]>
Commit | Line | Data |
---|---|---|
18080541 | 1 | /* Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014, 2015 Nicira, Inc. |
9583bc14 EJ |
2 | * |
3 | * Licensed under the Apache License, Version 2.0 (the "License"); | |
4 | * you may not use this file except in compliance with the License. | |
5 | * You may obtain a copy of the License at: | |
6 | * | |
7 | * http://www.apache.org/licenses/LICENSE-2.0 | |
8 | * | |
9 | * Unless required by applicable law or agreed to in writing, software | |
10 | * distributed under the License is distributed on an "AS IS" BASIS, | |
11 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |
12 | * See the License for the specific language governing permissions and | |
13 | * limitations under the License. */ | |
14 | ||
15 | #include <config.h> | |
16 | ||
17 | #include "ofproto/ofproto-dpif-xlate.h" | |
18 | ||
8449c4d6 | 19 | #include <errno.h> |
a36de779 PS |
20 | #include <arpa/inet.h> |
21 | #include <net/if.h> | |
22 | #include <sys/socket.h> | |
23 | #include <netinet/in.h> | |
8449c4d6 | 24 | |
a36de779 | 25 | #include "tnl-arp-cache.h" |
db7d4e46 | 26 | #include "bfd.h" |
9583bc14 EJ |
27 | #include "bitmap.h" |
28 | #include "bond.h" | |
29 | #include "bundle.h" | |
30 | #include "byte-order.h" | |
db7d4e46 | 31 | #include "cfm.h" |
9583bc14 EJ |
32 | #include "connmgr.h" |
33 | #include "coverage.h" | |
e14deea0 | 34 | #include "dp-packet.h" |
9583bc14 EJ |
35 | #include "dpif.h" |
36 | #include "dynamic-string.h" | |
f7f1ea29 | 37 | #include "in-band.h" |
db7d4e46 | 38 | #include "lacp.h" |
9583bc14 | 39 | #include "learn.h" |
46c88433 | 40 | #include "list.h" |
0477baa9 | 41 | #include "ovs-lldp.h" |
9583bc14 | 42 | #include "mac-learning.h" |
6d95c4e8 | 43 | #include "mcast-snooping.h" |
9583bc14 EJ |
44 | #include "meta-flow.h" |
45 | #include "multipath.h" | |
46 | #include "netdev-vport.h" | |
47 | #include "netlink.h" | |
48 | #include "nx-match.h" | |
49 | #include "odp-execute.h" | |
50 | #include "ofp-actions.h" | |
51 | #include "ofproto/ofproto-dpif-ipfix.h" | |
ec7ceaed | 52 | #include "ofproto/ofproto-dpif-mirror.h" |
60d02c72 | 53 | #include "ofproto/ofproto-dpif-monitor.h" |
9583bc14 EJ |
54 | #include "ofproto/ofproto-dpif-sflow.h" |
55 | #include "ofproto/ofproto-dpif.h" | |
6f00e29b | 56 | #include "ofproto/ofproto-provider.h" |
a36de779 PS |
57 | #include "ovs-router.h" |
58 | #include "tnl-ports.h" | |
9583bc14 | 59 | #include "tunnel.h" |
e6211adc | 60 | #include "openvswitch/vlog.h" |
9583bc14 | 61 | |
46c88433 | 62 | COVERAGE_DEFINE(xlate_actions); |
0f032e95 | 63 | COVERAGE_DEFINE(xlate_actions_oversize); |
7d031d7e | 64 | COVERAGE_DEFINE(xlate_actions_too_many_output); |
9583bc14 EJ |
65 | |
66 | VLOG_DEFINE_THIS_MODULE(ofproto_dpif_xlate); | |
67 | ||
8a553e9a EJ |
68 | /* Maximum depth of flow table recursion (due to resubmit actions) in a |
69 | * flow translation. */ | |
70 | #define MAX_RESUBMIT_RECURSION 64 | |
adcf00ba AZ |
71 | #define MAX_INTERNAL_RESUBMITS 1 /* Max resbmits allowed using rules in |
72 | internal table. */ | |
8a553e9a | 73 | |
7bbdd84f SH |
74 | /* Timeout for internal rules created to handle recirculation */ |
75 | #define RECIRC_TIMEOUT 60 | |
76 | ||
98b07853 BP |
77 | /* Maximum number of resubmit actions in a flow translation, whether they are |
78 | * recursive or not. */ | |
79 | #define MAX_RESUBMITS (MAX_RESUBMIT_RECURSION * MAX_RESUBMIT_RECURSION) | |
80 | ||
46c88433 EJ |
81 | struct xbridge { |
82 | struct hmap_node hmap_node; /* Node in global 'xbridges' map. */ | |
83 | struct ofproto_dpif *ofproto; /* Key in global 'xbridges' map. */ | |
84 | ||
ca6ba700 | 85 | struct ovs_list xbundles; /* Owned xbundles. */ |
46c88433 EJ |
86 | struct hmap xports; /* Indexed by ofp_port. */ |
87 | ||
88 | char *name; /* Name used in log messages. */ | |
89a8a7f0 | 89 | struct dpif *dpif; /* Datapath interface. */ |
46c88433 | 90 | struct mac_learning *ml; /* Mac learning handle. */ |
6d95c4e8 | 91 | struct mcast_snooping *ms; /* Multicast Snooping handle. */ |
46c88433 EJ |
92 | struct mbridge *mbridge; /* Mirroring. */ |
93 | struct dpif_sflow *sflow; /* SFlow handle, or null. */ | |
94 | struct dpif_ipfix *ipfix; /* Ipfix handle, or null. */ | |
ce3955be | 95 | struct netflow *netflow; /* Netflow handle, or null. */ |
9d189a50 | 96 | struct stp *stp; /* STP or null if disabled. */ |
9efd308e | 97 | struct rstp *rstp; /* RSTP or null if disabled. */ |
46c88433 | 98 | |
46c88433 EJ |
99 | bool has_in_band; /* Bridge has in band control? */ |
100 | bool forward_bpdu; /* Bridge forwards STP BPDUs? */ | |
4b97b70d | 101 | |
adcf00ba AZ |
102 | /* True if the datapath supports recirculation. */ |
103 | bool enable_recirc; | |
104 | ||
4b97b70d BP |
105 | /* True if the datapath supports variable-length |
106 | * OVS_USERSPACE_ATTR_USERDATA in OVS_ACTION_ATTR_USERSPACE actions. | |
107 | * False if the datapath supports only 8-byte (or shorter) userdata. */ | |
108 | bool variable_length_userdata; | |
8bfd0fda BP |
109 | |
110 | /* Number of MPLS label stack entries that the datapath supports | |
111 | * in matches. */ | |
112 | size_t max_mpls_depth; | |
53477c2c JR |
113 | |
114 | /* True if the datapath supports masked data in OVS_ACTION_ATTR_SET | |
115 | * actions. */ | |
116 | bool masked_set_action; | |
46c88433 EJ |
117 | }; |
118 | ||
119 | struct xbundle { | |
120 | struct hmap_node hmap_node; /* In global 'xbundles' map. */ | |
121 | struct ofbundle *ofbundle; /* Key in global 'xbundles' map. */ | |
122 | ||
ca6ba700 | 123 | struct ovs_list list_node; /* In parent 'xbridges' list. */ |
46c88433 EJ |
124 | struct xbridge *xbridge; /* Parent xbridge. */ |
125 | ||
ca6ba700 | 126 | struct ovs_list xports; /* Contains "struct xport"s. */ |
46c88433 EJ |
127 | |
128 | char *name; /* Name used in log messages. */ | |
129 | struct bond *bond; /* Nonnull iff more than one port. */ | |
130 | struct lacp *lacp; /* LACP handle or null. */ | |
131 | ||
132 | enum port_vlan_mode vlan_mode; /* VLAN mode. */ | |
133 | int vlan; /* -1=trunk port, else a 12-bit VLAN ID. */ | |
134 | unsigned long *trunks; /* Bitmap of trunked VLANs, if 'vlan' == -1. | |
135 | * NULL if all VLANs are trunked. */ | |
136 | bool use_priority_tags; /* Use 802.1p tag for frames in VLAN 0? */ | |
137 | bool floodable; /* No port has OFPUTIL_PC_NO_FLOOD set? */ | |
138 | }; | |
139 | ||
140 | struct xport { | |
141 | struct hmap_node hmap_node; /* Node in global 'xports' map. */ | |
142 | struct ofport_dpif *ofport; /* Key in global 'xports map. */ | |
143 | ||
144 | struct hmap_node ofp_node; /* Node in parent xbridge 'xports' map. */ | |
145 | ofp_port_t ofp_port; /* Key in parent xbridge 'xports' map. */ | |
146 | ||
147 | odp_port_t odp_port; /* Datapath port number or ODPP_NONE. */ | |
148 | ||
ca6ba700 | 149 | struct ovs_list bundle_node; /* In parent xbundle (if it exists). */ |
46c88433 EJ |
150 | struct xbundle *xbundle; /* Parent xbundle or null. */ |
151 | ||
152 | struct netdev *netdev; /* 'ofport''s netdev. */ | |
153 | ||
154 | struct xbridge *xbridge; /* Parent bridge. */ | |
155 | struct xport *peer; /* Patch port peer or null. */ | |
156 | ||
157 | enum ofputil_port_config config; /* OpenFlow port configuration. */ | |
dd8cd4b4 | 158 | enum ofputil_port_state state; /* OpenFlow port state. */ |
92cf817b | 159 | int stp_port_no; /* STP port number or -1 if not in use. */ |
f025bcb7 | 160 | struct rstp_port *rstp_port; /* RSTP port or null. */ |
46c88433 | 161 | |
55954f6e EJ |
162 | struct hmap skb_priorities; /* Map of 'skb_priority_to_dscp's. */ |
163 | ||
46c88433 EJ |
164 | bool may_enable; /* May be enabled in bonds. */ |
165 | bool is_tunnel; /* Is a tunnel port. */ | |
166 | ||
167 | struct cfm *cfm; /* CFM handle or null. */ | |
168 | struct bfd *bfd; /* BFD handle or null. */ | |
0477baa9 | 169 | struct lldp *lldp; /* LLDP handle or null. */ |
46c88433 EJ |
170 | }; |
171 | ||
4d0acc70 EJ |
172 | struct xlate_ctx { |
173 | struct xlate_in *xin; | |
174 | struct xlate_out *xout; | |
175 | ||
46c88433 | 176 | const struct xbridge *xbridge; |
4d0acc70 EJ |
177 | |
178 | /* Flow at the last commit. */ | |
179 | struct flow base_flow; | |
180 | ||
181 | /* Tunnel IP destination address as received. This is stored separately | |
182 | * as the base_flow.tunnel is cleared on init to reflect the datapath | |
183 | * behavior. Used to make sure not to send tunneled output to ourselves, | |
184 | * which might lead to an infinite loop. This could happen easily | |
185 | * if a tunnel is marked as 'ip_remote=flow', and the flow does not | |
186 | * actually set the tun_dst field. */ | |
187 | ovs_be32 orig_tunnel_ip_dst; | |
188 | ||
189 | /* Stack for the push and pop actions. Each stack element is of type | |
190 | * "union mf_subvalue". */ | |
191 | union mf_subvalue init_stack[1024 / sizeof(union mf_subvalue)]; | |
192 | struct ofpbuf stack; | |
193 | ||
194 | /* The rule that we are currently translating, or NULL. */ | |
195 | struct rule_dpif *rule; | |
196 | ||
98b07853 BP |
197 | /* Resubmit statistics, via xlate_table_action(). */ |
198 | int recurse; /* Current resubmit nesting depth. */ | |
199 | int resubmits; /* Total number of resubmits. */ | |
5a070238 | 200 | bool in_group; /* Currently translating ofgroup, if true. */ |
029ca940 | 201 | bool in_action_set; /* Currently translating action_set, if true. */ |
98b07853 | 202 | |
4d0acc70 EJ |
203 | uint32_t orig_skb_priority; /* Priority when packet arrived. */ |
204 | uint8_t table_id; /* OpenFlow table ID where flow was found. */ | |
205 | uint32_t sflow_n_outputs; /* Number of output ports. */ | |
4e022ec0 | 206 | odp_port_t sflow_odp_port; /* Output port for composing sFlow action. */ |
4d0acc70 EJ |
207 | uint16_t user_cookie_offset;/* Used for user_action_cookie fixup. */ |
208 | bool exit; /* No further actions should be processed. */ | |
7fdb60a7 | 209 | |
92c08f09 SH |
210 | bool use_recirc; /* Should generate recirc? */ |
211 | struct xlate_recirc recirc; /* Information used for generating | |
212 | * recirculation actions */ | |
213 | ||
7bbdd84f SH |
214 | /* True if a packet was but is no longer MPLS (due to an MPLS pop action). |
215 | * This is a trigger for recirculation in cases where translating an action | |
216 | * or looking up a flow requires access to the fields of the packet after | |
217 | * the MPLS label stack that was originally present. */ | |
218 | bool was_mpls; | |
219 | ||
7fdb60a7 SH |
220 | /* OpenFlow 1.1+ action set. |
221 | * | |
222 | * 'action_set' accumulates "struct ofpact"s added by OFPACT_WRITE_ACTIONS. | |
223 | * When translation is otherwise complete, ofpacts_execute_action_set() | |
224 | * converts it to a set of "struct ofpact"s that can be translated into | |
225 | * datapath actions. */ | |
c61f3870 | 226 | bool action_set_has_group; /* Action set contains OFPACT_GROUP? */ |
7fdb60a7 SH |
227 | struct ofpbuf action_set; /* Action set. */ |
228 | uint64_t action_set_stub[1024 / 8]; | |
4d0acc70 EJ |
229 | }; |
230 | ||
9583bc14 EJ |
231 | /* A controller may use OFPP_NONE as the ingress port to indicate that |
232 | * it did not arrive on a "real" port. 'ofpp_none_bundle' exists for | |
233 | * when an input bundle is needed for validation (e.g., mirroring or | |
234 | * OFPP_NORMAL processing). It is not connected to an 'ofproto' or have | |
3548d242 BP |
235 | * any 'port' structs, so care must be taken when dealing with it. */ |
236 | static struct xbundle ofpp_none_bundle = { | |
237 | .name = "OFPP_NONE", | |
238 | .vlan_mode = PORT_VLAN_TRUNK | |
239 | }; | |
9583bc14 | 240 | |
55954f6e EJ |
241 | /* Node in 'xport''s 'skb_priorities' map. Used to maintain a map from |
242 | * 'priority' (the datapath's term for QoS queue) to the dscp bits which all | |
243 | * traffic egressing the 'ofport' with that priority should be marked with. */ | |
244 | struct skb_priority_to_dscp { | |
245 | struct hmap_node hmap_node; /* Node in 'ofport_dpif''s 'skb_priorities'. */ | |
246 | uint32_t skb_priority; /* Priority of this queue (see struct flow). */ | |
247 | ||
248 | uint8_t dscp; /* DSCP bits to mark outgoing traffic with. */ | |
249 | }; | |
250 | ||
b256dc52 JS |
251 | enum xc_type { |
252 | XC_RULE, | |
253 | XC_BOND, | |
254 | XC_NETDEV, | |
255 | XC_NETFLOW, | |
256 | XC_MIRROR, | |
257 | XC_LEARN, | |
258 | XC_NORMAL, | |
259 | XC_FIN_TIMEOUT, | |
1e684d7d | 260 | XC_GROUP, |
a36de779 | 261 | XC_TNL_ARP, |
b256dc52 JS |
262 | }; |
263 | ||
264 | /* xlate_cache entries hold enough information to perform the side effects of | |
265 | * xlate_actions() for a rule, without needing to perform rule translation | |
266 | * from scratch. The primary usage of these is to submit statistics to objects | |
267 | * that a flow relates to, although they may be used for other effects as well | |
268 | * (for instance, refreshing hard timeouts for learned flows). */ | |
269 | struct xc_entry { | |
270 | enum xc_type type; | |
271 | union { | |
272 | struct rule_dpif *rule; | |
273 | struct { | |
274 | struct netdev *tx; | |
275 | struct netdev *rx; | |
276 | struct bfd *bfd; | |
277 | } dev; | |
278 | struct { | |
279 | struct netflow *netflow; | |
280 | struct flow *flow; | |
281 | ofp_port_t iface; | |
282 | } nf; | |
283 | struct { | |
284 | struct mbridge *mbridge; | |
285 | mirror_mask_t mirrors; | |
286 | } mirror; | |
287 | struct { | |
288 | struct bond *bond; | |
289 | struct flow *flow; | |
290 | uint16_t vid; | |
291 | } bond; | |
292 | struct { | |
4165b5e0 JS |
293 | struct ofproto_dpif *ofproto; |
294 | struct ofputil_flow_mod *fm; | |
295 | struct ofpbuf *ofpacts; | |
b256dc52 JS |
296 | } learn; |
297 | struct { | |
298 | struct ofproto_dpif *ofproto; | |
299 | struct flow *flow; | |
300 | int vlan; | |
301 | } normal; | |
302 | struct { | |
303 | struct rule_dpif *rule; | |
304 | uint16_t idle; | |
305 | uint16_t hard; | |
306 | } fin; | |
1e684d7d RW |
307 | struct { |
308 | struct group_dpif *group; | |
309 | struct ofputil_bucket *bucket; | |
310 | } group; | |
a36de779 PS |
311 | struct { |
312 | char br_name[IFNAMSIZ]; | |
313 | ovs_be32 d_ip; | |
314 | } tnl_arp_cache; | |
b256dc52 JS |
315 | } u; |
316 | }; | |
317 | ||
318 | #define XC_ENTRY_FOR_EACH(entry, entries, xcache) \ | |
319 | entries = xcache->entries; \ | |
320 | for (entry = ofpbuf_try_pull(&entries, sizeof *entry); \ | |
321 | entry; \ | |
322 | entry = ofpbuf_try_pull(&entries, sizeof *entry)) | |
323 | ||
324 | struct xlate_cache { | |
325 | struct ofpbuf entries; | |
326 | }; | |
327 | ||
84f0f298 RW |
328 | /* Xlate config contains hash maps of all bridges, bundles and ports. |
329 | * Xcfgp contains the pointer to the current xlate configuration. | |
330 | * When the main thread needs to change the configuration, it copies xcfgp to | |
331 | * new_xcfg and edits new_xcfg. This enables the use of RCU locking which | |
332 | * does not block handler and revalidator threads. */ | |
333 | struct xlate_cfg { | |
334 | struct hmap xbridges; | |
335 | struct hmap xbundles; | |
336 | struct hmap xports; | |
337 | }; | |
b1b72f2d | 338 | static OVSRCU_TYPE(struct xlate_cfg *) xcfgp = OVSRCU_INITIALIZER(NULL); |
f439f23b | 339 | static struct xlate_cfg *new_xcfg = NULL; |
46c88433 EJ |
340 | |
341 | static bool may_receive(const struct xport *, struct xlate_ctx *); | |
9583bc14 EJ |
342 | static void do_xlate_actions(const struct ofpact *, size_t ofpacts_len, |
343 | struct xlate_ctx *); | |
adcf00ba | 344 | static void xlate_normal(struct xlate_ctx *); |
34dd0d78 | 345 | static inline void xlate_report(struct xlate_ctx *, const char *); |
6d328fa2 SH |
346 | static void xlate_table_action(struct xlate_ctx *, ofp_port_t in_port, |
347 | uint8_t table_id, bool may_packet_in, | |
348 | bool honor_table_miss); | |
46c88433 EJ |
349 | static bool input_vid_is_valid(uint16_t vid, struct xbundle *, bool warn); |
350 | static uint16_t input_vid_to_vlan(const struct xbundle *, uint16_t vid); | |
351 | static void output_normal(struct xlate_ctx *, const struct xbundle *, | |
9583bc14 | 352 | uint16_t vlan); |
4e022ec0 | 353 | static void compose_output_action(struct xlate_ctx *, ofp_port_t ofp_port); |
9583bc14 | 354 | |
84f0f298 RW |
355 | static struct xbridge *xbridge_lookup(struct xlate_cfg *, |
356 | const struct ofproto_dpif *); | |
357 | static struct xbundle *xbundle_lookup(struct xlate_cfg *, | |
358 | const struct ofbundle *); | |
359 | static struct xport *xport_lookup(struct xlate_cfg *, | |
360 | const struct ofport_dpif *); | |
46c88433 | 361 | static struct xport *get_ofp_port(const struct xbridge *, ofp_port_t ofp_port); |
55954f6e EJ |
362 | static struct skb_priority_to_dscp *get_skb_priority(const struct xport *, |
363 | uint32_t skb_priority); | |
364 | static void clear_skb_priorities(struct xport *); | |
16194afd | 365 | static size_t count_skb_priorities(const struct xport *); |
55954f6e EJ |
366 | static bool dscp_from_skb_priority(const struct xport *, uint32_t skb_priority, |
367 | uint8_t *dscp); | |
46c88433 | 368 | |
b256dc52 JS |
369 | static struct xc_entry *xlate_cache_add_entry(struct xlate_cache *xc, |
370 | enum xc_type type); | |
84f0f298 RW |
371 | static void xlate_xbridge_init(struct xlate_cfg *, struct xbridge *); |
372 | static void xlate_xbundle_init(struct xlate_cfg *, struct xbundle *); | |
373 | static void xlate_xport_init(struct xlate_cfg *, struct xport *); | |
9efd308e | 374 | static void xlate_xbridge_set(struct xbridge *, struct dpif *, |
9efd308e DV |
375 | const struct mac_learning *, struct stp *, |
376 | struct rstp *, const struct mcast_snooping *, | |
377 | const struct mbridge *, | |
378 | const struct dpif_sflow *, | |
379 | const struct dpif_ipfix *, | |
2f47cdf4 | 380 | const struct netflow *, |
84f0f298 RW |
381 | bool forward_bpdu, bool has_in_band, |
382 | bool enable_recirc, | |
383 | bool variable_length_userdata, | |
53477c2c JR |
384 | size_t max_mpls_depth, |
385 | bool masked_set_action); | |
84f0f298 RW |
386 | static void xlate_xbundle_set(struct xbundle *xbundle, |
387 | enum port_vlan_mode vlan_mode, int vlan, | |
388 | unsigned long *trunks, bool use_priority_tags, | |
389 | const struct bond *bond, const struct lacp *lacp, | |
390 | bool floodable); | |
391 | static void xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
392 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 DF |
393 | const struct bfd *bfd, const struct lldp *lldp, |
394 | int stp_port_no, const struct rstp_port *rstp_port, | |
84f0f298 RW |
395 | enum ofputil_port_config config, |
396 | enum ofputil_port_state state, bool is_tunnel, | |
397 | bool may_enable); | |
398 | static void xlate_xbridge_remove(struct xlate_cfg *, struct xbridge *); | |
399 | static void xlate_xbundle_remove(struct xlate_cfg *, struct xbundle *); | |
400 | static void xlate_xport_remove(struct xlate_cfg *, struct xport *); | |
401 | static void xlate_xbridge_copy(struct xbridge *); | |
402 | static void xlate_xbundle_copy(struct xbridge *, struct xbundle *); | |
403 | static void xlate_xport_copy(struct xbridge *, struct xbundle *, | |
404 | struct xport *); | |
405 | static void xlate_xcfg_free(struct xlate_cfg *); | |
b256dc52 | 406 | |
34dd0d78 JR |
407 | static inline void |
408 | xlate_report(struct xlate_ctx *ctx, const char *s) | |
409 | { | |
410 | if (OVS_UNLIKELY(ctx->xin->report_hook)) { | |
411 | ctx->xin->report_hook(ctx->xin, s, ctx->recurse); | |
412 | } | |
413 | } | |
84f0f298 RW |
414 | |
415 | static void | |
416 | xlate_xbridge_init(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
417 | { | |
418 | list_init(&xbridge->xbundles); | |
419 | hmap_init(&xbridge->xports); | |
420 | hmap_insert(&xcfg->xbridges, &xbridge->hmap_node, | |
421 | hash_pointer(xbridge->ofproto, 0)); | |
422 | } | |
423 | ||
424 | static void | |
425 | xlate_xbundle_init(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
426 | { | |
427 | list_init(&xbundle->xports); | |
428 | list_insert(&xbundle->xbridge->xbundles, &xbundle->list_node); | |
429 | hmap_insert(&xcfg->xbundles, &xbundle->hmap_node, | |
430 | hash_pointer(xbundle->ofbundle, 0)); | |
431 | } | |
432 | ||
433 | static void | |
434 | xlate_xport_init(struct xlate_cfg *xcfg, struct xport *xport) | |
435 | { | |
436 | hmap_init(&xport->skb_priorities); | |
437 | hmap_insert(&xcfg->xports, &xport->hmap_node, | |
438 | hash_pointer(xport->ofport, 0)); | |
439 | hmap_insert(&xport->xbridge->xports, &xport->ofp_node, | |
440 | hash_ofp_port(xport->ofp_port)); | |
441 | } | |
442 | ||
443 | static void | |
444 | xlate_xbridge_set(struct xbridge *xbridge, | |
445 | struct dpif *dpif, | |
ec89fc6f | 446 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 447 | struct rstp *rstp, const struct mcast_snooping *ms, |
ec89fc6f | 448 | const struct mbridge *mbridge, |
46c88433 | 449 | const struct dpif_sflow *sflow, |
ce3955be | 450 | const struct dpif_ipfix *ipfix, |
2f47cdf4 | 451 | const struct netflow *netflow, |
4b97b70d | 452 | bool forward_bpdu, bool has_in_band, |
adcf00ba | 453 | bool enable_recirc, |
8bfd0fda | 454 | bool variable_length_userdata, |
53477c2c JR |
455 | size_t max_mpls_depth, |
456 | bool masked_set_action) | |
46c88433 | 457 | { |
46c88433 EJ |
458 | if (xbridge->ml != ml) { |
459 | mac_learning_unref(xbridge->ml); | |
460 | xbridge->ml = mac_learning_ref(ml); | |
461 | } | |
462 | ||
6d95c4e8 FL |
463 | if (xbridge->ms != ms) { |
464 | mcast_snooping_unref(xbridge->ms); | |
465 | xbridge->ms = mcast_snooping_ref(ms); | |
466 | } | |
467 | ||
46c88433 EJ |
468 | if (xbridge->mbridge != mbridge) { |
469 | mbridge_unref(xbridge->mbridge); | |
470 | xbridge->mbridge = mbridge_ref(mbridge); | |
471 | } | |
472 | ||
473 | if (xbridge->sflow != sflow) { | |
474 | dpif_sflow_unref(xbridge->sflow); | |
475 | xbridge->sflow = dpif_sflow_ref(sflow); | |
476 | } | |
477 | ||
478 | if (xbridge->ipfix != ipfix) { | |
479 | dpif_ipfix_unref(xbridge->ipfix); | |
480 | xbridge->ipfix = dpif_ipfix_ref(ipfix); | |
481 | } | |
482 | ||
9d189a50 EJ |
483 | if (xbridge->stp != stp) { |
484 | stp_unref(xbridge->stp); | |
485 | xbridge->stp = stp_ref(stp); | |
486 | } | |
487 | ||
9efd308e DV |
488 | if (xbridge->rstp != rstp) { |
489 | rstp_unref(xbridge->rstp); | |
490 | xbridge->rstp = rstp_ref(rstp); | |
491 | } | |
492 | ||
ce3955be EJ |
493 | if (xbridge->netflow != netflow) { |
494 | netflow_unref(xbridge->netflow); | |
495 | xbridge->netflow = netflow_ref(netflow); | |
496 | } | |
497 | ||
89a8a7f0 | 498 | xbridge->dpif = dpif; |
46c88433 EJ |
499 | xbridge->forward_bpdu = forward_bpdu; |
500 | xbridge->has_in_band = has_in_band; | |
adcf00ba | 501 | xbridge->enable_recirc = enable_recirc; |
4b97b70d | 502 | xbridge->variable_length_userdata = variable_length_userdata; |
8bfd0fda | 503 | xbridge->max_mpls_depth = max_mpls_depth; |
53477c2c | 504 | xbridge->masked_set_action = masked_set_action; |
46c88433 EJ |
505 | } |
506 | ||
84f0f298 RW |
507 | static void |
508 | xlate_xbundle_set(struct xbundle *xbundle, | |
509 | enum port_vlan_mode vlan_mode, int vlan, | |
510 | unsigned long *trunks, bool use_priority_tags, | |
511 | const struct bond *bond, const struct lacp *lacp, | |
512 | bool floodable) | |
513 | { | |
514 | ovs_assert(xbundle->xbridge); | |
515 | ||
516 | xbundle->vlan_mode = vlan_mode; | |
517 | xbundle->vlan = vlan; | |
518 | xbundle->trunks = trunks; | |
519 | xbundle->use_priority_tags = use_priority_tags; | |
520 | xbundle->floodable = floodable; | |
521 | ||
522 | if (xbundle->bond != bond) { | |
523 | bond_unref(xbundle->bond); | |
524 | xbundle->bond = bond_ref(bond); | |
525 | } | |
526 | ||
527 | if (xbundle->lacp != lacp) { | |
528 | lacp_unref(xbundle->lacp); | |
529 | xbundle->lacp = lacp_ref(lacp); | |
530 | } | |
531 | } | |
532 | ||
533 | static void | |
534 | xlate_xport_set(struct xport *xport, odp_port_t odp_port, | |
535 | const struct netdev *netdev, const struct cfm *cfm, | |
0477baa9 | 536 | const struct bfd *bfd, const struct lldp *lldp, int stp_port_no, |
f025bcb7 | 537 | const struct rstp_port* rstp_port, |
84f0f298 RW |
538 | enum ofputil_port_config config, enum ofputil_port_state state, |
539 | bool is_tunnel, bool may_enable) | |
540 | { | |
541 | xport->config = config; | |
542 | xport->state = state; | |
543 | xport->stp_port_no = stp_port_no; | |
544 | xport->is_tunnel = is_tunnel; | |
545 | xport->may_enable = may_enable; | |
546 | xport->odp_port = odp_port; | |
547 | ||
f025bcb7 JR |
548 | if (xport->rstp_port != rstp_port) { |
549 | rstp_port_unref(xport->rstp_port); | |
550 | xport->rstp_port = rstp_port_ref(rstp_port); | |
551 | } | |
552 | ||
84f0f298 RW |
553 | if (xport->cfm != cfm) { |
554 | cfm_unref(xport->cfm); | |
