP |
7840 | { |
7841 | struct dp_packet packet; | |
4d617a87 BP |
7842 | dp_packet_use_const(&packet, pin->base.packet, |
7843 | pin->base.packet_len); | |
77ab5fd2 | 7844 | |
07706c50 | 7845 | pkt_metadata_from_flow(&packet.md, &pin->base.flow_metadata.flow); |
2f355bff | 7846 | flow_extract(&packet, flow); |
77ab5fd2 BP |
7847 | |
7848 | struct xlate_in xin; | |
1f4a8933 | 7849 | xlate_in_init(&xin, ofproto, ofproto_dpif_get_tables_version(ofproto), |
2f355bff | 7850 | flow, 0, NULL, ntohs(flow->tcp_flags), |
77ab5fd2 | 7851 | &packet, NULL, odp_actions); |
2f355bff | 7852 | xin.xcache = xcache; |
77ab5fd2 BP |
7853 | |
7854 | struct ofpact_note noop; | |
7855 | ofpact_init_NOTE(&noop); | |
7856 | noop.length = 0; | |
7857 | ||
7858 | bool any_actions = pin->actions_len > 0; | |
7859 | struct frozen_state state = { | |
7860 | .table_id = 0, /* Not the table where NXAST_PAUSE was executed. */ | |
7861 | .ofproto_uuid = pin->bridge, | |
7862 | .stack = pin->stack, | |
84cf3c1f | 7863 | .stack_size = pin->stack_size, |
77ab5fd2 BP |
7864 | .mirrors = pin->mirrors, |
7865 | .conntracked = pin->conntracked, | |
00135b86 | 7866 | .xport_uuid = UUID_ZERO, |
77ab5fd2 BP |
7867 | |
7868 | /* When there are no actions, xlate_actions() will search the flow | |
7869 | * table. We don't want it to do that (we want it to resume), so | |
7870 | * supply a no-op action if there aren't any. | |
7871 | * | |
7872 | * (We can't necessarily avoid translating actions entirely if there | |
7873 | * aren't any actions, because there might be some finishing-up to do | |
7874 | * at the end of the pipeline, and we don't check for those | |
7875 | * conditions.) */ | |
7876 | .ofpacts = any_actions ? pin->actions : &noop.ofpact, | |
7877 | .ofpacts_len = any_actions ? pin->actions_len : sizeof noop, | |
7878 | ||
7879 | .action_set = pin->action_set, | |
7880 | .action_set_len = pin->action_set_len, | |
7881 | }; | |
7882 | frozen_metadata_from_flow(&state.metadata, | |
4d617a87 | 7883 | &pin->base.flow_metadata.flow); |
77ab5fd2 BP |
7884 | xin.frozen_state = &state; |
7885 | ||
7886 | struct xlate_out xout; | |
7887 | enum xlate_error error = xlate_actions(&xin, &xout); | |
7888 | *slow = xout.slow; | |
7889 | xlate_out_uninit(&xout); | |
7890 | ||
7891 | /* xlate_actions() can generate a number of errors, but only | |
7892 | * XLATE_BRIDGE_NOT_FOUND really stands out to me as one that we should be | |
7893 | * sure to report over OpenFlow. The others could come up in packet-outs | |
7894 | * or regular flow translation and I don't think that it's going to be too | |
7895 | * useful to report them to the controller. */ | |
7896 | return error == XLATE_BRIDGE_NOT_FOUND ? OFPERR_NXR_STALE : 0; | |
7897 | } | |
7898 | ||
2eb79142 JG |
7899 | /* Sends 'packet' out 'ofport'. If 'port' is a tunnel and that tunnel type |
7900 | * supports a notion of an OAM flag, sets it if 'oam' is true. | |
91d6cd12 AW |
7901 | * May modify 'packet'. |
7902 | * Returns 0 if successful, otherwise a positive errno value. */ | |
7903 | int | |
2eb79142 JG |
7904 | xlate_send_packet(const struct ofport_dpif *ofport, bool oam, |
7905 | struct dp_packet *packet) | |
91d6cd12 | 7906 | { |
84f0f298 | 7907 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
91d6cd12 | 7908 | struct xport *xport; |
2eb79142 JG |
7909 | uint64_t ofpacts_stub[1024 / 8]; |
7910 | struct ofpbuf ofpacts; | |
91d6cd12 | 7911 | struct flow flow; |
91d6cd12 | 7912 | |
2eb79142 | 7913 | ofpbuf_use_stack(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); |
91d6cd12 | 7914 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ |
cf62fa4c | 7915 | flow_extract(packet, &flow); |
b5e7e61a | 7916 | flow.in_port.ofp_port = OFPP_NONE; |
91d6cd12 | 7917 | |
84f0f298 | 7918 | xport = xport_lookup(xcfg, ofport); |
91d6cd12 | 7919 | if (!xport) { |
02ea2703 | 7920 | return EINVAL; |
91d6cd12 | 7921 | } |
2eb79142 JG |
7922 | |
7923 | if (oam) { | |
71f21279 BP |
7924 | const ovs_be16 flag = htons(NX_TUN_FLAG_OAM); |
7925 | ofpact_put_set_field(&ofpacts, mf_from_id(MFF_TUN_FLAGS), | |
7926 | &flag, &flag); | |
2eb79142 JG |
7927 | } |
7928 | ||
7929 | ofpact_put_OUTPUT(&ofpacts)->port = xport->ofp_port; | |
e491a67a | 7930 | |
1f4a8933 JR |
7931 | /* Actions here are not referring to anything versionable (flow tables or |
7932 | * groups) so we don't need to worry about the version here. */ | |
7933 | return ofproto_dpif_execute_actions(xport->xbridge->ofproto, | |
7934 | OVS_VERSION_MAX, &flow, NULL, | |
2eb79142 | 7935 | ofpacts.data, ofpacts.size, packet); |
9583bc14 | 7936 | } |
b256dc52 | 7937 | |
901a517e | 7938 | void |
064799a1 JR |
7939 | xlate_mac_learning_update(const struct ofproto_dpif *ofproto, |
7940 | ofp_port_t in_port, struct eth_addr dl_src, | |
7941 | int vlan, bool is_grat_arp) | |
b256dc52 | 7942 | { |
84f0f298 | 7943 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
b256dc52 JS |
7944 | struct xbridge *xbridge; |
7945 | struct xbundle *xbundle; | |
b256dc52 | 7946 | |
84f0f298 | 7947 | xbridge = xbridge_lookup(xcfg, ofproto); |
b256dc52 JS |
7948 | if (!xbridge) { |
7949 | return; | |
7950 | } | |
7951 | ||
2d9b49dd | 7952 | xbundle = lookup_input_bundle__(xbridge, in_port, NULL); |
b256dc52 JS |
7953 | if (!xbundle) { |
7954 | return; | |
7955 | } | |
7956 | ||
2d9b49dd | 7957 | update_learning_table__(xbridge, xbundle, dl_src, vlan, is_grat_arp); |
b256dc52 | 7958 | } |
bef503e8 | 7959 | |
88186383 AZ |
7960 | void |
7961 | xlate_set_support(const struct ofproto_dpif *ofproto, | |
7962 | const struct dpif_backer_support *support) | |
7963 | { | |
7964 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
7965 | struct xbridge *xbridge = xbridge_lookup(xcfg, ofproto); | |
7966 | ||
7967 | if (xbridge) { | |
7968 | xbridge->support = *support; | |
7969 | } | |
7970 | } |