]>
Commit | Line | Data |
---|---|---|
064af421 | 1 | /* |
db5ce514 | 2 | * Copyright (c) 2009, 2010, 2011 Nicira Networks. |
43253595 | 3 | * Copyright (c) 2010 Jean Tourrilhes - HP-Labs. |
064af421 | 4 | * |
a14bc59f BP |
5 | * Licensed under the Apache License, Version 2.0 (the "License"); |
6 | * you may not use this file except in compliance with the License. | |
7 | * You may obtain a copy of the License at: | |
064af421 | 8 | * |
a14bc59f BP |
9 | * http://www.apache.org/licenses/LICENSE-2.0 |
10 | * | |
11 | * Unless required by applicable law or agreed to in writing, software | |
12 | * distributed under the License is distributed on an "AS IS" BASIS, | |
13 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |
14 | * See the License for the specific language governing permissions and | |
15 | * limitations under the License. | |
064af421 BP |
16 | */ |
17 | ||
18 | #include <config.h> | |
19 | #include "ofproto.h" | |
20 | #include <errno.h> | |
21 | #include <inttypes.h> | |
9d82ec47 | 22 | #include <sys/socket.h> |
064af421 BP |
23 | #include <net/if.h> |
24 | #include <netinet/in.h> | |
25 | #include <stdbool.h> | |
26 | #include <stdlib.h> | |
10a24935 | 27 | #include "byte-order.h" |
064af421 BP |
28 | #include "classifier.h" |
29 | #include "coverage.h" | |
064af421 | 30 | #include "dpif.h" |
4f2cad2c | 31 | #include "dynamic-string.h" |
064af421 | 32 | #include "fail-open.h" |
ca0f572c BP |
33 | #include "hash.h" |
34 | #include "hmap.h" | |
064af421 BP |
35 | #include "in-band.h" |
36 | #include "mac-learning.h" | |
53ddd40a | 37 | #include "multipath.h" |
064af421 BP |
38 | #include "netdev.h" |
39 | #include "netflow.h" | |
cdee00fd | 40 | #include "netlink.h" |
09246b99 | 41 | #include "nx-match.h" |
064af421 BP |
42 | #include "odp-util.h" |
43 | #include "ofp-print.h" | |
fa37b408 | 44 | #include "ofp-util.h" |
72b06300 | 45 | #include "ofproto-sflow.h" |
064af421 BP |
46 | #include "ofpbuf.h" |
47 | #include "openflow/nicira-ext.h" | |
48 | #include "openflow/openflow.h" | |
064af421 BP |
49 | #include "openvswitch/datapath-protocol.h" |
50 | #include "packets.h" | |
51 | #include "pinsched.h" | |
52 | #include "pktbuf.h" | |
53 | #include "poll-loop.h" | |
064af421 BP |
54 | #include "rconn.h" |
55 | #include "shash.h" | |
fe55ad15 | 56 | #include "stream-ssl.h" |
064af421 BP |
57 | #include "svec.h" |
58 | #include "tag.h" | |
59 | #include "timeval.h" | |
c4617b3c | 60 | #include "unaligned.h" |
4f2cad2c | 61 | #include "unixctl.h" |
064af421 | 62 | #include "vconn.h" |
5136ce49 | 63 | #include "vlog.h" |
064af421 | 64 | |
d98e6007 | 65 | VLOG_DEFINE_THIS_MODULE(ofproto); |
064af421 | 66 | |
cc01d0bb BP |
67 | COVERAGE_DEFINE(facet_changed_rule); |
68 | COVERAGE_DEFINE(facet_revalidate); | |
d76f09ea | 69 | COVERAGE_DEFINE(odp_overflow); |
d76f09ea BP |
70 | COVERAGE_DEFINE(ofproto_agg_request); |
71 | COVERAGE_DEFINE(ofproto_costly_flags); | |
72 | COVERAGE_DEFINE(ofproto_ctlr_action); | |
cc01d0bb | 73 | COVERAGE_DEFINE(ofproto_del_rule); |
d76f09ea BP |
74 | COVERAGE_DEFINE(ofproto_error); |
75 | COVERAGE_DEFINE(ofproto_expiration); | |
76 | COVERAGE_DEFINE(ofproto_expired); | |
77 | COVERAGE_DEFINE(ofproto_flows_req); | |
78 | COVERAGE_DEFINE(ofproto_flush); | |
79 | COVERAGE_DEFINE(ofproto_invalidated); | |
d76f09ea | 80 | COVERAGE_DEFINE(ofproto_no_packet_in); |
d76f09ea BP |
81 | COVERAGE_DEFINE(ofproto_ofconn_stuck); |
82 | COVERAGE_DEFINE(ofproto_ofp2odp); | |
83 | COVERAGE_DEFINE(ofproto_packet_in); | |
84 | COVERAGE_DEFINE(ofproto_packet_out); | |
85 | COVERAGE_DEFINE(ofproto_queue_req); | |
86 | COVERAGE_DEFINE(ofproto_recv_openflow); | |
87 | COVERAGE_DEFINE(ofproto_reinit_ports); | |
d76f09ea BP |
88 | COVERAGE_DEFINE(ofproto_unexpected_rule); |
89 | COVERAGE_DEFINE(ofproto_uninstallable); | |
90 | COVERAGE_DEFINE(ofproto_update_port); | |
91 | ||
878ae780 EJ |
92 | /* Maximum depth of flow table recursion (due to NXAST_RESUBMIT actions) in a |
93 | * flow translation. */ | |
94 | #define MAX_RESUBMIT_RECURSION 16 | |
95 | ||
f29152ca BP |
96 | struct rule; |
97 | ||
064af421 | 98 | struct ofport { |
ca0f572c | 99 | struct hmap_node hmap_node; /* In struct ofproto's "ports" hmap. */ |
064af421 BP |
100 | struct netdev *netdev; |
101 | struct ofp_phy_port opp; /* In host byte order. */ | |
ca0f572c | 102 | uint16_t odp_port; |
064af421 BP |
103 | }; |
104 | ||
105 | static void ofport_free(struct ofport *); | |
106 | static void hton_ofp_phy_port(struct ofp_phy_port *); | |
107 | ||
f29152ca BP |
108 | struct action_xlate_ctx { |
109 | /* action_xlate_ctx_init() initializes these members. */ | |
110 | ||
111 | /* The ofproto. */ | |
112 | struct ofproto *ofproto; | |
113 | ||
114 | /* Flow to which the OpenFlow actions apply. xlate_actions() will modify | |
115 | * this flow when actions change header fields. */ | |
116 | struct flow flow; | |
117 | ||
118 | /* The packet corresponding to 'flow', or a null pointer if we are | |
119 | * revalidating without a packet to refer to. */ | |
120 | const struct ofpbuf *packet; | |
121 | ||
7aa697dd BP |
122 | /* If nonnull, called just before executing a resubmit action. |
123 | * | |
124 | * This is normally null so the client has to set it manually after | |
125 | * calling action_xlate_ctx_init(). */ | |
878ae780 | 126 | void (*resubmit_hook)(struct action_xlate_ctx *, struct rule *); |
7aa697dd | 127 | |
ebe482fd EJ |
128 | /* If true, the speciality of 'flow' should be checked before executing |
129 | * its actions. If special_cb returns false on 'flow' rendered | |
130 | * uninstallable and no actions will be executed. */ | |
131 | bool check_special; | |
132 | ||
f29152ca BP |
133 | /* xlate_actions() initializes and uses these members. The client might want |
134 | * to look at them after it returns. */ | |
135 | ||
cdee00fd | 136 | struct ofpbuf *odp_actions; /* Datapath actions. */ |
f29152ca BP |
137 | tag_type tags; /* Tags associated with OFPP_NORMAL actions. */ |
138 | bool may_set_up_flow; /* True ordinarily; false if the actions must | |
139 | * be reassessed for every packet. */ | |
140 | uint16_t nf_output_iface; /* Output interface index for NetFlow. */ | |
141 | ||
142 | /* xlate_actions() initializes and uses these members, but the client has no | |
143 | * reason to look at them. */ | |
144 | ||
145 | int recurse; /* Recursion level, via xlate_table_action. */ | |
cdee00fd | 146 | int last_pop_priority; /* Offset in 'odp_actions' just past most |
7aec165d | 147 | * recent ODP_ACTION_ATTR_SET_PRIORITY. */ |
f29152ca BP |
148 | }; |
149 | ||
150 | static void action_xlate_ctx_init(struct action_xlate_ctx *, | |
151 | struct ofproto *, const struct flow *, | |
152 | const struct ofpbuf *); | |
cdee00fd BP |
153 | static struct ofpbuf *xlate_actions(struct action_xlate_ctx *, |
154 | const union ofp_action *in, size_t n_in); | |
064af421 | 155 | |
bcf84111 | 156 | /* An OpenFlow flow. */ |
064af421 | 157 | struct rule { |
0c43ad9a | 158 | long long int used; /* Time last used; time created if not used. */ |
064af421 | 159 | long long int created; /* Creation time. */ |
064af421 | 160 | |
bcf84111 | 161 | /* These statistics: |
064af421 | 162 | * |
bcf84111 BP |
163 | * - Do include packets and bytes from facets that have been deleted or |
164 | * whose own statistics have been folded into the rule. | |
79eee1eb | 165 | * |
bcf84111 BP |
166 | * - Do include packets and bytes sent "by hand" that were accounted to |
167 | * the rule without any facet being involved (this is a rare corner | |
168 | * case in rule_execute()). | |
064af421 | 169 | * |
bcf84111 BP |
170 | * - Do not include packet or bytes that can be obtained from any facet's |
171 | * packet_count or byte_count member or that can be obtained from the | |
172 | * datapath by, e.g., dpif_flow_get() for any facet. | |
173 | */ | |
174 | uint64_t packet_count; /* Number of packets received. */ | |
175 | uint64_t byte_count; /* Number of bytes received. */ | |
176 | ||
177 | ovs_be64 flow_cookie; /* Controller-issued identifier. */ | |
178 | ||
179 | struct cls_rule cr; /* In owning ofproto's classifier. */ | |
180 | uint16_t idle_timeout; /* In seconds from time of last use. */ | |
181 | uint16_t hard_timeout; /* In seconds from time of creation. */ | |
182 | bool send_flow_removed; /* Send a flow removed message? */ | |
183 | int n_actions; /* Number of elements in actions[]. */ | |
184 | union ofp_action *actions; /* OpenFlow actions. */ | |
185 | struct list facets; /* List of "struct facet"s. */ | |
064af421 BP |
186 | }; |
187 | ||
bcf84111 BP |
188 | static struct rule *rule_from_cls_rule(const struct cls_rule *); |
189 | static bool rule_is_hidden(const struct rule *); | |
064af421 | 190 | |
bcf84111 | 191 | static struct rule *rule_create(const struct cls_rule *, |
0193b2af | 192 | const union ofp_action *, size_t n_actions, |
ca069229 | 193 | uint16_t idle_timeout, uint16_t hard_timeout, |
8054fc48 | 194 | ovs_be64 flow_cookie, bool send_flow_removed); |
064af421 | 195 | static void rule_destroy(struct ofproto *, struct rule *); |
bcf84111 BP |
196 | static void rule_free(struct rule *); |
197 | ||
198 | static struct rule *rule_lookup(struct ofproto *, const struct flow *); | |
afe75089 | 199 | static void rule_insert(struct ofproto *, struct rule *); |
064af421 | 200 | static void rule_remove(struct ofproto *, struct rule *); |
bcf84111 BP |
201 | |
202 | static void rule_send_removed(struct ofproto *, struct rule *, uint8_t reason); | |
3394b5b6 EJ |
203 | static void rule_get_stats(const struct rule *, uint64_t *packets, |
204 | uint64_t *bytes); | |
bcf84111 BP |
205 | |
206 | /* An exact-match instantiation of an OpenFlow flow. */ | |
207 | struct facet { | |
208 | long long int used; /* Time last used; time created if not used. */ | |
209 | ||
210 | /* These statistics: | |
211 | * | |
212 | * - Do include packets and bytes sent "by hand", e.g. with | |
213 | * dpif_execute(). | |
214 | * | |
215 | * - Do include packets and bytes that were obtained from the datapath | |
216 | * when a flow was deleted (e.g. dpif_flow_del()) or when its | |
ba25b8f4 BP |
217 | * statistics were reset (e.g. dpif_flow_put() with |
218 | * DPIF_FP_ZERO_STATS). | |
bcf84111 BP |
219 | * |
220 | * - Do not include any packets or bytes that can currently be obtained | |
221 | * from the datapath by, e.g., dpif_flow_get(). | |
222 | */ | |
223 | uint64_t packet_count; /* Number of packets received. */ | |
224 | uint64_t byte_count; /* Number of bytes received. */ | |
225 | ||
3394b5b6 EJ |
226 | uint64_t dp_packet_count; /* Last known packet count in the datapath. */ |
227 | uint64_t dp_byte_count; /* Last known byte count in the datapath. */ | |
228 | ||
878ae780 EJ |
229 | uint64_t rs_packet_count; /* Packets pushed to resubmit children. */ |
230 | uint64_t rs_byte_count; /* Bytes pushed to resubmit children. */ | |