555 | xport->cfm = cfm_ref(cfm); | |
556 | } | |
557 | ||
558 | if (xport->bfd != bfd) { | |
559 | bfd_unref(xport->bfd); | |
560 | xport->bfd = bfd_ref(bfd); | |
561 | } | |
562 | ||
0477baa9 DF |
563 | if (xport->lldp != lldp) { |
564 | lldp_unref(xport->lldp); | |
565 | xport->lldp = lldp_ref(lldp); | |
566 | } | |
567 | ||
84f0f298 RW |
568 | if (xport->netdev != netdev) { |
569 | netdev_close(xport->netdev); | |
570 | xport->netdev = netdev_ref(netdev); | |
571 | } | |
572 | } | |
573 | ||
574 | static void | |
575 | xlate_xbridge_copy(struct xbridge *xbridge) | |
576 | { | |
577 | struct xbundle *xbundle; | |
578 | struct xport *xport; | |
579 | struct xbridge *new_xbridge = xzalloc(sizeof *xbridge); | |
580 | new_xbridge->ofproto = xbridge->ofproto; | |
581 | new_xbridge->name = xstrdup(xbridge->name); | |
582 | xlate_xbridge_init(new_xcfg, new_xbridge); | |
583 | ||
584 | xlate_xbridge_set(new_xbridge, | |
34dd0d78 | 585 | xbridge->dpif, xbridge->ml, xbridge->stp, |
9efd308e DV |
586 | xbridge->rstp, xbridge->ms, xbridge->mbridge, |
587 | xbridge->sflow, xbridge->ipfix, xbridge->netflow, | |
2f47cdf4 | 588 | xbridge->forward_bpdu, |
9efd308e DV |
589 | xbridge->has_in_band, xbridge->enable_recirc, |
590 | xbridge->variable_length_userdata, | |
53477c2c | 591 | xbridge->max_mpls_depth, xbridge->masked_set_action); |
84f0f298 RW |
592 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
593 | xlate_xbundle_copy(new_xbridge, xbundle); | |
594 | } | |
595 | ||
596 | /* Copy xports which are not part of a xbundle */ | |
597 | HMAP_FOR_EACH (xport, ofp_node, &xbridge->xports) { | |
598 | if (!xport->xbundle) { | |
599 | xlate_xport_copy(new_xbridge, NULL, xport); | |
600 | } | |
601 | } | |
602 | } | |
603 | ||
604 | static void | |
605 | xlate_xbundle_copy(struct xbridge *xbridge, struct xbundle *xbundle) | |
606 | { | |
607 | struct xport *xport; | |
608 | struct xbundle *new_xbundle = xzalloc(sizeof *xbundle); | |
609 | new_xbundle->ofbundle = xbundle->ofbundle; | |
610 | new_xbundle->xbridge = xbridge; | |
611 | new_xbundle->name = xstrdup(xbundle->name); | |
612 | xlate_xbundle_init(new_xcfg, new_xbundle); | |
613 | ||
614 | xlate_xbundle_set(new_xbundle, xbundle->vlan_mode, | |
615 | xbundle->vlan, xbundle->trunks, | |
616 | xbundle->use_priority_tags, xbundle->bond, xbundle->lacp, | |
617 | xbundle->floodable); | |
618 | LIST_FOR_EACH (xport, bundle_node, &xbundle->xports) { | |
619 | xlate_xport_copy(xbridge, new_xbundle, xport); | |
620 | } | |
621 | } | |
622 | ||
623 | static void | |
624 | xlate_xport_copy(struct xbridge *xbridge, struct xbundle *xbundle, | |
625 | struct xport *xport) | |
626 | { | |
627 | struct skb_priority_to_dscp *pdscp, *new_pdscp; | |
628 | struct xport *new_xport = xzalloc(sizeof *xport); | |
629 | new_xport->ofport = xport->ofport; | |
630 | new_xport->ofp_port = xport->ofp_port; | |
631 | new_xport->xbridge = xbridge; | |
632 | xlate_xport_init(new_xcfg, new_xport); | |
633 | ||
634 | xlate_xport_set(new_xport, xport->odp_port, xport->netdev, xport->cfm, | |
0477baa9 DF |
635 | xport->bfd, xport->lldp, xport->stp_port_no, |
636 | xport->rstp_port, xport->config, xport->state, | |
637 | xport->is_tunnel, xport->may_enable); | |
84f0f298 RW |
638 | |
639 | if (xport->peer) { | |
640 | struct xport *peer = xport_lookup(new_xcfg, xport->peer->ofport); | |
641 | if (peer) { | |
642 | new_xport->peer = peer; | |
643 | new_xport->peer->peer = new_xport; | |
644 | } | |
645 | } | |
646 | ||
647 | if (xbundle) { | |
648 | new_xport->xbundle = xbundle; | |
649 | list_insert(&new_xport->xbundle->xports, &new_xport->bundle_node); | |
650 | } | |
651 | ||
652 | HMAP_FOR_EACH (pdscp, hmap_node, &xport->skb_priorities) { | |
653 | new_pdscp = xmalloc(sizeof *pdscp); | |
654 | new_pdscp->skb_priority = pdscp->skb_priority; | |
655 | new_pdscp->dscp = pdscp->dscp; | |
656 | hmap_insert(&new_xport->skb_priorities, &new_pdscp->hmap_node, | |
657 | hash_int(new_pdscp->skb_priority, 0)); | |
658 | } | |
659 | } | |
660 | ||
661 | /* Sets the current xlate configuration to new_xcfg and frees the old xlate | |
662 | * configuration in xcfgp. | |
663 | * | |
664 | * This needs to be called after editing the xlate configuration. | |
665 | * | |
666 | * Functions that edit the new xlate configuration are | |
667 | * xlate_<ofport/bundle/ofport>_set and xlate_<ofport/bundle/ofport>_remove. | |
668 | * | |
669 | * A sample workflow: | |
670 | * | |
671 | * xlate_txn_start(); | |
672 | * ... | |
673 | * edit_xlate_configuration(); | |
674 | * ... | |
675 | * xlate_txn_commit(); */ | |
46c88433 | 676 | void |
84f0f298 RW |
677 | xlate_txn_commit(void) |
678 | { | |
679 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
680 | ||
681 | ovsrcu_set(&xcfgp, new_xcfg); | |
40a9c4c2 AW |
682 | ovsrcu_synchronize(); |
683 | xlate_xcfg_free(xcfg); | |
84f0f298 RW |
684 | new_xcfg = NULL; |
685 | } | |
686 | ||
687 | /* Copies the current xlate configuration in xcfgp to new_xcfg. | |
688 | * | |
689 | * This needs to be called prior to editing the xlate configuration. */ | |
690 | void | |
691 | xlate_txn_start(void) | |
692 | { | |
693 | struct xbridge *xbridge; | |
694 | struct xlate_cfg *xcfg; | |
695 | ||
696 | ovs_assert(!new_xcfg); | |
697 | ||
698 | new_xcfg = xmalloc(sizeof *new_xcfg); | |
699 | hmap_init(&new_xcfg->xbridges); | |
700 | hmap_init(&new_xcfg->xbundles); | |
701 | hmap_init(&new_xcfg->xports); | |
702 | ||
703 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
704 | if (!xcfg) { | |
705 | return; | |
706 | } | |
707 | ||
708 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
709 | xlate_xbridge_copy(xbridge); | |
710 | } | |
711 | } | |
712 | ||
713 | ||
714 | static void | |
715 | xlate_xcfg_free(struct xlate_cfg *xcfg) | |
716 | { | |
717 | struct xbridge *xbridge, *next_xbridge; | |
718 | ||
719 | if (!xcfg) { | |
720 | return; | |
721 | } | |
722 | ||
723 | HMAP_FOR_EACH_SAFE (xbridge, next_xbridge, hmap_node, &xcfg->xbridges) { | |
724 | xlate_xbridge_remove(xcfg, xbridge); | |
725 | } | |
726 | ||
727 | hmap_destroy(&xcfg->xbridges); | |
728 | hmap_destroy(&xcfg->xbundles); | |
729 | hmap_destroy(&xcfg->xports); | |
730 | free(xcfg); | |
731 | } | |
732 | ||
733 | void | |
734 | xlate_ofproto_set(struct ofproto_dpif *ofproto, const char *name, | |
34dd0d78 | 735 | struct dpif *dpif, |
84f0f298 | 736 | const struct mac_learning *ml, struct stp *stp, |
9efd308e | 737 | struct rstp *rstp, const struct mcast_snooping *ms, |
84f0f298 RW |
738 | const struct mbridge *mbridge, |
739 | const struct dpif_sflow *sflow, | |
740 | const struct dpif_ipfix *ipfix, | |
2f47cdf4 | 741 | const struct netflow *netflow, |
9efd308e | 742 | bool forward_bpdu, bool has_in_band, bool enable_recirc, |
53477c2c JR |
743 | bool variable_length_userdata, size_t max_mpls_depth, |
744 | bool masked_set_action) | |
84f0f298 RW |
745 | { |
746 | struct xbridge *xbridge; | |
747 | ||
748 | ovs_assert(new_xcfg); | |
749 | ||
750 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
751 | if (!xbridge) { | |
752 | xbridge = xzalloc(sizeof *xbridge); | |
753 | xbridge->ofproto = ofproto; | |
754 | ||
755 | xlate_xbridge_init(new_xcfg, xbridge); | |
756 | } | |
757 | ||
758 | free(xbridge->name); | |
759 | xbridge->name = xstrdup(name); | |
760 | ||
34dd0d78 JR |
761 | xlate_xbridge_set(xbridge, dpif, ml, stp, rstp, ms, mbridge, sflow, ipfix, |
762 | netflow, forward_bpdu, has_in_band, enable_recirc, | |
53477c2c JR |
763 | variable_length_userdata, max_mpls_depth, |
764 | masked_set_action); | |
84f0f298 RW |
765 | } |
766 | ||
767 | static void | |
768 | xlate_xbridge_remove(struct xlate_cfg *xcfg, struct xbridge *xbridge) | |
46c88433 | 769 | { |
46c88433 EJ |
770 | struct xbundle *xbundle, *next_xbundle; |
771 | struct xport *xport, *next_xport; | |
772 | ||
773 | if (!xbridge) { | |
774 | return; | |
775 | } | |
776 | ||
777 | HMAP_FOR_EACH_SAFE (xport, next_xport, ofp_node, &xbridge->xports) { | |
84f0f298 | 778 | xlate_xport_remove(xcfg, xport); |
46c88433 EJ |
779 | } |
780 | ||
781 | LIST_FOR_EACH_SAFE (xbundle, next_xbundle, list_node, &xbridge->xbundles) { | |
84f0f298 | 782 | xlate_xbundle_remove(xcfg, xbundle); |
46c88433 EJ |
783 | } |
784 | ||
84f0f298 | 785 | hmap_remove(&xcfg->xbridges, &xbridge->hmap_node); |
795cc5c1 | 786 | mac_learning_unref(xbridge->ml); |
6d95c4e8 | 787 | mcast_snooping_unref(xbridge->ms); |
795cc5c1 EJ |
788 | mbridge_unref(xbridge->mbridge); |
789 | dpif_sflow_unref(xbridge->sflow); | |
790 | dpif_ipfix_unref(xbridge->ipfix); | |
791 | stp_unref(xbridge->stp); | |
9efd308e | 792 | rstp_unref(xbridge->rstp); |
795cc5c1 | 793 | hmap_destroy(&xbridge->xports); |
46c88433 EJ |
794 | free(xbridge->name); |
795 | free(xbridge); | |
796 | } | |
797 | ||
84f0f298 RW |
798 | void |
799 | xlate_remove_ofproto(struct ofproto_dpif *ofproto) | |
800 | { | |
801 | struct xbridge *xbridge; | |
802 | ||
803 | ovs_assert(new_xcfg); | |
804 | ||
805 | xbridge = xbridge_lookup(new_xcfg, ofproto); | |
806 | xlate_xbridge_remove(new_xcfg, xbridge); | |
807 | } | |
808 | ||
46c88433 EJ |
809 | void |
810 | xlate_bundle_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
811 | const char *name, enum port_vlan_mode vlan_mode, int vlan, | |
812 | unsigned long *trunks, bool use_priority_tags, | |
813 | const struct bond *bond, const struct lacp *lacp, | |
814 | bool floodable) | |
815 | { | |
84f0f298 | 816 | struct xbundle *xbundle; |
46c88433 | 817 | |
84f0f298 RW |
818 | ovs_assert(new_xcfg); |
819 | ||
820 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
46c88433 EJ |
821 | if (!xbundle) { |
822 | xbundle = xzalloc(sizeof *xbundle); | |
823 | xbundle->ofbundle = ofbundle; | |
84f0f298 | 824 | xbundle->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 | 825 | |
84f0f298 | 826 | xlate_xbundle_init(new_xcfg, xbundle); |
46c88433 EJ |
827 | } |
828 | ||
46c88433 EJ |
829 | free(xbundle->name); |
830 | xbundle->name = xstrdup(name); | |
831 | ||
84f0f298 RW |
832 | xlate_xbundle_set(xbundle, vlan_mode, vlan, trunks, |
833 | use_priority_tags, bond, lacp, floodable); | |
46c88433 EJ |
834 | } |
835 | ||
84f0f298 RW |
836 | static void |
837 | xlate_xbundle_remove(struct xlate_cfg *xcfg, struct xbundle *xbundle) | |
46c88433 | 838 | { |
46c88433 EJ |
839 | struct xport *xport, *next; |
840 | ||
841 | if (!xbundle) { | |
842 | return; | |
843 | } | |
844 | ||
845 | LIST_FOR_EACH_SAFE (xport, next, bundle_node, &xbundle->xports) { | |
846 | list_remove(&xport->bundle_node); | |
847 | xport->xbundle = NULL; | |
848 | } | |
849 | ||
84f0f298 | 850 | hmap_remove(&xcfg->xbundles, &xbundle->hmap_node); |
46c88433 EJ |
851 | list_remove(&xbundle->list_node); |
852 | bond_unref(xbundle->bond); | |
853 | lacp_unref(xbundle->lacp); | |
854 | free(xbundle->name); | |
855 | free(xbundle); | |
856 | } | |
857 | ||
84f0f298 RW |
858 | void |
859 | xlate_bundle_remove(struct ofbundle *ofbundle) | |
860 | { | |
861 | struct xbundle *xbundle; | |
862 | ||
863 | ovs_assert(new_xcfg); | |
864 | ||
865 | xbundle = xbundle_lookup(new_xcfg, ofbundle); | |
866 | xlate_xbundle_remove(new_xcfg, xbundle); | |
867 | } | |
868 | ||
46c88433 EJ |
869 | void |
870 | xlate_ofport_set(struct ofproto_dpif *ofproto, struct ofbundle *ofbundle, | |
871 | struct ofport_dpif *ofport, ofp_port_t ofp_port, | |
872 | odp_port_t odp_port, const struct netdev *netdev, | |
873 | const struct cfm *cfm, const struct bfd *bfd, | |
0477baa9 DF |
874 | const struct lldp *lldp, struct ofport_dpif *peer, |
875 | int stp_port_no, const struct rstp_port *rstp_port, | |
55954f6e | 876 | const struct ofproto_port_queue *qdscp_list, size_t n_qdscp, |
dd8cd4b4 SH |
877 | enum ofputil_port_config config, |
878 | enum ofputil_port_state state, bool is_tunnel, | |
9d189a50 | 879 | bool may_enable) |
46c88433 | 880 | { |
55954f6e | 881 | size_t i; |
84f0f298 RW |
882 | struct xport *xport; |
883 | ||
884 | ovs_assert(new_xcfg); | |
46c88433 | 885 | |
84f0f298 | 886 | xport = xport_lookup(new_xcfg, ofport); |
46c88433 EJ |
887 | if (!xport) { |
888 | xport = xzalloc(sizeof *xport); | |
889 | xport->ofport = ofport; | |
84f0f298 | 890 | xport->xbridge = xbridge_lookup(new_xcfg, ofproto); |
46c88433 EJ |
891 | xport->ofp_port = ofp_port; |
892 | ||
84f0f298 | 893 | xlate_xport_init(new_xcfg, xport); |
46c88433 EJ |
894 | } |
895 | ||
896 | ovs_assert(xport->ofp_port == ofp_port); | |
897 | ||
0477baa9 DF |
898 | xlate_xport_set(xport, odp_port, netdev, cfm, bfd, lldp, |
899 | stp_port_no, rstp_port, config, state, is_tunnel, | |
900 | may_enable); | |
46c88433 EJ |
901 | |
902 | if (xport->peer) { | |
903 | xport->peer->peer = NULL; | |
904 | } | |
84f0f298 | 905 | xport->peer = xport_lookup(new_xcfg, peer); |
46c88433 EJ |
906 | if (xport->peer) { |
907 | xport->peer->peer = xport; | |
908 | } | |
909 | ||
910 | if (xport->xbundle) { | |
911 | list_remove(&xport->bundle_node); | |
912 | } | |
84f0f298 | 913 | xport->xbundle = xbundle_lookup(new_xcfg, ofbundle); |
46c88433 EJ |
914 | if (xport->xbundle) { |
915 | list_insert(&xport->xbundle->xports, &xport->bundle_node); | |
916 | } | |
55954f6e EJ |
917 | |
918 | clear_skb_priorities(xport); | |
919 | for (i = 0; i < n_qdscp; i++) { | |
920 | struct skb_priority_to_dscp *pdscp; | |
921 | uint32_t skb_priority; | |
922 | ||
89a8a7f0 EJ |
923 | if (dpif_queue_to_priority(xport->xbridge->dpif, qdscp_list[i].queue, |
924 | &skb_priority)) { | |
55954f6e EJ |
925 | continue; |
926 | } | |
927 | ||
928 | pdscp = xmalloc(sizeof *pdscp); | |
929 | pdscp->skb_priority = skb_priority; | |
930 | pdscp->dscp = (qdscp_list[i].dscp << 2) & IP_DSCP_MASK; | |
931 | hmap_insert(&xport->skb_priorities, &pdscp->hmap_node, | |
932 | hash_int(pdscp->skb_priority, 0)); | |
933 | } | |
46c88433 EJ |
934 | } |
935 | ||
84f0f298 RW |
936 | static void |
937 | xlate_xport_remove(struct xlate_cfg *xcfg, struct xport *xport) | |
46c88433 | 938 | { |
46c88433 EJ |
939 | if (!xport) { |
940 | return; | |
941 | } | |
942 | ||
943 | if (xport->peer) { | |
944 | xport->peer->peer = NULL; | |
945 | xport->peer = NULL; | |
946 | } | |
947 | ||
e621a12d EJ |
948 | if (xport->xbundle) { |
949 | list_remove(&xport->bundle_node); | |
950 | } | |
951 | ||
55954f6e EJ |
952 | clear_skb_priorities(xport); |
953 | hmap_destroy(&xport->skb_priorities); | |
954 | ||
84f0f298 | 955 | hmap_remove(&xcfg->xports, &xport->hmap_node); |
46c88433 EJ |
956 | hmap_remove(&xport->xbridge->xports, &xport->ofp_node); |
957 | ||
958 | netdev_close(xport->netdev); | |
f025bcb7 | 959 | rstp_port_unref(xport->rstp_port); |
46c88433 EJ |
960 | cfm_unref(xport->cfm); |
961 | bfd_unref(xport->bfd); | |
0477baa9 | 962 | lldp_unref(xport->lldp); |
46c88433 EJ |
963 | free(xport); |
964 | } | |
965 | ||
84f0f298 RW |
966 | void |
967 | xlate_ofport_remove(struct ofport_dpif *ofport) | |
968 | { | |
969 | struct xport *xport; | |
970 | ||
971 | ovs_assert(new_xcfg); | |
972 | ||
973 | xport = xport_lookup(new_xcfg, ofport); | |
974 | xlate_xport_remove(new_xcfg, xport); | |
975 | } | |
976 | ||
ef377a58 JR |
977 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively) |
978 | * returns the corresponding struct xport, or NULL if none is found. */ | |
979 | static struct xport * | |
980 | xlate_lookup_xport(const struct dpif_backer *backer, const struct flow *flow) | |
981 | { | |
982 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
983 | ||
984 | return xport_lookup(xcfg, tnl_port_should_receive(flow) | |
985 | ? tnl_port_receive(flow) | |
986 | : odp_port_to_ofport(backer, flow->in_port.odp_port)); | |
987 | } | |
988 | ||
989 | static struct ofproto_dpif * | |
990 | xlate_lookup_ofproto_(const struct dpif_backer *backer, const struct flow *flow, | |
991 | ofp_port_t *ofp_in_port, const struct xport **xportp) | |
992 | { | |
f9038ef6 AW |
993 | struct ofproto_dpif *recv_ofproto = NULL; |
994 | struct ofproto_dpif *recirc_ofproto = NULL; | |
ef377a58 | 995 | const struct xport *xport; |
f9038ef6 | 996 | ofp_port_t in_port = OFPP_NONE; |
ef377a58 JR |
997 | |
998 | *xportp = xport = xlate_lookup_xport(backer, flow); | |
999 | ||
1000 | if (xport) { | |
f9038ef6 AW |
1001 | recv_ofproto = xport->xbridge->ofproto; |
1002 | in_port = xport->ofp_port; | |
1003 | } | |
1004 | ||
1005 | /* When recirc_id is set in 'flow', checks whether the ofproto_dpif that | |
1006 | * corresponds to the recirc_id is same as the receiving bridge. If they | |
1007 | * are the same, uses the 'recv_ofproto' and keeps the 'ofp_in_port' as | |
1008 | * assigned. Otherwise, uses the 'recirc_ofproto' that owns recirc_id and | |
1009 | * assigns OFPP_NONE to 'ofp_in_port'. Doing this is in that, the | |
1010 | * recirculated flow must be processced by the ofproto which originates | |
1011 | * the recirculation, and as bridges can only see their own ports, the | |
1012 | * in_port of the 'recv_ofproto' should not be passed to the | |
1013 | * 'recirc_ofproto'. | |
1014 | * | |
1015 | * Admittedly, setting the 'ofp_in_port' to OFPP_NONE limits the | |
1016 | * 'recirc_ofproto' from meaningfully matching on in_port of recirculated | |
1017 | * flow, and should be fixed in the near future. | |
1018 | * | |
1019 | * TODO: Restore the original patch port. | |
1020 | */ | |
1021 | if (recv_ofproto && flow->recirc_id) { | |
1022 | recirc_ofproto = ofproto_dpif_recirc_get_ofproto(backer, | |
1023 | flow->recirc_id); | |
1024 | if (recv_ofproto != recirc_ofproto) { | |
1025 | *xportp = xport = NULL; | |
1026 | in_port = OFPP_NONE; | |
ef377a58 | 1027 | } |
ef377a58 JR |
1028 | } |
1029 | ||
f9038ef6 AW |
1030 | if (ofp_in_port) { |
1031 | *ofp_in_port = in_port; | |
1032 | } | |
1033 | ||
1034 | return xport ? recv_ofproto : recirc_ofproto; | |
ef377a58 JR |
1035 | } |
1036 | ||
1037 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively) | |
1038 | * returns the corresponding struct ofproto_dpif and OpenFlow port number. */ | |
1039 | struct ofproto_dpif * | |
1040 | xlate_lookup_ofproto(const struct dpif_backer *backer, const struct flow *flow, | |
1041 | ofp_port_t *ofp_in_port) | |
1042 | { | |
1043 | const struct xport *xport; | |
1044 | ||
1045 | return xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); | |
1046 | } | |
1047 | ||
cc377352 | 1048 | /* Given a datapath and flow metadata ('backer', and 'flow' respectively), |
ef377a58 | 1049 | * optionally populates 'ofproto' with the ofproto_dpif, 'ofp_in_port' with the |
cc377352 | 1050 | * openflow in_port, and 'ipfix', 'sflow', and 'netflow' with the appropriate |
dcc2c6cd JR |
1051 | * handles for those protocols if they're enabled. Caller may use the returned |
1052 | * pointers until quiescing, for longer term use additional references must | |
1053 | * be taken. | |
8449c4d6 | 1054 | * |
f9038ef6 | 1055 | * Returns 0 if successful, ENODEV if the parsed flow has no associated ofproto. |
ef377a58 | 1056 | */ |
8449c4d6 | 1057 | int |
5c476ea3 JR |
1058 | xlate_lookup(const struct dpif_backer *backer, const struct flow *flow, |
1059 | struct ofproto_dpif **ofprotop, struct dpif_ipfix **ipfix, | |
1060 | struct dpif_sflow **sflow, struct netflow **netflow, | |
1061 | ofp_port_t *ofp_in_port) | |
8449c4d6 | 1062 | { |
ef377a58 | 1063 | struct ofproto_dpif *ofproto; |
84f0f298 | 1064 | const struct xport *xport; |
8449c4d6 | 1065 | |
ef377a58 | 1066 | ofproto = xlate_lookup_ofproto_(backer, flow, ofp_in_port, &xport); |
8449c4d6 | 1067 | |
f9038ef6 | 1068 | if (!ofproto) { |
cc377352 | 1069 | return ENODEV; |
8449c4d6 | 1070 | } |
8449c4d6 | 1071 | |
ef377a58 JR |
1072 | if (ofprotop) { |
1073 | *ofprotop = ofproto; | |
8449c4d6 EJ |
1074 | } |
1075 | ||
1dfdb9b3 | 1076 | if (ipfix) { |
f9038ef6 | 1077 | *ipfix = xport ? xport->xbridge->ipfix : NULL; |
1dfdb9b3 EJ |
1078 | } |
1079 | ||
1080 | if (sflow) { | |
f9038ef6 | 1081 | *sflow = xport ? xport->xbridge->sflow : NULL; |
1dfdb9b3 EJ |
1082 | } |
1083 | ||
1084 | if (netflow) { | |
f9038ef6 | 1085 | *netflow = xport ? xport->xbridge->netflow : NULL; |
1dfdb9b3 | 1086 | } |
f9038ef6 | 1087 | |
cc377352 | 1088 | return 0; |
8449c4d6 EJ |
1089 | } |
1090 | ||
46c88433 | 1091 | static struct xbridge * |
84f0f298 | 1092 | xbridge_lookup(struct xlate_cfg *xcfg, const struct ofproto_dpif *ofproto) |
46c88433 | 1093 | { |
84f0f298 | 1094 | struct hmap *xbridges; |
46c88433 EJ |
1095 | struct xbridge *xbridge; |
1096 | ||
84f0f298 | 1097 | if (!ofproto || !xcfg) { |
5e6af486 EJ |
1098 | return NULL; |
1099 | } | |
1100 | ||
84f0f298 RW |
1101 | xbridges = &xcfg->xbridges; |
1102 | ||
46c88433 | 1103 | HMAP_FOR_EACH_IN_BUCKET (xbridge, hmap_node, hash_pointer(ofproto, 0), |
84f0f298 | 1104 | xbridges) { |
46c88433 EJ |
1105 | if (xbridge->ofproto == ofproto) { |
1106 | return xbridge; | |
1107 | } | |
1108 | } | |
1109 | return NULL; | |
1110 | } | |
1111 | ||
1112 | static struct xbundle * | |
84f0f298 | 1113 | xbundle_lookup(struct xlate_cfg *xcfg, const struct ofbundle *ofbundle) |
46c88433 | 1114 | { |
84f0f298 | 1115 | struct hmap *xbundles; |
46c88433 EJ |
1116 | struct xbundle *xbundle; |
1117 | ||
84f0f298 | 1118 | if (!ofbundle || !xcfg) { |
5e6af486 EJ |
1119 | return NULL; |
1120 | } | |
1121 | ||
84f0f298 RW |
1122 | xbundles = &xcfg->xbundles; |
1123 | ||
46c88433 | 1124 | HMAP_FOR_EACH_IN_BUCKET (xbundle, hmap_node, hash_pointer(ofbundle, 0), |
84f0f298 | 1125 | xbundles) { |
46c88433 EJ |
1126 | if (xbundle->ofbundle == ofbundle) { |
1127 | return xbundle; | |
1128 | } | |
1129 | } | |
1130 | return NULL; | |
1131 | } | |
1132 | ||
1133 | static struct xport * | |
84f0f298 | 1134 | xport_lookup(struct xlate_cfg *xcfg, const struct ofport_dpif *ofport) |
46c88433 | 1135 | { |
84f0f298 | 1136 | struct hmap *xports; |
46c88433 EJ |
1137 | struct xport *xport; |
1138 | ||
84f0f298 | 1139 | if (!ofport || !xcfg) { |
5e6af486 EJ |
1140 | return NULL; |
1141 | } | |
1142 | ||
84f0f298 RW |
1143 | xports = &xcfg->xports; |
1144 | ||
46c88433 | 1145 | HMAP_FOR_EACH_IN_BUCKET (xport, hmap_node, hash_pointer(ofport, 0), |
84f0f298 | 1146 | xports) { |
46c88433 EJ |
1147 | if (xport->ofport == ofport) { |
1148 | return xport; | |
1149 | } | |
1150 | } | |
1151 | return NULL; | |
1152 | } | |
1153 | ||
40085e56 EJ |
1154 | static struct stp_port * |
1155 | xport_get_stp_port(const struct xport *xport) | |
1156 | { | |
92cf817b | 1157 | return xport->xbridge->stp && xport->stp_port_no != -1 |
40085e56 EJ |
1158 | ? stp_get_port(xport->xbridge->stp, xport->stp_port_no) |
1159 | : NULL; | |
1160 | } | |
9d189a50 | 1161 | |
0d1cee12 | 1162 | static bool |
9d189a50 EJ |
1163 | xport_stp_learn_state(const struct xport *xport) |
1164 | { | |
40085e56 | 1165 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1166 | return sp |
1167 | ? stp_learn_in_state(stp_port_get_state(sp)) | |
1168 | : true; | |
9d189a50 EJ |
1169 | } |
1170 | ||
1171 | static bool | |
1172 | xport_stp_forward_state(const struct xport *xport) | |
1173 | { | |
40085e56 | 1174 | struct stp_port *sp = xport_get_stp_port(xport); |
4b5f1996 DV |
1175 | return sp |
1176 | ? stp_forward_in_state(stp_port_get_state(sp)) | |
1177 | : true; | |
9d189a50 EJ |
1178 | } |
1179 | ||
0d1cee12 | 1180 | static bool |
bacdb85a | 1181 | xport_stp_should_forward_bpdu(const struct xport *xport) |
0d1cee12 K |
1182 | { |
1183 | struct stp_port *sp = xport_get_stp_port(xport); | |
bacdb85a | 1184 | return stp_should_forward_bpdu(sp ? stp_port_get_state(sp) : STP_DISABLED); |
0d1cee12 K |
1185 | } |
1186 | ||
9d189a50 EJ |
1187 | /* Returns true if STP should process 'flow'. Sets fields in 'wc' that |
1188 | * were used to make the determination.*/ | |
1189 | static bool | |
1190 | stp_should_process_flow(const struct flow *flow, struct flow_wildcards *wc) | |
1191 | { | |
bbbca389 | 1192 | /* is_stp() also checks dl_type, but dl_type is always set in 'wc'. */ |
9d189a50 | 1193 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
bbbca389 | 1194 | return is_stp(flow); |
9d189a50 EJ |
1195 | } |
1196 | ||
1197 | static void | |
cf62fa4c | 1198 | stp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9d189a50 | 1199 | { |
40085e56 | 1200 | struct stp_port *sp = xport_get_stp_port(xport); |
cf62fa4c PS |
1201 | struct dp_packet payload = *packet; |
1202 | struct eth_header *eth = dp_packet_data(&payload); | |
9d189a50 EJ |
1203 | |
1204 | /* Sink packets on ports that have STP disabled when the bridge has | |
1205 | * STP enabled. */ | |
1206 | if (!sp || stp_port_get_state(sp) == STP_DISABLED) { | |
1207 | return; | |
1208 | } | |
1209 | ||
1210 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1211 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1212 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9d189a50 EJ |
1213 | } |
1214 | ||
cf62fa4c PS |
1215 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1216 | stp_received_bpdu(sp, dp_packet_data(&payload), dp_packet_size(&payload)); | |
9d189a50 EJ |
1217 | } |
1218 | } | |
1219 | ||
f025bcb7 JR |
1220 | static enum rstp_state |
1221 | xport_get_rstp_port_state(const struct xport *xport) | |
9efd308e | 1222 | { |
f025bcb7 JR |
1223 | return xport->rstp_port |
1224 | ? rstp_port_get_state(xport->rstp_port) | |
1225 | : RSTP_DISABLED; | |
9efd308e DV |
1226 | } |
1227 | ||
1228 | static bool | |
1229 | xport_rstp_learn_state(const struct xport *xport) | |
1230 | { | |
4b5f1996 DV |
1231 | return xport->xbridge->rstp && xport->rstp_port |
1232 | ? rstp_learn_in_state(xport_get_rstp_port_state(xport)) | |
1233 | : true; | |
9efd308e DV |
1234 | } |
1235 | ||
1236 | static bool | |
1237 | xport_rstp_forward_state(const struct xport *xport) | |
1238 | { | |
4b5f1996 DV |
1239 | return xport->xbridge->rstp && xport->rstp_port |
1240 | ? rstp_forward_in_state(xport_get_rstp_port_state(xport)) | |
1241 | : true; | |
9efd308e DV |
1242 | } |
1243 | ||
1244 | static bool | |
1245 | xport_rstp_should_manage_bpdu(const struct xport *xport) | |
1246 | { | |
f025bcb7 | 1247 | return rstp_should_manage_bpdu(xport_get_rstp_port_state(xport)); |
9efd308e DV |
1248 | } |
1249 | ||
1250 | static void | |
cf62fa4c | 1251 | rstp_process_packet(const struct xport *xport, const struct dp_packet *packet) |
9efd308e | 1252 | { |
cf62fa4c PS |
1253 | struct dp_packet payload = *packet; |
1254 | struct eth_header *eth = dp_packet_data(&payload); | |
9efd308e | 1255 | |
f025bcb7 JR |
1256 | /* Sink packets on ports that have no RSTP. */ |
1257 | if (!xport->rstp_port) { | |
9efd308e DV |
1258 | return; |
1259 | } | |
1260 | ||
1261 | /* Trim off padding on payload. */ | |
cf62fa4c PS |
1262 | if (dp_packet_size(&payload) > ntohs(eth->eth_type) + ETH_HEADER_LEN) { |
1263 | dp_packet_set_size(&payload, ntohs(eth->eth_type) + ETH_HEADER_LEN); | |
9efd308e DV |
1264 | } |
1265 | ||
cf62fa4c PS |
1266 | if (dp_packet_try_pull(&payload, ETH_HEADER_LEN + LLC_HEADER_LEN)) { |
1267 | rstp_port_received_bpdu(xport->rstp_port, dp_packet_data(&payload), | |
1268 | dp_packet_size(&payload)); | |
9efd308e DV |
1269 | } |
1270 | } | |
1271 | ||
46c88433 EJ |
1272 | static struct xport * |
1273 | get_ofp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1274 | { | |
1275 | struct xport *xport; | |
1276 | ||
1277 | HMAP_FOR_EACH_IN_BUCKET (xport, ofp_node, hash_ofp_port(ofp_port), | |
1278 | &xbridge->xports) { | |
1279 | if (xport->ofp_port == ofp_port) { | |
1280 | return xport; | |
1281 | } | |
1282 | } | |
1283 | return NULL; | |
1284 | } | |
1285 | ||
1286 | static odp_port_t | |
1287 | ofp_port_to_odp_port(const struct xbridge *xbridge, ofp_port_t ofp_port) | |
1288 | { | |
1289 | const struct xport *xport = get_ofp_port(xbridge, ofp_port); | |
1290 | return xport ? xport->odp_port : ODPP_NONE; | |
1291 | } | |
1292 | ||
dd8cd4b4 SH |
1293 | static bool |
1294 | odp_port_is_alive(const struct xlate_ctx *ctx, ofp_port_t ofp_port) | |
1295 | { | |
086fa873 BP |
1296 | struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1297 | return xport && xport->may_enable; | |
dd8cd4b4 SH |
1298 | } |
1299 | ||
1e684d7d | 1300 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1301 | group_first_live_bucket(const struct xlate_ctx *, const struct group_dpif *, |
1302 | int depth); | |
1303 | ||
1304 | static bool | |
1305 | group_is_alive(const struct xlate_ctx *ctx, uint32_t group_id, int depth) | |
1306 | { | |
1307 | struct group_dpif *group; | |
dd8cd4b4 | 1308 | |
dc25893e AZ |
1309 | if (group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group)) { |
1310 | struct ofputil_bucket *bucket; | |
dd8cd4b4 | 1311 | |
dc25893e AZ |
1312 | bucket = group_first_live_bucket(ctx, group, depth); |
1313 | group_dpif_unref(group); | |
1314 | return bucket == NULL; | |
1315 | } | |
dd8cd4b4 | 1316 | |
dc25893e | 1317 | return false; |
dd8cd4b4 SH |
1318 | } |
1319 | ||
1320 | #define MAX_LIVENESS_RECURSION 128 /* Arbitrary limit */ | |
1321 | ||
1322 | static bool | |
1323 | bucket_is_alive(const struct xlate_ctx *ctx, | |
1e684d7d | 1324 | struct ofputil_bucket *bucket, int depth) |
dd8cd4b4 SH |
1325 | { |
1326 | if (depth >= MAX_LIVENESS_RECURSION) { | |
1327 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
1328 | ||
1329 | VLOG_WARN_RL(&rl, "bucket chaining exceeded %d links", | |
1330 | MAX_LIVENESS_RECURSION); | |
1331 | return false; | |
1332 | } | |
1333 | ||
fdb1999b AZ |
1334 | return (!ofputil_bucket_has_liveness(bucket) |
1335 | || (bucket->watch_port != OFPP_ANY | |
1336 | && odp_port_is_alive(ctx, bucket->watch_port)) | |
1337 | || (bucket->watch_group != OFPG_ANY | |
1338 | && group_is_alive(ctx, bucket->watch_group, depth + 1))); | |
dd8cd4b4 SH |
1339 | } |
1340 | ||
1e684d7d | 1341 | static struct ofputil_bucket * |
dd8cd4b4 SH |
1342 | group_first_live_bucket(const struct xlate_ctx *ctx, |
1343 | const struct group_dpif *group, int depth) | |
1344 | { | |
1345 | struct ofputil_bucket *bucket; | |
ca6ba700 | 1346 | const struct ovs_list *buckets; |
dd8cd4b4 SH |
1347 | |
1348 | group_dpif_get_buckets(group, &buckets); | |
1349 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1350 | if (bucket_is_alive(ctx, bucket, depth)) { | |
1351 | return bucket; | |
1352 | } | |
1353 | } | |
1354 | ||
1355 | return NULL; | |
1356 | } | |
1357 | ||
1e684d7d | 1358 | static struct ofputil_bucket * |
fe7e5749 SH |
1359 | group_best_live_bucket(const struct xlate_ctx *ctx, |
1360 | const struct group_dpif *group, | |
1361 | uint32_t basis) | |
1362 | { | |
1e684d7d | 1363 | struct ofputil_bucket *best_bucket = NULL; |
fe7e5749 SH |
1364 | uint32_t best_score = 0; |
1365 | int i = 0; | |
1366 | ||
1e684d7d | 1367 | struct ofputil_bucket *bucket; |
ca6ba700 | 1368 | const struct ovs_list *buckets; |
fe7e5749 SH |
1369 | |
1370 | group_dpif_get_buckets(group, &buckets); | |
1371 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
1372 | if (bucket_is_alive(ctx, bucket, 0)) { | |
7cb279c2 | 1373 | uint32_t score = (hash_int(i, basis) & 0xffff) * bucket->weight; |
fe7e5749 SH |
1374 | if (score >= best_score) { |
1375 | best_bucket = bucket; | |
1376 | best_score = score; | |
1377 | } | |
1378 | } | |
1379 | i++; | |
1380 | } | |
1381 | ||
1382 | return best_bucket; | |
1383 | } | |
1384 | ||
9583bc14 | 1385 | static bool |
46c88433 | 1386 | xbundle_trunks_vlan(const struct xbundle *bundle, uint16_t vlan) |
9583bc14 EJ |
1387 | { |
1388 | return (bundle->vlan_mode != PORT_VLAN_ACCESS | |
1389 | && (!bundle->trunks || bitmap_is_set(bundle->trunks, vlan))); | |
1390 | } | |
1391 | ||
1392 | static bool | |
46c88433 EJ |
1393 | xbundle_includes_vlan(const struct xbundle *xbundle, uint16_t vlan) |
1394 | { | |
1395 | return vlan == xbundle->vlan || xbundle_trunks_vlan(xbundle, vlan); | |
1396 | } | |
1397 | ||
1398 | static mirror_mask_t | |
1399 | xbundle_mirror_out(const struct xbridge *xbridge, struct xbundle *xbundle) | |
1400 | { | |
1401 | return xbundle != &ofpp_none_bundle | |
1402 | ? mirror_bundle_out(xbridge->mbridge, xbundle->ofbundle) | |
1403 | : 0; | |
1404 | } | |
1405 | ||
1406 | static mirror_mask_t | |
1407 | xbundle_mirror_src(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1408 | { |
46c88433 EJ |
1409 | return xbundle != &ofpp_none_bundle |
1410 | ? mirror_bundle_src(xbridge->mbridge, xbundle->ofbundle) | |
1411 | : 0; | |
9583bc14 EJ |
1412 | } |
1413 | ||
46c88433 EJ |
1414 | static mirror_mask_t |
1415 | xbundle_mirror_dst(const struct xbridge *xbridge, struct xbundle *xbundle) | |
9583bc14 | 1416 | { |
46c88433 EJ |
1417 | return xbundle != &ofpp_none_bundle |
1418 | ? mirror_bundle_dst(xbridge->mbridge, xbundle->ofbundle) | |
1419 | : 0; | |
1420 | } | |
1421 | ||
1422 | static struct xbundle * | |
1423 | lookup_input_bundle(const struct xbridge *xbridge, ofp_port_t in_port, | |
1424 | bool warn, struct xport **in_xportp) | |
1425 | { | |
1426 | struct xport *xport; | |
9583bc14 EJ |
1427 | |
1428 | /* Find the port and bundle for the received packet. */ | |
46c88433 EJ |
1429 | xport = get_ofp_port(xbridge, in_port); |
1430 | if (in_xportp) { | |
1431 | *in_xportp = xport; | |
9583bc14 | 1432 | } |
46c88433 EJ |
1433 | if (xport && xport->xbundle) { |
1434 | return xport->xbundle; | |
9583bc14 EJ |
1435 | } |
1436 | ||
6362203b YT |
1437 | /* Special-case OFPP_NONE (OF1.0) and OFPP_CONTROLLER (OF1.1+), |
1438 | * which a controller may use as the ingress port for traffic that | |
1439 | * it is sourcing. */ | |
1440 | if (in_port == OFPP_CONTROLLER || in_port == OFPP_NONE) { | |
9583bc14 EJ |
1441 | return &ofpp_none_bundle; |
1442 | } | |
1443 | ||
1444 | /* Odd. A few possible reasons here: | |
1445 | * | |
1446 | * - We deleted a port but there are still a few packets queued up | |
1447 | * from it. | |
1448 | * | |
1449 | * - Someone externally added a port (e.g. "ovs-dpctl add-if") that | |
1450 | * we don't know about. | |
1451 | * | |
1452 | * - The ofproto client didn't configure the port as part of a bundle. | |
1453 | * This is particularly likely to happen if a packet was received on the | |
1454 | * port after it was created, but before the client had a chance to | |
1455 | * configure its bundle. | |
1456 | */ | |
1457 | if (warn) { | |
1458 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1459 | ||
1460 | VLOG_WARN_RL(&rl, "bridge %s: received packet on unknown " | |
46c88433 | 1461 | "port %"PRIu16, xbridge->name, in_port); |
9583bc14 EJ |
1462 | } |
1463 | return NULL; | |
1464 | } | |
1465 | ||
1466 | static void | |
1467 | add_mirror_actions(struct xlate_ctx *ctx, const struct flow *orig_flow) | |
1468 | { | |
46c88433 | 1469 | const struct xbridge *xbridge = ctx->xbridge; |
9583bc14 | 1470 | mirror_mask_t mirrors; |
46c88433 | 1471 | struct xbundle *in_xbundle; |
9583bc14 EJ |
1472 | uint16_t vlan; |
1473 | uint16_t vid; | |
cdf5d3a5 EJ |
1474 | |
1475 | mirrors = ctx->xout->mirrors; | |
1476 | ctx->xout->mirrors = 0; | |
9583bc14 | 1477 | |
46c88433 EJ |
1478 | in_xbundle = lookup_input_bundle(xbridge, orig_flow->in_port.ofp_port, |
1479 | ctx->xin->packet != NULL, NULL); | |
1480 | if (!in_xbundle) { | |
9583bc14 EJ |
1481 | return; |
1482 | } | |
46c88433 | 1483 | mirrors |= xbundle_mirror_src(xbridge, in_xbundle); |
9583bc14 EJ |
1484 | |
1485 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 1486 | if (xbundle_mirror_out(xbridge, in_xbundle)) { |
9583bc14 EJ |
1487 | if (ctx->xin->packet != NULL) { |
1488 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
1489 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
1490 | "%s, which is reserved exclusively for mirroring", | |
46c88433 | 1491 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 | 1492 | } |
cc377352 | 1493 | ofpbuf_clear(ctx->xout->odp_actions); |
9583bc14 EJ |
1494 | return; |
1495 | } | |
1496 | ||
1497 | /* Check VLAN. */ | |
1498 | vid = vlan_tci_to_vid(orig_flow->vlan_tci); | |
46c88433 | 1499 | if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { |
9583bc14 EJ |
1500 | return; |
1501 | } | |
46c88433 | 1502 | vlan = input_vid_to_vlan(in_xbundle, vid); |
9583bc14 | 1503 | |
9583bc14 EJ |
1504 | if (!mirrors) { |
1505 | return; | |
1506 | } | |
1507 | ||
1508 | /* Restore the original packet before adding the mirror actions. */ | |
1509 | ctx->xin->flow = *orig_flow; | |
1510 | ||
1511 | while (mirrors) { | |
ec7ceaed EJ |
1512 | mirror_mask_t dup_mirrors; |
1513 | struct ofbundle *out; | |
1514 | unsigned long *vlans; | |
1515 | bool vlan_mirrored; | |
1516 | bool has_mirror; | |
1517 | int out_vlan; | |
1518 | ||
4b6ab2b0 | 1519 | has_mirror = mirror_get(xbridge->mbridge, raw_ctz(mirrors), |
ec7ceaed EJ |
1520 | &vlans, &dup_mirrors, &out, &out_vlan); |
1521 | ovs_assert(has_mirror); | |
1522 | ||
1523 | if (vlans) { | |
9583bc14 EJ |
1524 | ctx->xout->wc.masks.vlan_tci |= htons(VLAN_CFI | VLAN_VID_MASK); |
1525 | } | |
ec7ceaed EJ |
1526 | vlan_mirrored = !vlans || bitmap_is_set(vlans, vlan); |
1527 | free(vlans); | |
9583bc14 | 1528 | |
ec7ceaed | 1529 | if (!vlan_mirrored) { |
9583bc14 EJ |
1530 | mirrors = zero_rightmost_1bit(mirrors); |
1531 | continue; | |
1532 | } | |
1533 | ||
ec7ceaed EJ |
1534 | mirrors &= ~dup_mirrors; |
1535 | ctx->xout->mirrors |= dup_mirrors; | |
1536 | if (out) { | |
84f0f298 RW |
1537 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
1538 | struct xbundle *out_xbundle = xbundle_lookup(xcfg, out); | |
46c88433 EJ |
1539 | if (out_xbundle) { |
1540 | output_normal(ctx, out_xbundle, vlan); | |
1541 | } | |
ec7ceaed | 1542 | } else if (vlan != out_vlan |
9583bc14 | 1543 | && !eth_addr_is_reserved(orig_flow->dl_dst)) { |
46c88433 | 1544 | struct xbundle *xbundle; |
9583bc14 | 1545 | |
46c88433 EJ |
1546 | LIST_FOR_EACH (xbundle, list_node, &xbridge->xbundles) { |
1547 | if (xbundle_includes_vlan(xbundle, out_vlan) | |
1548 | && !xbundle_mirror_out(xbridge, xbundle)) { | |
1549 | output_normal(ctx, xbundle, out_vlan); | |
9583bc14 EJ |
1550 | } |
1551 | } | |
1552 | } | |
1553 | } | |
1554 | } | |
1555 | ||
1556 | /* Given 'vid', the VID obtained from the 802.1Q header that was received as | |
46c88433 | 1557 | * part of a packet (specify 0 if there was no 802.1Q header), and 'in_xbundle', |
9583bc14 EJ |
1558 | * the bundle on which the packet was received, returns the VLAN to which the |
1559 | * packet belongs. | |
1560 | * | |
1561 | * Both 'vid' and the return value are in the range 0...4095. */ | |
1562 | static uint16_t | |
46c88433 | 1563 | input_vid_to_vlan(const struct xbundle *in_xbundle, uint16_t vid) |
9583bc14 | 1564 | { |
46c88433 | 1565 | switch (in_xbundle->vlan_mode) { |
9583bc14 | 1566 | case PORT_VLAN_ACCESS: |
46c88433 | 1567 | return in_xbundle->vlan; |
9583bc14 EJ |
1568 | break; |
1569 | ||
1570 | case PORT_VLAN_TRUNK: | |
1571 | return vid; | |
1572 | ||
1573 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1574 | case PORT_VLAN_NATIVE_TAGGED: | |
46c88433 | 1575 | return vid ? vid : in_xbundle->vlan; |
9583bc14 EJ |
1576 | |
1577 | default: | |
428b2edd | 1578 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1579 | } |
1580 | } | |
1581 | ||
46c88433 | 1582 | /* Checks whether a packet with the given 'vid' may ingress on 'in_xbundle'. |
9583bc14 EJ |
1583 | * If so, returns true. Otherwise, returns false and, if 'warn' is true, logs |
1584 | * a warning. | |
1585 | * | |
1586 | * 'vid' should be the VID obtained from the 802.1Q header that was received as | |
1587 | * part of a packet (specify 0 if there was no 802.1Q header), in the range | |
1588 | * 0...4095. */ | |
1589 | static bool | |
46c88433 | 1590 | input_vid_is_valid(uint16_t vid, struct xbundle *in_xbundle, bool warn) |
9583bc14 EJ |
1591 | { |
1592 | /* Allow any VID on the OFPP_NONE port. */ | |
46c88433 | 1593 | if (in_xbundle == &ofpp_none_bundle) { |
9583bc14 EJ |
1594 | return true; |
1595 | } | |
1596 | ||
46c88433 | 1597 | switch (in_xbundle->vlan_mode) { |
9583bc14 EJ |
1598 | case PORT_VLAN_ACCESS: |
1599 | if (vid) { | |
1600 | if (warn) { | |
1601 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1602 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" tagged " |
9583bc14 | 1603 | "packet received on port %s configured as VLAN " |
46c88433 EJ |
1604 | "%"PRIu16" access port", vid, in_xbundle->name, |
1605 | in_xbundle->vlan); | |
9583bc14 EJ |
1606 | } |
1607 | return false; | |
1608 | } | |
1609 | return true; | |
1610 | ||
1611 | case PORT_VLAN_NATIVE_UNTAGGED: | |
1612 | case PORT_VLAN_NATIVE_TAGGED: | |
1613 | if (!vid) { | |
1614 | /* Port must always carry its native VLAN. */ | |
1615 | return true; | |
1616 | } | |
1617 | /* Fall through. */ | |
1618 | case PORT_VLAN_TRUNK: | |
46c88433 | 1619 | if (!xbundle_includes_vlan(in_xbundle, vid)) { |
9583bc14 EJ |
1620 | if (warn) { |
1621 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
46c88433 | 1622 | VLOG_WARN_RL(&rl, "dropping VLAN %"PRIu16" packet " |
9583bc14 | 1623 | "received on port %s not configured for trunking " |
46c88433 | 1624 | "VLAN %"PRIu16, vid, in_xbundle->name, vid); |
9583bc14 EJ |
1625 | } |
1626 | return false; | |
1627 | } | |
1628 | return true; | |
1629 | ||
1630 | default: | |
428b2edd | 1631 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1632 | } |
1633 | ||
1634 | } | |
1635 | ||
1636 | /* Given 'vlan', the VLAN that a packet belongs to, and | |
46c88433 | 1637 | * 'out_xbundle', a bundle on which the packet is to be output, returns the VID |
9583bc14 EJ |
1638 | * that should be included in the 802.1Q header. (If the return value is 0, |
1639 | * then the 802.1Q header should only be included in the packet if there is a | |
1640 | * nonzero PCP.) | |
1641 | * | |
1642 | * Both 'vlan' and the return value are in the range 0...4095. */ | |
1643 | static uint16_t | |
46c88433 | 1644 | output_vlan_to_vid(const struct xbundle *out_xbundle, uint16_t vlan) |
9583bc14 | 1645 | { |
46c88433 | 1646 | switch (out_xbundle->vlan_mode) { |
9583bc14 EJ |
1647 | case PORT_VLAN_ACCESS: |
1648 | return 0; | |
1649 | ||
1650 | case PORT_VLAN_TRUNK: | |
1651 | case PORT_VLAN_NATIVE_TAGGED: | |
1652 | return vlan; | |
1653 | ||
1654 | case PORT_VLAN_NATIVE_UNTAGGED: | |
46c88433 | 1655 | return vlan == out_xbundle->vlan ? 0 : vlan; |
9583bc14 EJ |
1656 | |
1657 | default: | |
428b2edd | 1658 | OVS_NOT_REACHED(); |
9583bc14 EJ |
1659 | } |
1660 | } | |
1661 | ||
1662 | static void | |
46c88433 | 1663 | output_normal(struct xlate_ctx *ctx, const struct xbundle *out_xbundle, |
9583bc14 EJ |
1664 | uint16_t vlan) |
1665 | { | |
33bf9176 | 1666 | ovs_be16 *flow_tci = &ctx->xin->flow.vlan_tci; |
9583bc14 EJ |
1667 | uint16_t vid; |
1668 | ovs_be16 tci, old_tci; | |
46c88433 | 1669 | struct xport *xport; |
9583bc14 | 1670 | |
46c88433 EJ |
1671 | vid = output_vlan_to_vid(out_xbundle, vlan); |
1672 | if (list_is_empty(&out_xbundle->xports)) { | |
1673 | /* Partially configured bundle with no slaves. Drop the packet. */ | |
1674 | return; | |
1675 | } else if (!out_xbundle->bond) { | |
92c08f09 | 1676 | ctx->use_recirc = false; |
46c88433 EJ |
1677 | xport = CONTAINER_OF(list_front(&out_xbundle->xports), struct xport, |
1678 | bundle_node); | |
9583bc14 | 1679 | } else { |
84f0f298 | 1680 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
54ecb5a2 | 1681 | struct flow_wildcards *wc = &ctx->xout->wc; |
84f0f298 RW |
1682 | struct xlate_recirc *xr = &ctx->recirc; |
1683 | struct ofport_dpif *ofport; | |
adcf00ba AZ |
1684 | |
1685 | if (ctx->xbridge->enable_recirc) { | |
92c08f09 | 1686 | ctx->use_recirc = bond_may_recirc( |
62ac1f20 | 1687 | out_xbundle->bond, &xr->recirc_id, &xr->hash_basis); |
adcf00ba | 1688 | |
92c08f09 | 1689 | if (ctx->use_recirc) { |
adcf00ba | 1690 | /* Only TCP mode uses recirculation. */ |
347bf289 | 1691 | xr->hash_alg = OVS_HASH_ALG_L4; |
adcf00ba | 1692 | bond_update_post_recirc_rules(out_xbundle->bond, false); |
54ecb5a2 AZ |
1693 | |
1694 | /* Recirculation does not require unmasking hash fields. */ | |
1695 | wc = NULL; | |
adcf00ba AZ |
1696 | } |
1697 | } | |
46c88433 | 1698 | |
54ecb5a2 AZ |
1699 | ofport = bond_choose_output_slave(out_xbundle->bond, |
1700 | &ctx->xin->flow, wc, vid); | |
84f0f298 | 1701 | xport = xport_lookup(xcfg, ofport); |
46c88433 EJ |
1702 | |
1703 | if (!xport) { | |
9583bc14 EJ |
1704 | /* No slaves enabled, so drop packet. */ |
1705 | return; | |
1706 | } | |
d6fc5f57 | 1707 | |
b256dc52 JS |
1708 | /* If ctx->xout->use_recirc is set, the main thread will handle stats |
1709 | * accounting for this bond. */ | |
92c08f09 | 1710 | if (!ctx->use_recirc) { |
b256dc52 JS |
1711 | if (ctx->xin->resubmit_stats) { |
1712 | bond_account(out_xbundle->bond, &ctx->xin->flow, vid, | |
1713 | ctx->xin->resubmit_stats->n_bytes); | |
1714 | } | |
1715 | if (ctx->xin->xcache) { | |
1716 | struct xc_entry *entry; | |
1717 | struct flow *flow; | |
1718 | ||
1719 | flow = &ctx->xin->flow; | |
1720 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_BOND); | |
1721 | entry->u.bond.bond = bond_ref(out_xbundle->bond); | |
1722 | entry->u.bond.flow = xmemdup(flow, sizeof *flow); | |
1723 | entry->u.bond.vid = vid; | |
1724 | } | |
d6fc5f57 | 1725 | } |
9583bc14 EJ |
1726 | } |
1727 | ||
33bf9176 | 1728 | old_tci = *flow_tci; |
9583bc14 | 1729 | tci = htons(vid); |
46c88433 | 1730 | if (tci || out_xbundle->use_priority_tags) { |
33bf9176 | 1731 | tci |= *flow_tci & htons(VLAN_PCP_MASK); |
9583bc14 EJ |
1732 | if (tci) { |
1733 | tci |= htons(VLAN_CFI); | |
1734 | } | |
1735 | } | |
33bf9176 | 1736 | *flow_tci = tci; |
9583bc14 | 1737 | |
46c88433 | 1738 | compose_output_action(ctx, xport->ofp_port); |
33bf9176 | 1739 | *flow_tci = old_tci; |
9583bc14 EJ |
1740 | } |
1741 | ||
1742 | /* A VM broadcasts a gratuitous ARP to indicate that it has resumed after | |
1743 | * migration. Older Citrix-patched Linux DomU used gratuitous ARP replies to | |
1744 | * indicate this; newer upstream kernels use gratuitous ARP requests. */ | |
1745 | static bool | |
1746 | is_gratuitous_arp(const struct flow *flow, struct flow_wildcards *wc) | |
1747 | { | |
1748 | if (flow->dl_type != htons(ETH_TYPE_ARP)) { | |
1749 | return false; | |
1750 | } | |
1751 | ||
1752 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1753 | if (!eth_addr_is_broadcast(flow->dl_dst)) { | |
1754 | return false; | |
1755 | } | |
1756 | ||
1757 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
1758 | if (flow->nw_proto == ARP_OP_REPLY) { | |
1759 | return true; | |
1760 | } else if (flow->nw_proto == ARP_OP_REQUEST) { | |
1761 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
1762 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
1763 | ||
1764 | return flow->nw_src == flow->nw_dst; | |
1765 | } else { | |
1766 | return false; | |
1767 | } | |
1768 | } | |
1769 | ||
ff69c24a FL |
1770 | /* Determines whether packets in 'flow' within 'xbridge' should be forwarded or |
1771 | * dropped. Returns true if they may be forwarded, false if they should be | |
1772 | * dropped. | |
1773 | * | |
1774 | * 'in_port' must be the xport that corresponds to flow->in_port. | |
1775 | * 'in_port' must be part of a bundle (e.g. in_port->bundle must be nonnull). | |
1776 | * | |
1777 | * 'vlan' must be the VLAN that corresponds to flow->vlan_tci on 'in_port', as | |
1778 | * returned by input_vid_to_vlan(). It must be a valid VLAN for 'in_port', as | |
1779 | * checked by input_vid_is_valid(). | |
1780 | * | |
1781 | * May also add tags to '*tags', although the current implementation only does | |
1782 | * so in one special case. | |
1783 | */ | |
1784 | static bool | |
1785 | is_admissible(struct xlate_ctx *ctx, struct xport *in_port, | |
1786 | uint16_t vlan) | |
1787 | { | |
1788 | struct xbundle *in_xbundle = in_port->xbundle; | |
1789 | const struct xbridge *xbridge = ctx->xbridge; | |
1790 | struct flow *flow = &ctx->xin->flow; | |
1791 | ||
1792 | /* Drop frames for reserved multicast addresses | |
1793 | * only if forward_bpdu option is absent. */ | |
1794 | if (!xbridge->forward_bpdu && eth_addr_is_reserved(flow->dl_dst)) { | |
1795 | xlate_report(ctx, "packet has reserved destination MAC, dropping"); | |
1796 | return false; | |
1797 | } | |
1798 | ||
1799 | if (in_xbundle->bond) { | |
1800 | struct mac_entry *mac; | |
1801 | ||
1802 | switch (bond_check_admissibility(in_xbundle->bond, in_port->ofport, | |
1803 | flow->dl_dst)) { | |
1804 | case BV_ACCEPT: | |
1805 | break; | |
1806 | ||
1807 | case BV_DROP: | |
1808 | xlate_report(ctx, "bonding refused admissibility, dropping"); | |
1809 | return false; | |
1810 | ||
1811 | case BV_DROP_IF_MOVED: | |
1812 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
1813 | mac = mac_learning_lookup(xbridge->ml, flow->dl_src, vlan); | |
9d078ec2 BP |
1814 | if (mac |
1815 | && mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle | |
1816 | && (!is_gratuitous_arp(flow, &ctx->xout->wc) | |
1817 | || mac_entry_is_grat_arp_locked(mac))) { | |
ff69c24a FL |
1818 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
1819 | xlate_report(ctx, "SLB bond thinks this packet looped back, " | |
1820 | "dropping"); | |
1821 | return false; | |
1822 | } | |
1823 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
1824 | break; | |
1825 | } | |
1826 | } | |
1827 | ||
1828 | return true; | |
1829 | } | |
1830 | ||
ee047520 BP |
1831 | /* Checks whether a MAC learning update is necessary for MAC learning table |
1832 | * 'ml' given that a packet matching 'flow' was received on 'in_xbundle' in | |
1833 | * 'vlan'. | |
1834 | * | |
1835 | * Most packets processed through the MAC learning table do not actually | |
1836 | * change it in any way. This function requires only a read lock on the MAC | |