231 | long long int rs_used; /* Used time pushed to resubmit children. */ | |
232 | ||
bcf84111 BP |
233 | /* Number of bytes passed to account_cb. This may include bytes that can |
234 | * currently obtained from the datapath (thus, it can be greater than | |
235 | * byte_count). */ | |
236 | uint64_t accounted_bytes; | |
237 | ||
238 | struct hmap_node hmap_node; /* In owning ofproto's 'facets' hmap. */ | |
239 | struct list list_node; /* In owning rule's 'facets' list. */ | |
240 | struct rule *rule; /* Owning rule. */ | |
241 | struct flow flow; /* Exact-match flow. */ | |
242 | bool installed; /* Installed in datapath? */ | |
243 | bool may_install; /* True ordinarily; false if actions must | |
244 | * be reassessed for every packet. */ | |
cf22f8cb | 245 | size_t actions_len; /* Number of bytes in actions[]. */ |
cdee00fd | 246 | struct nlattr *actions; /* Datapath actions. */ |
bcf84111 BP |
247 | tag_type tags; /* Tags (set only by hooks). */ |
248 | struct netflow_flow nf_flow; /* Per-flow NetFlow tracking data. */ | |
249 | }; | |
250 | ||
251 | static struct facet *facet_create(struct ofproto *, struct rule *, | |
252 | const struct flow *, | |
253 | const struct ofpbuf *packet); | |
254 | static void facet_remove(struct ofproto *, struct facet *); | |
255 | static void facet_free(struct facet *); | |
256 | ||
257 | static struct facet *facet_lookup_valid(struct ofproto *, const struct flow *); | |
258 | static bool facet_revalidate(struct ofproto *, struct facet *); | |
259 | ||
260 | static void facet_install(struct ofproto *, struct facet *, bool zero_stats); | |
261 | static void facet_uninstall(struct ofproto *, struct facet *); | |
d530fcd2 | 262 | static void facet_flush_stats(struct ofproto *, struct facet *); |
bcf84111 | 263 | |
7f7ae89d BP |
264 | static void facet_make_actions(struct ofproto *, struct facet *, |
265 | const struct ofpbuf *packet); | |
bcf84111 | 266 | static void facet_update_stats(struct ofproto *, struct facet *, |
c97fb132 | 267 | const struct dpif_flow_stats *); |
878ae780 | 268 | static void facet_push_stats(struct ofproto *, struct facet *); |
064af421 | 269 | |
76ce9432 BP |
270 | /* ofproto supports two kinds of OpenFlow connections: |
271 | * | |
5899143f BP |
272 | * - "Primary" connections to ordinary OpenFlow controllers. ofproto |
273 | * maintains persistent connections to these controllers and by default | |
274 | * sends them asynchronous messages such as packet-ins. | |
76ce9432 | 275 | * |
5899143f | 276 | * - "Service" connections, e.g. from ovs-ofctl. When these connections |
76ce9432 BP |
277 | * drop, it is the other side's responsibility to reconnect them if |
278 | * necessary. ofproto does not send them asynchronous messages by default. | |
7d674866 BP |
279 | * |
280 | * Currently, active (tcp, ssl, unix) connections are always "primary" | |
281 | * connections and passive (ptcp, pssl, punix) connections are always "service" | |
282 | * connections. There is no inherent reason for this, but it reflects the | |
283 | * common case. | |
76ce9432 BP |
284 | */ |
285 | enum ofconn_type { | |
5899143f BP |
286 | OFCONN_PRIMARY, /* An ordinary OpenFlow controller. */ |
287 | OFCONN_SERVICE /* A service connection, e.g. "ovs-ofctl". */ | |
76ce9432 | 288 | }; |
064af421 | 289 | |
7d674866 BP |
290 | /* A listener for incoming OpenFlow "service" connections. */ |
291 | struct ofservice { | |
292 | struct hmap_node node; /* In struct ofproto's "services" hmap. */ | |
293 | struct pvconn *pvconn; /* OpenFlow connection listener. */ | |
294 | ||
295 | /* These are not used by ofservice directly. They are settings for | |
296 | * accepted "struct ofconn"s from the pvconn. */ | |
297 | int probe_interval; /* Max idle time before probing, in seconds. */ | |
298 | int rate_limit; /* Max packet-in rate in packets per second. */ | |
299 | int burst_limit; /* Limit on accumulating packet credits. */ | |
300 | }; | |
301 | ||
302 | static struct ofservice *ofservice_lookup(struct ofproto *, | |
303 | const char *target); | |
304 | static int ofservice_create(struct ofproto *, | |
305 | const struct ofproto_controller *); | |
306 | static void ofservice_reconfigure(struct ofservice *, | |
307 | const struct ofproto_controller *); | |
308 | static void ofservice_destroy(struct ofproto *, struct ofservice *); | |
309 | ||
76ce9432 BP |
310 | /* An OpenFlow connection. */ |
311 | struct ofconn { | |
312 | struct ofproto *ofproto; /* The ofproto that owns this connection. */ | |
313 | struct list node; /* In struct ofproto's "all_conns" list. */ | |
314 | struct rconn *rconn; /* OpenFlow connection. */ | |
315 | enum ofconn_type type; /* Type. */ | |
492f7572 | 316 | enum nx_flow_format flow_format; /* Currently selected flow format. */ |
76ce9432 BP |
317 | |
318 | /* OFPT_PACKET_IN related data. */ | |
319 | struct rconn_packet_counter *packet_in_counter; /* # queued on 'rconn'. */ | |
82272ede BP |
320 | #define N_SCHEDULERS 2 |
321 | struct pinsched *schedulers[N_SCHEDULERS]; | |
76ce9432 BP |
322 | struct pktbuf *pktbuf; /* OpenFlow packet buffers. */ |
323 | int miss_send_len; /* Bytes to send of buffered packets. */ | |
324 | ||
325 | /* Number of OpenFlow messages queued on 'rconn' as replies to OpenFlow | |
326 | * requests, and the maximum number before we stop reading OpenFlow | |
327 | * requests. */ | |
064af421 BP |
328 | #define OFCONN_REPLY_MAX 100 |
329 | struct rconn_packet_counter *reply_counter; | |
76ce9432 | 330 | |
5899143f | 331 | /* type == OFCONN_PRIMARY only. */ |
9deba63b | 332 | enum nx_role role; /* Role. */ |
76ce9432 | 333 | struct hmap_node hmap_node; /* In struct ofproto's "controllers" map. */ |
d2ede7bc | 334 | enum ofproto_band band; /* In-band or out-of-band? */ |
064af421 BP |
335 | }; |
336 | ||
76ce9432 BP |
337 | |
338 | static struct ofconn *ofconn_create(struct ofproto *, struct rconn *, | |
339 | enum ofconn_type); | |
c475ae67 | 340 | static void ofconn_destroy(struct ofconn *); |
3269c562 | 341 | static void ofconn_run(struct ofconn *); |
064af421 | 342 | static void ofconn_wait(struct ofconn *); |
c91248b3 | 343 | static bool ofconn_receives_async_msgs(const struct ofconn *); |
eb15cdbb | 344 | static char *ofconn_make_name(const struct ofproto *, const char *target); |
7d674866 | 345 | static void ofconn_set_rate_limit(struct ofconn *, int rate, int burst); |
c91248b3 | 346 | |
064af421 BP |
347 | static void queue_tx(struct ofpbuf *msg, const struct ofconn *ofconn, |
348 | struct rconn_packet_counter *counter); | |
349 | ||
856081f6 BP |
350 | static void send_packet_in(struct ofproto *, struct dpif_upcall *, |
351 | const struct flow *, bool clone); | |
352 | static void do_send_packet_in(struct ofpbuf *ofp_packet_in, void *ofconn); | |
76ce9432 | 353 | |
064af421 BP |
354 | struct ofproto { |
355 | /* Settings. */ | |
356 | uint64_t datapath_id; /* Datapath ID. */ | |
357 | uint64_t fallback_dpid; /* Datapath ID if no better choice found. */ | |
5a719c38 JP |
358 | char *mfr_desc; /* Manufacturer. */ |
359 | char *hw_desc; /* Hardware. */ | |
360 | char *sw_desc; /* Software version. */ | |
361 | char *serial_desc; /* Serial number. */ | |
8abc4ed7 | 362 | char *dp_desc; /* Datapath description. */ |
064af421 BP |
363 | |
364 | /* Datapath. */ | |
c228a364 | 365 | struct dpif *dpif; |
e9e28be3 | 366 | struct netdev_monitor *netdev_monitor; |
ca0f572c | 367 | struct hmap ports; /* Contains "struct ofport"s. */ |
064af421 BP |
368 | struct shash port_by_name; |
369 | uint32_t max_ports; | |
370 | ||
371 | /* Configuration. */ | |
064af421 | 372 | struct fail_open *fail_open; |
064af421 | 373 | struct netflow *netflow; |
72b06300 | 374 | struct ofproto_sflow *sflow; |
064af421 | 375 | |
d2ede7bc BP |
376 | /* In-band control. */ |
377 | struct in_band *in_band; | |
378 | long long int next_in_band_update; | |
917e50e1 BP |
379 | struct sockaddr_in *extra_in_band_remotes; |
380 | size_t n_extra_remotes; | |
b1da6250 | 381 | int in_band_queue; |
917e50e1 | 382 | |
064af421 BP |
383 | /* Flow table. */ |
384 | struct classifier cls; | |
064af421 | 385 | long long int next_expiration; |
bcf84111 BP |
386 | |
387 | /* Facets. */ | |
388 | struct hmap facets; | |
389 | bool need_revalidate; | |
064af421 BP |
390 | struct tag_set revalidate_set; |
391 | ||
392 | /* OpenFlow connections. */ | |
76ce9432 BP |
393 | struct hmap controllers; /* Controller "struct ofconn"s. */ |
394 | struct list all_conns; /* Contains "struct ofconn"s. */ | |
31681a5d | 395 | enum ofproto_fail_mode fail_mode; |
7d674866 BP |
396 | |
397 | /* OpenFlow listeners. */ | |
398 | struct hmap services; /* Contains "struct ofservice"s. */ | |
064af421 BP |
399 | struct pvconn **snoops; |
400 | size_t n_snoops; | |
401 | ||
402 | /* Hooks for ovs-vswitchd. */ | |
403 | const struct ofhooks *ofhooks; | |
404 | void *aux; | |
405 | ||
406 | /* Used by default ofhooks. */ | |
407 | struct mac_learning *ml; | |
408 | }; | |
409 | ||
7aa697dd BP |
410 | /* Map from dpif name to struct ofproto, for use by unixctl commands. */ |
411 | static struct shash all_ofprotos = SHASH_INITIALIZER(&all_ofprotos); | |
412 | ||
064af421 BP |
413 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); |
414 | ||
415 | static const struct ofhooks default_ofhooks; | |
416 | ||
fa60c019 | 417 | static uint64_t pick_datapath_id(const struct ofproto *); |
064af421 | 418 | static uint64_t pick_fallback_dpid(void); |
76ce9432 | 419 | |
0de7a4b4 | 420 | static int ofproto_expire(struct ofproto *); |
878ae780 EJ |
421 | static void flow_push_stats(struct ofproto *, const struct rule *, |
422 | struct flow *, uint64_t packets, uint64_t bytes, | |
423 | long long int used); | |
4a4cdb3b | 424 | |
856081f6 | 425 | static void handle_upcall(struct ofproto *, struct dpif_upcall *); |
064af421 | 426 | |
3269c562 | 427 | static void handle_openflow(struct ofconn *, struct ofpbuf *); |
064af421 | 428 | |
ca0f572c | 429 | static struct ofport *get_port(const struct ofproto *, uint16_t odp_port); |
064af421 BP |
430 | static void update_port(struct ofproto *, const char *devname); |
431 | static int init_ports(struct ofproto *); | |
432 | static void reinit_ports(struct ofproto *); | |
433 | ||
7aa697dd BP |
434 | static void ofproto_unixctl_init(void); |
435 | ||
064af421 | 436 | int |
1a6f1e2a JG |
437 | ofproto_create(const char *datapath, const char *datapath_type, |
438 | const struct ofhooks *ofhooks, void *aux, | |
064af421 BP |
439 | struct ofproto **ofprotop) |
440 | { | |
064af421 | 441 | struct ofproto *p; |
c228a364 | 442 | struct dpif *dpif; |
064af421 BP |
443 | int error; |
444 | ||
445 | *ofprotop = NULL; | |
446 | ||
7aa697dd BP |
447 | ofproto_unixctl_init(); |
448 | ||
064af421 | 449 | /* Connect to datapath and start listening for messages. */ |
1a6f1e2a | 450 | error = dpif_open(datapath, datapath_type, &dpif); |
064af421 BP |