1837 | * learning table, so it is much cheaper in this common case. | |
1838 | * | |
1839 | * Keep the code here synchronized with that in update_learning_table__() | |
1840 | * below. */ | |
1841 | static bool | |
1842 | is_mac_learning_update_needed(const struct mac_learning *ml, | |
1843 | const struct flow *flow, | |
1844 | struct flow_wildcards *wc, | |
1845 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 1846 | OVS_REQ_RDLOCK(ml->rwlock) |
9583bc14 EJ |
1847 | { |
1848 | struct mac_entry *mac; | |
1849 | ||
ee047520 BP |
1850 | if (!mac_learning_may_learn(ml, flow->dl_src, vlan)) { |
1851 | return false; | |
1852 | } | |
1853 | ||
1854 | mac = mac_learning_lookup(ml, flow->dl_src, vlan); | |
1855 | if (!mac || mac_entry_age(ml, mac)) { | |
1856 | return true; | |
9583bc14 EJ |
1857 | } |
1858 | ||
ee047520 BP |
1859 | if (is_gratuitous_arp(flow, wc)) { |
1860 | /* We don't want to learn from gratuitous ARP packets that are | |
1861 | * reflected back over bond slaves so we lock the learning table. */ | |
1862 | if (!in_xbundle->bond) { | |
1863 | return true; | |
1864 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
1865 | return false; | |
1866 | } | |
1867 | } | |
1868 | ||
9d078ec2 | 1869 | return mac_entry_get_port(ml, mac) != in_xbundle->ofbundle; |
ee047520 BP |
1870 | } |
1871 | ||
1872 | ||
1873 | /* Updates MAC learning table 'ml' given that a packet matching 'flow' was | |
1874 | * received on 'in_xbundle' in 'vlan'. | |
1875 | * | |
1876 | * This code repeats all the checks in is_mac_learning_update_needed() because | |
1877 | * the lock was released between there and here and thus the MAC learning state | |
1878 | * could have changed. | |
1879 | * | |
1880 | * Keep the code here synchronized with that in is_mac_learning_update_needed() | |
1881 | * above. */ | |
1882 | static void | |
1883 | update_learning_table__(const struct xbridge *xbridge, | |
1884 | const struct flow *flow, struct flow_wildcards *wc, | |
1885 | int vlan, struct xbundle *in_xbundle) | |
d6fc5f57 | 1886 | OVS_REQ_WRLOCK(xbridge->ml->rwlock) |
ee047520 BP |
1887 | { |
1888 | struct mac_entry *mac; | |
1889 | ||
46c88433 | 1890 | if (!mac_learning_may_learn(xbridge->ml, flow->dl_src, vlan)) { |
ee047520 | 1891 | return; |
9583bc14 EJ |
1892 | } |
1893 | ||
46c88433 | 1894 | mac = mac_learning_insert(xbridge->ml, flow->dl_src, vlan); |
9583bc14 EJ |
1895 | if (is_gratuitous_arp(flow, wc)) { |
1896 | /* We don't want to learn from gratuitous ARP packets that are | |
1897 | * reflected back over bond slaves so we lock the learning table. */ | |
46c88433 | 1898 | if (!in_xbundle->bond) { |
9583bc14 EJ |
1899 | mac_entry_set_grat_arp_lock(mac); |
1900 | } else if (mac_entry_is_grat_arp_locked(mac)) { | |
ee047520 | 1901 | return; |
9583bc14 EJ |
1902 | } |
1903 | } | |
1904 | ||
9d078ec2 | 1905 | if (mac_entry_get_port(xbridge->ml, mac) != in_xbundle->ofbundle) { |
9583bc14 EJ |
1906 | /* The log messages here could actually be useful in debugging, |
1907 | * so keep the rate limit relatively high. */ | |
1908 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
ee047520 | 1909 | |
9583bc14 EJ |
1910 | VLOG_DBG_RL(&rl, "bridge %s: learned that "ETH_ADDR_FMT" is " |
1911 | "on port %s in VLAN %d", | |
46c88433 EJ |
1912 | xbridge->name, ETH_ADDR_ARGS(flow->dl_src), |
1913 | in_xbundle->name, vlan); | |
9583bc14 | 1914 | |
9d078ec2 | 1915 | mac_entry_set_port(xbridge->ml, mac, in_xbundle->ofbundle); |
9583bc14 | 1916 | } |
ee047520 BP |
1917 | } |
1918 | ||
1919 | static void | |
1920 | update_learning_table(const struct xbridge *xbridge, | |
1921 | const struct flow *flow, struct flow_wildcards *wc, | |
1922 | int vlan, struct xbundle *in_xbundle) | |
1923 | { | |
1924 | bool need_update; | |
1925 | ||
1926 | /* Don't learn the OFPP_NONE port. */ | |
1927 | if (in_xbundle == &ofpp_none_bundle) { | |
1928 | return; | |
1929 | } | |
1930 | ||
1931 | /* First try the common case: no change to MAC learning table. */ | |
1932 | ovs_rwlock_rdlock(&xbridge->ml->rwlock); | |
1933 | need_update = is_mac_learning_update_needed(xbridge->ml, flow, wc, vlan, | |
1934 | in_xbundle); | |
509c0149 | 1935 | ovs_rwlock_unlock(&xbridge->ml->rwlock); |
ee047520 BP |
1936 | |
1937 | if (need_update) { | |
1938 | /* Slow path: MAC learning table might need an update. */ | |
1939 | ovs_rwlock_wrlock(&xbridge->ml->rwlock); | |
1940 | update_learning_table__(xbridge, flow, wc, vlan, in_xbundle); | |
1941 | ovs_rwlock_unlock(&xbridge->ml->rwlock); | |
1942 | } | |
9583bc14 EJ |
1943 | } |
1944 | ||
86e2dcdd FL |
1945 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' |
1946 | * was received on 'in_xbundle' in 'vlan' and is either Report or Query. */ | |
1947 | static void | |
1948 | update_mcast_snooping_table__(const struct xbridge *xbridge, | |
1949 | const struct flow *flow, | |
1950 | struct mcast_snooping *ms, | |
1951 | ovs_be32 ip4, int vlan, | |
1952 | struct xbundle *in_xbundle) | |
1953 | OVS_REQ_WRLOCK(ms->rwlock) | |
1954 | { | |
1955 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(60, 30); | |
1956 | ||
1957 | switch (ntohs(flow->tp_src)) { | |
1958 | case IGMP_HOST_MEMBERSHIP_REPORT: | |
1959 | case IGMPV2_HOST_MEMBERSHIP_REPORT: | |
1960 | if (mcast_snooping_add_group(ms, ip4, vlan, in_xbundle->ofbundle)) { | |
1961 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping learned that " | |
1962 | IP_FMT" is on port %s in VLAN %d", | |
1963 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
1964 | } | |
1965 | break; | |
1966 | case IGMP_HOST_LEAVE_MESSAGE: | |
1967 | if (mcast_snooping_leave_group(ms, ip4, vlan, in_xbundle->ofbundle)) { | |
1968 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping leaving " | |
1969 | IP_FMT" is on port %s in VLAN %d", | |
1970 | xbridge->name, IP_ARGS(ip4), in_xbundle->name, vlan); | |
1971 | } | |
1972 | break; | |
1973 | case IGMP_HOST_MEMBERSHIP_QUERY: | |
1974 | if (flow->nw_src && mcast_snooping_add_mrouter(ms, vlan, | |
1975 | in_xbundle->ofbundle)) { | |
1976 | VLOG_DBG_RL(&rl, "bridge %s: multicast snooping query from " | |
1977 | IP_FMT" is on port %s in VLAN %d", | |
1978 | xbridge->name, IP_ARGS(flow->nw_src), | |
1979 | in_xbundle->name, vlan); | |
1980 | } | |
1981 | break; | |
1982 | } | |
1983 | } | |
1984 | ||
1985 | /* Updates multicast snooping table 'ms' given that a packet matching 'flow' | |
1986 | * was received on 'in_xbundle' in 'vlan'. */ | |
1987 | static void | |
1988 | update_mcast_snooping_table(const struct xbridge *xbridge, | |
1989 | const struct flow *flow, int vlan, | |
1990 | struct xbundle *in_xbundle) | |
1991 | { | |
1992 | struct mcast_snooping *ms = xbridge->ms; | |
1993 | struct xlate_cfg *xcfg; | |
1994 | struct xbundle *mcast_xbundle; | |
f4ae6e23 | 1995 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
1996 | |
1997 | /* Don't learn the OFPP_NONE port. */ | |
1998 | if (in_xbundle == &ofpp_none_bundle) { | |
1999 | return; | |
2000 | } | |
2001 | ||
2002 | /* Don't learn from flood ports */ | |
2003 | mcast_xbundle = NULL; | |
2004 | ovs_rwlock_wrlock(&ms->rwlock); | |
2005 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2006 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2007 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2008 | if (mcast_xbundle == in_xbundle) { | |
2009 | break; | |
2010 | } | |
2011 | } | |
2012 | ||
2013 | if (!mcast_xbundle || mcast_xbundle != in_xbundle) { | |
2014 | update_mcast_snooping_table__(xbridge, flow, ms, flow->igmp_group_ip4, | |
2015 | vlan, in_xbundle); | |
2016 | } | |
2017 | ovs_rwlock_unlock(&ms->rwlock); | |
2018 | } | |
2019 | ||
2020 | /* send the packet to ports having the multicast group learned */ | |
2021 | static void | |
2022 | xlate_normal_mcast_send_group(struct xlate_ctx *ctx, | |
2023 | struct mcast_snooping *ms OVS_UNUSED, | |
2024 | struct mcast_group *grp, | |
2025 | struct xbundle *in_xbundle, uint16_t vlan) | |
2026 | OVS_REQ_RDLOCK(ms->rwlock) | |
2027 | { | |
2028 | struct xlate_cfg *xcfg; | |
2029 | struct mcast_group_bundle *b; | |
2030 | struct xbundle *mcast_xbundle; | |
2031 | ||
2032 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2033 | LIST_FOR_EACH(b, bundle_node, &grp->bundle_lru) { | |
2034 | mcast_xbundle = xbundle_lookup(xcfg, b->port); | |
2035 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2036 | xlate_report(ctx, "forwarding to mcast group port"); | |
2037 | output_normal(ctx, mcast_xbundle, vlan); | |
2038 | } else if (!mcast_xbundle) { | |
2039 | xlate_report(ctx, "mcast group port is unknown, dropping"); | |
2040 | } else { | |
2041 | xlate_report(ctx, "mcast group port is input port, dropping"); | |
2042 | } | |
2043 | } | |
2044 | } | |
2045 | ||
2046 | /* send the packet to ports connected to multicast routers */ | |
2047 | static void | |
2048 | xlate_normal_mcast_send_mrouters(struct xlate_ctx *ctx, | |
2049 | struct mcast_snooping *ms, | |
2050 | struct xbundle *in_xbundle, uint16_t vlan) | |
2051 | OVS_REQ_RDLOCK(ms->rwlock) | |
2052 | { | |
2053 | struct xlate_cfg *xcfg; | |
2054 | struct mcast_mrouter_bundle *mrouter; | |
2055 | struct xbundle *mcast_xbundle; | |
2056 | ||
2057 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2058 | LIST_FOR_EACH(mrouter, mrouter_node, &ms->mrouter_lru) { | |
2059 | mcast_xbundle = xbundle_lookup(xcfg, mrouter->port); | |
2060 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2061 | xlate_report(ctx, "forwarding to mcast router port"); | |
2062 | output_normal(ctx, mcast_xbundle, vlan); | |
2063 | } else if (!mcast_xbundle) { | |
2064 | xlate_report(ctx, "mcast router port is unknown, dropping"); | |
2065 | } else { | |
2066 | xlate_report(ctx, "mcast router port is input port, dropping"); | |
2067 | } | |
2068 | } | |
2069 | } | |
2070 | ||
2071 | /* send the packet to ports flagged to be flooded */ | |
2072 | static void | |
2073 | xlate_normal_mcast_send_fports(struct xlate_ctx *ctx, | |
2074 | struct mcast_snooping *ms, | |
2075 | struct xbundle *in_xbundle, uint16_t vlan) | |
2076 | OVS_REQ_RDLOCK(ms->rwlock) | |
2077 | { | |
2078 | struct xlate_cfg *xcfg; | |
f4ae6e23 | 2079 | struct mcast_port_bundle *fport; |
86e2dcdd FL |
2080 | struct xbundle *mcast_xbundle; |
2081 | ||
2082 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
f4ae6e23 | 2083 | LIST_FOR_EACH(fport, node, &ms->fport_list) { |
86e2dcdd FL |
2084 | mcast_xbundle = xbundle_lookup(xcfg, fport->port); |
2085 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2086 | xlate_report(ctx, "forwarding to mcast flood port"); | |
2087 | output_normal(ctx, mcast_xbundle, vlan); | |
2088 | } else if (!mcast_xbundle) { | |
2089 | xlate_report(ctx, "mcast flood port is unknown, dropping"); | |
2090 | } else { | |
2091 | xlate_report(ctx, "mcast flood port is input port, dropping"); | |
2092 | } | |
2093 | } | |
2094 | } | |
2095 | ||
8e04a33f FL |
2096 | /* forward the Reports to configured ports */ |
2097 | static void | |
2098 | xlate_normal_mcast_send_rports(struct xlate_ctx *ctx, | |
2099 | struct mcast_snooping *ms, | |
2100 | struct xbundle *in_xbundle, uint16_t vlan) | |
2101 | OVS_REQ_RDLOCK(ms->rwlock) | |
2102 | { | |
2103 | struct xlate_cfg *xcfg; | |
2104 | struct mcast_port_bundle *rport; | |
2105 | struct xbundle *mcast_xbundle; | |
2106 | ||
2107 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2108 | LIST_FOR_EACH(rport, node, &ms->rport_list) { | |
2109 | mcast_xbundle = xbundle_lookup(xcfg, rport->port); | |
2110 | if (mcast_xbundle && mcast_xbundle != in_xbundle) { | |
2111 | xlate_report(ctx, "forwarding Report to mcast flagged port"); | |
2112 | output_normal(ctx, mcast_xbundle, vlan); | |
2113 | } else if (!mcast_xbundle) { | |
2114 | xlate_report(ctx, "mcast port is unknown, dropping the Report"); | |
2115 | } else { | |
2116 | xlate_report(ctx, "mcast port is input port, dropping the Report"); | |
2117 | } | |
2118 | } | |
2119 | } | |
2120 | ||
682800a4 FL |
2121 | static void |
2122 | xlate_normal_flood(struct xlate_ctx *ctx, struct xbundle *in_xbundle, | |
2123 | uint16_t vlan) | |
2124 | { | |
2125 | struct xbundle *xbundle; | |
2126 | ||
2127 | LIST_FOR_EACH (xbundle, list_node, &ctx->xbridge->xbundles) { | |
2128 | if (xbundle != in_xbundle | |
2129 | && xbundle_includes_vlan(xbundle, vlan) | |
2130 | && xbundle->floodable | |
2131 | && !xbundle_mirror_out(ctx->xbridge, xbundle)) { | |
2132 | output_normal(ctx, xbundle, vlan); | |
2133 | } | |
2134 | } | |
2135 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
2136 | } | |
2137 | ||
9583bc14 EJ |
2138 | static void |
2139 | xlate_normal(struct xlate_ctx *ctx) | |
2140 | { | |
33bf9176 BP |
2141 | struct flow_wildcards *wc = &ctx->xout->wc; |
2142 | struct flow *flow = &ctx->xin->flow; | |
46c88433 EJ |
2143 | struct xbundle *in_xbundle; |
2144 | struct xport *in_port; | |
9583bc14 | 2145 | struct mac_entry *mac; |
d6d5bbc9 | 2146 | void *mac_port; |
9583bc14 EJ |
2147 | uint16_t vlan; |
2148 | uint16_t vid; | |
2149 | ||
2150 | ctx->xout->has_normal = true; | |
2151 | ||
33bf9176 BP |
2152 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
2153 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1dd35f8a | 2154 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
9583bc14 | 2155 | |
46c88433 EJ |
2156 | in_xbundle = lookup_input_bundle(ctx->xbridge, flow->in_port.ofp_port, |
2157 | ctx->xin->packet != NULL, &in_port); | |
2158 | if (!in_xbundle) { | |
9583bc14 EJ |
2159 | xlate_report(ctx, "no input bundle, dropping"); |
2160 | return; | |
2161 | } | |
2162 | ||
2163 | /* Drop malformed frames. */ | |
33bf9176 BP |
2164 | if (flow->dl_type == htons(ETH_TYPE_VLAN) && |
2165 | !(flow->vlan_tci & htons(VLAN_CFI))) { | |
9583bc14 EJ |
2166 | if (ctx->xin->packet != NULL) { |
2167 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2168 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet with partial " | |
2169 | "VLAN tag received on port %s", | |
46c88433 | 2170 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2171 | } |
2172 | xlate_report(ctx, "partial VLAN tag, dropping"); | |
2173 | return; | |
2174 | } | |
2175 | ||
2176 | /* Drop frames on bundles reserved for mirroring. */ | |
46c88433 | 2177 | if (xbundle_mirror_out(ctx->xbridge, in_xbundle)) { |
9583bc14 EJ |
2178 | if (ctx->xin->packet != NULL) { |
2179 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2180 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet received on port " | |
2181 | "%s, which is reserved exclusively for mirroring", | |
46c88433 | 2182 | ctx->xbridge->name, in_xbundle->name); |
9583bc14 EJ |
2183 | } |
2184 | xlate_report(ctx, "input port is mirror output port, dropping"); | |
2185 | return; | |
2186 | } | |
2187 | ||
2188 | /* Check VLAN. */ | |
33bf9176 | 2189 | vid = vlan_tci_to_vid(flow->vlan_tci); |
46c88433 | 2190 | if (!input_vid_is_valid(vid, in_xbundle, ctx->xin->packet != NULL)) { |
9583bc14 EJ |
2191 | xlate_report(ctx, "disallowed VLAN VID for this input port, dropping"); |
2192 | return; | |
2193 | } | |
46c88433 | 2194 | vlan = input_vid_to_vlan(in_xbundle, vid); |
9583bc14 EJ |
2195 | |
2196 | /* Check other admissibility requirements. */ | |
2197 | if (in_port && !is_admissible(ctx, in_port, vlan)) { | |
2198 | return; | |
2199 | } | |
2200 | ||
2201 | /* Learn source MAC. */ | |
2202 | if (ctx->xin->may_learn) { | |
46c88433 | 2203 | update_learning_table(ctx->xbridge, flow, wc, vlan, in_xbundle); |
9583bc14 | 2204 | } |
b256dc52 JS |
2205 | if (ctx->xin->xcache) { |
2206 | struct xc_entry *entry; | |
2207 | ||
2208 | /* Save enough info to update mac learning table later. */ | |
2209 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NORMAL); | |
9edf6b48 | 2210 | entry->u.normal.ofproto = ctx->xbridge->ofproto; |
b256dc52 JS |
2211 | entry->u.normal.flow = xmemdup(flow, sizeof *flow); |
2212 | entry->u.normal.vlan = vlan; | |
2213 | } | |
9583bc14 EJ |
2214 | |
2215 | /* Determine output bundle. */ | |
86e2dcdd FL |
2216 | if (mcast_snooping_enabled(ctx->xbridge->ms) |
2217 | && !eth_addr_is_broadcast(flow->dl_dst) | |
2218 | && eth_addr_is_multicast(flow->dl_dst) | |
2219 | && flow->dl_type == htons(ETH_TYPE_IP)) { | |
2220 | struct mcast_snooping *ms = ctx->xbridge->ms; | |
2221 | struct mcast_group *grp; | |
2222 | ||
2223 | if (flow->nw_proto == IPPROTO_IGMP) { | |
2224 | if (ctx->xin->may_learn) { | |
2225 | if (mcast_snooping_is_membership(flow->tp_src) || | |
2226 | mcast_snooping_is_query(flow->tp_src)) { | |
2227 | update_mcast_snooping_table(ctx->xbridge, flow, vlan, | |
2228 | in_xbundle); | |
2229 | } | |
2230 | } | |
d6d5bbc9 | 2231 | |
86e2dcdd FL |
2232 | if (mcast_snooping_is_membership(flow->tp_src)) { |
2233 | ovs_rwlock_rdlock(&ms->rwlock); | |
2234 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
8e04a33f FL |
2235 | /* RFC4541: section 2.1.1, item 1: A snooping switch should |
2236 | * forward IGMP Membership Reports only to those ports where | |
2237 | * multicast routers are attached. Alternatively stated: a | |
2238 | * snooping switch should not forward IGMP Membership Reports | |
2239 | * to ports on which only hosts are attached. | |
2240 | * An administrative control may be provided to override this | |
2241 | * restriction, allowing the report messages to be flooded to | |
2242 | * other ports. */ | |
2243 | xlate_normal_mcast_send_rports(ctx, ms, in_xbundle, vlan); | |
86e2dcdd FL |
2244 | ovs_rwlock_unlock(&ms->rwlock); |
2245 | } else { | |
2246 | xlate_report(ctx, "multicast traffic, flooding"); | |
2247 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2248 | } | |
2249 | return; | |
2250 | } else { | |
2251 | if (ip_is_local_multicast(flow->nw_dst)) { | |
2252 | /* RFC4541: section 2.1.2, item 2: Packets with a dst IP | |
2253 | * address in the 224.0.0.x range which are not IGMP must | |
2254 | * be forwarded on all ports */ | |
2255 | xlate_report(ctx, "RFC4541: section 2.1.2, item 2, flooding"); | |
2256 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2257 | return; | |
2258 | } | |
2259 | } | |
2260 | ||
2261 | /* forwarding to group base ports */ | |
2262 | ovs_rwlock_rdlock(&ms->rwlock); | |
2263 | grp = mcast_snooping_lookup(ms, flow->nw_dst, vlan); | |
2264 | if (grp) { | |
2265 | xlate_normal_mcast_send_group(ctx, ms, grp, in_xbundle, vlan); | |
2266 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2267 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
9583bc14 | 2268 | } else { |
86e2dcdd FL |
2269 | if (mcast_snooping_flood_unreg(ms)) { |
2270 | xlate_report(ctx, "unregistered multicast, flooding"); | |
2271 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2272 | } else { | |
2273 | xlate_normal_mcast_send_mrouters(ctx, ms, in_xbundle, vlan); | |
2274 | xlate_normal_mcast_send_fports(ctx, ms, in_xbundle, vlan); | |
2275 | } | |
9583bc14 | 2276 | } |
86e2dcdd | 2277 | ovs_rwlock_unlock(&ms->rwlock); |
9583bc14 | 2278 | } else { |
86e2dcdd FL |
2279 | ovs_rwlock_rdlock(&ctx->xbridge->ml->rwlock); |
2280 | mac = mac_learning_lookup(ctx->xbridge->ml, flow->dl_dst, vlan); | |
9d078ec2 | 2281 | mac_port = mac ? mac_entry_get_port(ctx->xbridge->ml, mac) : NULL; |
86e2dcdd FL |
2282 | ovs_rwlock_unlock(&ctx->xbridge->ml->rwlock); |
2283 | ||
2284 | if (mac_port) { | |
2285 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2286 | struct xbundle *mac_xbundle = xbundle_lookup(xcfg, mac_port); | |
2287 | if (mac_xbundle && mac_xbundle != in_xbundle) { | |
2288 | xlate_report(ctx, "forwarding to learned port"); | |
2289 | output_normal(ctx, mac_xbundle, vlan); | |
2290 | } else if (!mac_xbundle) { | |
2291 | xlate_report(ctx, "learned port is unknown, dropping"); | |
2292 | } else { | |
2293 | xlate_report(ctx, "learned port is input port, dropping"); | |
2294 | } | |
2295 | } else { | |
2296 | xlate_report(ctx, "no learned MAC for destination, flooding"); | |
2297 | xlate_normal_flood(ctx, in_xbundle, vlan); | |
2298 | } | |
9583bc14 EJ |
2299 | } |
2300 | } | |
2301 | ||
2302 | /* Compose SAMPLE action for sFlow or IPFIX. The given probability is | |
2303 | * the number of packets out of UINT32_MAX to sample. The given | |
2304 | * cookie is passed back in the callback for each sampled packet. | |
2305 | */ | |
2306 | static size_t | |
46c88433 | 2307 | compose_sample_action(const struct xbridge *xbridge, |
9583bc14 EJ |
2308 | struct ofpbuf *odp_actions, |
2309 | const struct flow *flow, | |
2310 | const uint32_t probability, | |
2311 | const union user_action_cookie *cookie, | |
8b7ea2d4 WZ |
2312 | const size_t cookie_size, |
2313 | const odp_port_t tunnel_out_port) | |
9583bc14 EJ |
2314 | { |
2315 | size_t sample_offset, actions_offset; | |
89a8a7f0 | 2316 | odp_port_t odp_port; |
9583bc14 | 2317 | int cookie_offset; |
89a8a7f0 | 2318 | uint32_t pid; |
9583bc14 EJ |
2319 | |
2320 | sample_offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SAMPLE); | |
2321 | ||
2322 | nl_msg_put_u32(odp_actions, OVS_SAMPLE_ATTR_PROBABILITY, probability); | |
2323 | ||
2324 | actions_offset = nl_msg_start_nested(odp_actions, OVS_SAMPLE_ATTR_ACTIONS); | |
89a8a7f0 EJ |
2325 | |
2326 | odp_port = ofp_port_to_odp_port(xbridge, flow->in_port.ofp_port); | |
9a159f74 AW |
2327 | pid = dpif_port_get_pid(xbridge->dpif, odp_port, |
2328 | flow_hash_5tuple(flow, 0)); | |
2329 | cookie_offset = odp_put_userspace_action(pid, cookie, cookie_size, | |
8b7ea2d4 | 2330 | tunnel_out_port, odp_actions); |
9583bc14 EJ |
2331 | |
2332 | nl_msg_end_nested(odp_actions, actions_offset); | |
2333 | nl_msg_end_nested(odp_actions, sample_offset); | |
2334 | return cookie_offset; | |
2335 | } | |
2336 | ||
2337 | static void | |
46c88433 EJ |
2338 | compose_sflow_cookie(const struct xbridge *xbridge, ovs_be16 vlan_tci, |
2339 | odp_port_t odp_port, unsigned int n_outputs, | |
2340 | union user_action_cookie *cookie) | |
9583bc14 EJ |
2341 | { |
2342 | int ifindex; | |
2343 | ||
2344 | cookie->type = USER_ACTION_COOKIE_SFLOW; | |
2345 | cookie->sflow.vlan_tci = vlan_tci; | |
2346 | ||
2347 | /* See http://www.sflow.org/sflow_version_5.txt (search for "Input/output | |
2348 | * port information") for the interpretation of cookie->output. */ | |
2349 | switch (n_outputs) { | |
2350 | case 0: | |
2351 | /* 0x40000000 | 256 means "packet dropped for unknown reason". */ | |
2352 | cookie->sflow.output = 0x40000000 | 256; | |
2353 | break; | |
2354 | ||
2355 | case 1: | |
46c88433 | 2356 | ifindex = dpif_sflow_odp_port_to_ifindex(xbridge->sflow, odp_port); |
9583bc14 EJ |
2357 | if (ifindex) { |
2358 | cookie->sflow.output = ifindex; | |
2359 | break; | |
2360 | } | |
2361 | /* Fall through. */ | |
2362 | default: | |
2363 | /* 0x80000000 means "multiple output ports. */ | |
2364 | cookie->sflow.output = 0x80000000 | n_outputs; | |
2365 | break; | |
2366 | } | |
2367 | } | |
2368 | ||
2369 | /* Compose SAMPLE action for sFlow bridge sampling. */ | |
2370 | static size_t | |
46c88433 | 2371 | compose_sflow_action(const struct xbridge *xbridge, |
9583bc14 EJ |
2372 | struct ofpbuf *odp_actions, |
2373 | const struct flow *flow, | |
4e022ec0 | 2374 | odp_port_t odp_port) |
9583bc14 EJ |
2375 | { |
2376 | uint32_t probability; | |
2377 | union user_action_cookie cookie; | |
2378 | ||
46c88433 | 2379 | if (!xbridge->sflow || flow->in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
2380 | return 0; |
2381 | } | |
2382 | ||
46c88433 EJ |
2383 | probability = dpif_sflow_get_probability(xbridge->sflow); |
2384 | compose_sflow_cookie(xbridge, htons(0), odp_port, | |
4e022ec0 | 2385 | odp_port == ODPP_NONE ? 0 : 1, &cookie); |
9583bc14 | 2386 | |
46c88433 | 2387 | return compose_sample_action(xbridge, odp_actions, flow, probability, |
8b7ea2d4 | 2388 | &cookie, sizeof cookie.sflow, ODPP_NONE); |
9583bc14 EJ |
2389 | } |
2390 | ||
2391 | static void | |
2392 | compose_flow_sample_cookie(uint16_t probability, uint32_t collector_set_id, | |
2393 | uint32_t obs_domain_id, uint32_t obs_point_id, | |
2394 | union user_action_cookie *cookie) | |
2395 | { | |
2396 | cookie->type = USER_ACTION_COOKIE_FLOW_SAMPLE; | |
2397 | cookie->flow_sample.probability = probability; | |
2398 | cookie->flow_sample.collector_set_id = collector_set_id; | |
2399 | cookie->flow_sample.obs_domain_id = obs_domain_id; | |
2400 | cookie->flow_sample.obs_point_id = obs_point_id; | |
2401 | } | |
2402 | ||
2403 | static void | |
8b7ea2d4 WZ |
2404 | compose_ipfix_cookie(union user_action_cookie *cookie, |
2405 | odp_port_t output_odp_port) | |
9583bc14 EJ |
2406 | { |
2407 | cookie->type = USER_ACTION_COOKIE_IPFIX; | |
8b7ea2d4 | 2408 | cookie->ipfix.output_odp_port = output_odp_port; |
9583bc14 EJ |
2409 | } |
2410 | ||
2411 | /* Compose SAMPLE action for IPFIX bridge sampling. */ | |
2412 | static void | |
46c88433 | 2413 | compose_ipfix_action(const struct xbridge *xbridge, |