451 | if (error) { |
452 | VLOG_ERR("failed to open datapath %s: %s", datapath, strerror(error)); | |
453 | return error; | |
454 | } | |
82272ede BP |
455 | error = dpif_recv_set_mask(dpif, |
456 | ((1u << DPIF_UC_MISS) | | |
457 | (1u << DPIF_UC_ACTION) | | |
458 | (1u << DPIF_UC_SAMPLE))); | |
064af421 BP |
459 | if (error) { |
460 | VLOG_ERR("failed to listen on datapath %s: %s", | |
461 | datapath, strerror(error)); | |
c228a364 | 462 | dpif_close(dpif); |
064af421 BP |
463 | return error; |
464 | } | |
c228a364 | 465 | dpif_flow_flush(dpif); |
8f24562a | 466 | dpif_recv_purge(dpif); |
064af421 BP |
467 | |
468 | /* Initialize settings. */ | |
ec6fde61 | 469 | p = xzalloc(sizeof *p); |
064af421 | 470 | p->fallback_dpid = pick_fallback_dpid(); |
fa60c019 | 471 | p->datapath_id = p->fallback_dpid; |
5a719c38 JP |
472 | p->mfr_desc = xstrdup(DEFAULT_MFR_DESC); |
473 | p->hw_desc = xstrdup(DEFAULT_HW_DESC); | |
474 | p->sw_desc = xstrdup(DEFAULT_SW_DESC); | |
475 | p->serial_desc = xstrdup(DEFAULT_SERIAL_DESC); | |
23ff2821 | 476 | p->dp_desc = xstrdup(DEFAULT_DP_DESC); |
064af421 BP |
477 | |
478 | /* Initialize datapath. */ | |
479 | p->dpif = dpif; | |
8b61709d | 480 | p->netdev_monitor = netdev_monitor_create(); |
ca0f572c | 481 | hmap_init(&p->ports); |
064af421 | 482 | shash_init(&p->port_by_name); |
996c1b3d | 483 | p->max_ports = dpif_get_max_ports(dpif); |
064af421 BP |
484 | |
485 | /* Initialize submodules. */ | |
064af421 | 486 | p->fail_open = NULL; |
064af421 | 487 | p->netflow = NULL; |
72b06300 | 488 | p->sflow = NULL; |
064af421 | 489 | |
b1da6250 BP |
490 | /* Initialize in-band control. */ |
491 | p->in_band = NULL; | |
492 | p->in_band_queue = -1; | |
493 | ||
064af421 BP |
494 | /* Initialize flow table. */ |
495 | classifier_init(&p->cls); | |
064af421 | 496 | p->next_expiration = time_msec() + 1000; |
bcf84111 BP |
497 | |
498 | /* Initialize facet table. */ | |
499 | hmap_init(&p->facets); | |
500 | p->need_revalidate = false; | |
064af421 BP |
501 | tag_set_init(&p->revalidate_set); |
502 | ||
503 | /* Initialize OpenFlow connections. */ | |
504 | list_init(&p->all_conns); | |
76ce9432 | 505 | hmap_init(&p->controllers); |
7d674866 | 506 | hmap_init(&p->services); |
064af421 BP |
507 | p->snoops = NULL; |
508 | p->n_snoops = 0; | |
509 | ||
510 | /* Initialize hooks. */ | |
511 | if (ofhooks) { | |
512 | p->ofhooks = ofhooks; | |
513 | p->aux = aux; | |
514 | p->ml = NULL; | |
515 | } else { | |
516 | p->ofhooks = &default_ofhooks; | |
517 | p->aux = p; | |
518 | p->ml = mac_learning_create(); | |
519 | } | |
520 | ||
fa60c019 BP |
521 | /* Pick final datapath ID. */ |
522 | p->datapath_id = pick_datapath_id(p); | |
b123cc3c | 523 | VLOG_INFO("using datapath ID %016"PRIx64, p->datapath_id); |
fa60c019 | 524 | |
7aa697dd BP |
525 | shash_add_once(&all_ofprotos, dpif_name(p->dpif), p); |
526 | ||
064af421 BP |
527 | *ofprotop = p; |
528 | return 0; | |
529 | } | |
530 | ||
531 | void | |
532 | ofproto_set_datapath_id(struct ofproto *p, uint64_t datapath_id) | |
533 | { | |
534 | uint64_t old_dpid = p->datapath_id; | |
fa60c019 | 535 | p->datapath_id = datapath_id ? datapath_id : pick_datapath_id(p); |
064af421 | 536 | if (p->datapath_id != old_dpid) { |
b123cc3c | 537 | VLOG_INFO("datapath ID changed to %016"PRIx64, p->datapath_id); |
76ce9432 BP |
538 | |
539 | /* Force all active connections to reconnect, since there is no way to | |
540 | * notify a controller that the datapath ID has changed. */ | |
fa05809b | 541 | ofproto_reconnect_controllers(p); |
064af421 BP |
542 | } |
543 | } | |
544 | ||
76ce9432 BP |
545 | /* Creates a new controller in 'ofproto'. Some of the settings are initially |
546 | * drawn from 'c', but update_controller() needs to be called later to finish | |
547 | * the new ofconn's configuration. */ | |
548 | static void | |
549 | add_controller(struct ofproto *ofproto, const struct ofproto_controller *c) | |
550 | { | |
195c8086 | 551 | char *name = ofconn_make_name(ofproto, c->target); |
76ce9432 BP |
552 | struct ofconn *ofconn; |
553 | ||
5899143f | 554 | ofconn = ofconn_create(ofproto, rconn_create(5, 8), OFCONN_PRIMARY); |
76ce9432 BP |
555 | ofconn->pktbuf = pktbuf_create(); |
556 | ofconn->miss_send_len = OFP_DEFAULT_MISS_SEND_LEN; | |
195c8086 | 557 | rconn_connect(ofconn->rconn, c->target, name); |
76ce9432 BP |
558 | hmap_insert(&ofproto->controllers, &ofconn->hmap_node, |
559 | hash_string(c->target, 0)); | |
195c8086 BP |
560 | |
561 | free(name); | |
76ce9432 BP |
562 | } |
563 | ||
564 | /* Reconfigures 'ofconn' to match 'c'. This function cannot update an ofconn's | |
195c8086 BP |
565 | * target (this is done by creating new ofconns and deleting old ones), but it |
566 | * can update the rest of an ofconn's settings. */ | |
76ce9432 BP |
567 | static void |
568 | update_controller(struct ofconn *ofconn, const struct ofproto_controller *c) | |
064af421 | 569 | { |
76ce9432 | 570 | int probe_interval; |
79c9f2ee | 571 | |
195c8086 | 572 | ofconn->band = c->band; |
d2ede7bc | 573 | |
76ce9432 | 574 | rconn_set_max_backoff(ofconn->rconn, c->max_backoff); |
79c9f2ee | 575 | |
76ce9432 BP |
576 | probe_interval = c->probe_interval ? MAX(c->probe_interval, 5) : 0; |
577 | rconn_set_probe_interval(ofconn->rconn, probe_interval); | |
79c9f2ee | 578 | |
7d674866 | 579 | ofconn_set_rate_limit(ofconn, c->rate_limit, c->burst_limit); |
76ce9432 | 580 | } |
79c9f2ee | 581 | |
76ce9432 BP |
582 | static const char * |
583 | ofconn_get_target(const struct ofconn *ofconn) | |
584 | { | |
195c8086 | 585 | return rconn_get_target(ofconn->rconn); |
76ce9432 BP |
586 | } |
587 | ||
588 | static struct ofconn * | |
589 | find_controller_by_target(struct ofproto *ofproto, const char *target) | |
590 | { | |
591 | struct ofconn *ofconn; | |
592 | ||
4e8e4213 | 593 | HMAP_FOR_EACH_WITH_HASH (ofconn, hmap_node, |
76ce9432 BP |
594 | hash_string(target, 0), &ofproto->controllers) { |
595 | if (!strcmp(ofconn_get_target(ofconn), target)) { | |
596 | return ofconn; | |
79c9f2ee | 597 | } |
064af421 | 598 | } |
76ce9432 BP |
599 | return NULL; |
600 | } | |
064af421 | 601 | |
d2ede7bc BP |
602 | static void |
603 | update_in_band_remotes(struct ofproto *ofproto) | |
604 | { | |
605 | const struct ofconn *ofconn; | |
606 | struct sockaddr_in *addrs; | |
917e50e1 | 607 | size_t max_addrs, n_addrs; |
917e50e1 | 608 | size_t i; |
d2ede7bc | 609 | |
917e50e1 BP |
610 | /* Allocate enough memory for as many remotes as we could possibly have. */ |
611 | max_addrs = ofproto->n_extra_remotes + hmap_count(&ofproto->controllers); | |
612 | addrs = xmalloc(max_addrs * sizeof *addrs); | |
d2ede7bc BP |
613 | n_addrs = 0; |
614 | ||
615 | /* Add all the remotes. */ | |
4e8e4213 | 616 | HMAP_FOR_EACH (ofconn, hmap_node, &ofproto->controllers) { |
d2ede7bc BP |
617 | struct sockaddr_in *sin = &addrs[n_addrs]; |
618 | ||
487ec65f BP |
619 | if (ofconn->band == OFPROTO_OUT_OF_BAND) { |
620 | continue; | |
621 | } | |
622 | ||
d2ede7bc BP |
623 | sin->sin_addr.s_addr = rconn_get_remote_ip(ofconn->rconn); |
624 | if (sin->sin_addr.s_addr) { | |
625 | sin->sin_port = rconn_get_remote_port(ofconn->rconn); | |
626 | n_addrs++; | |
627 | } | |
d2ede7bc | 628 | } |
917e50e1 BP |
629 | for (i = 0; i < ofproto->n_extra_remotes; i++) { |
630 | addrs[n_addrs++] = ofproto->extra_in_band_remotes[i]; | |
631 | } | |
d2ede7bc | 632 | |
195c8086 BP |
633 | /* Create or update or destroy in-band. */ |
634 | if (n_addrs) { | |
d2ede7bc | 635 | if (!ofproto->in_band) { |
9b45d7f5 | 636 | in_band_create(ofproto, ofproto->dpif, &ofproto->in_band); |
d2ede7bc | 637 | } |
40cae670 BP |
638 | if (ofproto->in_band) { |
639 | in_band_set_remotes(ofproto->in_band, addrs, n_addrs); | |
640 | } | |
b1da6250 | 641 | in_band_set_queue(ofproto->in_band, ofproto->in_band_queue); |
d2ede7bc BP |
642 | ofproto->next_in_band_update = time_msec() + 1000; |
643 | } else { | |
644 | in_band_destroy(ofproto->in_band); | |
645 | ofproto->in_band = NULL; | |
646 | } | |
647 | ||
648 | /* Clean up. */ | |
649 | free(addrs); | |
650 | } | |
651 | ||
31681a5d JP |
652 | static void |
653 | update_fail_open(struct ofproto *p) | |
654 | { | |
655 | struct ofconn *ofconn; | |
656 | ||
657 | if (!hmap_is_empty(&p->controllers) | |
658 | && p->fail_mode == OFPROTO_FAIL_STANDALONE) { | |
659 | struct rconn **rconns; | |
660 | size_t n; | |
661 | ||
662 | if (!p->fail_open) { | |
9b45d7f5 | 663 | p->fail_open = fail_open_create(p); |
31681a5d JP |
664 | } |
665 | ||
666 | n = 0; | |
667 | rconns = xmalloc(hmap_count(&p->controllers) * sizeof *rconns); | |
4e8e4213 | 668 | HMAP_FOR_EACH (ofconn, hmap_node, &p->controllers) { |
31681a5d JP |
669 | rconns[n++] = ofconn->rconn; |
670 | } | |
671 | ||
672 | fail_open_set_controllers(p->fail_open, rconns, n); | |
673 | /* p->fail_open takes ownership of 'rconns'. */ | |
674 | } else { | |
675 | fail_open_destroy(p->fail_open); | |
676 | p->fail_open = NULL; | |
677 | } | |
678 | } | |
679 | ||
76ce9432 BP |
680 | void |
681 | ofproto_set_controllers(struct ofproto *p, | |
682 | const struct ofproto_controller *controllers, | |
683 | size_t n_controllers) | |
684 | { | |
685 | struct shash new_controllers; | |
7d674866 BP |
686 | struct ofconn *ofconn, *next_ofconn; |
687 | struct ofservice *ofservice, *next_ofservice; | |
76ce9432 | 688 | size_t i; |
79c9f2ee | 689 | |
7d674866 BP |
690 | /* Create newly configured controllers and services. |
691 | * Create a name to ofproto_controller mapping in 'new_controllers'. */ | |
76ce9432 BP |
692 | shash_init(&new_controllers); |
693 | for (i = 0; i < n_controllers; i++) { | |
694 | const struct ofproto_controller *c = &controllers[i]; | |
695 | ||
195c8086 | 696 | if (!vconn_verify_name(c->target)) { |
7d674866 BP |
697 | if (!find_controller_by_target(p, c->target)) { |
698 | add_controller(p, c); | |
699 | } | |
700 | } else if (!pvconn_verify_name(c->target)) { | |
701 | if (!ofservice_lookup(p, c->target) && ofservice_create(p, c)) { | |
702 | continue; | |
703 | } | |
704 | } else { | |
705 | VLOG_WARN_RL(&rl, "%s: unsupported controller \"%s\"", | |
706 | dpif_name(p->dpif), c->target); | |
707 | continue; | |
76ce9432 | 708 | } |
7d674866 BP |
709 | |
710 | shash_add_once(&new_controllers, c->target, &controllers[i]); | |
76ce9432 BP |
711 | } |
712 | ||
7d674866 BP |
713 | /* Delete controllers that are no longer configured. |
714 | * Update configuration of all now-existing controllers. */ | |
4e8e4213 | 715 | HMAP_FOR_EACH_SAFE (ofconn, next_ofconn, hmap_node, &p->controllers) { |
76ce9432 BP |
716 | struct ofproto_controller *c; |
717 | ||
718 | c = shash_find_data(&new_controllers, ofconn_get_target(ofconn)); | |
719 | if (!c) { | |
720 | ofconn_destroy(ofconn); | |
79c9f2ee | 721 | } else { |
76ce9432 | 722 | update_controller(ofconn, c); |
76ce9432 BP |
723 | } |
724 | } | |
7d674866 BP |
725 | |