9583bc14 | 2414 | struct ofpbuf *odp_actions, |
8b7ea2d4 WZ |
2415 | const struct flow *flow, |
2416 | odp_port_t output_odp_port) | |
9583bc14 EJ |
2417 | { |
2418 | uint32_t probability; | |
2419 | union user_action_cookie cookie; | |
8b7ea2d4 | 2420 | odp_port_t tunnel_out_port = ODPP_NONE; |
9583bc14 | 2421 | |
46c88433 | 2422 | if (!xbridge->ipfix || flow->in_port.ofp_port == OFPP_NONE) { |
9583bc14 EJ |
2423 | return; |
2424 | } | |
2425 | ||
8b7ea2d4 WZ |
2426 | /* For input case, output_odp_port is ODPP_NONE, which is an invalid port |
2427 | * number. */ | |
2428 | if (output_odp_port == ODPP_NONE && | |
2429 | !dpif_ipfix_get_bridge_exporter_input_sampling(xbridge->ipfix)) { | |
2430 | return; | |
2431 | } | |
2432 | ||
2433 | /* For output case, output_odp_port is valid*/ | |
2434 | if (output_odp_port != ODPP_NONE) { | |
2435 | if (!dpif_ipfix_get_bridge_exporter_output_sampling(xbridge->ipfix)) { | |
2436 | return; | |
2437 | } | |
2438 | /* If tunnel sampling is enabled, put an additional option attribute: | |
2439 | * OVS_USERSPACE_ATTR_TUNNEL_OUT_PORT | |
2440 | */ | |
2441 | if (dpif_ipfix_get_bridge_exporter_tunnel_sampling(xbridge->ipfix) && | |
2442 | dpif_ipfix_get_tunnel_port(xbridge->ipfix, output_odp_port) ) { | |
2443 | tunnel_out_port = output_odp_port; | |
2444 | } | |
2445 | } | |
2446 | ||
46c88433 | 2447 | probability = dpif_ipfix_get_bridge_exporter_probability(xbridge->ipfix); |
8b7ea2d4 | 2448 | compose_ipfix_cookie(&cookie, output_odp_port); |
9583bc14 | 2449 | |
46c88433 | 2450 | compose_sample_action(xbridge, odp_actions, flow, probability, |
8b7ea2d4 | 2451 | &cookie, sizeof cookie.ipfix, tunnel_out_port); |
9583bc14 EJ |
2452 | } |
2453 | ||
2454 | /* SAMPLE action for sFlow must be first action in any given list of | |
2455 | * actions. At this point we do not have all information required to | |
2456 | * build it. So try to build sample action as complete as possible. */ | |
2457 | static void | |
2458 | add_sflow_action(struct xlate_ctx *ctx) | |
2459 | { | |
46c88433 | 2460 | ctx->user_cookie_offset = compose_sflow_action(ctx->xbridge, |
cc377352 | 2461 | ctx->xout->odp_actions, |
4e022ec0 | 2462 | &ctx->xin->flow, ODPP_NONE); |
9583bc14 EJ |
2463 | ctx->sflow_odp_port = 0; |
2464 | ctx->sflow_n_outputs = 0; | |
2465 | } | |
2466 | ||
2467 | /* SAMPLE action for IPFIX must be 1st or 2nd action in any given list | |
2468 | * of actions, eventually after the SAMPLE action for sFlow. */ | |
2469 | static void | |
2470 | add_ipfix_action(struct xlate_ctx *ctx) | |
2471 | { | |
cc377352 | 2472 | compose_ipfix_action(ctx->xbridge, ctx->xout->odp_actions, |
8b7ea2d4 WZ |
2473 | &ctx->xin->flow, ODPP_NONE); |
2474 | } | |
2475 | ||
2476 | static void | |
2477 | add_ipfix_output_action(struct xlate_ctx *ctx, odp_port_t port) | |
2478 | { | |
2479 | compose_ipfix_action(ctx->xbridge, ctx->xout->odp_actions, | |
2480 | &ctx->xin->flow, port); | |
9583bc14 EJ |
2481 | } |
2482 | ||
2483 | /* Fix SAMPLE action according to data collected while composing ODP actions. | |
2484 | * We need to fix SAMPLE actions OVS_SAMPLE_ATTR_ACTIONS attribute, i.e. nested | |
2485 | * USERSPACE action's user-cookie which is required for sflow. */ | |
2486 | static void | |
2487 | fix_sflow_action(struct xlate_ctx *ctx) | |
2488 | { | |
2489 | const struct flow *base = &ctx->base_flow; | |
2490 | union user_action_cookie *cookie; | |
2491 | ||
2492 | if (!ctx->user_cookie_offset) { | |
2493 | return; | |
2494 | } | |
2495 | ||
cc377352 | 2496 | cookie = ofpbuf_at(ctx->xout->odp_actions, ctx->user_cookie_offset, |
9583bc14 EJ |
2497 | sizeof cookie->sflow); |
2498 | ovs_assert(cookie->type == USER_ACTION_COOKIE_SFLOW); | |
2499 | ||
46c88433 | 2500 | compose_sflow_cookie(ctx->xbridge, base->vlan_tci, |
9583bc14 EJ |
2501 | ctx->sflow_odp_port, ctx->sflow_n_outputs, cookie); |
2502 | } | |
2503 | ||
db7d4e46 | 2504 | static enum slow_path_reason |
642dc74d | 2505 | process_special(struct xlate_ctx *ctx, const struct flow *flow, |
cf62fa4c | 2506 | const struct xport *xport, const struct dp_packet *packet) |
db7d4e46 | 2507 | { |
642dc74d | 2508 | struct flow_wildcards *wc = &ctx->xout->wc; |
46c88433 | 2509 | const struct xbridge *xbridge = ctx->xbridge; |
642dc74d | 2510 | |
46c88433 | 2511 | if (!xport) { |
db7d4e46 | 2512 | return 0; |
46c88433 | 2513 | } else if (xport->cfm && cfm_should_process_flow(xport->cfm, flow, wc)) { |
db7d4e46 | 2514 | if (packet) { |
46c88433 | 2515 | cfm_process_heartbeat(xport->cfm, packet); |
db7d4e46 JP |
2516 | } |
2517 | return SLOW_CFM; | |
fab52e16 | 2518 | } else if (xport->bfd && bfd_should_process_flow(xport->bfd, flow, wc)) { |
db7d4e46 | 2519 | if (packet) { |
46c88433 | 2520 | bfd_process_packet(xport->bfd, flow, packet); |
60d02c72 AW |
2521 | /* If POLL received, immediately sends FINAL back. */ |
2522 | if (bfd_should_send_packet(xport->bfd)) { | |
6d308b28 | 2523 | ofproto_dpif_monitor_port_send_soon(xport->ofport); |
60d02c72 | 2524 | } |
db7d4e46 JP |
2525 | } |
2526 | return SLOW_BFD; | |
46c88433 | 2527 | } else if (xport->xbundle && xport->xbundle->lacp |
db7d4e46 JP |
2528 | && flow->dl_type == htons(ETH_TYPE_LACP)) { |
2529 | if (packet) { | |
46c88433 | 2530 | lacp_process_packet(xport->xbundle->lacp, xport->ofport, packet); |
db7d4e46 JP |
2531 | } |
2532 | return SLOW_LACP; | |
9efd308e DV |
2533 | } else if ((xbridge->stp || xbridge->rstp) && |
2534 | stp_should_process_flow(flow, wc)) { | |
db7d4e46 | 2535 | if (packet) { |
f025bcb7 JR |
2536 | xbridge->stp |
2537 | ? stp_process_packet(xport, packet) | |
2538 | : rstp_process_packet(xport, packet); | |
db7d4e46 JP |
2539 | } |
2540 | return SLOW_STP; | |
0477baa9 DF |
2541 | } else if (xport->lldp && lldp_should_process_flow(flow)) { |
2542 | if (packet) { | |
2543 | lldp_process_packet(xport->lldp, packet); | |
2544 | } | |
2545 | return SLOW_LLDP; | |
db7d4e46 JP |
2546 | } else { |
2547 | return 0; | |
2548 | } | |
2549 | } | |
2550 | ||
a36de779 PS |
2551 | static int |
2552 | tnl_route_lookup_flow(const struct flow *oflow, | |
2553 | ovs_be32 *ip, struct xport **out_port) | |
2554 | { | |
2555 | char out_dev[IFNAMSIZ]; | |
2556 | struct xbridge *xbridge; | |
2557 | struct xlate_cfg *xcfg; | |
2558 | ovs_be32 gw; | |
2559 | ||
2560 | if (!ovs_router_lookup(oflow->tunnel.ip_dst, out_dev, &gw)) { | |
2561 | return -ENOENT; | |
2562 | } | |
2563 | ||
2564 | if (gw) { | |
2565 | *ip = gw; | |
2566 | } else { | |
2567 | *ip = oflow->tunnel.ip_dst; | |
2568 | } | |
2569 | ||
2570 | xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); | |
2571 | ovs_assert(xcfg); | |
2572 | ||
2573 | HMAP_FOR_EACH (xbridge, hmap_node, &xcfg->xbridges) { | |
2574 | if (!strncmp(xbridge->name, out_dev, IFNAMSIZ)) { | |
2575 | struct xport *port; | |
2576 | ||
2577 | HMAP_FOR_EACH (port, ofp_node, &xbridge->xports) { | |
2578 | if (!strncmp(netdev_get_name(port->netdev), out_dev, IFNAMSIZ)) { | |
2579 | *out_port = port; | |
2580 | return 0; | |
2581 | } | |
2582 | } | |
2583 | } | |
2584 | } | |
2585 | return -ENOENT; | |
2586 | } | |
2587 | ||
2588 | static int | |
cf62fa4c | 2589 | xlate_flood_packet(struct xbridge *xbridge, struct dp_packet *packet) |
a36de779 PS |
2590 | { |
2591 | struct ofpact_output output; | |
2592 | struct flow flow; | |
2593 | ||
2594 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); | |
2595 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ | |
cf62fa4c | 2596 | flow_extract(packet, &flow); |
a36de779 PS |
2597 | flow.in_port.ofp_port = OFPP_NONE; |
2598 | output.port = OFPP_FLOOD; | |
2599 | output.max_len = 0; | |
2600 | ||
2601 | return ofproto_dpif_execute_actions(xbridge->ofproto, &flow, NULL, | |
2602 | &output.ofpact, sizeof output, | |
2603 | packet); | |
2604 | } | |
2605 | ||
2606 | static void | |
2607 | tnl_send_arp_request(const struct xport *out_dev, const uint8_t eth_src[ETH_ADDR_LEN], | |
2608 | ovs_be32 ip_src, ovs_be32 ip_dst) | |
2609 | { | |
2610 | struct xbridge *xbridge = out_dev->xbridge; | |
cf62fa4c | 2611 | struct dp_packet packet; |
a36de779 | 2612 | |
cf62fa4c | 2613 | dp_packet_init(&packet, 0); |
a36de779 PS |
2614 | compose_arp(&packet, eth_src, ip_src, ip_dst); |
2615 | ||
2616 | xlate_flood_packet(xbridge, &packet); | |
cf62fa4c | 2617 | dp_packet_uninit(&packet); |
a36de779 PS |
2618 | } |
2619 | ||
2620 | static int | |
2621 | build_tunnel_send(const struct xlate_ctx *ctx, const struct xport *xport, | |
2622 | const struct flow *flow, odp_port_t tunnel_odp_port) | |
2623 | { | |
2624 | struct ovs_action_push_tnl tnl_push_data; | |
2625 | struct xport *out_dev = NULL; | |
2626 | ovs_be32 s_ip, d_ip = 0; | |
2627 | uint8_t smac[ETH_ADDR_LEN]; | |
2628 | uint8_t dmac[ETH_ADDR_LEN]; | |
2629 | int err; | |
2630 | ||
2631 | err = tnl_route_lookup_flow(flow, &d_ip, &out_dev); | |
2632 | if (err) { | |
2633 | return err; | |
2634 | } | |
2635 | ||
2636 | /* Use mac addr of bridge port of the peer. */ | |
2637 | err = netdev_get_etheraddr(out_dev->netdev, smac); | |
2638 | if (err) { | |
2639 | return err; | |
2640 | } | |
2641 | ||
2642 | err = netdev_get_in4(out_dev->netdev, (struct in_addr *) &s_ip, NULL); | |
2643 | if (err) { | |
2644 | return err; | |
2645 | } | |
2646 | ||
2647 | err = tnl_arp_lookup(out_dev->xbridge->name, d_ip, dmac); | |
2648 | if (err) { | |
2649 | tnl_send_arp_request(out_dev, smac, s_ip, d_ip); | |
2650 | return err; | |
2651 | } | |
2652 | if (ctx->xin->xcache) { | |
2653 | struct xc_entry *entry; | |
2654 | ||
2655 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_TNL_ARP); | |
8742957c BP |
2656 | ovs_strlcpy(entry->u.tnl_arp_cache.br_name, out_dev->xbridge->name, |
2657 | sizeof entry->u.tnl_arp_cache.br_name); | |
a36de779 PS |
2658 | entry->u.tnl_arp_cache.d_ip = d_ip; |
2659 | } | |
2660 | err = tnl_port_build_header(xport->ofport, flow, | |
2661 | dmac, smac, s_ip, &tnl_push_data); | |
2662 | if (err) { | |
2663 | return err; | |
2664 | } | |
2665 | tnl_push_data.tnl_port = odp_to_u32(tunnel_odp_port); | |
2666 | tnl_push_data.out_port = odp_to_u32(out_dev->odp_port); | |
2667 | odp_put_tnl_push_action(ctx->xout->odp_actions, &tnl_push_data); | |
2668 | return 0; | |
2669 | } | |
2670 | ||
9583bc14 | 2671 | static void |
4e022ec0 | 2672 | compose_output_action__(struct xlate_ctx *ctx, ofp_port_t ofp_port, |
9583bc14 EJ |
2673 | bool check_stp) |
2674 | { | |
46c88433 | 2675 | const struct xport *xport = get_ofp_port(ctx->xbridge, ofp_port); |
1dd35f8a | 2676 | struct flow_wildcards *wc = &ctx->xout->wc; |
33bf9176 | 2677 | struct flow *flow = &ctx->xin->flow; |
a36de779 | 2678 | struct flow_tnl flow_tnl; |
9583bc14 | 2679 | ovs_be16 flow_vlan_tci; |
1362e248 | 2680 | uint32_t flow_pkt_mark; |
9583bc14 | 2681 | uint8_t flow_nw_tos; |
4e022ec0 | 2682 | odp_port_t out_port, odp_port; |
a36de779 | 2683 | bool tnl_push_pop_send = false; |
ca077186 | 2684 | uint8_t dscp; |
9583bc14 EJ |
2685 | |
2686 | /* If 'struct flow' gets additional metadata, we'll need to zero it out | |
2687 | * before traversing a patch port. */ | |
ac6073e3 | 2688 | BUILD_ASSERT_DECL(FLOW_WC_SEQ == 31); |
a36de779 | 2689 | memset(&flow_tnl, 0, sizeof flow_tnl); |
9583bc14 | 2690 | |
46c88433 | 2691 | if (!xport) { |
9583bc14 EJ |
2692 | xlate_report(ctx, "Nonexistent output port"); |
2693 | return; | |
46c88433 | 2694 | } else if (xport->config & OFPUTIL_PC_NO_FWD) { |
9583bc14 EJ |
2695 | xlate_report(ctx, "OFPPC_NO_FWD set, skipping output"); |
2696 | return; | |
0d1cee12 | 2697 | } else if (check_stp) { |
bbbca389 | 2698 | if (is_stp(&ctx->base_flow)) { |
9efd308e DV |
2699 | if (!xport_stp_should_forward_bpdu(xport) && |
2700 | !xport_rstp_should_manage_bpdu(xport)) { | |
2701 | if (ctx->xbridge->stp != NULL) { | |
2702 | xlate_report(ctx, "STP not in listening state, " | |
2703 | "skipping bpdu output"); | |
2704 | } else if (ctx->xbridge->rstp != NULL) { | |
2705 | xlate_report(ctx, "RSTP not managing BPDU in this state, " | |
2706 | "skipping bpdu output"); | |
2707 | } | |
0d1cee12 K |
2708 | return; |
2709 | } | |
9efd308e DV |
2710 | } else if (!xport_stp_forward_state(xport) || |
2711 | !xport_rstp_forward_state(xport)) { | |
2712 | if (ctx->xbridge->stp != NULL) { | |
2713 | xlate_report(ctx, "STP not in forwarding state, " | |
2714 | "skipping output"); | |
2715 | } else if (ctx->xbridge->rstp != NULL) { | |
2716 | xlate_report(ctx, "RSTP not in forwarding state, " | |
2717 | "skipping output"); | |
2718 | } | |
0d1cee12 K |
2719 | return; |
2720 | } | |
9583bc14 EJ |
2721 | } |
2722 | ||
46c88433 EJ |
2723 | if (mbridge_has_mirrors(ctx->xbridge->mbridge) && xport->xbundle) { |
2724 | ctx->xout->mirrors |= xbundle_mirror_dst(xport->xbundle->xbridge, | |
2725 | xport->xbundle); | |
cdf5d3a5 EJ |
2726 | } |
2727 | ||
46c88433 EJ |
2728 | if (xport->peer) { |
2729 | const struct xport *peer = xport->peer; | |
9583bc14 | 2730 | struct flow old_flow = ctx->xin->flow; |
9583bc14 | 2731 | enum slow_path_reason special; |
0c7812e5 | 2732 | uint8_t table_id = rule_dpif_lookup_get_init_table_id(&ctx->xin->flow); |
9583bc14 | 2733 | |
46c88433 EJ |
2734 | ctx->xbridge = peer->xbridge; |
2735 | flow->in_port.ofp_port = peer->ofp_port; | |
33bf9176 BP |
2736 | flow->metadata = htonll(0); |
2737 | memset(&flow->tunnel, 0, sizeof flow->tunnel); | |
2738 | memset(flow->regs, 0, sizeof flow->regs); | |
c61f3870 | 2739 | flow->actset_output = OFPP_UNSET; |
9583bc14 | 2740 | |
642dc74d | 2741 | special = process_special(ctx, &ctx->xin->flow, peer, |
9583bc14 EJ |
2742 | ctx->xin->packet); |
2743 | if (special) { | |
04594cd5 | 2744 | ctx->xout->slow |= special; |
ddd3c975 | 2745 | } else if (may_receive(peer, ctx)) { |
9efd308e | 2746 | if (xport_stp_forward_state(peer) && xport_rstp_forward_state(peer)) { |
0c7812e5 AW |
2747 | xlate_table_action(ctx, flow->in_port.ofp_port, table_id, |
2748 | true, true); | |
9583bc14 | 2749 | } else { |
9efd308e DV |
2750 | /* Forwarding is disabled by STP and RSTP. Let OFPP_NORMAL and |
2751 | * the learning action look at the packet, then drop it. */ | |
9583bc14 | 2752 | struct flow old_base_flow = ctx->base_flow; |
6fd6ed71 | 2753 | size_t old_size = ctx->xout->odp_actions->size; |
cdf5d3a5 | 2754 | mirror_mask_t old_mirrors = ctx->xout->mirrors; |
0c7812e5 AW |
2755 | xlate_table_action(ctx, flow->in_port.ofp_port, table_id, |
2756 | true, true); | |
cdf5d3a5 | 2757 | ctx->xout->mirrors = old_mirrors; |
9583bc14 | 2758 | ctx->base_flow = old_base_flow; |
6fd6ed71 | 2759 | ctx->xout->odp_actions->size = old_size; |
9583bc14 EJ |
2760 | } |
2761 | } | |
2762 | ||
2763 | ctx->xin->flow = old_flow; | |
832554e3 | 2764 | ctx->xbridge = xport->xbridge; |
9583bc14 EJ |
2765 | |
2766 | if (ctx->xin->resubmit_stats) { | |
46c88433 EJ |
2767 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
2768 | netdev_vport_inc_rx(peer->netdev, ctx->xin->resubmit_stats); | |
a1aeea86 AW |
2769 | if (peer->bfd) { |
2770 | bfd_account_rx(peer->bfd, ctx->xin->resubmit_stats); | |
2771 | } | |
9583bc14 | 2772 | } |
b256dc52 JS |
2773 | if (ctx->xin->xcache) { |
2774 | struct xc_entry *entry; | |
2775 | ||
2776 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
2777 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
2778 | entry->u.dev.rx = netdev_ref(peer->netdev); | |
2779 | entry->u.dev.bfd = bfd_ref(peer->bfd); | |
2780 | } | |
9583bc14 EJ |
2781 | return; |
2782 | } | |
2783 | ||
33bf9176 | 2784 | flow_vlan_tci = flow->vlan_tci; |
1362e248 | 2785 | flow_pkt_mark = flow->pkt_mark; |
33bf9176 | 2786 | flow_nw_tos = flow->nw_tos; |
9583bc14 | 2787 | |
16194afd DDP |
2788 | if (count_skb_priorities(xport)) { |
2789 | memset(&wc->masks.skb_priority, 0xff, sizeof wc->masks.skb_priority); | |
2790 | if (dscp_from_skb_priority(xport, flow->skb_priority, &dscp)) { | |
2791 | wc->masks.nw_tos |= IP_DSCP_MASK; | |
2792 | flow->nw_tos &= ~IP_DSCP_MASK; | |
2793 | flow->nw_tos |= dscp; | |
2794 | } | |
9583bc14 EJ |
2795 | } |
2796 | ||
46c88433 | 2797 | if (xport->is_tunnel) { |
9583bc14 EJ |
2798 | /* Save tunnel metadata so that changes made due to |
2799 | * the Logical (tunnel) Port are not visible for any further | |
2800 | * matches, while explicit set actions on tunnel metadata are. | |
2801 | */ | |
a36de779 | 2802 | flow_tnl = flow->tunnel; |
46c88433 | 2803 | odp_port = tnl_port_send(xport->ofport, flow, &ctx->xout->wc); |
4e022ec0 | 2804 | if (odp_port == ODPP_NONE) { |
9583bc14 EJ |
2805 | xlate_report(ctx, "Tunneling decided against output"); |
2806 | goto out; /* restore flow_nw_tos */ | |
2807 | } | |
33bf9176 | 2808 | if (flow->tunnel.ip_dst == ctx->orig_tunnel_ip_dst) { |
9583bc14 EJ |
2809 | xlate_report(ctx, "Not tunneling to our own address"); |
2810 | goto out; /* restore flow_nw_tos */ | |
2811 | } | |
2812 | if (ctx->xin->resubmit_stats) { | |
46c88433 | 2813 | netdev_vport_inc_tx(xport->netdev, ctx->xin->resubmit_stats); |
9583bc14 | 2814 | } |
b256dc52 JS |
2815 | if (ctx->xin->xcache) { |
2816 | struct xc_entry *entry; | |
2817 | ||
2818 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_NETDEV); | |
2819 | entry->u.dev.tx = netdev_ref(xport->netdev); | |
2820 | } | |
9583bc14 | 2821 | out_port = odp_port; |
a36de779 PS |
2822 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
2823 | tnl_push_pop_send = true; | |
2824 | } else { | |
2825 | commit_odp_tunnel_action(flow, &ctx->base_flow, | |
2826 | ctx->xout->odp_actions); | |
2827 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ | |
2828 | } | |
9583bc14 | 2829 | } else { |
46c88433 | 2830 | odp_port = xport->odp_port; |
7614e5d0 | 2831 | out_port = odp_port; |
46c88433 | 2832 | if (ofproto_has_vlan_splinters(ctx->xbridge->ofproto)) { |
7614e5d0 JR |
2833 | ofp_port_t vlandev_port; |
2834 | ||
1dd35f8a | 2835 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
7614e5d0 JR |
2836 | vlandev_port = vsp_realdev_to_vlandev(ctx->xbridge->ofproto, |
2837 | ofp_port, flow->vlan_tci); | |
2838 | if (vlandev_port != ofp_port) { | |
2839 | out_port = ofp_port_to_odp_port(ctx->xbridge, vlandev_port); | |
2840 | flow->vlan_tci = htons(0); | |
2841 | } | |
9583bc14 | 2842 | } |
9583bc14 | 2843 | } |
9583bc14 | 2844 | |
4e022ec0 | 2845 | if (out_port != ODPP_NONE) { |
7fd91025 | 2846 | ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow, |
cc377352 | 2847 | ctx->xout->odp_actions, |
a36de779 | 2848 | wc, |
d23df9a8 | 2849 | ctx->xbridge->masked_set_action); |
adcf00ba | 2850 | |
92c08f09 | 2851 | if (ctx->use_recirc) { |
347bf289 | 2852 | struct ovs_action_hash *act_hash; |
92c08f09 | 2853 | struct xlate_recirc *xr = &ctx->recirc; |
adcf00ba | 2854 | |
347bf289 | 2855 | /* Hash action. */ |
cc377352 | 2856 | act_hash = nl_msg_put_unspec_uninit(ctx->xout->odp_actions, |
347bf289 AZ |
2857 | OVS_ACTION_ATTR_HASH, |
2858 | sizeof *act_hash); | |
2859 | act_hash->hash_alg = xr->hash_alg; | |
62ac1f20 | 2860 | act_hash->hash_basis = xr->hash_basis; |
347bf289 AZ |
2861 | |
2862 | /* Recirc action. */ | |
cc377352 | 2863 | nl_msg_put_u32(ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC, |
347bf289 | 2864 | xr->recirc_id); |
adcf00ba | 2865 | } else { |
a36de779 PS |
2866 | |
2867 | if (tnl_push_pop_send) { | |
2868 | build_tunnel_send(ctx, xport, flow, odp_port); | |
2869 | flow->tunnel = flow_tnl; /* Restore tunnel metadata */ | |
2870 | } else { | |
2871 | odp_port_t odp_tnl_port = ODPP_NONE; | |
2872 | ||
2873 | /* XXX: Write better Filter for tunnel port. We can use inport | |
2874 | * int tunnel-port flow to avoid these checks completely. */ | |
2875 | if (ofp_port == OFPP_LOCAL && | |
2876 | ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { | |
2877 | ||
2878 | odp_tnl_port = tnl_port_map_lookup(flow, wc); | |
2879 | } | |
2880 | ||
2881 | if (odp_tnl_port != ODPP_NONE) { | |
2882 | nl_msg_put_odp_port(ctx->xout->odp_actions, | |
2883 | OVS_ACTION_ATTR_TUNNEL_POP, | |
2884 | odp_tnl_port); | |
2885 | } else { | |
2886 | /* Tunnel push-pop action is not compatible with | |
2887 | * IPFIX action. */ | |
2888 | add_ipfix_output_action(ctx, out_port); | |
2889 | nl_msg_put_odp_port(ctx->xout->odp_actions, | |
2890 | OVS_ACTION_ATTR_OUTPUT, | |
2891 | out_port); | |
2892 | } | |
2893 | } | |
adcf00ba | 2894 | } |
9583bc14 | 2895 | |
6cbbf4fa EJ |
2896 | ctx->sflow_odp_port = odp_port; |
2897 | ctx->sflow_n_outputs++; | |
2898 | ctx->xout->nf_output_iface = ofp_port; | |
2899 | } | |
2900 | ||
2901 | out: | |
9583bc14 | 2902 | /* Restore flow */ |
33bf9176 | 2903 | flow->vlan_tci = flow_vlan_tci; |
1362e248 | 2904 | flow->pkt_mark = flow_pkt_mark; |
33bf9176 | 2905 | flow->nw_tos = flow_nw_tos; |
9583bc14 EJ |
2906 | } |
2907 | ||
2908 | static void | |
4e022ec0 | 2909 | compose_output_action(struct xlate_ctx *ctx, ofp_port_t ofp_port) |
9583bc14 EJ |
2910 | { |
2911 | compose_output_action__(ctx, ofp_port, true); | |
2912 | } | |
2913 | ||
bb61b33d BP |
2914 | static void |
2915 | xlate_recursively(struct xlate_ctx *ctx, struct rule_dpif *rule) | |
bb61b33d BP |
2916 | { |
2917 | struct rule_dpif *old_rule = ctx->rule; | |
dc723c44 | 2918 | const struct rule_actions *actions; |
bb61b33d BP |
2919 | |
2920 | if (ctx->xin->resubmit_stats) { | |
70742c7f | 2921 | rule_dpif_credit_stats(rule, ctx->xin->resubmit_stats); |
bb61b33d BP |
2922 | } |
2923 | ||
98b07853 | 2924 | ctx->resubmits++; |
bb61b33d BP |
2925 | ctx->recurse++; |
2926 | ctx->rule = rule; | |
6f00e29b BP |
2927 | actions = rule_dpif_get_actions(rule); |
2928 | do_xlate_actions(actions->ofpacts, actions->ofpacts_len, ctx); | |
bb61b33d BP |
2929 | ctx->rule = old_rule; |
2930 | ctx->recurse--; | |
bb61b33d BP |
2931 | } |
2932 | ||
bd3240ba SH |
2933 | static bool |
2934 | xlate_resubmit_resource_check(struct xlate_ctx *ctx) | |
9583bc14 | 2935 | { |
98b07853 BP |
2936 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); |
2937 | ||
adcf00ba | 2938 | if (ctx->recurse >= MAX_RESUBMIT_RECURSION + MAX_INTERNAL_RESUBMITS) { |
98b07853 BP |
2939 | VLOG_ERR_RL(&rl, "resubmit actions recursed over %d times", |
2940 | MAX_RESUBMIT_RECURSION); | |
adcf00ba | 2941 | } else if (ctx->resubmits >= MAX_RESUBMITS + MAX_INTERNAL_RESUBMITS) { |
98b07853 | 2942 | VLOG_ERR_RL(&rl, "over %d resubmit actions", MAX_RESUBMITS); |
6fd6ed71 | 2943 | } else if (ctx->xout->odp_actions->size > UINT16_MAX) { |
98b07853 | 2944 | VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of actions"); |
6fd6ed71 | 2945 | } else if (ctx->stack.size >= 65536) { |
98b07853 BP |
2946 | VLOG_ERR_RL(&rl, "resubmits yielded over 64 kB of stack"); |
2947 | } else { | |
bd3240ba SH |
2948 | return true; |
2949 | } | |
2950 | ||
2951 | return false; | |
2952 | } | |
2953 | ||
2954 | static void | |
6d328fa2 SH |
2955 | xlate_table_action(struct xlate_ctx *ctx, ofp_port_t in_port, uint8_t table_id, |
2956 | bool may_packet_in, bool honor_table_miss) | |
bd3240ba SH |
2957 | { |
2958 | if (xlate_resubmit_resource_check(ctx)) { | |
34dd0d78 | 2959 | struct flow_wildcards *wc; |
9583bc14 | 2960 | uint8_t old_table_id = ctx->table_id; |
3f207910 | 2961 | struct rule_dpif *rule; |
9583bc14 EJ |
2962 | |
2963 | ctx->table_id = table_id; | |
34dd0d78 | 2964 | wc = (ctx->xin->skip_wildcards) ? NULL : &ctx->xout->wc; |
9583bc14 | 2965 | |
34dd0d78 JR |
2966 | rule = rule_dpif_lookup_from_table(ctx->xbridge->ofproto, |
2967 | &ctx->xin->flow, wc, | |
2968 | ctx->xin->xcache != NULL, | |
2969 | ctx->xin->resubmit_stats, | |
2970 | &ctx->table_id, in_port, | |
2971 | may_packet_in, honor_table_miss); | |
9583bc14 | 2972 | |
a8c31348 BP |
2973 | if (OVS_UNLIKELY(ctx->xin->resubmit_hook)) { |
2974 | ctx->xin->resubmit_hook(ctx->xin, rule, ctx->recurse + 1); | |
ad3efdcb EJ |
2975 | } |
2976 | ||
a2143702 | 2977 | if (rule) { |
83709dfa JR |
2978 | /* Fill in the cache entry here instead of xlate_recursively |
2979 | * to make the reference counting more explicit. We take a | |
2980 | * reference in the lookups above if we are going to cache the | |
2981 | * rule. */ | |
2982 | if (ctx->xin->xcache) { | |
2983 | struct xc_entry *entry; | |
2984 | ||
2985 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_RULE); | |
2986 | entry->u.rule = rule; | |
2987 | } | |
bb61b33d | 2988 | xlate_recursively(ctx, rule); |
ad3efdcb EJ |
2989 | } |
2990 | ||
9583bc14 | 2991 | ctx->table_id = old_table_id; |
98b07853 | 2992 | return; |
9583bc14 | 2993 | } |
98b07853 BP |
2994 | |
2995 | ctx->exit = true; | |
9583bc14 EJ |
2996 | } |
2997 | ||
f4fb341b | 2998 | static void |
1e684d7d RW |
2999 | xlate_group_stats(struct xlate_ctx *ctx, struct group_dpif *group, |
3000 | struct ofputil_bucket *bucket) | |
3001 | { | |
3002 | if (ctx->xin->resubmit_stats) { | |
3003 | group_dpif_credit_stats(group, bucket, ctx->xin->resubmit_stats); | |
3004 | } | |
3005 | if (ctx->xin->xcache) { | |
3006 | struct xc_entry *entry; | |
3007 | ||
3008 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_GROUP); | |
3009 | entry->u.group.group = group_dpif_ref(group); | |
3010 | entry->u.group.bucket = bucket; | |
3011 | } | |
3012 | } | |
3013 | ||
3014 | static void | |
3015 | xlate_group_bucket(struct xlate_ctx *ctx, struct ofputil_bucket *bucket) | |
f4fb341b SH |
3016 | { |
3017 | uint64_t action_list_stub[1024 / 8]; | |
3018 | struct ofpbuf action_list, action_set; | |
3019 | ||
3020 | ofpbuf_use_const(&action_set, bucket->ofpacts, bucket->ofpacts_len); | |
3021 | ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub); | |
3022 | ||
3023 | ofpacts_execute_action_set(&action_list, &action_set); | |
3024 | ctx->recurse++; | |
6fd6ed71 | 3025 | do_xlate_actions(action_list.data, action_list.size, ctx); |
f4fb341b SH |
3026 | ctx->recurse--; |
3027 | ||
3028 | ofpbuf_uninit(&action_set); | |
3029 | ofpbuf_uninit(&action_list); | |
3030 | } | |
3031 | ||
3032 | static void | |
3033 | xlate_all_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3034 | { | |
1e684d7d | 3035 | struct ofputil_bucket *bucket; |
ca6ba700 | 3036 | const struct ovs_list *buckets; |
f4fb341b SH |
3037 | struct flow old_flow = ctx->xin->flow; |
3038 | ||
3039 | group_dpif_get_buckets(group, &buckets); | |
3040 | ||
3041 | LIST_FOR_EACH (bucket, list_node, buckets) { | |
3042 | xlate_group_bucket(ctx, bucket); | |
3043 | /* Roll back flow to previous state. | |
3044 | * This is equivalent to cloning the packet for each bucket. | |
3045 | * | |
3046 | * As a side effect any subsequently applied actions will | |
3047 | * also effectively be applied to a clone of the packet taken | |
3048 | * just before applying the all or indirect group. */ | |
3049 | ctx->xin->flow = old_flow; | |
3050 | } | |
1e684d7d | 3051 | xlate_group_stats(ctx, group, NULL); |
f4fb341b SH |
3052 | } |
3053 | ||
dd8cd4b4 SH |
3054 | static void |
3055 | xlate_ff_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3056 | { | |
1e684d7d | 3057 | struct ofputil_bucket *bucket; |
dd8cd4b4 SH |
3058 | |
3059 | bucket = group_first_live_bucket(ctx, group, 0); | |
3060 | if (bucket) { | |
3061 | xlate_group_bucket(ctx, bucket); | |
1e684d7d | 3062 | xlate_group_stats(ctx, group, bucket); |
dd8cd4b4 SH |
3063 | } |
3064 | } | |
3065 | ||
fe7e5749 SH |
3066 | static void |
3067 | xlate_select_group(struct xlate_ctx *ctx, struct group_dpif *group) | |
3068 | { | |
3069 | struct flow_wildcards *wc = &ctx->xout->wc; | |
1e684d7d | 3070 | struct ofputil_bucket *bucket; |
fe7e5749 SH |
3071 | uint32_t basis; |
3072 | ||
1d1aae0b | 3073 | basis = flow_hash_symmetric_l4(&ctx->xin->flow, 0); |
fe7e5749 SH |
3074 | bucket = group_best_live_bucket(ctx, group, basis); |
3075 | if (bucket) { | |
3076 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); | |
1d1aae0b SS |
3077 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
3078 | memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type); | |
3079 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); | |
3080 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); | |
3081 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); | |
3082 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
3083 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
3084 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); | |
3085 | ||
fe7e5749 | 3086 | xlate_group_bucket(ctx, bucket); |
1e684d7d | 3087 | xlate_group_stats(ctx, group, bucket); |
fe7e5749 SH |
3088 | } |
3089 | } | |
3090 | ||
f4fb341b SH |
3091 | static void |
3092 | xlate_group_action__(struct xlate_ctx *ctx, struct group_dpif *group) | |
3093 | { | |
5a070238 BP |
3094 | ctx->in_group = true; |
3095 | ||
f4fb341b SH |
3096 | switch (group_dpif_get_type(group)) { |
3097 | case OFPGT11_ALL: | |
3098 | case OFPGT11_INDIRECT: | |
3099 | xlate_all_group(ctx, group); | |
3100 | break; | |
3101 | case OFPGT11_SELECT: | |
fe7e5749 | 3102 | xlate_select_group(ctx, group); |
f4fb341b | 3103 | break; |
dd8cd4b4 SH |
3104 | case OFPGT11_FF: |
3105 | xlate_ff_group(ctx, group); | |
3106 | break; | |
f4fb341b | 3107 | default: |
428b2edd | 3108 | OVS_NOT_REACHED(); |
f4fb341b | 3109 | } |
809c7548 | 3110 | group_dpif_unref(group); |
5a070238 BP |
3111 | |
3112 | ctx->in_group = false; | |
3113 | } | |
3114 | ||
3115 | static bool | |
3116 | xlate_group_resource_check(struct xlate_ctx *ctx) | |
3117 | { | |
3118 | if (!xlate_resubmit_resource_check(ctx)) { | |
3119 | return false; | |
3120 | } else if (ctx->in_group) { | |
3121 | /* Prevent nested translation of OpenFlow groups. | |
3122 | * | |
3123 | * OpenFlow allows this restriction. We enforce this restriction only | |
3124 | * because, with the current architecture, we would otherwise have to | |
3125 | * take a possibly recursive read lock on the ofgroup rwlock, which is | |
3126 | * unsafe given that POSIX allows taking a read lock to block if there | |
3127 | * is a thread blocked on taking the write lock. Other solutions | |
3128 | * without this restriction are also possible, but seem unwarranted | |
3129 | * given the current limited use of groups. */ | |
3130 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
3131 | ||
3132 | VLOG_ERR_RL(&rl, "cannot recursively translate OpenFlow group"); | |
3133 | return false; | |
3134 | } else { | |
3135 | return true; | |
3136 | } | |
f4fb341b SH |
3137 | } |
3138 | ||
3139 | static bool | |
3140 | xlate_group_action(struct xlate_ctx *ctx, uint32_t group_id) | |
3141 | { | |
5a070238 | 3142 | if (xlate_group_resource_check(ctx)) { |
f4fb341b SH |
3143 | struct group_dpif *group; |
3144 | bool got_group; | |
3145 | ||
3146 | got_group = group_dpif_lookup(ctx->xbridge->ofproto, group_id, &group); | |
3147 | if (got_group) { | |
3148 | xlate_group_action__(ctx, group); | |
3149 | } else { | |
3150 | return true; | |
3151 | } | |
3152 | } | |
3153 | ||
3154 | return false; | |
3155 | } | |
3156 | ||
9583bc14 EJ |
3157 | static void |
3158 | xlate_ofpact_resubmit(struct xlate_ctx *ctx, | |
3159 | const struct ofpact_resubmit *resubmit) | |
3160 | { | |
4e022ec0 | 3161 | ofp_port_t in_port; |
9583bc14 | 3162 | uint8_t table_id; |
adcf00ba AZ |
3163 | bool may_packet_in = false; |
3164 | bool honor_table_miss = false; | |
3165 | ||
3166 | if (ctx->rule && rule_dpif_is_internal(ctx->rule)) { | |
3167 | /* Still allow missed packets to be sent to the controller | |
3168 | * if resubmitting from an internal table. */ | |
3169 | may_packet_in = true; | |
3170 | honor_table_miss = true; | |
3171 | } | |
9583bc14 EJ |
3172 | |
3173 | in_port = resubmit->in_port; | |
3174 | if (in_port == OFPP_IN_PORT) { | |
4e022ec0 | 3175 | in_port = ctx->xin->flow.in_port.ofp_port; |
9583bc14 EJ |
3176 | } |
3177 | ||
3178 | table_id = resubmit->table_id; | |
3179 | if (table_id == 255) { | |
3180 | table_id = ctx->table_id; | |
3181 | } | |
3182 | ||
adcf00ba AZ |
3183 | xlate_table_action(ctx, in_port, table_id, may_packet_in, |
3184 | honor_table_miss); | |
9583bc14 EJ |
3185 | } |
3186 | ||
3187 | static void | |
3188 | flood_packets(struct xlate_ctx *ctx, bool all) | |
3189 | { | |
46c88433 | 3190 | const struct xport *xport; |
9583bc14 | 3191 | |
46c88433 EJ |
3192 | HMAP_FOR_EACH (xport, ofp_node, &ctx->xbridge->xports) { |
3193 | if (xport->ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3194 | continue; |
3195 | } | |
3196 | ||
3197 | if (all) { | |
46c88433 EJ |
3198 | compose_output_action__(ctx, xport->ofp_port, false); |
3199 | } else if (!(xport->config & OFPUTIL_PC_NO_FLOOD)) { | |
3200 | compose_output_action(ctx, xport->ofp_port); | |
9583bc14 EJ |
3201 | } |
3202 | } | |
3203 | ||
3204 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
3205 | } | |
3206 | ||
3207 | static void | |
3208 | execute_controller_action(struct xlate_ctx *ctx, int len, | |
3209 | enum ofp_packet_in_reason reason, | |
3210 | uint16_t controller_id) | |
3211 | { | |
0fb7792a | 3212 | struct ofproto_packet_in *pin; |
e14deea0 | 3213 | struct dp_packet *packet; |
9583bc14 | 3214 | |
04594cd5 | 3215 | ctx->xout->slow |= SLOW_CONTROLLER; |
9583bc14 EJ |
3216 | if (!ctx->xin->packet) { |
3217 | return; | |
3218 | } | |
3219 | ||
cf62fa4c | 3220 | packet = dp_packet_clone(ctx->xin->packet); |
9583bc14 | 3221 | |
7fd91025 | 3222 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, |
cc377352 | 3223 | ctx->xout->odp_actions, |
d23df9a8 JR |
3224 | &ctx->xout->wc, |
3225 | ctx->xbridge->masked_set_action); | |
9583bc14 | 3226 | |
41ccaa24 | 3227 | odp_execute_actions(NULL, &packet, 1, false, |
6fd6ed71 PS |
3228 | ctx->xout->odp_actions->data, |
3229 | ctx->xout->odp_actions->size, NULL); | |
9583bc14 | 3230 | |
ada3a58d | 3231 | pin = xmalloc(sizeof *pin); |
cf62fa4c PS |
3232 | pin->up.packet_len = dp_packet_size(packet); |
3233 | pin->up.packet = dp_packet_steal_data(packet); | |
0fb7792a | 3234 | pin->up.reason = reason; |
0fb7792a | 3235 | pin->up.table_id = ctx->table_id; |
d4fa4e79 BP |
3236 | pin->up.cookie = (ctx->rule |
3237 | ? rule_dpif_get_flow_cookie(ctx->rule) | |
3238 | : OVS_BE64_MAX); | |
0fb7792a | 3239 | |
0fb7792a | 3240 | flow_get_metadata(&ctx->xin->flow, &pin->up.fmd); |
9583bc14 | 3241 | |
f11c7538 | 3242 | pin->controller_id = controller_id; |
d38a3c7b | 3243 | pin->send_len = len; |
32260212 SH |
3244 | /* If a rule is a table-miss rule then this is |
3245 | * a table-miss handled by a table-miss rule. | |
3246 | * | |
3247 | * Else, if rule is internal and has a controller action, | |
3248 | * the later being implied by the rule being processed here, | |
3249 | * then this is a table-miss handled without a table-miss rule. | |
3250 | * | |
3251 | * Otherwise this is not a table-miss. */ | |
3252 | pin->miss_type = OFPROTO_PACKET_IN_NO_MISS; | |
3253 | if (ctx->rule) { | |
3254 | if (rule_dpif_is_table_miss(ctx->rule)) { | |
3255 | pin->miss_type = OFPROTO_PACKET_IN_MISS_FLOW; | |
3256 | } else if (rule_dpif_is_internal(ctx->rule)) { | |
3257 | pin->miss_type = OFPROTO_PACKET_IN_MISS_WITHOUT_FLOW; | |
3258 | } | |
3259 | } | |
ada3a58d | 3260 | ofproto_dpif_send_packet_in(ctx->xbridge->ofproto, pin); |
e14deea0 | 3261 | dp_packet_delete(packet); |
9583bc14 EJ |
3262 | } |
3263 | ||
7bbdd84f SH |
3264 | static void |
3265 | compose_recirculate_action(struct xlate_ctx *ctx, | |
3266 | const struct ofpact *ofpacts_base, | |
3267 | const struct ofpact *ofpact_current, | |
3268 | size_t ofpacts_base_len) | |
3269 | { | |
3270 | uint32_t id; | |
3271 | int error; | |
3272 | unsigned ofpacts_len; | |
3273 | struct match match; | |
3274 | struct rule *rule; | |
3275 | struct ofpbuf ofpacts; | |
3276 | ||
3277 | ctx->exit = true; | |
3278 | ||
3279 | ofpacts_len = ofpacts_base_len - | |
3280 | ((uint8_t *)ofpact_current - (uint8_t *)ofpacts_base); | |
3281 | ||
3282 | if (ctx->rule) { | |
3283 | id = rule_dpif_get_recirc_id(ctx->rule); | |
3284 | } else { | |
3285 | /* In the case where ctx has no rule then allocate a recirc id. | |
3286 | * The life-cycle of this recirc id is managed by associating it | |
3287 | * with the internal rule that is created to to handle | |
3288 | * recirculation below. | |
3289 | * | |
3290 | * The known use-case of this is packet_out which | |
3291 | * translates actions without a rule */ | |
3292 | id = ofproto_dpif_alloc_recirc_id(ctx->xbridge->ofproto); | |
3293 | } | |
3294 | if (!id) { | |
3295 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3296 | VLOG_ERR_RL(&rl, "Failed to allocate recirculation id"); | |
3297 | return; | |
3298 | } | |
3299 | ||
3300 | match_init_catchall(&match); | |
3301 | match_set_recirc_id(&match, id); | |
3302 | ofpbuf_use_const(&ofpacts, ofpact_current, ofpacts_len); | |
3303 | error = ofproto_dpif_add_internal_flow(ctx->xbridge->ofproto, &match, | |
3304 | RECIRC_RULE_PRIORITY, | |
3305 | RECIRC_TIMEOUT, &ofpacts, &rule); | |
3306 | if (error) { | |
3307 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3308 | VLOG_ERR_RL(&rl, "Failed to add post recirculation flow %s", | |
3309 | match_to_string(&match, 0)); | |
3310 | return; | |
3311 | } | |
3312 | /* If ctx has no rule then associate the recirc id, which | |
3313 | * was allocated above, with the internal rule. This allows | |
3314 | * the recirc id to be released when the internal rule times out. */ | |
3315 | if (!ctx->rule) { | |
3316 | rule_set_recirc_id(rule, id); | |
3317 | } | |
3318 | ||
3319 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, | |
cc377352 | 3320 | ctx->xout->odp_actions, |
d23df9a8 JR |
3321 | &ctx->xout->wc, |
3322 | ctx->xbridge->masked_set_action); | |
cc377352 | 3323 | nl_msg_put_u32(ctx->xout->odp_actions, OVS_ACTION_ATTR_RECIRC, id); |
7bbdd84f SH |
3324 | } |
3325 | ||
8bfd0fda BP |
3326 | static void |
3327 | compose_mpls_push_action(struct xlate_ctx *ctx, struct ofpact_push_mpls *mpls) | |
9583bc14 | 3328 | { |
33bf9176 BP |
3329 | struct flow_wildcards *wc = &ctx->xout->wc; |
3330 | struct flow *flow = &ctx->xin->flow; | |
8bfd0fda | 3331 | int n; |
33bf9176 | 3332 | |
8bfd0fda | 3333 | ovs_assert(eth_type_mpls(mpls->ethertype)); |
b0a17866 | 3334 | |
8bfd0fda BP |
3335 | n = flow_count_mpls_labels(flow, wc); |
3336 | if (!n) { | |
8bfd0fda | 3337 | ctx->xout->slow |= commit_odp_actions(flow, &ctx->base_flow, |
cc377352 | 3338 | ctx->xout->odp_actions, |
d23df9a8 JR |
3339 | &ctx->xout->wc, |
3340 | ctx->xbridge->masked_set_action); | |
8bfd0fda BP |
3341 | } else if (n >= FLOW_MAX_MPLS_LABELS) { |
3342 | if (ctx->xin->packet != NULL) { | |
3343 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3344 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an " | |
3345 | "MPLS push action can't be performed as it would " | |
3346 | "have more MPLS LSEs than the %d supported.", | |
3347 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
9583bc14 | 3348 | } |
8bfd0fda BP |
3349 | ctx->exit = true; |
3350 | return; | |
9583bc14 | 3351 | } |
b0a17866 | 3352 | |
8bfd0fda | 3353 | flow_push_mpls(flow, n, mpls->ethertype, wc); |
9583bc14 EJ |
3354 | } |
3355 | ||
8bfd0fda | 3356 | static void |
9cfef3d0 | 3357 | compose_mpls_pop_action(struct xlate_ctx *ctx, ovs_be16 eth_type) |
9583bc14 | 3358 | { |
33bf9176 | 3359 | struct flow_wildcards *wc = &ctx->xout->wc; |
8bfd0fda BP |
3360 | struct flow *flow = &ctx->xin->flow; |
3361 | int n = flow_count_mpls_labels(flow, wc); | |
33bf9176 | 3362 | |
7bbdd84f | 3363 | if (flow_pop_mpls(flow, n, eth_type, wc)) { |
5af43325 | 3364 | if (ctx->xbridge->enable_recirc) { |
7bbdd84f SH |
3365 | ctx->was_mpls = true; |
3366 | } | |
3367 | } else if (n >= FLOW_MAX_MPLS_LABELS) { | |
8bfd0fda BP |
3368 | if (ctx->xin->packet != NULL) { |
3369 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
3370 | VLOG_WARN_RL(&rl, "bridge %s: dropping packet on which an " | |
3371 | "MPLS pop action can't be performed as it has " | |
3372 | "more MPLS LSEs than the %d supported.", | |
3373 | ctx->xbridge->name, FLOW_MAX_MPLS_LABELS); | |
3374 | } | |
3375 | ctx->exit = true; | |
cc377352 | 3376 | ofpbuf_clear(ctx->xout->odp_actions); |
9583bc14 EJ |
3377 | } |
3378 | } | |
3379 | ||
3380 | static bool | |
3381 | compose_dec_ttl(struct xlate_ctx *ctx, struct ofpact_cnt_ids *ids) | |
3382 | { | |
33bf9176 BP |
3383 | struct flow *flow = &ctx->xin->flow; |
3384 | ||
3385 | if (!is_ip_any(flow)) { | |
9583bc14 EJ |
3386 | return false; |
3387 | } | |
3388 | ||
1dd35f8a | 3389 | ctx->xout->wc.masks.nw_ttl = 0xff; |
33bf9176 BP |
3390 | if (flow->nw_ttl > 1) { |
3391 | flow->nw_ttl--; | |
9583bc14 EJ |
3392 | return false; |
3393 | } else { | |
3394 | size_t i; | |
3395 | ||
3396 | for (i = 0; i < ids->n_controllers; i++) { | |
3397 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, | |
3398 | ids->cnt_ids[i]); | |
3399 | } | |
3400 | ||
3401 | /* Stop processing for current table. */ | |
3402 | return true; | |
3403 | } | |
3404 | } | |
3405 | ||
8bfd0fda | 3406 | static void |
097d4939 JR |
3407 | compose_set_mpls_label_action(struct xlate_ctx *ctx, ovs_be32 label) |
3408 | { | |
8bfd0fda BP |
3409 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
3410 | ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_LABEL_MASK); | |
3411 | set_mpls_lse_label(&ctx->xin->flow.mpls_lse[0], label); | |
097d4939 | 3412 | } |
097d4939 JR |
3413 | } |
3414 | ||
8bfd0fda | 3415 | static void |
097d4939 JR |
3416 | compose_set_mpls_tc_action(struct xlate_ctx *ctx, uint8_t tc) |
3417 | { | |
8bfd0fda BP |
3418 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
3419 | ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_TC_MASK); | |
3420 | set_mpls_lse_tc(&ctx->xin->flow.mpls_lse[0], tc); | |
097d4939 | 3421 | } |
097d4939 JR |
3422 | } |
3423 | ||
8bfd0fda | 3424 | static void |
9cfef3d0 | 3425 | compose_set_mpls_ttl_action(struct xlate_ctx *ctx, uint8_t ttl) |
9583bc14 | 3426 | { |
8bfd0fda BP |
3427 | if (eth_type_mpls(ctx->xin->flow.dl_type)) { |
3428 | ctx->xout->wc.masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); | |
3429 | set_mpls_lse_ttl(&ctx->xin->flow.mpls_lse[0], ttl); | |
b0a17866 | 3430 | } |
9583bc14 EJ |
3431 | } |
3432 | ||
3433 | static bool | |
9cfef3d0 | 3434 | compose_dec_mpls_ttl_action(struct xlate_ctx *ctx) |
9583bc14 | 3435 | { |
33bf9176 | 3436 | struct flow *flow = &ctx->xin->flow; |
1dd35f8a JP |
3437 | struct flow_wildcards *wc = &ctx->xout->wc; |
3438 | ||
8bfd0fda | 3439 | if (eth_type_mpls(flow->dl_type)) { |
22d38fca JR |
3440 | uint8_t ttl = mpls_lse_to_ttl(flow->mpls_lse[0]); |
3441 | ||
3442 | wc->masks.mpls_lse[0] |= htonl(MPLS_TTL_MASK); | |
8bfd0fda BP |
3443 | if (ttl > 1) { |
3444 | ttl--; | |
3445 | set_mpls_lse_ttl(&flow->mpls_lse[0], ttl); | |
3446 | return false; | |
3447 | } else { | |
3448 | execute_controller_action(ctx, UINT16_MAX, OFPR_INVALID_TTL, 0); | |
8bfd0fda | 3449 | } |
9583bc14 | 3450 | } |
22d38fca JR |
3451 | |
3452 | /* Stop processing for current table. */ | |
3453 | return true; | |
9583bc14 EJ |
3454 | } |
3455 | ||
3456 | static void | |
3457 | xlate_output_action(struct xlate_ctx *ctx, | |
4e022ec0 | 3458 | ofp_port_t port, uint16_t max_len, bool may_packet_in) |
9583bc14 | 3459 | { |
4e022ec0 | 3460 | ofp_port_t prev_nf_output_iface = ctx->xout->nf_output_iface; |
9583bc14 EJ |
3461 | |
3462 | ctx->xout->nf_output_iface = NF_OUT_DROP; | |
3463 | ||
3464 | switch (port) { | |
3465 | case OFPP_IN_PORT: | |
4e022ec0 | 3466 | compose_output_action(ctx, ctx->xin->flow.in_port.ofp_port); |
9583bc14 EJ |
3467 | break; |
3468 | case OFPP_TABLE: | |
4e022ec0 | 3469 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 3470 | 0, may_packet_in, true); |
9583bc14 EJ |
3471 | break; |
3472 | case OFPP_NORMAL: | |
3473 | xlate_normal(ctx); | |
3474 | break; | |
3475 | case OFPP_FLOOD: | |
3476 | flood_packets(ctx, false); | |
3477 | break; | |
3478 | case OFPP_ALL: | |
3479 | flood_packets(ctx, true); | |
3480 | break; | |
3481 | case OFPP_CONTROLLER: | |
3a11fd5b | 3482 | execute_controller_action(ctx, max_len, |
029ca940 SS |
3483 | (ctx->in_group ? OFPR_GROUP |
3484 | : ctx->in_action_set ? OFPR_ACTION_SET | |
3485 | : OFPR_ACTION), | |
3486 | 0); | |
9583bc14 EJ |
3487 | break; |
3488 | case OFPP_NONE: | |
3489 | break; | |
3490 | case OFPP_LOCAL: | |
3491 | default: | |
4e022ec0 | 3492 | if (port != ctx->xin->flow.in_port.ofp_port) { |
9583bc14 EJ |
3493 | compose_output_action(ctx, port); |
3494 | } else { | |
3495 | xlate_report(ctx, "skipping output to input port"); | |
3496 | } | |
3497 | break; | |
3498 | } | |
3499 | ||
3500 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
3501 | ctx->xout->nf_output_iface = NF_OUT_FLOOD; | |
3502 | } else if (ctx->xout->nf_output_iface == NF_OUT_DROP) { | |
3503 | ctx->xout->nf_output_iface = prev_nf_output_iface; | |
3504 | } else if (prev_nf_output_iface != NF_OUT_DROP && | |
3505 | ctx->xout->nf_output_iface != NF_OUT_FLOOD) { | |
3506 | ctx->xout->nf_output_iface = NF_OUT_MULTI; | |
3507 | } | |
3508 | } | |
3509 | ||
3510 | static void | |
3511 | xlate_output_reg_action(struct xlate_ctx *ctx, | |
3512 | const struct ofpact_output_reg *or) | |
3513 | { | |
3514 | uint64_t port = mf_get_subfield(&or->src, &ctx->xin->flow); | |
3515 | if (port <= UINT16_MAX) { | |
3516 | union mf_subvalue value; | |
3517 | ||
3518 | memset(&value, 0xff, sizeof value); | |
3519 | mf_write_subfield_flow(&or->src, &value, &ctx->xout->wc.masks); | |
4e022ec0 AW |
3520 | xlate_output_action(ctx, u16_to_ofp(port), |
3521 | or->max_len, false); | |
9583bc14 EJ |
3522 | } |
3523 | } | |
3524 | ||
3525 | static void | |
3526 | xlate_enqueue_action(struct xlate_ctx *ctx, | |
3527 | const struct ofpact_enqueue *enqueue) | |
3528 | { | |
4e022ec0 | 3529 | ofp_port_t ofp_port = enqueue->port; |
9583bc14 EJ |
3530 | uint32_t queue_id = enqueue->queue; |
3531 | uint32_t flow_priority, priority; | |
3532 | int error; | |
3533 | ||
3534 | /* Translate queue to priority. */ | |
89a8a7f0 | 3535 | error = dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &priority); |
9583bc14 EJ |
3536 | if (error) { |
3537 | /* Fall back to ordinary output action. */ | |
3538 | xlate_output_action(ctx, enqueue->port, 0, false); | |
3539 | return; | |
3540 | } | |
3541 | ||
3542 | /* Check output port. */ | |
3543 | if (ofp_port == OFPP_IN_PORT) { | |
4e022ec0 AW |
3544 | ofp_port = ctx->xin->flow.in_port.ofp_port; |
3545 | } else if (ofp_port == ctx->xin->flow.in_port.ofp_port) { | |
9583bc14 EJ |
3546 | return; |
3547 | } | |
3548 | ||
3549 | /* Add datapath actions. */ | |
3550 | flow_priority = ctx->xin->flow.skb_priority; | |
3551 | ctx->xin->flow.skb_priority = priority; | |
3552 | compose_output_action(ctx, ofp_port); | |
3553 | ctx->xin->flow.skb_priority = flow_priority; | |
3554 | ||
3555 | /* Update NetFlow output port. */ | |
3556 | if (ctx->xout->nf_output_iface == NF_OUT_DROP) { | |
3557 | ctx->xout->nf_output_iface = ofp_port; | |
3558 | } else if (ctx->xout->nf_output_iface != NF_OUT_FLOOD) { | |
3559 | ctx->xout->nf_output_iface = NF_OUT_MULTI; | |
3560 | } | |
3561 | } | |
3562 | ||
3563 | static void | |
3564 | xlate_set_queue_action(struct xlate_ctx *ctx, uint32_t queue_id) | |
3565 | { | |
3566 | uint32_t skb_priority; | |
3567 | ||
89a8a7f0 | 3568 | if (!dpif_queue_to_priority(ctx->xbridge->dpif, queue_id, &skb_priority)) { |
9583bc14 EJ |
3569 | ctx->xin->flow.skb_priority = skb_priority; |
3570 | } else { | |
3571 | /* Couldn't translate queue to a priority. Nothing to do. A warning | |
3572 | * has already been logged. */ | |
3573 | } | |
3574 | } | |
3575 | ||
3576 | static bool | |
46c88433 | 3577 | slave_enabled_cb(ofp_port_t ofp_port, void *xbridge_) |
9583bc14 | 3578 | { |
46c88433 EJ |
3579 | const struct xbridge *xbridge = xbridge_; |
3580 | struct xport *port; | |
9583bc14 EJ |
3581 | |
3582 | switch (ofp_port) { | |
3583 | case OFPP_IN_PORT: | |
3584 | case OFPP_TABLE: | |
3585 | case OFPP_NORMAL: | |
3586 | case OFPP_FLOOD: | |
3587 | case OFPP_ALL: | |
3588 | case OFPP_NONE: | |
3589 | return true; | |
3590 | case OFPP_CONTROLLER: /* Not supported by the bundle action. */ | |
3591 | return false; | |
3592 | default: | |
46c88433 | 3593 | port = get_ofp_port(xbridge, ofp_port); |
9583bc14 EJ |
3594 | return port ? port->may_enable : false; |
3595 | } | |
3596 | } | |
3597 | ||
3598 | static void | |
3599 | xlate_bundle_action(struct xlate_ctx *ctx, | |
3600 | const struct ofpact_bundle *bundle) | |
3601 | { | |
4e022ec0 | 3602 | ofp_port_t port; |
9583bc14 EJ |
3603 | |
3604 | port = bundle_execute(bundle, &ctx->xin->flow, &ctx->xout->wc, | |
46c88433 EJ |
3605 | slave_enabled_cb, |
3606 | CONST_CAST(struct xbridge *, ctx->xbridge)); | |
9583bc14 | 3607 | if (bundle->dst.field) { |
f74e7df7 JP |
3608 | nxm_reg_load(&bundle->dst, ofp_to_u16(port), &ctx->xin->flow, |
3609 | &ctx->xout->wc); | |
9583bc14 EJ |
3610 | } else { |
3611 | xlate_output_action(ctx, port, 0, false); | |
3612 | } | |
3613 | } | |
3614 | ||
3615 | static void | |
4165b5e0 JS |
3616 | xlate_learn_action__(struct xlate_ctx *ctx, const struct ofpact_learn *learn, |
3617 | struct ofputil_flow_mod *fm, struct ofpbuf *ofpacts) | |
9583bc14 | 3618 | { |
4165b5e0 JS |
3619 | learn_execute(learn, &ctx->xin->flow, fm, ofpacts); |
3620 | if (ctx->xin->may_learn) { | |
3621 | ofproto_dpif_flow_mod(ctx->xbridge->ofproto, fm); | |
3622 | } | |
3623 | } | |
9583bc14 | 3624 | |
4165b5e0 JS |
3625 | static void |
3626 | xlate_learn_action(struct xlate_ctx *ctx, const struct ofpact_learn *learn) | |
3627 | { | |
9583bc14 | 3628 | ctx->xout->has_learn = true; |
9583bc14 EJ |
3629 | learn_mask(learn, &ctx->xout->wc); |
3630 | ||
b256dc52 JS |
3631 | if (ctx->xin->xcache) { |
3632 | struct xc_entry *entry; | |
3633 | ||
3634 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_LEARN); | |