726 | /* Delete services that are no longer configured. | |
727 | * Update configuration of all now-existing services. */ | |
4e8e4213 | 728 | HMAP_FOR_EACH_SAFE (ofservice, next_ofservice, node, &p->services) { |
7d674866 BP |
729 | struct ofproto_controller *c; |
730 | ||
731 | c = shash_find_data(&new_controllers, | |
732 | pvconn_get_name(ofservice->pvconn)); | |
733 | if (!c) { | |
734 | ofservice_destroy(p, ofservice); | |
735 | } else { | |
736 | ofservice_reconfigure(ofservice, c); | |
737 | } | |
738 | } | |
739 | ||
76ce9432 BP |
740 | shash_destroy(&new_controllers); |
741 | ||
d2ede7bc | 742 | update_in_band_remotes(p); |
31681a5d | 743 | update_fail_open(p); |
064af421 BP |
744 | } |
745 | ||
31681a5d JP |
746 | void |
747 | ofproto_set_fail_mode(struct ofproto *p, enum ofproto_fail_mode fail_mode) | |
748 | { | |
749 | p->fail_mode = fail_mode; | |
750 | update_fail_open(p); | |
751 | } | |
752 | ||
fa05809b BP |
753 | /* Drops the connections between 'ofproto' and all of its controllers, forcing |
754 | * them to reconnect. */ | |
755 | void | |
756 | ofproto_reconnect_controllers(struct ofproto *ofproto) | |
757 | { | |
758 | struct ofconn *ofconn; | |
759 | ||
4e8e4213 | 760 | LIST_FOR_EACH (ofconn, node, &ofproto->all_conns) { |
fa05809b BP |
761 | rconn_reconnect(ofconn->rconn); |
762 | } | |
763 | } | |
764 | ||
917e50e1 BP |
765 | static bool |
766 | any_extras_changed(const struct ofproto *ofproto, | |
767 | const struct sockaddr_in *extras, size_t n) | |
768 | { | |
769 | size_t i; | |
770 | ||
771 | if (n != ofproto->n_extra_remotes) { | |
772 | return true; | |
773 | } | |
774 | ||
775 | for (i = 0; i < n; i++) { | |
776 | const struct sockaddr_in *old = &ofproto->extra_in_band_remotes[i]; | |
777 | const struct sockaddr_in *new = &extras[i]; | |
778 | ||
779 | if (old->sin_addr.s_addr != new->sin_addr.s_addr || | |
780 | old->sin_port != new->sin_port) { | |
781 | return true; | |
782 | } | |
783 | } | |
784 | ||
785 | return false; | |
786 | } | |
787 | ||
788 | /* Sets the 'n' TCP port addresses in 'extras' as ones to which 'ofproto''s | |
789 | * in-band control should guarantee access, in the same way that in-band | |
790 | * control guarantees access to OpenFlow controllers. */ | |
791 | void | |
792 | ofproto_set_extra_in_band_remotes(struct ofproto *ofproto, | |
793 | const struct sockaddr_in *extras, size_t n) | |
794 | { | |
795 | if (!any_extras_changed(ofproto, extras, n)) { | |
796 | return; | |
797 | } | |
798 | ||
799 | free(ofproto->extra_in_band_remotes); | |
800 | ofproto->n_extra_remotes = n; | |
801 | ofproto->extra_in_band_remotes = xmemdup(extras, n * sizeof *extras); | |
802 | ||
803 | update_in_band_remotes(ofproto); | |
804 | } | |
805 | ||
b1da6250 BP |
806 | /* Sets the OpenFlow queue used by flows set up by in-band control on |
807 | * 'ofproto' to 'queue_id'. If 'queue_id' is negative, then in-band control | |
808 | * flows will use the default queue. */ | |
809 | void | |
810 | ofproto_set_in_band_queue(struct ofproto *ofproto, int queue_id) | |
811 | { | |
812 | if (queue_id != ofproto->in_band_queue) { | |
813 | ofproto->in_band_queue = queue_id; | |
814 | update_in_band_remotes(ofproto); | |
815 | } | |
816 | } | |
817 | ||
064af421 BP |
818 | void |
819 | ofproto_set_desc(struct ofproto *p, | |
5a719c38 JP |
820 | const char *mfr_desc, const char *hw_desc, |
821 | const char *sw_desc, const char *serial_desc, | |
8abc4ed7 | 822 | const char *dp_desc) |
064af421 | 823 | { |
5a719c38 JP |
824 | struct ofp_desc_stats *ods; |
825 | ||
826 | if (mfr_desc) { | |
827 | if (strlen(mfr_desc) >= sizeof ods->mfr_desc) { | |
828 | VLOG_WARN("truncating mfr_desc, must be less than %zu characters", | |
829 | sizeof ods->mfr_desc); | |
830 | } | |
831 | free(p->mfr_desc); | |
832 | p->mfr_desc = xstrdup(mfr_desc); | |
064af421 | 833 | } |
5a719c38 JP |
834 | if (hw_desc) { |
835 | if (strlen(hw_desc) >= sizeof ods->hw_desc) { | |
836 | VLOG_WARN("truncating hw_desc, must be less than %zu characters", | |
837 | sizeof ods->hw_desc); | |
838 | } | |
839 | free(p->hw_desc); | |
840 | p->hw_desc = xstrdup(hw_desc); | |
064af421 | 841 | } |
5a719c38 JP |
842 | if (sw_desc) { |
843 | if (strlen(sw_desc) >= sizeof ods->sw_desc) { | |
844 | VLOG_WARN("truncating sw_desc, must be less than %zu characters", | |
845 | sizeof ods->sw_desc); | |
846 | } | |
847 | free(p->sw_desc); | |
848 | p->sw_desc = xstrdup(sw_desc); | |
849 | } | |
850 | if (serial_desc) { | |
851 | if (strlen(serial_desc) >= sizeof ods->serial_num) { | |
852 | VLOG_WARN("truncating serial_desc, must be less than %zu " | |
853 | "characters", | |
854 | sizeof ods->serial_num); | |
855 | } | |
856 | free(p->serial_desc); | |
857 | p->serial_desc = xstrdup(serial_desc); | |
064af421 | 858 | } |
8abc4ed7 | 859 | if (dp_desc) { |
5a719c38 JP |
860 | if (strlen(dp_desc) >= sizeof ods->dp_desc) { |
861 | VLOG_WARN("truncating dp_desc, must be less than %zu characters", | |
862 | sizeof ods->dp_desc); | |
863 | } | |
8abc4ed7 JP |
864 | free(p->dp_desc); |
865 | p->dp_desc = xstrdup(dp_desc); | |
866 | } | |
064af421 BP |
867 | } |
868 | ||
064af421 BP |
869 | static int |
870 | set_pvconns(struct pvconn ***pvconnsp, size_t *n_pvconnsp, | |
871 | const struct svec *svec) | |
872 | { | |
873 | struct pvconn **pvconns = *pvconnsp; | |
874 | size_t n_pvconns = *n_pvconnsp; | |
875 | int retval = 0; | |
876 | size_t i; | |
877 | ||
878 | for (i = 0; i < n_pvconns; i++) { | |
879 | pvconn_close(pvconns[i]); | |
880 | } | |
881 | free(pvconns); | |
882 | ||
883 | pvconns = xmalloc(svec->n * sizeof *pvconns); | |
884 | n_pvconns = 0; | |
885 | for (i = 0; i < svec->n; i++) { | |
886 | const char *name = svec->names[i]; | |
887 | struct pvconn *pvconn; | |
888 | int error; | |
889 | ||
890 | error = pvconn_open(name, &pvconn); | |
891 | if (!error) { | |
892 | pvconns[n_pvconns++] = pvconn; | |
893 | } else { | |
894 | VLOG_ERR("failed to listen on %s: %s", name, strerror(error)); | |
895 | if (!retval) { | |
896 | retval = error; | |
897 | } | |
898 | } | |
899 | } | |
900 | ||
901 | *pvconnsp = pvconns; | |
902 | *n_pvconnsp = n_pvconns; | |
903 | ||
904 | return retval; | |
905 | } | |
906 | ||
064af421 BP |
907 | int |
908 | ofproto_set_snoops(struct ofproto *ofproto, const struct svec *snoops) | |
909 | { | |
910 | return set_pvconns(&ofproto->snoops, &ofproto->n_snoops, snoops); | |
911 | } | |
912 | ||
913 | int | |
0193b2af JG |
914 | ofproto_set_netflow(struct ofproto *ofproto, |
915 | const struct netflow_options *nf_options) | |
064af421 | 916 | { |
76343538 | 917 | if (nf_options && nf_options->collectors.n) { |
064af421 BP |
918 | if (!ofproto->netflow) { |
919 | ofproto->netflow = netflow_create(); | |
920 | } | |
0193b2af | 921 | return netflow_set_options(ofproto->netflow, nf_options); |
064af421 BP |
922 | } else { |
923 | netflow_destroy(ofproto->netflow); | |
924 | ofproto->netflow = NULL; | |
925 | return 0; | |
926 | } | |
927 | } | |
928 | ||
72b06300 BP |
929 | void |
930 | ofproto_set_sflow(struct ofproto *ofproto, | |
931 | const struct ofproto_sflow_options *oso) | |
932 | { | |
933 | struct ofproto_sflow *os = ofproto->sflow; | |
934 | if (oso) { | |
935 | if (!os) { | |
936 | struct ofport *ofport; | |
72b06300 BP |
937 | |
938 | os = ofproto->sflow = ofproto_sflow_create(ofproto->dpif); | |
4e8e4213 | 939 | HMAP_FOR_EACH (ofport, hmap_node, &ofproto->ports) { |
ca0f572c | 940 | ofproto_sflow_add_port(os, ofport->odp_port, |
72b06300 BP |
941 | netdev_get_name(ofport->netdev)); |
942 | } | |
943 | } | |
944 | ofproto_sflow_set_options(os, oso); | |
945 | } else { | |
946 | ofproto_sflow_destroy(os); | |
947 | ofproto->sflow = NULL; | |
948 | } | |
949 | } | |
950 | ||
064af421 BP |
951 | uint64_t |
952 | ofproto_get_datapath_id(const struct ofproto *ofproto) | |
953 | { | |
954 | return ofproto->datapath_id; | |
955 | } | |
956 | ||
76ce9432 | 957 | bool |
7d674866 | 958 | ofproto_has_primary_controller(const struct ofproto *ofproto) |
064af421 | 959 | { |
76ce9432 | 960 | return !hmap_is_empty(&ofproto->controllers); |
064af421 BP |
961 | } |
962 | ||
abdfe474 JP |
963 | enum ofproto_fail_mode |
964 | ofproto_get_fail_mode(const struct ofproto *p) | |
965 | { | |
966 | return p->fail_mode; | |
967 | } | |
968 | ||
064af421 BP |
969 | void |
970 | ofproto_get_snoops(const struct ofproto *ofproto, struct svec *snoops) | |
971 | { | |
972 | size_t i; | |
973 | ||
974 | for (i = 0; i < ofproto->n_snoops; i++) { | |
975 | svec_add(snoops, pvconn_get_name(ofproto->snoops[i])); | |
976 | } | |
977 | } | |
978 | ||
979 | void | |
980 | ofproto_destroy(struct ofproto *p) | |
981 | { | |
7d674866 | 982 | struct ofservice *ofservice, *next_ofservice; |
064af421 | 983 | struct ofconn *ofconn, *next_ofconn; |
ca0f572c | 984 | struct ofport *ofport, *next_ofport; |
064af421 BP |
985 | size_t i; |
986 | ||
987 | if (!p) { | |
988 | return; | |
989 | } | |
990 | ||
7aa697dd BP |
991 | shash_find_and_delete(&all_ofprotos, dpif_name(p->dpif)); |
992 | ||
f7de2cdf | 993 | /* Destroy fail-open and in-band early, since they touch the classifier. */ |
79c9f2ee BP |
994 | fail_open_destroy(p->fail_open); |
995 | p->fail_open = NULL; | |
996 | ||
997 | in_band_destroy(p->in_band); | |
998 | p->in_band = NULL; | |
917e50e1 | 999 | free(p->extra_in_band_remotes); |
2f6d3445 | 1000 | |
064af421 BP |
1001 | ofproto_flush_flows(p); |
1002 | classifier_destroy(&p->cls); | |
bcf84111 | 1003 | hmap_destroy(&p->facets); |
064af421 | 1004 | |
4e8e4213 | 1005 | LIST_FOR_EACH_SAFE (ofconn, next_ofconn, node, &p->all_conns) { |
c475ae67 | 1006 | ofconn_destroy(ofconn); |
064af421 | 1007 | } |
76ce9432 | 1008 | hmap_destroy(&p->controllers); |
064af421 | 1009 | |
c228a364 | 1010 | dpif_close(p->dpif); |
e9e28be3 | 1011 | netdev_monitor_destroy(p->netdev_monitor); |
4e8e4213 | 1012 | HMAP_FOR_EACH_SAFE (ofport, next_ofport, hmap_node, &p->ports) { |
ca0f572c | 1013 | hmap_remove(&p->ports, &ofport->hmap_node); |
064af421 BP |
1014 | ofport_free(ofport); |
1015 | } | |
1016 | shash_destroy(&p->port_by_name); | |
1017 | ||
064af421 | 1018 | netflow_destroy(p->netflow); |
72b06300 | 1019 | ofproto_sflow_destroy(p->sflow); |
064af421 | 1020 | |
4e8e4213 | 1021 | HMAP_FOR_EACH_SAFE (ofservice, next_ofservice, node, &p->services) { |
7d674866 | 1022 | ofservice_destroy(p, ofservice); |
064af421 | 1023 | } |
7d674866 | 1024 | hmap_destroy(&p->services); |
064af421 BP |
1025 | |
1026 | for (i = 0; i < p->n_snoops; i++) { | |
1027 | pvconn_close(p->snoops[i]); | |
1028 | } | |
1029 | free(p->snoops); | |
1030 | ||
1031 | mac_learning_destroy(p->ml); | |
1032 | ||
5a719c38 JP |
1033 | free(p->mfr_desc); |
1034 | free(p->hw_desc); | |
1035 | free(p->sw_desc); | |
1036 | free(p->serial_desc); | |
cb871ae0 JP |
1037 | free(p->dp_desc); |
1038 | ||
ca0f572c | 1039 | hmap_destroy(&p->ports); |
3b917492 | 1040 | |
064af421 BP |
1041 | free(p); |
1042 | } | |
1043 | ||
1044 | int | |
1045 | ofproto_run(struct ofproto *p) | |
1046 | { | |
1047 | int error = ofproto_run1(p); | |
1048 | if (!error) { | |
1049 | error = ofproto_run2(p, false); | |
1050 | } | |
1051 | return error; | |
1052 | } | |
1053 | ||
e9e28be3 BP |
1054 | static void |
1055 | process_port_change(struct ofproto *ofproto, int error, char *devname) | |
1056 | { | |
1057 | if (error == ENOBUFS) { | |
1058 | reinit_ports(ofproto); | |
1059 | } else if (!error) { | |
1060 | update_port(ofproto, devname); | |
1061 | free(devname); | |
1062 | } | |
1063 | } | |
1064 | ||
e2bfacb6 BP |
1065 | /* Returns a "preference level" for snooping 'ofconn'. A higher return value |
1066 | * means that 'ofconn' is more interesting for monitoring than a lower return | |
1067 | * value. */ | |
1068 | static int | |
1069 | snoop_preference(const struct ofconn *ofconn) | |
1070 | { | |
1071 | switch (ofconn->role) { | |
1072 | case NX_ROLE_MASTER: | |
1073 | return 3; | |
1074 | case NX_ROLE_OTHER: | |
1075 | return 2; | |
1076 | case NX_ROLE_SLAVE: | |
1077 | return 1; | |
1078 | default: | |
1079 | /* Shouldn't happen. */ | |
1080 | return 0; | |
1081 | } | |
1082 | } | |
1083 | ||
76ce9432 BP |
1084 | /* One of ofproto's "snoop" pvconns has accepted a new connection on 'vconn'. |
1085 | * Connects this vconn to a controller. */ | |
1086 | static void | |
1087 | add_snooper(struct ofproto *ofproto, struct vconn *vconn) | |
1088 | { | |
e2bfacb6 | 1089 | struct ofconn *ofconn, *best; |
76ce9432 | 1090 | |
e2bfacb6 BP |
1091 | /* Pick a controller for monitoring. */ |
1092 | best = NULL; | |
4e8e4213 | 1093 | LIST_FOR_EACH (ofconn, node, &ofproto->all_conns) { |
5899143f | 1094 | if (ofconn->type == OFCONN_PRIMARY |
e2bfacb6 BP |
1095 | && (!best || snoop_preference(ofconn) > snoop_preference(best))) { |
1096 | best = ofconn; | |
76ce9432 | 1097 | } |
e2bfacb6 | 1098 | } |
76ce9432 | 1099 | |
e2bfacb6 BP |
1100 | if (best) { |
1101 | rconn_add_monitor(best->rconn, vconn); | |
1102 | } else { | |
1103 | VLOG_INFO_RL(&rl, "no controller connection to snoop"); | |
1104 | vconn_close(vconn); | |
76ce9432 | 1105 | } |
76ce9432 BP |
1106 | } |
1107 | ||
064af421 BP |
1108 | int |
1109 | ofproto_run1(struct ofproto *p) | |
1110 | { | |
1111 | struct ofconn *ofconn, *next_ofconn; | |
7d674866 | 1112 | struct ofservice *ofservice; |
064af421 BP |
1113 | char *devname; |
1114 | int error; | |
1115 | int i; | |
1116 | ||
149f577a JG |
1117 | if (shash_is_empty(&p->port_by_name)) { |
1118 | init_ports(p); | |
1119 | } | |
1120 | ||
064af421 | 1121 | for (i = 0; i < 50; i++) { |
856081f6 | 1122 | struct dpif_upcall packet; |
064af421 | 1123 | |
856081f6 | 1124 | error = dpif_recv(p->dpif, &packet); |
064af421 BP |
1125 | if (error) { |
1126 | if (error == ENODEV) { | |
1127 | /* Someone destroyed the datapath behind our back. The caller | |
1128 | * better destroy us and give up, because we're just going to | |
1129 | * spin from here on out. */ | |
39a559f2 BP |
1130 | static struct vlog_rate_limit rl2 = VLOG_RATE_LIMIT_INIT(1, 5); |
1131 | VLOG_ERR_RL(&rl2, "%s: datapath was destroyed externally", | |
c228a364 | 1132 | dpif_name(p->dpif)); |
064af421 BP |
1133 | return ENODEV; |
1134 | } | |
1135 | break; | |
1136 | } | |
1137 | ||
856081f6 | 1138 | handle_upcall(p, &packet); |
064af421 BP |
1139 | } |
1140 | ||
e9e28be3 BP |
1141 | while ((error = dpif_port_poll(p->dpif, &devname)) != EAGAIN) { |
1142 | process_port_change(p, error, devname); | |
1143 | } | |
1144 | while ((error = netdev_monitor_poll(p->netdev_monitor, | |
1145 | &devname)) != EAGAIN) { | |
1146 | process_port_change(p, error, devname); | |
064af421 BP |
1147 | } |
1148 | ||
1149 | if (p->in_band) { | |
d2ede7bc BP |
1150 | if (time_msec() >= p->next_in_band_update) { |
1151 | update_in_band_remotes(p); | |
1152 | } | |
064af421 BP |
1153 | in_band_run(p->in_band); |
1154 | } | |
064af421 | 1155 | |
4e8e4213 | 1156 | LIST_FOR_EACH_SAFE (ofconn, next_ofconn, node, &p->all_conns) { |
3269c562 | 1157 | ofconn_run(ofconn); |
064af421 BP |
1158 | } |
1159 | ||
7778bd15 BP |
1160 | /* Fail-open maintenance. Do this after processing the ofconns since |
1161 | * fail-open checks the status of the controller rconn. */ | |
1162 | if (p->fail_open) { | |
1163 | fail_open_run(p->fail_open); | |
1164 | } | |
1165 | ||
4e8e4213 | 1166 | HMAP_FOR_EACH (ofservice, node, &p->services) { |
064af421 BP |
1167 | struct vconn *vconn; |
1168 | int retval; | |
1169 | ||
7d674866 | 1170 | retval = pvconn_accept(ofservice->pvconn, OFP_VERSION, &vconn); |
064af421 | 1171 | if (!retval) { |
9794e806 | 1172 | struct rconn *rconn; |
eb15cdbb | 1173 | char *name; |
9794e806 | 1174 | |
7d674866 | 1175 | rconn = rconn_create(ofservice->probe_interval, 0); |
eb15cdbb BP |
1176 | name = ofconn_make_name(p, vconn_get_name(vconn)); |
1177 | rconn_connect_unreliably(rconn, vconn, name); | |
1178 | free(name); | |
1179 | ||
7d674866 BP |
1180 | ofconn = ofconn_create(p, rconn, OFCONN_SERVICE); |
1181 | ofconn_set_rate_limit(ofconn, ofservice->rate_limit, | |
1182 | ofservice->burst_limit); | |
064af421 BP |
1183 | } else if (retval != EAGAIN) { |
1184 | VLOG_WARN_RL(&rl, "accept failed (%s)", strerror(retval)); | |
1185 | } | |
1186 | } | |
1187 | ||
1188 | for (i = 0; i < p->n_snoops; i++) { | |
1189 | struct vconn *vconn; | |
1190 | int retval; | |
1191 | ||
1192 | retval = pvconn_accept(p->snoops[i], OFP_VERSION, &vconn); | |
1193 | if (!retval) { | |
76ce9432 | 1194 | add_snooper(p, vconn); |
064af421 BP |
1195 | } else if (retval != EAGAIN) { |
1196 | VLOG_WARN_RL(&rl, "accept failed (%s)", strerror(retval)); | |
1197 | } | |
1198 | } | |
1199 | ||
1200 | if (time_msec() >= p->next_expiration) { | |
0de7a4b4 BP |
1201 | int delay = ofproto_expire(p); |
1202 | p->next_expiration = time_msec() + delay; | |
064af421 | 1203 | COVERAGE_INC(ofproto_expiration); |
064af421 BP |
1204 | } |
1205 | ||
1206 | if (p->netflow) { | |
1207 | netflow_run(p->netflow); | |
1208 | } | |
72b06300 BP |
1209 | if (p->sflow) { |
1210 | ofproto_sflow_run(p->sflow); | |
1211 | } | |
064af421 BP |
1212 | |
1213 | return 0; | |
1214 | } | |
1215 | ||
064af421 BP |
1216 | int |
1217 | ofproto_run2(struct ofproto *p, bool revalidate_all) | |
1218 | { | |
bcf84111 BP |
1219 | /* Figure out what we need to revalidate now, if anything. */ |
1220 | struct tag_set revalidate_set = p->revalidate_set; | |
1221 | if (p->need_revalidate) { | |
1222 | revalidate_all = true; | |
1223 | } | |
1224 | ||
1225 | /* Clear the revalidation flags. */ | |
1226 | tag_set_init(&p->revalidate_set); | |
1227 | p->need_revalidate = false; | |
1228 | ||
1229 | /* Now revalidate if there's anything to do. */ | |
1230 | if (revalidate_all || !tag_set_is_empty(&revalidate_set)) { | |
1231 | struct facet *facet, *next; | |
1232 | ||
1233 | HMAP_FOR_EACH_SAFE (facet, next, hmap_node, &p->facets) { | |
1234 | if (revalidate_all | |
1235 | || tag_set_intersects(&revalidate_set, facet->tags)) { | |
1236 | facet_revalidate(p, facet); | |
1237 | } | |
1238 | } | |
064af421 BP |
1239 | } |
1240 | ||
1241 | return 0; | |
1242 | } | |
1243 | ||
1244 | void | |
1245 | ofproto_wait(struct ofproto *p) | |
1246 | { | |
7d674866 | 1247 | struct ofservice *ofservice; |
064af421 BP |
1248 | struct ofconn *ofconn; |
1249 | size_t i; | |
1250 | ||
c228a364 | 1251 | dpif_recv_wait(p->dpif); |
e9e28be3 BP |
1252 | dpif_port_poll_wait(p->dpif); |
1253 | netdev_monitor_poll_wait(p->netdev_monitor); | |
4e8e4213 | 1254 | LIST_FOR_EACH (ofconn, node, &p->all_conns) { |
064af421 BP |
1255 | ofconn_wait(ofconn); |
1256 | } | |
1257 | if (p->in_band) { | |
7cf8b266 | 1258 | poll_timer_wait_until(p->next_in_band_update); |
064af421 BP |
1259 | in_band_wait(p->in_band); |
1260 | } | |
064af421 BP |
1261 | if (p->fail_open) { |
1262 | fail_open_wait(p->fail_open); | |
1263 | } | |
72b06300 BP |
1264 | if (p->sflow) { |
1265 | ofproto_sflow_wait(p->sflow); | |
1266 | } | |
064af421 BP |
1267 | if (!tag_set_is_empty(&p->revalidate_set)) { |
1268 | poll_immediate_wake(); | |
1269 | } | |
1270 | if (p->need_revalidate) { | |
1271 | /* Shouldn't happen, but if it does just go around again. */ | |
1272 | VLOG_DBG_RL(&rl, "need revalidate in ofproto_wait_cb()"); | |
1273 | poll_immediate_wake(); | |
1274 | } else if (p->next_expiration != LLONG_MAX) { | |
7cf8b266 | 1275 | poll_timer_wait_until(p->next_expiration); |
064af421 | 1276 | } |
4e8e4213 | 1277 | HMAP_FOR_EACH (ofservice, node, &p->services) { |
7d674866 | 1278 | pvconn_wait(ofservice->pvconn); |
064af421 BP |
1279 | } |
1280 | for (i = 0; i < p->n_snoops; i++) { | |
1281 | pvconn_wait(p->snoops[i]); | |
1282 | } | |
1283 | } | |
1284 | ||
1285 | void | |
1286 | ofproto_revalidate(struct ofproto *ofproto, tag_type tag) | |
1287 | { | |
1288 | tag_set_add(&ofproto->revalidate_set, tag); | |
1289 | } | |
1290 | ||
1291 | struct tag_set * | |
1292 | ofproto_get_revalidate_set(struct ofproto *ofproto) | |
1293 | { | |
1294 | return &ofproto->revalidate_set; | |
1295 | } | |
1296 | ||
1297 | bool | |
1298 | ofproto_is_alive(const struct ofproto *p) | |
1299 | { | |
76ce9432 | 1300 | return !hmap_is_empty(&p->controllers); |
064af421 BP |
1301 | } |
1302 | ||
bffc0589 | 1303 | void |
2cdcb898 | 1304 | ofproto_get_ofproto_controller_info(const struct ofproto *ofproto, |
bffc0589 AE |
1305 | struct shash *info) |
1306 | { | |
1307 | const struct ofconn *ofconn; | |
1308 | ||
1309 | shash_init(info); | |
1310 | ||
1311 | HMAP_FOR_EACH (ofconn, hmap_node, &ofproto->controllers) { | |
1312 | const struct rconn *rconn = ofconn->rconn; | |
2cdcb898 AE |
1313 | time_t now = time_now(); |
1314 | time_t last_connection = rconn_get_last_connection(rconn); | |
1315 | time_t last_disconnect = rconn_get_last_disconnect(rconn); | |
bffc0589 AE |
1316 | const int last_error = rconn_get_last_error(rconn); |
1317 | struct ofproto_controller_info *cinfo = xmalloc(sizeof *cinfo); | |
1318 | ||
1319 | shash_add(info, rconn_get_target(rconn), cinfo); | |
1320 | ||
1321 | cinfo->is_connected = rconn_is_connected(rconn); | |
1322 | cinfo->role = ofconn->role; | |
1323 | ||
1324 | cinfo->pairs.n = 0; | |
1325 | ||
c18ea70d | 1326 | if (last_error) { |
bffc0589 AE |
1327 | cinfo->pairs.keys[cinfo->pairs.n] = "last_error"; |
1328 | cinfo->pairs.values[cinfo->pairs.n++] = | |
c18ea70d | 1329 | xstrdup(ovs_retval_to_string(last_error)); |
bffc0589 AE |
1330 | } |
1331 | ||
1332 | cinfo->pairs.keys[cinfo->pairs.n] = "state"; | |
1333 | cinfo->pairs.values[cinfo->pairs.n++] = | |
1334 | xstrdup(rconn_get_state(rconn)); | |
1335 | ||
2cdcb898 AE |
1336 | if (last_connection != TIME_MIN) { |
1337 | cinfo->pairs.keys[cinfo->pairs.n] = "sec_since_connect"; | |
1338 | cinfo->pairs.values[cinfo->pairs.n++] | |
1339 | = xasprintf("%ld", (long int) (now - last_connection)); | |