4165b5e0 JS |
3635 | entry->u.learn.ofproto = ctx->xbridge->ofproto; |
3636 | entry->u.learn.fm = xmalloc(sizeof *entry->u.learn.fm); | |
3637 | entry->u.learn.ofpacts = ofpbuf_new(64); | |
3638 | xlate_learn_action__(ctx, learn, entry->u.learn.fm, | |
3639 | entry->u.learn.ofpacts); | |
3640 | } else if (ctx->xin->may_learn) { | |
3641 | uint64_t ofpacts_stub[1024 / 8]; | |
3642 | struct ofputil_flow_mod fm; | |
3643 | struct ofpbuf ofpacts; | |
3644 | ||
3645 | ofpbuf_use_stub(&ofpacts, ofpacts_stub, sizeof ofpacts_stub); | |
3646 | xlate_learn_action__(ctx, learn, &fm, &ofpacts); | |
3647 | ofpbuf_uninit(&ofpacts); | |
b256dc52 JS |
3648 | } |
3649 | } | |
3650 | ||
3651 | static void | |
3652 | xlate_fin_timeout__(struct rule_dpif *rule, uint16_t tcp_flags, | |
3653 | uint16_t idle_timeout, uint16_t hard_timeout) | |
3654 | { | |
3655 | if (tcp_flags & (TCP_FIN | TCP_RST)) { | |
3656 | rule_dpif_reduce_timeouts(rule, idle_timeout, hard_timeout); | |
3657 | } | |
9583bc14 EJ |
3658 | } |
3659 | ||
9583bc14 EJ |
3660 | static void |
3661 | xlate_fin_timeout(struct xlate_ctx *ctx, | |
3662 | const struct ofpact_fin_timeout *oft) | |
3663 | { | |
b256dc52 JS |
3664 | if (ctx->rule) { |
3665 | xlate_fin_timeout__(ctx->rule, ctx->xin->tcp_flags, | |
3666 | oft->fin_idle_timeout, oft->fin_hard_timeout); | |
3667 | if (ctx->xin->xcache) { | |
3668 | struct xc_entry *entry; | |
3669 | ||
3670 | entry = xlate_cache_add_entry(ctx->xin->xcache, XC_FIN_TIMEOUT); | |
83709dfa JR |
3671 | /* XC_RULE already holds a reference on the rule, none is taken |
3672 | * here. */ | |
b256dc52 JS |
3673 | entry->u.fin.rule = ctx->rule; |
3674 | entry->u.fin.idle = oft->fin_idle_timeout; | |
3675 | entry->u.fin.hard = oft->fin_hard_timeout; | |
b256dc52 | 3676 | } |
9583bc14 EJ |
3677 | } |
3678 | } | |
3679 | ||
3680 | static void | |
3681 | xlate_sample_action(struct xlate_ctx *ctx, | |
3682 | const struct ofpact_sample *os) | |
3683 | { | |
3684 | union user_action_cookie cookie; | |
3685 | /* Scale the probability from 16-bit to 32-bit while representing | |
3686 | * the same percentage. */ | |
3687 | uint32_t probability = (os->probability << 16) | os->probability; | |
3688 | ||
4b97b70d BP |
3689 | if (!ctx->xbridge->variable_length_userdata) { |
3690 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
3691 | ||
3692 | VLOG_ERR_RL(&rl, "ignoring NXAST_SAMPLE action because datapath " | |
3693 | "lacks support (needs Linux 3.10+ or kernel module from " | |
3694 | "OVS 1.11+)"); | |
3695 | return; | |
3696 | } | |
3697 | ||
7fd91025 | 3698 | ctx->xout->slow |= commit_odp_actions(&ctx->xin->flow, &ctx->base_flow, |
cc377352 | 3699 | ctx->xout->odp_actions, |
d23df9a8 JR |
3700 | &ctx->xout->wc, |
3701 | ctx->xbridge->masked_set_action); | |
9583bc14 EJ |
3702 | |
3703 | compose_flow_sample_cookie(os->probability, os->collector_set_id, | |
3704 | os->obs_domain_id, os->obs_point_id, &cookie); | |
cc377352 | 3705 | compose_sample_action(ctx->xbridge, ctx->xout->odp_actions, &ctx->xin->flow, |
8b7ea2d4 WZ |
3706 | probability, &cookie, sizeof cookie.flow_sample, |
3707 | ODPP_NONE); | |
9583bc14 EJ |
3708 | } |
3709 | ||
3710 | static bool | |
46c88433 | 3711 | may_receive(const struct xport *xport, struct xlate_ctx *ctx) |
9583bc14 | 3712 | { |
bbbca389 | 3713 | if (xport->config & (is_stp(&ctx->xin->flow) |
46c88433 EJ |
3714 | ? OFPUTIL_PC_NO_RECV_STP |
3715 | : OFPUTIL_PC_NO_RECV)) { | |
9583bc14 EJ |
3716 | return false; |
3717 | } | |
3718 | ||
3719 | /* Only drop packets here if both forwarding and learning are | |
3720 | * disabled. If just learning is enabled, we need to have | |
3721 | * OFPP_NORMAL and the learning action have a look at the packet | |
3722 | * before we can drop it. */ | |
9efd308e DV |
3723 | if ((!xport_stp_forward_state(xport) && !xport_stp_learn_state(xport)) || |
3724 | (!xport_rstp_forward_state(xport) && !xport_rstp_learn_state(xport))) { | |
9583bc14 EJ |
3725 | return false; |
3726 | } | |
3727 | ||
3728 | return true; | |
3729 | } | |
3730 | ||
7fdb60a7 SH |
3731 | static void |
3732 | xlate_write_actions(struct xlate_ctx *ctx, const struct ofpact *a) | |
3733 | { | |
c61f3870 BP |
3734 | const struct ofpact_nest *on = ofpact_get_WRITE_ACTIONS(a); |
3735 | size_t on_len = ofpact_nest_get_action_len(on); | |
3736 | const struct ofpact *inner; | |
3737 | ||
3738 | /* Maintain actset_output depending on the contents of the action set: | |
3739 | * | |
3740 | * - OFPP_UNSET, if there is no "output" action. | |
3741 | * | |
3742 | * - The output port, if there is an "output" action and no "group" | |
3743 | * action. | |
3744 | * | |
3745 | * - OFPP_UNSET, if there is a "group" action. | |
3746 | */ | |
3747 | if (!ctx->action_set_has_group) { | |
3748 | OFPACT_FOR_EACH (inner, on->actions, on_len) { | |
3749 | if (inner->type == OFPACT_OUTPUT) { | |
3750 | ctx->xin->flow.actset_output = ofpact_get_OUTPUT(inner)->port; | |
3751 | } else if (inner->type == OFPACT_GROUP) { | |
3752 | ctx->xin->flow.actset_output = OFPP_UNSET; | |
3753 | ctx->action_set_has_group = true; | |
3754 | } | |
3755 | } | |
3756 | } | |
3757 | ||
3758 | ofpbuf_put(&ctx->action_set, on->actions, on_len); | |
7fdb60a7 SH |
3759 | ofpact_pad(&ctx->action_set); |
3760 | } | |
3761 | ||
3762 | static void | |
3763 | xlate_action_set(struct xlate_ctx *ctx) | |
3764 | { | |
3765 | uint64_t action_list_stub[1024 / 64]; | |
3766 | struct ofpbuf action_list; | |
3767 | ||
029ca940 | 3768 | ctx->in_action_set = true; |
7fdb60a7 SH |
3769 | ofpbuf_use_stub(&action_list, action_list_stub, sizeof action_list_stub); |
3770 | ofpacts_execute_action_set(&action_list, &ctx->action_set); | |
6fd6ed71 | 3771 | do_xlate_actions(action_list.data, action_list.size, ctx); |
029ca940 | 3772 | ctx->in_action_set = false; |
7fdb60a7 SH |
3773 | ofpbuf_uninit(&action_list); |
3774 | } | |
3775 | ||
7bbdd84f | 3776 | static bool |
5af43325 | 3777 | ofpact_needs_recirculation_after_mpls(const struct ofpact *a, struct xlate_ctx *ctx) |
7bbdd84f SH |
3778 | { |
3779 | struct flow_wildcards *wc = &ctx->xout->wc; | |
3780 | struct flow *flow = &ctx->xin->flow; | |
3781 | ||
5af43325 PS |
3782 | if (!ctx->was_mpls) { |
3783 | return false; | |
3784 | } | |
3785 | ||
7bbdd84f SH |
3786 | switch (a->type) { |
3787 | case OFPACT_OUTPUT: | |
3788 | case OFPACT_GROUP: | |
3789 | case OFPACT_CONTROLLER: | |
3790 | case OFPACT_STRIP_VLAN: | |
3791 | case OFPACT_SET_VLAN_PCP: | |
3792 | case OFPACT_SET_VLAN_VID: | |
3793 | case OFPACT_ENQUEUE: | |
3794 | case OFPACT_PUSH_VLAN: | |
3795 | case OFPACT_SET_ETH_SRC: | |
3796 | case OFPACT_SET_ETH_DST: | |
3797 | case OFPACT_SET_TUNNEL: | |
3798 | case OFPACT_SET_QUEUE: | |
3799 | case OFPACT_POP_QUEUE: | |
18080541 | 3800 | case OFPACT_CONJUNCTION: |
7bbdd84f SH |
3801 | case OFPACT_NOTE: |
3802 | case OFPACT_OUTPUT_REG: | |
3803 | case OFPACT_EXIT: | |
3804 | case OFPACT_METER: | |
3805 | case OFPACT_WRITE_METADATA: | |
3806 | case OFPACT_WRITE_ACTIONS: | |
3807 | case OFPACT_CLEAR_ACTIONS: | |
3808 | case OFPACT_SAMPLE: | |
3809 | return false; | |
3810 | ||
5af43325 PS |
3811 | case OFPACT_POP_MPLS: |
3812 | case OFPACT_DEC_MPLS_TTL: | |
3813 | case OFPACT_SET_MPLS_TTL: | |
3814 | case OFPACT_SET_MPLS_TC: | |
3815 | case OFPACT_SET_MPLS_LABEL: | |
7bbdd84f SH |
3816 | case OFPACT_SET_IPV4_SRC: |
3817 | case OFPACT_SET_IPV4_DST: | |
3818 | case OFPACT_SET_IP_DSCP: | |
3819 | case OFPACT_SET_IP_ECN: | |
3820 | case OFPACT_SET_IP_TTL: | |
3821 | case OFPACT_SET_L4_SRC_PORT: | |
3822 | case OFPACT_SET_L4_DST_PORT: | |
3823 | case OFPACT_RESUBMIT: | |
3824 | case OFPACT_STACK_PUSH: | |
3825 | case OFPACT_STACK_POP: | |
3826 | case OFPACT_DEC_TTL: | |
3827 | case OFPACT_MULTIPATH: | |
3828 | case OFPACT_BUNDLE: | |
3829 | case OFPACT_LEARN: | |
3830 | case OFPACT_FIN_TIMEOUT: | |
3831 | case OFPACT_GOTO_TABLE: | |
3832 | return true; | |
3833 | ||
3834 | case OFPACT_REG_MOVE: | |
3835 | return (mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->dst.field) || | |
3836 | mf_is_l3_or_higher(ofpact_get_REG_MOVE(a)->src.field)); | |
3837 | ||
7bbdd84f SH |
3838 | case OFPACT_SET_FIELD: |
3839 | return mf_is_l3_or_higher(ofpact_get_SET_FIELD(a)->field); | |
3840 | ||
3841 | case OFPACT_PUSH_MPLS: | |
3842 | /* Recirculate if it is an IP packet with a zero ttl. This may | |
3843 | * indicate that the packet was previously MPLS and an MPLS pop action | |
3844 | * converted it to IP. In this case recirculating should reveal the IP | |
3845 | * TTL which is used as the basis for a new MPLS LSE. */ | |
3846 | return (!flow_count_mpls_labels(flow, wc) | |
3847 | && flow->nw_ttl == 0 | |
3848 | && is_ip_any(flow)); | |
3849 | } | |
3850 | ||
3851 | OVS_NOT_REACHED(); | |
3852 | } | |
3853 | ||
9583bc14 EJ |
3854 | static void |
3855 | do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, | |
3856 | struct xlate_ctx *ctx) | |
3857 | { | |
33bf9176 BP |
3858 | struct flow_wildcards *wc = &ctx->xout->wc; |
3859 | struct flow *flow = &ctx->xin->flow; | |
9583bc14 EJ |
3860 | const struct ofpact *a; |
3861 | ||
a36de779 PS |
3862 | if (ovs_native_tunneling_is_on(ctx->xbridge->ofproto)) { |
3863 | tnl_arp_snoop(flow, wc, ctx->xbridge->name); | |
3864 | } | |
f47ea021 JR |
3865 | /* dl_type already in the mask, not set below. */ |
3866 | ||
9583bc14 EJ |
3867 | OFPACT_FOR_EACH (a, ofpacts, ofpacts_len) { |
3868 | struct ofpact_controller *controller; | |
3869 | const struct ofpact_metadata *metadata; | |
b2dd70be JR |
3870 | const struct ofpact_set_field *set_field; |
3871 | const struct mf_field *mf; | |
9583bc14 EJ |
3872 | |
3873 | if (ctx->exit) { | |
3874 | break; | |
3875 | } | |
3876 | ||
5af43325 | 3877 | if (ofpact_needs_recirculation_after_mpls(a, ctx)) { |
7bbdd84f SH |
3878 | compose_recirculate_action(ctx, ofpacts, a, ofpacts_len); |
3879 | return; | |
3880 | } | |
3881 | ||
9583bc14 EJ |
3882 | switch (a->type) { |
3883 | case OFPACT_OUTPUT: | |
3884 | xlate_output_action(ctx, ofpact_get_OUTPUT(a)->port, | |
3885 | ofpact_get_OUTPUT(a)->max_len, true); | |
3886 | break; | |
3887 | ||
7395c052 | 3888 | case OFPACT_GROUP: |
f4fb341b SH |
3889 | if (xlate_group_action(ctx, ofpact_get_GROUP(a)->group_id)) { |
3890 | return; | |
3891 | } | |
7395c052 NZ |
3892 | break; |
3893 | ||
9583bc14 EJ |
3894 | case OFPACT_CONTROLLER: |
3895 | controller = ofpact_get_CONTROLLER(a); | |
3896 | execute_controller_action(ctx, controller->max_len, | |
3897 | controller->reason, | |
3898 | controller->controller_id); | |
3899 | break; | |
3900 | ||
3901 | case OFPACT_ENQUEUE: | |
16194afd DDP |
3902 | memset(&wc->masks.skb_priority, 0xff, |
3903 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
3904 | xlate_enqueue_action(ctx, ofpact_get_ENQUEUE(a)); |
3905 | break; | |
3906 | ||
3907 | case OFPACT_SET_VLAN_VID: | |
f74e7df7 | 3908 | wc->masks.vlan_tci |= htons(VLAN_VID_MASK | VLAN_CFI); |
ca287d20 JR |
3909 | if (flow->vlan_tci & htons(VLAN_CFI) || |
3910 | ofpact_get_SET_VLAN_VID(a)->push_vlan_if_needed) { | |
3911 | flow->vlan_tci &= ~htons(VLAN_VID_MASK); | |
3912 | flow->vlan_tci |= (htons(ofpact_get_SET_VLAN_VID(a)->vlan_vid) | |
3913 | | htons(VLAN_CFI)); | |
3914 | } | |
9583bc14 EJ |
3915 | break; |
3916 | ||
3917 | case OFPACT_SET_VLAN_PCP: | |
f74e7df7 | 3918 | wc->masks.vlan_tci |= htons(VLAN_PCP_MASK | VLAN_CFI); |
ca287d20 JR |
3919 | if (flow->vlan_tci & htons(VLAN_CFI) || |
3920 | ofpact_get_SET_VLAN_PCP(a)->push_vlan_if_needed) { | |
3921 | flow->vlan_tci &= ~htons(VLAN_PCP_MASK); | |
3922 | flow->vlan_tci |= htons((ofpact_get_SET_VLAN_PCP(a)->vlan_pcp | |
3923 | << VLAN_PCP_SHIFT) | VLAN_CFI); | |
3924 | } | |
9583bc14 EJ |
3925 | break; |
3926 | ||
3927 | case OFPACT_STRIP_VLAN: | |
f74e7df7 | 3928 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 3929 | flow->vlan_tci = htons(0); |
9583bc14 EJ |
3930 | break; |
3931 | ||
3932 | case OFPACT_PUSH_VLAN: | |
3933 | /* XXX 802.1AD(QinQ) */ | |
f74e7df7 | 3934 | memset(&wc->masks.vlan_tci, 0xff, sizeof wc->masks.vlan_tci); |
33bf9176 | 3935 | flow->vlan_tci = htons(VLAN_CFI); |
9583bc14 EJ |
3936 | break; |
3937 | ||
3938 | case OFPACT_SET_ETH_SRC: | |
f74e7df7 | 3939 | memset(&wc->masks.dl_src, 0xff, sizeof wc->masks.dl_src); |
33bf9176 | 3940 | memcpy(flow->dl_src, ofpact_get_SET_ETH_SRC(a)->mac, ETH_ADDR_LEN); |
9583bc14 EJ |
3941 | break; |
3942 | ||
3943 | case OFPACT_SET_ETH_DST: | |
f74e7df7 | 3944 | memset(&wc->masks.dl_dst, 0xff, sizeof wc->masks.dl_dst); |
33bf9176 | 3945 | memcpy(flow->dl_dst, ofpact_get_SET_ETH_DST(a)->mac, ETH_ADDR_LEN); |
9583bc14 EJ |
3946 | break; |
3947 | ||
3948 | case OFPACT_SET_IPV4_SRC: | |
33bf9176 | 3949 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 3950 | memset(&wc->masks.nw_src, 0xff, sizeof wc->masks.nw_src); |
33bf9176 | 3951 | flow->nw_src = ofpact_get_SET_IPV4_SRC(a)->ipv4; |
9583bc14 EJ |
3952 | } |
3953 | break; | |
3954 | ||
3955 | case OFPACT_SET_IPV4_DST: | |
33bf9176 | 3956 | if (flow->dl_type == htons(ETH_TYPE_IP)) { |
f47ea021 | 3957 | memset(&wc->masks.nw_dst, 0xff, sizeof wc->masks.nw_dst); |
33bf9176 | 3958 | flow->nw_dst = ofpact_get_SET_IPV4_DST(a)->ipv4; |
9583bc14 EJ |
3959 | } |
3960 | break; | |
3961 | ||
04f01c24 BP |
3962 | case OFPACT_SET_IP_DSCP: |
3963 | if (is_ip_any(flow)) { | |
f47ea021 | 3964 | wc->masks.nw_tos |= IP_DSCP_MASK; |
33bf9176 | 3965 | flow->nw_tos &= ~IP_DSCP_MASK; |
04f01c24 | 3966 | flow->nw_tos |= ofpact_get_SET_IP_DSCP(a)->dscp; |
9583bc14 EJ |
3967 | } |
3968 | break; | |
3969 | ||
ff14eb7a JR |
3970 | case OFPACT_SET_IP_ECN: |
3971 | if (is_ip_any(flow)) { | |
3972 | wc->masks.nw_tos |= IP_ECN_MASK; | |
3973 | flow->nw_tos &= ~IP_ECN_MASK; | |
3974 | flow->nw_tos |= ofpact_get_SET_IP_ECN(a)->ecn; | |
3975 | } | |
3976 | break; | |
3977 | ||
0c20dbe4 JR |
3978 | case OFPACT_SET_IP_TTL: |
3979 | if (is_ip_any(flow)) { | |
3980 | wc->masks.nw_ttl = 0xff; | |
3981 | flow->nw_ttl = ofpact_get_SET_IP_TTL(a)->ttl; | |
3982 | } | |
3983 | break; | |
3984 | ||
9583bc14 | 3985 | case OFPACT_SET_L4_SRC_PORT: |
b8778a0d | 3986 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
3987 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
3988 | memset(&wc->masks.tp_src, 0xff, sizeof wc->masks.tp_src); | |
33bf9176 | 3989 | flow->tp_src = htons(ofpact_get_SET_L4_SRC_PORT(a)->port); |
9583bc14 EJ |
3990 | } |
3991 | break; | |
3992 | ||
3993 | case OFPACT_SET_L4_DST_PORT: | |
b8778a0d | 3994 | if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { |
f47ea021 JR |
3995 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
3996 | memset(&wc->masks.tp_dst, 0xff, sizeof wc->masks.tp_dst); | |
33bf9176 | 3997 | flow->tp_dst = htons(ofpact_get_SET_L4_DST_PORT(a)->port); |
9583bc14 EJ |
3998 | } |
3999 | break; | |
4000 | ||
4001 | case OFPACT_RESUBMIT: | |
4002 | xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a)); | |
4003 | break; | |
4004 | ||
4005 | case OFPACT_SET_TUNNEL: | |
33bf9176 | 4006 | flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id); |
9583bc14 EJ |
4007 | break; |
4008 | ||
4009 | case OFPACT_SET_QUEUE: | |
16194afd DDP |
4010 | memset(&wc->masks.skb_priority, 0xff, |
4011 | sizeof wc->masks.skb_priority); | |
9583bc14 EJ |
4012 | xlate_set_queue_action(ctx, ofpact_get_SET_QUEUE(a)->queue_id); |
4013 | break; | |
4014 | ||
4015 | case OFPACT_POP_QUEUE: | |
16194afd DDP |
4016 | memset(&wc->masks.skb_priority, 0xff, |
4017 | sizeof wc->masks.skb_priority); | |
33bf9176 | 4018 | flow->skb_priority = ctx->orig_skb_priority; |
9583bc14 EJ |
4019 | break; |
4020 | ||
4021 | case OFPACT_REG_MOVE: | |
33bf9176 | 4022 | nxm_execute_reg_move(ofpact_get_REG_MOVE(a), flow, wc); |
9583bc14 EJ |
4023 | break; |
4024 | ||
b2dd70be JR |
4025 | case OFPACT_SET_FIELD: |
4026 | set_field = ofpact_get_SET_FIELD(a); | |
4027 | mf = set_field->field; | |
b2dd70be JR |
4028 | |
4029 | /* Set field action only ever overwrites packet's outermost | |
4030 | * applicable header fields. Do nothing if no header exists. */ | |
4479846e JR |
4031 | if (mf->id == MFF_VLAN_VID) { |
4032 | wc->masks.vlan_tci |= htons(VLAN_CFI); | |
4033 | if (!(flow->vlan_tci & htons(VLAN_CFI))) { | |
4034 | break; | |
4035 | } | |
4036 | } else if ((mf->id == MFF_MPLS_LABEL || mf->id == MFF_MPLS_TC) | |
4037 | /* 'dl_type' is already unwildcarded. */ | |
4038 | && !eth_type_mpls(flow->dl_type)) { | |
4039 | break; | |
b2dd70be | 4040 | } |
b8778a0d JR |
4041 | /* A flow may wildcard nw_frag. Do nothing if setting a trasport |
4042 | * header field on a packet that does not have them. */ | |
4479846e | 4043 | mf_mask_field_and_prereqs(mf, &wc->masks); |
b8778a0d JR |
4044 | if (mf_are_prereqs_ok(mf, flow)) { |
4045 | mf_set_flow_value_masked(mf, &set_field->value, | |
4046 | &set_field->mask, flow); | |
4047 | } | |
b2dd70be JR |
4048 | break; |
4049 | ||
9583bc14 | 4050 | case OFPACT_STACK_PUSH: |
33bf9176 BP |
4051 | nxm_execute_stack_push(ofpact_get_STACK_PUSH(a), flow, wc, |
4052 | &ctx->stack); | |
9583bc14 EJ |
4053 | break; |
4054 | ||
4055 | case OFPACT_STACK_POP: | |
f74e7df7 JP |
4056 | nxm_execute_stack_pop(ofpact_get_STACK_POP(a), flow, wc, |
4057 | &ctx->stack); | |
9583bc14 EJ |
4058 | break; |
4059 | ||
4060 | case OFPACT_PUSH_MPLS: | |
8bfd0fda | 4061 | compose_mpls_push_action(ctx, ofpact_get_PUSH_MPLS(a)); |
9583bc14 EJ |
4062 | break; |
4063 | ||
4064 | case OFPACT_POP_MPLS: | |
8bfd0fda | 4065 | compose_mpls_pop_action(ctx, ofpact_get_POP_MPLS(a)->ethertype); |
9583bc14 EJ |
4066 | break; |
4067 | ||
097d4939 | 4068 | case OFPACT_SET_MPLS_LABEL: |
8bfd0fda BP |
4069 | compose_set_mpls_label_action( |
4070 | ctx, ofpact_get_SET_MPLS_LABEL(a)->label); | |
4071 | break; | |
097d4939 JR |
4072 | |
4073 | case OFPACT_SET_MPLS_TC: | |
8bfd0fda | 4074 | compose_set_mpls_tc_action(ctx, ofpact_get_SET_MPLS_TC(a)->tc); |
097d4939 JR |
4075 | break; |
4076 | ||
9583bc14 | 4077 | case OFPACT_SET_MPLS_TTL: |
8bfd0fda | 4078 | compose_set_mpls_ttl_action(ctx, ofpact_get_SET_MPLS_TTL(a)->ttl); |
9583bc14 EJ |
4079 | break; |
4080 | ||
4081 | case OFPACT_DEC_MPLS_TTL: | |
9cfef3d0 | 4082 | if (compose_dec_mpls_ttl_action(ctx)) { |
ad3efdcb | 4083 | return; |
9583bc14 EJ |
4084 | } |
4085 | break; | |
4086 | ||
4087 | case OFPACT_DEC_TTL: | |
f74e7df7 | 4088 | wc->masks.nw_ttl = 0xff; |
9583bc14 | 4089 | if (compose_dec_ttl(ctx, ofpact_get_DEC_TTL(a))) { |
ad3efdcb | 4090 | return; |
9583bc14 EJ |
4091 | } |
4092 | break; | |
4093 | ||
4094 | case OFPACT_NOTE: | |
4095 | /* Nothing to do. */ | |
4096 | break; | |
4097 | ||
4098 | case OFPACT_MULTIPATH: | |
33bf9176 | 4099 | multipath_execute(ofpact_get_MULTIPATH(a), flow, wc); |
9583bc14 EJ |
4100 | break; |
4101 | ||
4102 | case OFPACT_BUNDLE: | |
9583bc14 EJ |
4103 | xlate_bundle_action(ctx, ofpact_get_BUNDLE(a)); |
4104 | break; | |
4105 | ||
4106 | case OFPACT_OUTPUT_REG: | |
4107 | xlate_output_reg_action(ctx, ofpact_get_OUTPUT_REG(a)); | |
4108 | break; | |
4109 | ||
4110 | case OFPACT_LEARN: | |
4111 | xlate_learn_action(ctx, ofpact_get_LEARN(a)); | |
4112 | break; | |
4113 | ||
afc3987b BP |
4114 | case OFPACT_CONJUNCTION: { |
4115 | /* A flow with a "conjunction" action represents part of a special | |
4116 | * kind of "set membership match". Such a flow should not actually | |
4117 | * get executed, but it could via, say, a "packet-out", even though | |
4118 | * that wouldn't be useful. Log it to help debugging. */ | |
4119 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); | |
4120 | VLOG_INFO_RL(&rl, "executing no-op conjunction action"); | |
18080541 | 4121 | break; |
afc3987b | 4122 | } |
18080541 | 4123 | |
9583bc14 EJ |
4124 | case OFPACT_EXIT: |
4125 | ctx->exit = true; | |
4126 | break; | |
4127 | ||
4128 | case OFPACT_FIN_TIMEOUT: | |
33bf9176 | 4129 | memset(&wc->masks.nw_proto, 0xff, sizeof wc->masks.nw_proto); |
9583bc14 EJ |
4130 | ctx->xout->has_fin_timeout = true; |
4131 | xlate_fin_timeout(ctx, ofpact_get_FIN_TIMEOUT(a)); | |
4132 | break; | |
4133 | ||
4134 | case OFPACT_CLEAR_ACTIONS: | |
7fdb60a7 | 4135 | ofpbuf_clear(&ctx->action_set); |
c61f3870 BP |
4136 | ctx->xin->flow.actset_output = OFPP_UNSET; |
4137 | ctx->action_set_has_group = false; | |
7fdb60a7 SH |
4138 | break; |
4139 | ||
4140 | case OFPACT_WRITE_ACTIONS: | |
4141 | xlate_write_actions(ctx, a); | |
9583bc14 EJ |
4142 | break; |
4143 | ||
4144 | case OFPACT_WRITE_METADATA: | |
4145 | metadata = ofpact_get_WRITE_METADATA(a); | |
33bf9176 BP |
4146 | flow->metadata &= ~metadata->mask; |
4147 | flow->metadata |= metadata->metadata & metadata->mask; | |
9583bc14 EJ |
4148 | break; |
4149 | ||
638a19b0 JR |
4150 | case OFPACT_METER: |
4151 | /* Not implemented yet. */ | |
4152 | break; | |
4153 | ||
9583bc14 | 4154 | case OFPACT_GOTO_TABLE: { |
9583bc14 | 4155 | struct ofpact_goto_table *ogt = ofpact_get_GOTO_TABLE(a); |
9583bc14 | 4156 | |
7bbdd84f SH |
4157 | /* Allow ctx->table_id == TBL_INTERNAL, which will be greater |
4158 | * than ogt->table_id. This is to allow goto_table actions that | |
4159 | * triggered recirculation: ctx->table_id will be TBL_INTERNAL | |
4160 | * after recirculation. */ | |
4161 | ovs_assert(ctx->table_id == TBL_INTERNAL | |
4162 | || ctx->table_id < ogt->table_id); | |
4468099e | 4163 | xlate_table_action(ctx, ctx->xin->flow.in_port.ofp_port, |
6d328fa2 | 4164 | ogt->table_id, true, true); |
9583bc14 EJ |
4165 | break; |
4166 | } | |
4167 | ||
4168 | case OFPACT_SAMPLE: | |
4169 | xlate_sample_action(ctx, ofpact_get_SAMPLE(a)); | |
4170 | break; | |
4171 | } | |
4172 | } | |
9583bc14 EJ |
4173 | } |
4174 | ||
4175 | void | |
4176 | xlate_in_init(struct xlate_in *xin, struct ofproto_dpif *ofproto, | |
cc377352 EJ |
4177 | const struct flow *flow, ofp_port_t in_port, |
4178 | struct rule_dpif *rule, uint16_t tcp_flags, | |
cf62fa4c | 4179 | const struct dp_packet *packet) |
9583bc14 EJ |
4180 | { |
4181 | xin->ofproto = ofproto; | |
4182 | xin->flow = *flow; | |
cc377352 | 4183 | xin->flow.in_port.ofp_port = in_port; |
c61f3870 | 4184 | xin->flow.actset_output = OFPP_UNSET; |
9583bc14 EJ |
4185 | xin->packet = packet; |
4186 | xin->may_learn = packet != NULL; | |
4187 | xin->rule = rule; | |
b256dc52 | 4188 | xin->xcache = NULL; |
9583bc14 EJ |
4189 | xin->ofpacts = NULL; |
4190 | xin->ofpacts_len = 0; | |
4191 | xin->tcp_flags = tcp_flags; | |
4192 | xin->resubmit_hook = NULL; | |
4193 | xin->report_hook = NULL; | |
4194 | xin->resubmit_stats = NULL; | |
3f207910 | 4195 | xin->skip_wildcards = false; |
cc377352 | 4196 | xin->odp_actions = NULL; |
9583bc14 EJ |
4197 | } |
4198 | ||
4199 | void | |
4200 | xlate_out_uninit(struct xlate_out *xout) | |
4201 | { | |
cc377352 EJ |
4202 | if (xout && xout->odp_actions == &xout->odp_actions_buf) { |
4203 | ofpbuf_uninit(xout->odp_actions); | |
9583bc14 EJ |
4204 | } |
4205 | } | |
4206 | ||
4207 | /* Translates the 'ofpacts_len' bytes of "struct ofpact"s starting at 'ofpacts' | |
4208 | * into datapath actions, using 'ctx', and discards the datapath actions. */ | |
4209 | void | |
4210 | xlate_actions_for_side_effects(struct xlate_in *xin) | |
4211 | { | |
4212 | struct xlate_out xout; | |
4213 | ||
4214 | xlate_actions(xin, &xout); | |
4215 | xlate_out_uninit(&xout); | |
4216 | } | |
4217 | ||
9583bc14 EJ |
4218 | void |
4219 | xlate_out_copy(struct xlate_out *dst, const struct xlate_out *src) | |
4220 | { | |
4221 | dst->wc = src->wc; | |
9583bc14 EJ |
4222 | dst->slow = src->slow; |
4223 | dst->has_learn = src->has_learn; | |
4224 | dst->has_normal = src->has_normal; | |
4225 | dst->has_fin_timeout = src->has_fin_timeout; | |
4226 | dst->nf_output_iface = src->nf_output_iface; | |
4227 | dst->mirrors = src->mirrors; | |
4228 | ||
cc377352 EJ |
4229 | dst->odp_actions = &dst->odp_actions_buf; |
4230 | ofpbuf_use_stub(dst->odp_actions, dst->odp_actions_stub, | |
9583bc14 | 4231 | sizeof dst->odp_actions_stub); |
6fd6ed71 | 4232 | ofpbuf_put(dst->odp_actions, src->odp_actions->data, src->odp_actions->size); |
9583bc14 EJ |
4233 | } |
4234 | \f | |
55954f6e EJ |
4235 | static struct skb_priority_to_dscp * |
4236 | get_skb_priority(const struct xport *xport, uint32_t skb_priority) | |
4237 | { | |
4238 | struct skb_priority_to_dscp *pdscp; | |
4239 | uint32_t hash; | |
4240 | ||
4241 | hash = hash_int(skb_priority, 0); | |
4242 | HMAP_FOR_EACH_IN_BUCKET (pdscp, hmap_node, hash, &xport->skb_priorities) { | |
4243 | if (pdscp->skb_priority == skb_priority) { | |
4244 | return pdscp; | |
4245 | } | |
4246 | } | |
4247 | return NULL; | |
4248 | } | |
4249 | ||
4250 | static bool | |
4251 | dscp_from_skb_priority(const struct xport *xport, uint32_t skb_priority, | |
4252 | uint8_t *dscp) | |
4253 | { | |
4254 | struct skb_priority_to_dscp *pdscp = get_skb_priority(xport, skb_priority); | |