1340 | } | |
1341 | ||
1342 | if (last_disconnect != TIME_MIN) { | |
1343 | cinfo->pairs.keys[cinfo->pairs.n] = "sec_since_disconnect"; | |
1344 | cinfo->pairs.values[cinfo->pairs.n++] | |
1345 | = xasprintf("%ld", (long int) (now - last_disconnect)); | |
24190e86 | 1346 | } |
bffc0589 AE |
1347 | } |
1348 | } | |
1349 | ||
1350 | void | |
1351 | ofproto_free_ofproto_controller_info(struct shash *info) | |
1352 | { | |
1353 | struct shash_node *node; | |
1354 | ||
1355 | SHASH_FOR_EACH (node, info) { | |
1356 | struct ofproto_controller_info *cinfo = node->data; | |
1357 | while (cinfo->pairs.n) { | |
1358 | free((char *) cinfo->pairs.values[--cinfo->pairs.n]); | |
1359 | } | |
1360 | free(cinfo); | |
1361 | } | |
1362 | shash_destroy(info); | |
1363 | } | |
1364 | ||
3a6ccc8c BP |
1365 | /* Deletes port number 'odp_port' from the datapath for 'ofproto'. |
1366 | * | |
1367 | * This is almost the same as calling dpif_port_del() directly on the | |
1368 | * datapath, but it also makes 'ofproto' close its open netdev for the port | |
1369 | * (if any). This makes it possible to create a new netdev of a different | |
1370 | * type under the same name, which otherwise the netdev library would refuse | |
1371 | * to do because of the conflict. (The netdev would eventually get closed on | |
1372 | * the next trip through ofproto_run(), but this interface is more direct.) | |
1373 | * | |
3a6ccc8c BP |
1374 | * Returns 0 if successful, otherwise a positive errno. */ |
1375 | int | |
1376 | ofproto_port_del(struct ofproto *ofproto, uint16_t odp_port) | |
1377 | { | |
1378 | struct ofport *ofport = get_port(ofproto, odp_port); | |
0b61210e | 1379 | const char *name = ofport ? ofport->opp.name : "<unknown>"; |
3a6ccc8c BP |
1380 | int error; |
1381 | ||
1382 | error = dpif_port_del(ofproto->dpif, odp_port); | |
1383 | if (error) { | |
1384 | VLOG_ERR("%s: failed to remove port %"PRIu16" (%s) interface (%s)", | |
1385 | dpif_name(ofproto->dpif), odp_port, name, strerror(error)); | |
1386 | } else if (ofport) { | |
1387 | /* 'name' is ofport->opp.name and update_port() is going to destroy | |
1388 | * 'ofport'. Just in case update_port() refers to 'name' after it | |
1389 | * destroys 'ofport', make a copy of it around the update_port() | |
1390 | * call. */ | |
1391 | char *devname = xstrdup(name); | |
1392 | update_port(ofproto, devname); | |
1393 | free(devname); | |
1394 | } | |
1395 | return error; | |
1396 | } | |
1397 | ||
a4e2e1f2 EJ |
1398 | /* Checks if 'ofproto' thinks 'odp_port' should be included in floods. Returns |
1399 | * true if 'odp_port' exists and should be included, false otherwise. */ | |
1400 | bool | |
1401 | ofproto_port_is_floodable(struct ofproto *ofproto, uint16_t odp_port) | |
1402 | { | |
1403 | struct ofport *ofport = get_port(ofproto, odp_port); | |
1404 | return ofport && !(ofport->opp.config & OFPPC_NO_FLOOD); | |
1405 | } | |
1406 | ||
064af421 | 1407 | int |
ae412e7d | 1408 | ofproto_send_packet(struct ofproto *p, const struct flow *flow, |
064af421 BP |
1409 | const union ofp_action *actions, size_t n_actions, |
1410 | const struct ofpbuf *packet) | |
1411 | { | |
f29152ca | 1412 | struct action_xlate_ctx ctx; |
cdee00fd | 1413 | struct ofpbuf *odp_actions; |
064af421 | 1414 | |
f29152ca | 1415 | action_xlate_ctx_init(&ctx, p, flow, packet); |
ebe482fd EJ |
1416 | /* Always xlate packets originated in this function. */ |
1417 | ctx.check_special = false; | |
cdee00fd | 1418 | odp_actions = xlate_actions(&ctx, actions, n_actions); |
064af421 BP |
1419 | |
1420 | /* XXX Should we translate the dpif_execute() errno value into an OpenFlow | |
1421 | * error code? */ | |
cdee00fd BP |
1422 | dpif_execute(p->dpif, odp_actions->data, odp_actions->size, packet); |
1423 | ||
1424 | ofpbuf_delete(odp_actions); | |
1425 | ||
064af421 BP |
1426 | return 0; |
1427 | } | |
1428 | ||
fa8b054f BP |
1429 | /* Adds a flow to the OpenFlow flow table in 'p' that matches 'cls_rule' and |
1430 | * performs the 'n_actions' actions in 'actions'. The new flow will not | |
1431 | * timeout. | |
1432 | * | |
1433 | * If cls_rule->priority is in the range of priorities supported by OpenFlow | |
1434 | * (0...65535, inclusive) then the flow will be visible to OpenFlow | |
1435 | * controllers; otherwise, it will be hidden. | |
1436 | * | |
1437 | * The caller retains ownership of 'cls_rule' and 'actions'. */ | |
064af421 | 1438 | void |
cf3fad8a | 1439 | ofproto_add_flow(struct ofproto *p, const struct cls_rule *cls_rule, |
fa8b054f | 1440 | const union ofp_action *actions, size_t n_actions) |
064af421 BP |
1441 | { |
1442 | struct rule *rule; | |
bcf84111 | 1443 | rule = rule_create(cls_rule, actions, n_actions, 0, 0, 0, false); |
afe75089 | 1444 | rule_insert(p, rule); |
064af421 BP |
1445 | } |
1446 | ||
1447 | void | |
cf3fad8a | 1448 | ofproto_delete_flow(struct ofproto *ofproto, const struct cls_rule *target) |
064af421 BP |
1449 | { |
1450 | struct rule *rule; | |
1451 | ||
1452 | rule = rule_from_cls_rule(classifier_find_rule_exactly(&ofproto->cls, | |
cf3fad8a | 1453 | target)); |
064af421 BP |
1454 | if (rule) { |
1455 | rule_remove(ofproto, rule); | |
1456 | } | |
1457 | } | |
1458 | ||
064af421 BP |
1459 | void |
1460 | ofproto_flush_flows(struct ofproto *ofproto) | |
1461 | { | |
bcf84111 | 1462 | struct facet *facet, *next_facet; |
5ecc9d81 BP |
1463 | struct rule *rule, *next_rule; |
1464 | struct cls_cursor cursor; | |
bcf84111 | 1465 | |
064af421 | 1466 | COVERAGE_INC(ofproto_flush); |
bcf84111 BP |
1467 | |
1468 | HMAP_FOR_EACH_SAFE (facet, next_facet, hmap_node, &ofproto->facets) { | |
1469 | /* Mark the facet as not installed so that facet_remove() doesn't | |
1470 | * bother trying to uninstall it. There is no point in uninstalling it | |
1471 | * individually since we are about to blow away all the facets with | |
1472 | * dpif_flow_flush(). */ | |
1473 | facet->installed = false; | |
a2c6a63c EJ |
1474 | facet->dp_packet_count = 0; |
1475 | facet->dp_byte_count = 0; | |
bcf84111 BP |
1476 | facet_remove(ofproto, facet); |
1477 | } | |
5ecc9d81 BP |
1478 | |
1479 | cls_cursor_init(&cursor, &ofproto->cls, NULL); | |
1480 | CLS_CURSOR_FOR_EACH_SAFE (rule, next_rule, cr, &cursor) { | |
1481 | rule_remove(ofproto, rule); | |
1482 | } | |
1483 | ||
c228a364 | 1484 | dpif_flow_flush(ofproto->dpif); |
064af421 BP |
1485 | if (ofproto->in_band) { |
1486 | in_band_flushed(ofproto->in_band); | |
1487 | } | |
1488 | if (ofproto->fail_open) { | |
1489 | fail_open_flushed(ofproto->fail_open); | |
1490 | } | |
1491 | } | |
1492 | \f | |
1493 | static void | |
1494 | reinit_ports(struct ofproto *p) | |
1495 | { | |
b0ec0f27 | 1496 | struct dpif_port_dump dump; |
c77d9d13 BP |
1497 | struct shash_node *node; |
1498 | struct shash devnames; | |
064af421 | 1499 | struct ofport *ofport; |
4c738a8d | 1500 | struct dpif_port dpif_port; |
064af421 | 1501 | |
898bf89d JP |
1502 | COVERAGE_INC(ofproto_reinit_ports); |
1503 | ||
c77d9d13 | 1504 | shash_init(&devnames); |
4e8e4213 | 1505 | HMAP_FOR_EACH (ofport, hmap_node, &p->ports) { |
c77d9d13 | 1506 | shash_add_once (&devnames, ofport->opp.name, NULL); |
064af421 | 1507 | } |
4c738a8d BP |
1508 | DPIF_PORT_FOR_EACH (&dpif_port, &dump, p->dpif) { |
1509 | shash_add_once (&devnames, dpif_port.name, NULL); | |
064af421 | 1510 | } |
064af421 | 1511 | |
c77d9d13 BP |
1512 | SHASH_FOR_EACH (node, &devnames) { |
1513 | update_port(p, node->name); | |
064af421 | 1514 | } |
c77d9d13 | 1515 | shash_destroy(&devnames); |
064af421 BP |
1516 | } |
1517 | ||
064af421 | 1518 | static struct ofport * |
4c738a8d | 1519 | make_ofport(const struct dpif_port *dpif_port) |
064af421 | 1520 | { |
149f577a | 1521 | struct netdev_options netdev_options; |
064af421 BP |
1522 | enum netdev_flags flags; |
1523 | struct ofport *ofport; | |
1524 | struct netdev *netdev; | |
064af421 BP |
1525 | int error; |
1526 | ||
149f577a | 1527 | memset(&netdev_options, 0, sizeof netdev_options); |
4c738a8d BP |
1528 | netdev_options.name = dpif_port->name; |
1529 | netdev_options.type = dpif_port->type; | |
149f577a | 1530 | netdev_options.ethertype = NETDEV_ETH_TYPE_NONE; |
149f577a JG |
1531 | |
1532 | error = netdev_open(&netdev_options, &netdev); | |
064af421 BP |
1533 | if (error) { |
1534 | VLOG_WARN_RL(&rl, "ignoring port %s (%"PRIu16") because netdev %s " | |
1535 | "cannot be opened (%s)", | |
4c738a8d BP |
1536 | dpif_port->name, dpif_port->port_no, |
1537 | dpif_port->name, strerror(error)); | |
064af421 BP |
1538 | return NULL; |
1539 | } | |
1540 | ||
0a6f5542 | 1541 | ofport = xzalloc(sizeof *ofport); |
064af421 | 1542 | ofport->netdev = netdev; |
4c738a8d BP |
1543 | ofport->odp_port = dpif_port->port_no; |
1544 | ofport->opp.port_no = odp_port_to_ofp_port(dpif_port->port_no); | |
80992a35 | 1545 | netdev_get_etheraddr(netdev, ofport->opp.hw_addr); |
4c738a8d | 1546 | ovs_strlcpy(ofport->opp.name, dpif_port->name, sizeof ofport->opp.name); |
064af421 BP |
1547 | |
1548 | netdev_get_flags(netdev, &flags); | |
1549 | ofport->opp.config = flags & NETDEV_UP ? 0 : OFPPC_PORT_DOWN; | |
1550 | ||
85da620e | 1551 | ofport->opp.state = netdev_get_carrier(netdev) ? 0 : OFPPS_LINK_DOWN; |
064af421 BP |
1552 | |
1553 | netdev_get_features(netdev, | |
1554 | &ofport->opp.curr, &ofport->opp.advertised, | |
1555 | &ofport->opp.supported, &ofport->opp.peer); | |
1556 | return ofport; | |
1557 | } | |
1558 | ||
1559 | static bool | |
4c738a8d | 1560 | ofport_conflicts(const struct ofproto *p, const struct dpif_port *dpif_port) |
064af421 | 1561 | { |
4c738a8d | 1562 | if (get_port(p, dpif_port->port_no)) { |
064af421 | 1563 | VLOG_WARN_RL(&rl, "ignoring duplicate port %"PRIu16" in datapath", |
4c738a8d | 1564 | dpif_port->port_no); |
064af421 | 1565 | return true; |
4c738a8d | 1566 | } else if (shash_find(&p->port_by_name, dpif_port->name)) { |
064af421 | 1567 | VLOG_WARN_RL(&rl, "ignoring duplicate device %s in datapath", |
4c738a8d | 1568 | dpif_port->name); |
064af421 BP |
1569 | return true; |
1570 | } else { | |
1571 | return false; | |
1572 | } | |
1573 | } | |
1574 | ||
1575 | static int | |
1576 | ofport_equal(const struct ofport *a_, const struct ofport *b_) | |
1577 | { | |
1578 | const struct ofp_phy_port *a = &a_->opp; | |
1579 | const struct ofp_phy_port *b = &b_->opp; | |
1580 | ||
1581 | BUILD_ASSERT_DECL(sizeof *a == 48); /* Detect ofp_phy_port changes. */ | |
1582 | return (a->port_no == b->port_no | |
1583 | && !memcmp(a->hw_addr, b->hw_addr, sizeof a->hw_addr) | |
0b61210e | 1584 | && !strcmp(a->name, b->name) |
064af421 BP |
1585 | && a->state == b->state |
1586 | && a->config == b->config | |
1587 | && a->curr == b->curr | |
1588 | && a->advertised == b->advertised | |
1589 | && a->supported == b->supported | |
1590 | && a->peer == b->peer); | |
1591 | } | |
1592 | ||
1593 | static void | |
1594 | send_port_status(struct ofproto *p, const struct ofport *ofport, | |
1595 | uint8_t reason) | |
1596 | { | |
1597 | /* XXX Should limit the number of queued port status change messages. */ | |
1598 | struct ofconn *ofconn; | |
4e8e4213 | 1599 | LIST_FOR_EACH (ofconn, node, &p->all_conns) { |
064af421 BP |
1600 | struct ofp_port_status *ops; |
1601 | struct ofpbuf *b; | |
1602 | ||
197a992f BP |
1603 | /* Primary controllers, even slaves, should always get port status |
1604 | updates. Otherwise obey ofconn_receives_async_msgs(). */ | |
1605 | if (ofconn->type != OFCONN_PRIMARY | |
1606 | && !ofconn_receives_async_msgs(ofconn)) { | |
9deba63b BP |
1607 | continue; |
1608 | } | |
1609 | ||
064af421 BP |
1610 | ops = make_openflow_xid(sizeof *ops, OFPT_PORT_STATUS, 0, &b); |
1611 | ops->reason = reason; | |
1612 | ops->desc = ofport->opp; | |
1613 | hton_ofp_phy_port(&ops->desc); | |
1614 | queue_tx(b, ofconn, NULL); | |
1615 | } | |
064af421 BP |
1616 | } |
1617 | ||
1618 | static void | |
1619 | ofport_install(struct ofproto *p, struct ofport *ofport) | |
1620 | { | |
0b61210e | 1621 | const char *netdev_name = ofport->opp.name; |
72b06300 | 1622 | |
e9e28be3 | 1623 | netdev_monitor_add(p->netdev_monitor, ofport->netdev); |
ca0f572c | 1624 | hmap_insert(&p->ports, &ofport->hmap_node, hash_int(ofport->odp_port, 0)); |
72b06300 BP |
1625 | shash_add(&p->port_by_name, netdev_name, ofport); |
1626 | if (p->sflow) { | |
ca0f572c | 1627 | ofproto_sflow_add_port(p->sflow, ofport->odp_port, netdev_name); |
72b06300 | 1628 | } |
064af421 BP |
1629 | } |
1630 | ||
1631 | static void | |
1632 | ofport_remove(struct ofproto *p, struct ofport *ofport) | |
1633 | { | |
e9e28be3 | 1634 | netdev_monitor_remove(p->netdev_monitor, ofport->netdev); |
ca0f572c | 1635 | hmap_remove(&p->ports, &ofport->hmap_node); |
064af421 | 1636 | shash_delete(&p->port_by_name, |
0b61210e | 1637 | shash_find(&p->port_by_name, ofport->opp.name)); |
72b06300 | 1638 | if (p->sflow) { |
ca0f572c | 1639 | ofproto_sflow_del_port(p->sflow, ofport->odp_port); |
72b06300 | 1640 | } |
064af421 BP |
1641 | } |
1642 | ||
1643 | static void | |
1644 | ofport_free(struct ofport *ofport) | |
1645 | { | |
1646 | if (ofport) { | |
1647 | netdev_close(ofport->netdev); | |
1648 | free(ofport); | |
1649 | } | |
1650 | } | |
1651 | ||
ca0f572c BP |
1652 | static struct ofport * |
1653 | get_port(const struct ofproto *ofproto, uint16_t odp_port) | |
1654 | { | |
1655 | struct ofport *port; | |
1656 | ||
4e8e4213 | 1657 | HMAP_FOR_EACH_IN_BUCKET (port, hmap_node, |
ca0f572c BP |
1658 | hash_int(odp_port, 0), &ofproto->ports) { |
1659 | if (port->odp_port == odp_port) { | |
1660 | return port; | |
1661 | } | |
1662 | } | |
1663 | return NULL; | |
1664 | } | |
1665 | ||
064af421 BP |
1666 | static void |
1667 | update_port(struct ofproto *p, const char *devname) | |
1668 | { | |
4c738a8d | 1669 | struct dpif_port dpif_port; |
c874dc6d BP |
1670 | struct ofport *old_ofport; |
1671 | struct ofport *new_ofport; | |
064af421 BP |
1672 | int error; |
1673 | ||
1674 | COVERAGE_INC(ofproto_update_port); | |
c874dc6d BP |
1675 | |
1676 | /* Query the datapath for port information. */ | |
4c738a8d | 1677 | error = dpif_port_query_by_name(p->dpif, devname, &dpif_port); |
064af421 | 1678 | |
c874dc6d BP |
1679 | /* Find the old ofport. */ |
1680 | old_ofport = shash_find_data(&p->port_by_name, devname); | |
1681 | if (!error) { | |
1682 | if (!old_ofport) { | |
1683 | /* There's no port named 'devname' but there might be a port with | |
1684 | * the same port number. This could happen if a port is deleted | |
1685 | * and then a new one added in its place very quickly, or if a port | |
1686 | * is renamed. In the former case we want to send an OFPPR_DELETE | |
1687 | * and an OFPPR_ADD, and in the latter case we want to send a | |
1688 | * single OFPPR_MODIFY. We can distinguish the cases by comparing | |
1689 | * the old port's ifindex against the new port, or perhaps less | |
1690 | * reliably but more portably by comparing the old port's MAC | |
1691 | * against the new port's MAC. However, this code isn't that smart | |
1692 | * and always sends an OFPPR_MODIFY (XXX). */ | |
4c738a8d | 1693 | old_ofport = get_port(p, dpif_port.port_no); |
064af421 | 1694 | } |
c874dc6d | 1695 | } else if (error != ENOENT && error != ENODEV) { |
064af421 BP |
1696 | VLOG_WARN_RL(&rl, "dpif_port_query_by_name returned unexpected error " |
1697 | "%s", strerror(error)); | |
4c738a8d | 1698 | goto exit; |
064af421 | 1699 | } |
c874dc6d BP |
1700 | |
1701 | /* Create a new ofport. */ | |
4c738a8d | 1702 | new_ofport = !error ? make_ofport(&dpif_port) : NULL; |
c874dc6d BP |
1703 | |
1704 | /* Eliminate a few pathological cases. */ | |
1705 | if (!old_ofport && !new_ofport) { | |
4c738a8d | 1706 | goto exit; |
c874dc6d BP |
1707 | } else if (old_ofport && new_ofport) { |
1708 | /* Most of the 'config' bits are OpenFlow soft state, but | |
bc4a55cd BP |
1709 | * OFPPC_PORT_DOWN is maintained by the kernel. So transfer the |
1710 | * OpenFlow bits from old_ofport. (make_ofport() only sets | |
1711 | * OFPPC_PORT_DOWN and leaves the other bits 0.) */ | |
c874dc6d BP |
1712 | new_ofport->opp.config |= old_ofport->opp.config & ~OFPPC_PORT_DOWN; |
1713 | ||
1714 | if (ofport_equal(old_ofport, new_ofport)) { | |
1715 | /* False alarm--no change. */ | |
1716 | ofport_free(new_ofport); | |
4c738a8d | 1717 | goto exit; |
c874dc6d BP |
1718 | } |
1719 | } | |
1720 | ||
1721 | /* Now deal with the normal cases. */ | |
1722 | if (old_ofport) { | |
1723 | ofport_remove(p, old_ofport); | |
1724 | } | |
1725 | if (new_ofport) { | |
1726 | ofport_install(p, new_ofport); | |
1727 | } | |
1728 | send_port_status(p, new_ofport ? new_ofport : old_ofport, | |
1729 | (!old_ofport ? OFPPR_ADD | |
1730 | : !new_ofport ? OFPPR_DELETE | |
1731 | : OFPPR_MODIFY)); | |
1732 | ofport_free(old_ofport); | |
4c738a8d BP |
1733 | |
1734 | exit: | |
1735 | dpif_port_destroy(&dpif_port); | |
064af421 BP |
1736 | } |
1737 | ||
1738 | static int | |
1739 | init_ports(struct ofproto *p) | |
1740 | { | |
b0ec0f27 | 1741 | struct dpif_port_dump dump; |
4c738a8d | 1742 | struct dpif_port dpif_port; |
064af421 | 1743 | |
4c738a8d BP |
1744 | DPIF_PORT_FOR_EACH (&dpif_port, &dump, p->dpif) { |
1745 | if (!ofport_conflicts(p, &dpif_port)) { | |
1746 | struct ofport *ofport = make_ofport(&dpif_port); | |
064af421 BP |
1747 | if (ofport) { |
1748 | ofport_install(p, ofport); | |
1749 | } | |
1750 | } | |
1751 | } | |
b0ec0f27 | 1752 | |
064af421 BP |
1753 | return 0; |
1754 | } | |
1755 | \f | |
1756 | static struct ofconn * | |
76ce9432 | 1757 | ofconn_create(struct ofproto *p, struct rconn *rconn, enum ofconn_type type) |
064af421 | 1758 | { |
76ce9432 BP |
1759 | struct ofconn *ofconn = xzalloc(sizeof *ofconn); |
1760 | ofconn->ofproto = p; | |
064af421 BP |
1761 | list_push_back(&p->all_conns, &ofconn->node); |
1762 | ofconn->rconn = rconn; | |
76ce9432 | 1763 | ofconn->type = type; |
b70eac89 | 1764 | ofconn->flow_format = NXFF_OPENFLOW10; |
9deba63b | 1765 | ofconn->role = NX_ROLE_OTHER; |
76ce9432 | 1766 | ofconn->packet_in_counter = rconn_packet_counter_create (); |
064af421 | 1767 | ofconn->pktbuf = NULL; |
064af421 | 1768 | ofconn->miss_send_len = 0; |
064af421 BP |
1769 | ofconn->reply_counter = rconn_packet_counter_create (); |
1770 | return ofconn; | |
1771 | } | |
1772 | ||
1773 | static void | |
c475ae67 | 1774 | ofconn_destroy(struct ofconn *ofconn) |
064af421 | 1775 | { |
5899143f | 1776 | if (ofconn->type == OFCONN_PRIMARY) { |
76ce9432 BP |
1777 | hmap_remove(&ofconn->ofproto->controllers, &ofconn->hmap_node); |
1778 | } | |
76ce9432 | 1779 | |
064af421 BP |
1780 | list_remove(&ofconn->node); |
1781 | rconn_destroy(ofconn->rconn); | |
1782 | rconn_packet_counter_destroy(ofconn->packet_in_counter); | |
1783 | rconn_packet_counter_destroy(ofconn->reply_counter); | |
1784 | pktbuf_destroy(ofconn->pktbuf); | |
1785 | free(ofconn); | |
1786 | } | |
1787 | ||
1788 | static void | |
3269c562 | 1789 | ofconn_run(struct ofconn *ofconn) |
064af421 | 1790 | { |
3269c562 | 1791 | struct ofproto *p = ofconn->ofproto; |
064af421 | 1792 | int iteration; |
76ce9432 BP |
1793 | size_t i; |
1794 | ||
76ce9432 BP |
1795 | for (i = 0; i < N_SCHEDULERS; i++) { |
1796 | pinsched_run(ofconn->schedulers[i], do_send_packet_in, ofconn); | |
1797 | } | |
064af421 BP |
1798 | |
1799 | rconn_run(ofconn->rconn); | |
1800 | ||
1801 | if (rconn_packet_counter_read (ofconn->reply_counter) < OFCONN_REPLY_MAX) { | |
1802 | /* Limit the number of iterations to prevent other tasks from | |
1803 | * starving. */ | |
1804 | for (iteration = 0; iteration < 50; iteration++) { | |
1805 | struct ofpbuf *of_msg = rconn_recv(ofconn->rconn); | |
1806 | if (!of_msg) { | |
1807 | break; | |
1808 | } | |
7778bd15 BP |
1809 | if (p->fail_open) { |
1810 | fail_open_maybe_recover(p->fail_open); | |
1811 | } | |
3269c562 | 1812 | handle_openflow(ofconn, of_msg); |
064af421 BP |
1813 | ofpbuf_delete(of_msg); |
1814 | } | |
1815 | } | |
1816 | ||
195c8086 | 1817 | if (!rconn_is_alive(ofconn->rconn)) { |
c475ae67 | 1818 | ofconn_destroy(ofconn); |
064af421 BP |
1819 | } |
1820 | } | |
1821 | ||
1822 | static void | |
1823 | ofconn_wait(struct ofconn *ofconn) | |
1824 | { | |
76ce9432 BP |
1825 | int i; |
1826 | ||
76ce9432 BP |
1827 | for (i = 0; i < N_SCHEDULERS; i++) { |
1828 | pinsched_wait(ofconn->schedulers[i]); | |
1829 | } | |
064af421 BP |
1830 | rconn_run_wait(ofconn->rconn); |
1831 | if (rconn_packet_counter_read (ofconn->reply_counter) < OFCONN_REPLY_MAX) { | |
1832 | rconn_recv_wait(ofconn->rconn); | |
1833 | } else { | |
1834 | COVERAGE_INC(ofproto_ofconn_stuck); | |
1835 | } | |
1836 | } | |
c91248b3 BP |
1837 | |
1838 | /* Returns true if 'ofconn' should receive asynchronous messages. */ | |
1839 | static bool | |
1840 | ofconn_receives_async_msgs(const struct ofconn *ofconn) | |
1841 | { | |
5899143f BP |
1842 | if (ofconn->type == OFCONN_PRIMARY) { |
1843 | /* Primary controllers always get asynchronous messages unless they | |
c91248b3 BP |
1844 | * have configured themselves as "slaves". */ |
1845 | return ofconn->role != NX_ROLE_SLAVE; | |
1846 | } else { | |
5899143f BP |
1847 | /* Service connections don't get asynchronous messages unless they have |
1848 | * explicitly asked for them by setting a nonzero miss send length. */ | |
c91248b3 BP |
1849 | return ofconn->miss_send_len > 0; |
1850 | } | |
1851 | } | |
eb15cdbb BP |
1852 | |
1853 | /* Returns a human-readable name for an OpenFlow connection between 'ofproto' | |
1854 | * and 'target', suitable for use in log messages for identifying the | |
1855 | * connection. | |
1856 | * | |
1857 | * The name is dynamically allocated. The caller should free it (with free()) | |
1858 | * when it is no longer needed. */ | |
1859 | static char * | |
1860 | ofconn_make_name(const struct ofproto *ofproto, const char *target) | |