4255 | *dscp = pdscp ? pdscp->dscp : 0; | |
4256 | return pdscp != NULL; | |
4257 | } | |
4258 | ||
16194afd DDP |
4259 | static size_t |
4260 | count_skb_priorities(const struct xport *xport) | |
4261 | { | |
4262 | return hmap_count(&xport->skb_priorities); | |
4263 | } | |
4264 | ||
55954f6e EJ |
4265 | static void |
4266 | clear_skb_priorities(struct xport *xport) | |
4267 | { | |
4268 | struct skb_priority_to_dscp *pdscp, *next; | |
4269 | ||
4270 | HMAP_FOR_EACH_SAFE (pdscp, next, hmap_node, &xport->skb_priorities) { | |
4271 | hmap_remove(&xport->skb_priorities, &pdscp->hmap_node); | |
4272 | free(pdscp); | |
4273 | } | |
4274 | } | |
4275 | ||
ce4a6b76 BP |
4276 | static bool |
4277 | actions_output_to_local_port(const struct xlate_ctx *ctx) | |
4278 | { | |
46c88433 | 4279 | odp_port_t local_odp_port = ofp_port_to_odp_port(ctx->xbridge, OFPP_LOCAL); |
ce4a6b76 BP |
4280 | const struct nlattr *a; |
4281 | unsigned int left; | |
4282 | ||
6fd6ed71 PS |
4283 | NL_ATTR_FOR_EACH_UNSAFE (a, left, ctx->xout->odp_actions->data, |
4284 | ctx->xout->odp_actions->size) { | |
ce4a6b76 BP |
4285 | if (nl_attr_type(a) == OVS_ACTION_ATTR_OUTPUT |
4286 | && nl_attr_get_odp_port(a) == local_odp_port) { | |
4287 | return true; | |
4288 | } | |
4289 | } | |
4290 | return false; | |
4291 | } | |
9583bc14 | 4292 | |
5e2a6702 | 4293 | #if defined(__linux__) |
7d031d7e BP |
4294 | /* Returns the maximum number of packets that the Linux kernel is willing to |
4295 | * queue up internally to certain kinds of software-implemented ports, or the | |
4296 | * default (and rarely modified) value if it cannot be determined. */ | |
4297 | static int | |
4298 | netdev_max_backlog(void) | |
4299 | { | |
4300 | static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; | |
4301 | static int max_backlog = 1000; /* The normal default value. */ | |
4302 | ||
4303 | if (ovsthread_once_start(&once)) { | |
4304 | static const char filename[] = "/proc/sys/net/core/netdev_max_backlog"; | |
4305 | FILE *stream; | |
4306 | int n; | |
4307 | ||
4308 | stream = fopen(filename, "r"); | |
4309 | if (!stream) { | |
4310 | VLOG_WARN("%s: open failed (%s)", filename, ovs_strerror(errno)); | |
4311 | } else { | |
4312 | if (fscanf(stream, "%d", &n) != 1) { | |
4313 | VLOG_WARN("%s: read error", filename); | |
4314 | } else if (n <= 100) { | |
4315 | VLOG_WARN("%s: unexpectedly small value %d", filename, n); | |
4316 | } else { | |
4317 | max_backlog = n; | |
4318 | } | |
4319 | fclose(stream); | |
4320 | } | |
4321 | ovsthread_once_done(&once); | |
4322 | ||
4323 | VLOG_DBG("%s: using %d max_backlog", filename, max_backlog); | |
4324 | } | |
4325 | ||
4326 | return max_backlog; | |
4327 | } | |
4328 | ||
4329 | /* Counts and returns the number of OVS_ACTION_ATTR_OUTPUT actions in | |
4330 | * 'odp_actions'. */ | |
4331 | static int | |
4332 | count_output_actions(const struct ofpbuf *odp_actions) | |
4333 | { | |
4334 | const struct nlattr *a; | |
4335 | size_t left; | |
4336 | int n = 0; | |
4337 | ||
6fd6ed71 | 4338 | NL_ATTR_FOR_EACH_UNSAFE (a, left, odp_actions->data, odp_actions->size) { |
7d031d7e BP |
4339 | if (a->nla_type == OVS_ACTION_ATTR_OUTPUT) { |
4340 | n++; | |
4341 | } | |
4342 | } | |
4343 | return n; | |
4344 | } | |
5e2a6702 | 4345 | #endif /* defined(__linux__) */ |
7d031d7e BP |
4346 | |
4347 | /* Returns true if 'odp_actions' contains more output actions than the datapath | |
4348 | * can reliably handle in one go. On Linux, this is the value of the | |
4349 | * net.core.netdev_max_backlog sysctl, which limits the maximum number of | |
4350 | * packets that the kernel is willing to queue up for processing while the | |
4351 | * datapath is processing a set of actions. */ | |
4352 | static bool | |
5e2a6702 | 4353 | too_many_output_actions(const struct ofpbuf *odp_actions OVS_UNUSED) |
7d031d7e BP |
4354 | { |
4355 | #ifdef __linux__ | |
6fd6ed71 | 4356 | return (odp_actions->size / NL_A_U32_SIZE > netdev_max_backlog() |
7d031d7e BP |
4357 | && count_output_actions(odp_actions) > netdev_max_backlog()); |
4358 | #else | |
4359 | /* OSes other than Linux might have similar limits, but we don't know how | |
4360 | * to determine them.*/ | |
4361 | return false; | |
4362 | #endif | |
4363 | } | |
4364 | ||
9583bc14 | 4365 | /* Translates the 'ofpacts_len' bytes of "struct ofpacts" starting at 'ofpacts' |
56450a41 BP |
4366 | * into datapath actions in 'odp_actions', using 'ctx'. |
4367 | * | |
4368 | * The caller must take responsibility for eventually freeing 'xout', with | |
4369 | * xlate_out_uninit(). */ | |
84f0f298 RW |
4370 | void |
4371 | xlate_actions(struct xlate_in *xin, struct xlate_out *xout) | |
9583bc14 | 4372 | { |
84f0f298 | 4373 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
33bf9176 BP |
4374 | struct flow_wildcards *wc = &xout->wc; |
4375 | struct flow *flow = &xin->flow; | |
10c44245 | 4376 | struct rule_dpif *rule = NULL; |
33bf9176 | 4377 | |
9583bc14 EJ |
4378 | enum slow_path_reason special; |
4379 | const struct ofpact *ofpacts; | |
46c88433 | 4380 | struct xport *in_port; |
9583bc14 EJ |
4381 | struct flow orig_flow; |
4382 | struct xlate_ctx ctx; | |
4383 | size_t ofpacts_len; | |
62a7cc71 | 4384 | bool tnl_may_send; |
c56fac1b | 4385 | bool is_icmp; |
9583bc14 | 4386 | |
46c88433 | 4387 | COVERAGE_INC(xlate_actions); |
9583bc14 EJ |
4388 | |
4389 | /* Flow initialization rules: | |
4390 | * - 'base_flow' must match the kernel's view of the packet at the | |
4391 | * time that action processing starts. 'flow' represents any | |
4392 | * transformations we wish to make through actions. | |
4393 | * - By default 'base_flow' and 'flow' are the same since the input | |
4394 | * packet matches the output before any actions are applied. | |
4395 | * - When using VLAN splinters, 'base_flow''s VLAN is set to the value | |
4396 | * of the received packet as seen by the kernel. If we later output | |
4397 | * to another device without any modifications this will cause us to | |
4398 | * insert a new tag since the original one was stripped off by the | |
4399 | * VLAN device. | |
4400 | * - Tunnel metadata as received is retained in 'flow'. This allows | |
4401 | * tunnel metadata matching also in later tables. | |
4402 | * Since a kernel action for setting the tunnel metadata will only be | |
4403 | * generated with actual tunnel output, changing the tunnel metadata | |
4404 | * values in 'flow' (such as tun_id) will only have effect with a later | |
4405 | * tunnel output action. | |
4406 | * - Tunnel 'base_flow' is completely cleared since that is what the | |
4407 | * kernel does. If we wish to maintain the original values an action | |
4408 | * needs to be generated. */ | |
4409 | ||
4410 | ctx.xin = xin; | |
4411 | ctx.xout = xout; | |
46c88433 EJ |
4412 | ctx.xout->slow = 0; |
4413 | ctx.xout->has_learn = false; | |
4414 | ctx.xout->has_normal = false; | |
4415 | ctx.xout->has_fin_timeout = false; | |
4416 | ctx.xout->nf_output_iface = NF_OUT_DROP; | |
4417 | ctx.xout->mirrors = 0; | |
cc377352 EJ |
4418 | |
4419 | xout->odp_actions = xin->odp_actions; | |
4420 | if (!xout->odp_actions) { | |
4421 | xout->odp_actions = &xout->odp_actions_buf; | |
4422 | ofpbuf_use_stub(xout->odp_actions, xout->odp_actions_stub, | |
4423 | sizeof xout->odp_actions_stub); | |
4424 | } | |
4425 | ofpbuf_reserve(xout->odp_actions, NL_A_U32_SIZE); | |
46c88433 | 4426 | |
84f0f298 | 4427 | ctx.xbridge = xbridge_lookup(xcfg, xin->ofproto); |
46c88433 | 4428 | if (!ctx.xbridge) { |
83709dfa | 4429 | return; |
46c88433 | 4430 | } |
9583bc14 | 4431 | |
9583bc14 EJ |
4432 | ctx.rule = xin->rule; |
4433 | ||
33bf9176 | 4434 | ctx.base_flow = *flow; |
9583bc14 | 4435 | memset(&ctx.base_flow.tunnel, 0, sizeof ctx.base_flow.tunnel); |
33bf9176 | 4436 | ctx.orig_tunnel_ip_dst = flow->tunnel.ip_dst; |
9583bc14 | 4437 | |
33bf9176 BP |
4438 | flow_wildcards_init_catchall(wc); |
4439 | memset(&wc->masks.in_port, 0xff, sizeof wc->masks.in_port); | |
7431e171 | 4440 | memset(&wc->masks.dl_type, 0xff, sizeof wc->masks.dl_type); |
0934ebba BP |
4441 | if (is_ip_any(flow)) { |
4442 | wc->masks.nw_frag |= FLOW_NW_FRAG_MASK; | |
4443 | } | |
c56fac1b | 4444 | is_icmp = is_icmpv4(flow) || is_icmpv6(flow); |
9583bc14 | 4445 | |
62a7cc71 | 4446 | tnl_may_send = tnl_xlate_init(&ctx.base_flow, flow, wc); |
ce3955be | 4447 | if (ctx.xbridge->netflow) { |
9b658910 | 4448 | netflow_mask_wc(flow, wc); |
9583bc14 EJ |
4449 | } |
4450 | ||
9583bc14 | 4451 | ctx.recurse = 0; |
98b07853 | 4452 | ctx.resubmits = 0; |
5a070238 | 4453 | ctx.in_group = false; |
029ca940 | 4454 | ctx.in_action_set = false; |
33bf9176 | 4455 | ctx.orig_skb_priority = flow->skb_priority; |
9583bc14 EJ |
4456 | ctx.table_id = 0; |
4457 | ctx.exit = false; | |
92c08f09 | 4458 | ctx.use_recirc = false; |
7bbdd84f | 4459 | ctx.was_mpls = false; |
9583bc14 | 4460 | |
10c44245 | 4461 | if (!xin->ofpacts && !ctx.rule) { |
34dd0d78 JR |
4462 | rule = rule_dpif_lookup(ctx.xbridge->ofproto, flow, |
4463 | xin->skip_wildcards ? NULL : wc, | |
4464 | ctx.xin->xcache != NULL, | |
4465 | ctx.xin->resubmit_stats, &ctx.table_id); | |
10c44245 EJ |
4466 | if (ctx.xin->resubmit_stats) { |
4467 | rule_dpif_credit_stats(rule, ctx.xin->resubmit_stats); | |
4468 | } | |
b256dc52 JS |
4469 | if (ctx.xin->xcache) { |
4470 | struct xc_entry *entry; | |
4471 | ||
4472 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_RULE); | |
b256dc52 JS |
4473 | entry->u.rule = rule; |
4474 | } | |
10c44245 | 4475 | ctx.rule = rule; |
a8c31348 BP |
4476 | |
4477 | if (OVS_UNLIKELY(ctx.xin->resubmit_hook)) { | |
4478 | ctx.xin->resubmit_hook(ctx.xin, rule, 0); | |
4479 | } | |
10c44245 | 4480 | } |
ee1afdd5 | 4481 | xout->fail_open = ctx.rule && rule_dpif_is_fail_open(ctx.rule); |
10c44245 | 4482 | |
9583bc14 EJ |
4483 | if (xin->ofpacts) { |
4484 | ofpacts = xin->ofpacts; | |
4485 | ofpacts_len = xin->ofpacts_len; | |
10c44245 | 4486 | } else if (ctx.rule) { |
d8227eba JR |
4487 | const struct rule_actions *actions = rule_dpif_get_actions(ctx.rule); |
4488 | ||
6f00e29b BP |
4489 | ofpacts = actions->ofpacts; |
4490 | ofpacts_len = actions->ofpacts_len; | |
9583bc14 | 4491 | } else { |
428b2edd | 4492 | OVS_NOT_REACHED(); |
9583bc14 EJ |
4493 | } |
4494 | ||
4495 | ofpbuf_use_stub(&ctx.stack, ctx.init_stack, sizeof ctx.init_stack); | |
c61f3870 BP |
4496 | |
4497 | ctx.action_set_has_group = false; | |
7fdb60a7 SH |
4498 | ofpbuf_use_stub(&ctx.action_set, |
4499 | ctx.action_set_stub, sizeof ctx.action_set_stub); | |
9583bc14 | 4500 | |
ade6ad9c | 4501 | if (mbridge_has_mirrors(ctx.xbridge->mbridge)) { |
9583bc14 EJ |
4502 | /* Do this conditionally because the copy is expensive enough that it |
4503 | * shows up in profiles. */ | |
33bf9176 | 4504 | orig_flow = *flow; |
9583bc14 EJ |
4505 | } |
4506 | ||
46c88433 | 4507 | in_port = get_ofp_port(ctx.xbridge, flow->in_port.ofp_port); |
b256dc52 JS |
4508 | if (in_port && in_port->is_tunnel) { |
4509 | if (ctx.xin->resubmit_stats) { | |
4510 | netdev_vport_inc_rx(in_port->netdev, ctx.xin->resubmit_stats); | |
4511 | if (in_port->bfd) { | |
4512 | bfd_account_rx(in_port->bfd, ctx.xin->resubmit_stats); | |
4513 | } | |
4514 | } | |
4515 | if (ctx.xin->xcache) { | |
4516 | struct xc_entry *entry; | |
4517 | ||
4518 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETDEV); | |
4519 | entry->u.dev.rx = netdev_ref(in_port->netdev); | |
4520 | entry->u.dev.bfd = bfd_ref(in_port->bfd); | |
d6fc5f57 EJ |
4521 | } |
4522 | } | |
4523 | ||
642dc74d | 4524 | special = process_special(&ctx, flow, in_port, ctx.xin->packet); |
9583bc14 | 4525 | if (special) { |
04594cd5 | 4526 | ctx.xout->slow |= special; |
9583bc14 | 4527 | } else { |
9583bc14 | 4528 | size_t sample_actions_len; |
9583bc14 | 4529 | |
4e022ec0 | 4530 | if (flow->in_port.ofp_port |
46c88433 EJ |
4531 | != vsp_realdev_to_vlandev(ctx.xbridge->ofproto, |
4532 | flow->in_port.ofp_port, | |
33bf9176 | 4533 | flow->vlan_tci)) { |
9583bc14 EJ |
4534 | ctx.base_flow.vlan_tci = 0; |
4535 | } | |
4536 | ||
4537 | add_sflow_action(&ctx); | |
4538 | add_ipfix_action(&ctx); | |
6fd6ed71 | 4539 | sample_actions_len = ctx.xout->odp_actions->size; |
9583bc14 | 4540 | |
62a7cc71 | 4541 | if (tnl_may_send && (!in_port || may_receive(in_port, &ctx))) { |
9583bc14 EJ |
4542 | do_xlate_actions(ofpacts, ofpacts_len, &ctx); |
4543 | ||
4544 | /* We've let OFPP_NORMAL and the learning action look at the | |
4545 | * packet, so drop it now if forwarding is disabled. */ | |
9efd308e DV |
4546 | if (in_port && (!xport_stp_forward_state(in_port) || |
4547 | !xport_rstp_forward_state(in_port))) { | |
8ec90813 | 4548 | /* Drop all actions added by do_xlate_actions() above. */ |
6fd6ed71 | 4549 | ctx.xout->odp_actions->size = sample_actions_len; |
8ec90813 JR |
4550 | } else if (ctx.action_set.size) { |
4551 | /* Translate action set only if not dropping the packet. */ | |
4552 | xlate_action_set(&ctx); | |
9583bc14 EJ |
4553 | } |
4554 | } | |
4555 | ||
46c88433 | 4556 | if (ctx.xbridge->has_in_band |
ce4a6b76 BP |
4557 | && in_band_must_output_to_local_port(flow) |
4558 | && !actions_output_to_local_port(&ctx)) { | |
9583bc14 EJ |
4559 | compose_output_action(&ctx, OFPP_LOCAL); |
4560 | } | |
aaa0fbae EJ |
4561 | |
4562 | fix_sflow_action(&ctx); | |
4563 | ||
46c88433 | 4564 | if (mbridge_has_mirrors(ctx.xbridge->mbridge)) { |
9583bc14 EJ |
4565 | add_mirror_actions(&ctx, &orig_flow); |
4566 | } | |
9583bc14 EJ |
4567 | } |
4568 | ||
6fd6ed71 | 4569 | if (nl_attr_oversized(ctx.xout->odp_actions->size)) { |
542024c4 | 4570 | /* These datapath actions are too big for a Netlink attribute, so we |
0f032e95 BP |
4571 | * can't hand them to the kernel directly. dpif_execute() can execute |
4572 | * them one by one with help, so just mark the result as SLOW_ACTION to | |
4573 | * prevent the flow from being installed. */ | |
4574 | COVERAGE_INC(xlate_actions_oversize); | |
4575 | ctx.xout->slow |= SLOW_ACTION; | |
7d031d7e BP |
4576 | } else if (too_many_output_actions(ctx.xout->odp_actions)) { |
4577 | COVERAGE_INC(xlate_actions_too_many_output); | |
4578 | ctx.xout->slow |= SLOW_ACTION; | |
542024c4 BP |
4579 | } |
4580 | ||
b256dc52 JS |
4581 | if (mbridge_has_mirrors(ctx.xbridge->mbridge)) { |
4582 | if (ctx.xin->resubmit_stats) { | |
4583 | mirror_update_stats(ctx.xbridge->mbridge, xout->mirrors, | |
4584 | ctx.xin->resubmit_stats->n_packets, | |
4585 | ctx.xin->resubmit_stats->n_bytes); | |
4586 | } | |
4587 | if (ctx.xin->xcache) { | |
4588 | struct xc_entry *entry; | |
4589 | ||
4590 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_MIRROR); | |
4591 | entry->u.mirror.mbridge = mbridge_ref(ctx.xbridge->mbridge); | |
4592 | entry->u.mirror.mirrors = xout->mirrors; | |
4593 | } | |
4594 | } | |
4595 | ||
4596 | if (ctx.xbridge->netflow) { | |
b256dc52 JS |
4597 | /* Only update netflow if we don't have controller flow. We don't |
4598 | * report NetFlow expiration messages for such facets because they | |
4599 | * are just part of the control logic for the network, not real | |
4600 | * traffic. */ | |
4601 | if (ofpacts_len == 0 | |
4602 | || ofpacts->type != OFPACT_CONTROLLER | |
4603 | || ofpact_next(ofpacts) < ofpact_end(ofpacts, ofpacts_len)) { | |
4604 | if (ctx.xin->resubmit_stats) { | |
d6fc5f57 EJ |
4605 | netflow_flow_update(ctx.xbridge->netflow, flow, |
4606 | xout->nf_output_iface, | |
4607 | ctx.xin->resubmit_stats); | |
4608 | } | |
b256dc52 JS |
4609 | if (ctx.xin->xcache) { |
4610 | struct xc_entry *entry; | |
4611 | ||
4612 | entry = xlate_cache_add_entry(ctx.xin->xcache, XC_NETFLOW); | |
4613 | entry->u.nf.netflow = netflow_ref(ctx.xbridge->netflow); | |
4614 | entry->u.nf.flow = xmemdup(flow, sizeof *flow); | |
4615 | entry->u.nf.iface = xout->nf_output_iface; | |
4616 | } | |
d6fc5f57 EJ |
4617 | } |
4618 | } | |
4619 | ||
9583bc14 | 4620 | ofpbuf_uninit(&ctx.stack); |
7fdb60a7 | 4621 | ofpbuf_uninit(&ctx.action_set); |
9583bc14 EJ |
4622 | |
4623 | /* Clear the metadata and register wildcard masks, because we won't | |
4624 | * use non-header fields as part of the cache. */ | |
c11c6faa | 4625 | flow_wildcards_clear_non_packet_fields(wc); |
dc24a00f | 4626 | |
c56fac1b BP |
4627 | /* ICMPv4 and ICMPv6 have 8-bit "type" and "code" fields. struct flow uses |
4628 | * the low 8 bits of the 16-bit tp_src and tp_dst members to represent | |
4629 | * these fields. The datapath interface, on the other hand, represents | |
4630 | * them with just 8 bits each. This means that if the high 8 bits of the | |
4631 | * masks for these fields somehow become set, then they will get chopped | |
4632 | * off by a round trip through the datapath, and revalidation will spot | |
4633 | * that as an inconsistency and delete the flow. Avoid the problem here by | |
4634 | * making sure that only the low 8 bits of either field can be unwildcarded | |
4635 | * for ICMP. | |
4636 | */ | |
4637 | if (is_icmp) { | |
4638 | wc->masks.tp_src &= htons(UINT8_MAX); | |
4639 | wc->masks.tp_dst &= htons(UINT8_MAX); | |
4640 | } | |
91d6cd12 AW |
4641 | } |
4642 | ||
4643 | /* Sends 'packet' out 'ofport'. | |
4644 | * May modify 'packet'. | |
4645 | * Returns 0 if successful, otherwise a positive errno value. */ | |
4646 | int | |
cf62fa4c | 4647 | xlate_send_packet(const struct ofport_dpif *ofport, struct dp_packet *packet) |
91d6cd12 | 4648 | { |
84f0f298 | 4649 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
91d6cd12 | 4650 | struct xport *xport; |
91d6cd12 | 4651 | struct ofpact_output output; |
91d6cd12 | 4652 | struct flow flow; |
91d6cd12 | 4653 | |
91d6cd12 AW |
4654 | ofpact_init(&output.ofpact, OFPACT_OUTPUT, sizeof output); |
4655 | /* Use OFPP_NONE as the in_port to avoid special packet processing. */ | |
cf62fa4c | 4656 | flow_extract(packet, &flow); |
b5e7e61a | 4657 | flow.in_port.ofp_port = OFPP_NONE; |
91d6cd12 | 4658 | |
84f0f298 | 4659 | xport = xport_lookup(xcfg, ofport); |
91d6cd12 | 4660 | if (!xport) { |
02ea2703 | 4661 | return EINVAL; |
91d6cd12 | 4662 | } |
91d6cd12 AW |
4663 | output.port = xport->ofp_port; |
4664 | output.max_len = 0; | |
e491a67a YT |
4665 | |
4666 | return ofproto_dpif_execute_actions(xport->xbridge->ofproto, &flow, NULL, | |
4667 | &output.ofpact, sizeof output, | |
4668 | packet); | |
9583bc14 | 4669 | } |
b256dc52 JS |
4670 | |
4671 | struct xlate_cache * | |
4672 | xlate_cache_new(void) | |
4673 | { | |
4674 | struct xlate_cache *xcache = xmalloc(sizeof *xcache); | |
4675 | ||
4676 | ofpbuf_init(&xcache->entries, 512); | |
4677 | return xcache; | |
4678 | } | |
4679 | ||
4680 | static struct xc_entry * | |
4681 | xlate_cache_add_entry(struct xlate_cache *xcache, enum xc_type type) | |
4682 | { | |
4683 | struct xc_entry *entry; | |
4684 | ||
4685 | entry = ofpbuf_put_zeros(&xcache->entries, sizeof *entry); | |
4686 | entry->type = type; | |
4687 | ||
4688 | return entry; | |
4689 | } | |
4690 | ||
4691 | static void | |
4692 | xlate_cache_netdev(struct xc_entry *entry, const struct dpif_flow_stats *stats) | |
4693 | { | |
4694 | if (entry->u.dev.tx) { | |
4695 | netdev_vport_inc_tx(entry->u.dev.tx, stats); | |
4696 | } | |
4697 | if (entry->u.dev.rx) { | |
4698 | netdev_vport_inc_rx(entry->u.dev.rx, stats); | |
4699 | } | |
4700 | if (entry->u.dev.bfd) { | |
4701 | bfd_account_rx(entry->u.dev.bfd, stats); | |
4702 | } | |
4703 | } | |
4704 | ||
4705 | static void | |
4706 | xlate_cache_normal(struct ofproto_dpif *ofproto, struct flow *flow, int vlan) | |
4707 | { | |
84f0f298 | 4708 | struct xlate_cfg *xcfg = ovsrcu_get(struct xlate_cfg *, &xcfgp); |
b256dc52 JS |
4709 | struct xbridge *xbridge; |
4710 | struct xbundle *xbundle; | |
4711 | struct flow_wildcards wc; | |
4712 | ||
84f0f298 | 4713 | xbridge = xbridge_lookup(xcfg, ofproto); |
b256dc52 JS |
4714 | if (!xbridge) { |
4715 | return; | |
4716 | } | |
4717 | ||
4718 | xbundle = lookup_input_bundle(xbridge, flow->in_port.ofp_port, false, | |
4719 | NULL); | |
4720 | if (!xbundle) { | |
4721 | return; | |
4722 | } | |
4723 | ||
4724 | update_learning_table(xbridge, flow, &wc, vlan, xbundle); | |
4725 | } | |
4726 | ||
4727 | /* Push stats and perform side effects of flow translation. */ | |
4728 | void | |
0725e747 | 4729 | xlate_push_stats(struct xlate_cache *xcache, |
b256dc52 JS |
4730 | const struct dpif_flow_stats *stats) |
4731 | { | |
4732 | struct xc_entry *entry; | |
4733 | struct ofpbuf entries = xcache->entries; | |
a36de779 | 4734 | uint8_t dmac[ETH_ADDR_LEN]; |
b256dc52 | 4735 | |
bc388583 BP |
4736 | if (!stats->n_packets) { |
4737 | return; | |
4738 | } | |
4739 | ||
b256dc52 JS |
4740 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { |
4741 | switch (entry->type) { | |
4742 | case XC_RULE: | |
4743 | rule_dpif_credit_stats(entry->u.rule, stats); | |
4744 | break; | |
4745 | case XC_BOND: | |
4746 | bond_account(entry->u.bond.bond, entry->u.bond.flow, | |
4747 | entry->u.bond.vid, stats->n_bytes); | |
4748 | break; | |
4749 | case XC_NETDEV: | |
4750 | xlate_cache_netdev(entry, stats); | |
4751 | break; | |
4752 | case XC_NETFLOW: | |
4753 | netflow_flow_update(entry->u.nf.netflow, entry->u.nf.flow, | |
4754 | entry->u.nf.iface, stats); | |
4755 | break; | |
4756 | case XC_MIRROR: | |
4757 | mirror_update_stats(entry->u.mirror.mbridge, | |
4758 | entry->u.mirror.mirrors, | |
4759 | stats->n_packets, stats->n_bytes); | |
4760 | break; | |
4761 | case XC_LEARN: | |
0725e747 | 4762 | ofproto_dpif_flow_mod(entry->u.learn.ofproto, entry->u.learn.fm); |
b256dc52 JS |
4763 | break; |
4764 | case XC_NORMAL: | |
0725e747 BP |
4765 | xlate_cache_normal(entry->u.normal.ofproto, entry->u.normal.flow, |
4766 | entry->u.normal.vlan); | |
b256dc52 JS |
4767 | break; |
4768 | case XC_FIN_TIMEOUT: | |
4769 | xlate_fin_timeout__(entry->u.fin.rule, stats->tcp_flags, | |
4770 | entry->u.fin.idle, entry->u.fin.hard); | |
4771 | break; | |
1e684d7d RW |
4772 | case XC_GROUP: |
4773 | group_dpif_credit_stats(entry->u.group.group, entry->u.group.bucket, | |
4774 | stats); | |
4775 | break; | |
a36de779 PS |
4776 | case XC_TNL_ARP: |
4777 | /* Lookup arp to avoid arp timeout. */ | |
4778 | tnl_arp_lookup(entry->u.tnl_arp_cache.br_name, entry->u.tnl_arp_cache.d_ip, dmac); | |
4779 | break; | |
b256dc52 JS |
4780 | default: |
4781 | OVS_NOT_REACHED(); | |
4782 | } | |
4783 | } | |
4784 | } | |
4785 | ||
4786 | static void | |
4787 | xlate_dev_unref(struct xc_entry *entry) | |
4788 | { | |
4789 | if (entry->u.dev.tx) { | |
4790 | netdev_close(entry->u.dev.tx); | |
4791 | } | |
4792 | if (entry->u.dev.rx) { | |
4793 | netdev_close(entry->u.dev.rx); | |
4794 | } | |
4795 | if (entry->u.dev.bfd) { | |
4796 | bfd_unref(entry->u.dev.bfd); | |
4797 | } | |
4798 | } | |
4799 | ||
4800 | static void | |
4801 | xlate_cache_clear_netflow(struct netflow *netflow, struct flow *flow) | |
4802 | { | |
b256dc52 JS |
4803 | netflow_flow_clear(netflow, flow); |
4804 | netflow_unref(netflow); | |
4805 | free(flow); | |
4806 | } | |
4807 | ||
4808 | void | |
4809 | xlate_cache_clear(struct xlate_cache *xcache) | |
4810 | { | |
4811 | struct xc_entry *entry; | |
4812 | struct ofpbuf entries; | |
4813 | ||
4814 | if (!xcache) { | |
4815 | return; | |
4816 | } | |
4817 | ||
4818 | XC_ENTRY_FOR_EACH (entry, entries, xcache) { | |
4819 | switch (entry->type) { | |
4820 | case XC_RULE: | |
4821 | rule_dpif_unref(entry->u.rule); | |
4822 | break; | |
4823 | case XC_BOND: | |
4824 | free(entry->u.bond.flow); | |
4825 | bond_unref(entry->u.bond.bond); | |
4826 | break; | |
4827 | case XC_NETDEV: | |
4828 | xlate_dev_unref(entry); | |
4829 | break; | |
4830 | case XC_NETFLOW: | |
4831 | xlate_cache_clear_netflow(entry->u.nf.netflow, entry->u.nf.flow); | |
4832 | break; | |
4833 | case XC_MIRROR: | |
4834 | mbridge_unref(entry->u.mirror.mbridge); | |
4835 | break; | |
4836 | case XC_LEARN: | |
4165b5e0 JS |
4837 | free(entry->u.learn.fm); |
4838 | ofpbuf_delete(entry->u.learn.ofpacts); | |
b256dc52 JS |
4839 | break; |
4840 | case XC_NORMAL: | |
4841 | free(entry->u.normal.flow); | |
4842 | break; | |
4843 | case XC_FIN_TIMEOUT: | |
83709dfa JR |
4844 | /* 'u.fin.rule' is always already held as a XC_RULE, which |
4845 | * has already released it's reference above. */ | |
b256dc52 | 4846 | break; |
1e684d7d RW |
4847 | case XC_GROUP: |
4848 | group_dpif_unref(entry->u.group.group); | |
4849 | break; | |
a36de779 PS |
4850 | case XC_TNL_ARP: |
4851 | break; | |
b256dc52 JS |
4852 | default: |
4853 | OVS_NOT_REACHED(); | |
4854 | } | |
4855 | } | |
4856 | ||
4857 | ofpbuf_clear(&xcache->entries); | |
4858 | } | |
4859 | ||
4860 | void | |
4861 | xlate_cache_delete(struct xlate_cache *xcache) | |
4862 | { | |
4863 | xlate_cache_clear(xcache); | |
4864 | ofpbuf_uninit(&xcache->entries); | |
4865 | free(xcache); | |
4866 | } |