1861 | { | |
1862 | return xasprintf("%s<->%s", dpif_base_name(ofproto->dpif), target); | |
1863 | } | |
7d674866 BP |
1864 | |
1865 | static void | |
1866 | ofconn_set_rate_limit(struct ofconn *ofconn, int rate, int burst) | |
1867 | { | |
1868 | int i; | |
1869 | ||
1870 | for (i = 0; i < N_SCHEDULERS; i++) { | |
1871 | struct pinsched **s = &ofconn->schedulers[i]; | |
1872 | ||
1873 | if (rate > 0) { | |
1874 | if (!*s) { | |
9b45d7f5 | 1875 | *s = pinsched_create(rate, burst); |
7d674866 BP |
1876 | } else { |
1877 | pinsched_set_limits(*s, rate, burst); | |
1878 | } | |
1879 | } else { | |
1880 | pinsched_destroy(*s); | |
1881 | *s = NULL; | |
1882 | } | |
1883 | } | |
1884 | } | |
1885 | \f | |
1886 | static void | |
1887 | ofservice_reconfigure(struct ofservice *ofservice, | |
1888 | const struct ofproto_controller *c) | |
1889 | { | |
1890 | ofservice->probe_interval = c->probe_interval; | |
1891 | ofservice->rate_limit = c->rate_limit; | |
1892 | ofservice->burst_limit = c->burst_limit; | |
1893 | } | |
1894 | ||
1895 | /* Creates a new ofservice in 'ofproto'. Returns 0 if successful, otherwise a | |
1896 | * positive errno value. */ | |
1897 | static int | |
1898 | ofservice_create(struct ofproto *ofproto, const struct ofproto_controller *c) | |
1899 | { | |
1900 | struct ofservice *ofservice; | |
1901 | struct pvconn *pvconn; | |
1902 | int error; | |
1903 | ||
1904 | error = pvconn_open(c->target, &pvconn); | |
1905 | if (error) { | |
1906 | return error; | |
1907 | } | |
1908 | ||
1909 | ofservice = xzalloc(sizeof *ofservice); | |
1910 | hmap_insert(&ofproto->services, &ofservice->node, | |
1911 | hash_string(c->target, 0)); | |
1912 | ofservice->pvconn = pvconn; | |
1913 | ||
1914 | ofservice_reconfigure(ofservice, c); | |
1915 | ||
1916 | return 0; | |
1917 | } | |
1918 | ||
1919 | static void | |
1920 | ofservice_destroy(struct ofproto *ofproto, struct ofservice *ofservice) | |
1921 | { | |
1922 | hmap_remove(&ofproto->services, &ofservice->node); | |
1923 | pvconn_close(ofservice->pvconn); | |
1924 | free(ofservice); | |
1925 | } | |
1926 | ||
1927 | /* Finds and returns the ofservice within 'ofproto' that has the given | |
1928 | * 'target', or a null pointer if none exists. */ | |
1929 | static struct ofservice * | |
1930 | ofservice_lookup(struct ofproto *ofproto, const char *target) | |
1931 | { | |
1932 | struct ofservice *ofservice; | |
1933 | ||
4e8e4213 BP |
1934 | HMAP_FOR_EACH_WITH_HASH (ofservice, node, hash_string(target, 0), |
1935 | &ofproto->services) { | |
7d674866 BP |
1936 | if (!strcmp(pvconn_get_name(ofservice->pvconn), target)) { |
1937 | return ofservice; | |
1938 | } | |
1939 | } | |
1940 | return NULL; | |
1941 | } | |
064af421 | 1942 | \f |
bcf84111 BP |
1943 | /* Returns true if 'rule' should be hidden from the controller. |
1944 | * | |
1945 | * Rules with priority higher than UINT16_MAX are set up by ofproto itself | |
1946 | * (e.g. by in-band control) and are intentionally hidden from the | |
1947 | * controller. */ | |
1948 | static bool | |
1949 | rule_is_hidden(const struct rule *rule) | |
1950 | { | |
1951 | return rule->cr.priority > UINT16_MAX; | |
1952 | } | |
1953 | ||
1954 | /* Creates and returns a new rule initialized as specified. | |
1955 | * | |
1956 | * The caller is responsible for inserting the rule into the classifier (with | |
1957 | * rule_insert()). */ | |
064af421 | 1958 | static struct rule * |
bcf84111 | 1959 | rule_create(const struct cls_rule *cls_rule, |
064af421 | 1960 | const union ofp_action *actions, size_t n_actions, |
ca069229 | 1961 | uint16_t idle_timeout, uint16_t hard_timeout, |
8054fc48 | 1962 | ovs_be64 flow_cookie, bool send_flow_removed) |
064af421 | 1963 | { |
ec6fde61 | 1964 | struct rule *rule = xzalloc(sizeof *rule); |
bcf84111 | 1965 | rule->cr = *cls_rule; |
064af421 BP |
1966 | rule->idle_timeout = idle_timeout; |
1967 | rule->hard_timeout = hard_timeout; | |
39997502 | 1968 | rule->flow_cookie = flow_cookie; |
064af421 | 1969 | rule->used = rule->created = time_msec(); |
ca069229 | 1970 | rule->send_flow_removed = send_flow_removed; |
bcf84111 | 1971 | list_init(&rule->facets); |
3dffcf07 BP |
1972 | if (n_actions > 0) { |
1973 | rule->n_actions = n_actions; | |
1974 | rule->actions = xmemdup(actions, n_actions * sizeof *actions); | |
1975 | } | |
0193b2af | 1976 | |
064af421 BP |
1977 | return rule; |
1978 | } | |
1979 | ||
1980 | static struct rule * | |
1981 | rule_from_cls_rule(const struct cls_rule *cls_rule) | |
1982 | { | |
1983 | return cls_rule ? CONTAINER_OF(cls_rule, struct rule, cr) : NULL; | |
1984 | } | |
1985 | ||
1986 | static void | |
1987 | rule_free(struct rule *rule) | |
1988 | { | |
1989 | free(rule->actions); | |
064af421 BP |
1990 | free(rule); |
1991 | } | |
1992 | ||
bcf84111 BP |
1993 | /* Destroys 'rule' and iterates through all of its facets and revalidates them, |
1994 | * destroying any that no longer has a rule (which is probably all of them). | |
064af421 | 1995 | * |
bcf84111 | 1996 | * The caller must have already removed 'rule' from the classifier. */ |
064af421 BP |
1997 | static void |
1998 | rule_destroy(struct ofproto *ofproto, struct rule *rule) | |
1999 | { | |
bcf84111 BP |
2000 | struct facet *facet, *next_facet; |
2001 | LIST_FOR_EACH_SAFE (facet, next_facet, list_node, &rule->facets) { | |
2002 | facet_revalidate(ofproto, facet); | |
064af421 BP |
2003 | } |
2004 | rule_free(rule); | |
2005 | } | |
2006 | ||
bcf84111 BP |
2007 | /* Returns true if 'rule' has an OpenFlow OFPAT_OUTPUT or OFPAT_ENQUEUE action |
2008 | * that outputs to 'out_port' (output to OFPP_FLOOD and OFPP_ALL doesn't | |
2009 | * count). */ | |
064af421 | 2010 | static bool |
8054fc48 | 2011 | rule_has_out_port(const struct rule *rule, ovs_be16 out_port) |
064af421 BP |
2012 | { |
2013 | const union ofp_action *oa; | |
2014 | struct actions_iterator i; | |
2015 | ||
2016 | if (out_port == htons(OFPP_NONE)) { | |
2017 | return true; | |
2018 | } | |
2019 | for (oa = actions_first(&i, rule->actions, rule->n_actions); oa; | |
2020 | oa = actions_next(&i)) { | |
c1c9c9c4 | 2021 | if (action_outputs_to_port(oa, out_port)) { |
064af421 BP |
2022 | return true; |
2023 | } | |
2024 | } | |
2025 | return false; | |
2026 | } | |
2027 | ||
750638bb BP |
2028 | /* Executes, within 'ofproto', the 'n_actions' actions in 'actions' on |
2029 | * 'packet', which arrived on 'in_port'. | |
2030 | * | |
2031 | * Takes ownership of 'packet'. */ | |
9dbb9d5e | 2032 | static bool |
856081f6 | 2033 | execute_odp_actions(struct ofproto *ofproto, const struct flow *flow, |
cf22f8cb | 2034 | const struct nlattr *odp_actions, size_t actions_len, |
750638bb | 2035 | struct ofpbuf *packet) |
9dbb9d5e | 2036 | { |
b9298d3f | 2037 | if (actions_len == NLA_ALIGN(NLA_HDRLEN + sizeof(uint64_t)) |
7aec165d | 2038 | && odp_actions->nla_type == ODP_ACTION_ATTR_CONTROLLER) { |
9dbb9d5e BP |
2039 | /* As an optimization, avoid a round-trip from userspace to kernel to |
2040 | * userspace. This also avoids possibly filling up kernel packet | |
2041 | * buffers along the way. */ | |
856081f6 | 2042 | struct dpif_upcall upcall; |
9dbb9d5e | 2043 | |
82272ede | 2044 | upcall.type = DPIF_UC_ACTION; |
856081f6 BP |
2045 | upcall.packet = packet; |
2046 | upcall.key = NULL; | |
2047 | upcall.key_len = 0; | |
2048 | upcall.userdata = nl_attr_get_u64(odp_actions); | |
2049 | upcall.sample_pool = 0; | |
2050 | upcall.actions = NULL; | |
2051 | upcall.actions_len = 0; | |
9dbb9d5e | 2052 | |
856081f6 | 2053 | send_packet_in(ofproto, &upcall, flow, false); |
9dbb9d5e | 2054 | |
750638bb BP |
2055 | return true; |
2056 | } else { | |
2057 | int error; | |
9dbb9d5e | 2058 | |
cdee00fd | 2059 | error = dpif_execute(ofproto->dpif, odp_actions, actions_len, packet); |
750638bb BP |
2060 | ofpbuf_delete(packet); |
2061 | return !error; | |
2062 | } | |
9dbb9d5e BP |
2063 | } |
2064 | ||
bcf84111 BP |
2065 | /* Executes the actions indicated by 'facet' on 'packet' and credits 'facet''s |
2066 | * statistics appropriately. 'packet' must have at least sizeof(struct | |
2067 | * ofp_packet_in) bytes of headroom. | |
064af421 | 2068 | * |
bcf84111 BP |
2069 | * For correct results, 'packet' must actually be in 'facet''s flow; that is, |
2070 | * applying flow_extract() to 'packet' would yield the same flow as | |
2071 | * 'facet->flow'. | |
064af421 | 2072 | * |
bcf84111 BP |
2073 | * 'facet' must have accurately composed ODP actions; that is, it must not be |
2074 | * in need of revalidation. | |
750638bb BP |
2075 | * |
2076 | * Takes ownership of 'packet'. */ | |
064af421 | 2077 | static void |
bcf84111 BP |
2078 | facet_execute(struct ofproto *ofproto, struct facet *facet, |
2079 | struct ofpbuf *packet) | |
064af421 | 2080 | { |
c97fb132 | 2081 | struct dpif_flow_stats stats; |
bcf84111 BP |
2082 | |
2083 | assert(ofpbuf_headroom(packet) >= sizeof(struct ofp_packet_in)); | |
2084 | ||
2085 | flow_extract_stats(&facet->flow, packet, &stats); | |
9fe215f0 | 2086 | stats.used = time_msec(); |
856081f6 | 2087 | if (execute_odp_actions(ofproto, &facet->flow, |
cdee00fd | 2088 | facet->actions, facet->actions_len, packet)) { |
878ae780 | 2089 | facet_update_stats(ofproto, facet, &stats); |
bcf84111 BP |
2090 | } |
2091 | } | |
2092 | ||
2093 | /* Executes the actions indicated by 'rule' on 'packet' and credits 'rule''s | |
2094 | * statistics (or the statistics for one of its facets) appropriately. | |
2095 | * 'packet' must have at least sizeof(struct ofp_packet_in) bytes of headroom. | |
2096 | * | |
2097 | * 'packet' doesn't necessarily have to match 'rule'. 'rule' will be credited | |
2098 | * with statistics for 'packet' either way. | |
2099 | * | |
2100 | * Takes ownership of 'packet'. */ | |
2101 | static void | |
2102 | rule_execute(struct ofproto *ofproto, struct rule *rule, uint16_t in_port, | |
2103 | struct ofpbuf *packet) | |
2104 | { | |
f29152ca | 2105 | struct action_xlate_ctx ctx; |
cdee00fd | 2106 | struct ofpbuf *odp_actions; |
bcf84111 | 2107 | struct facet *facet; |
bcf84111 BP |
2108 | struct flow flow; |
2109 | size_t size; | |
064af421 | 2110 | |
750638bb BP |
2111 | assert(ofpbuf_headroom(packet) >= sizeof(struct ofp_packet_in)); |
2112 | ||
bcf84111 BP |
2113 | flow_extract(packet, 0, in_port, &flow); |
2114 | ||
2115 | /* First look for a related facet. If we find one, account it to that. */ | |
2116 | facet = facet_lookup_valid(ofproto, &flow); | |
2117 | if (facet && facet->rule == rule) { | |
2118 | facet_execute(ofproto, facet, packet); | |
2119 | return; | |
064af421 BP |
2120 | } |
2121 | ||
bcf84111 BP |
2122 | /* Otherwise, if 'rule' is in fact the correct rule for 'packet', then |
2123 | * create a new facet for it and use that. */ | |
2124 | if (rule_lookup(ofproto, &flow) == rule) { | |
2125 | facet = facet_create(ofproto, rule, &flow, packet); | |
2126 | facet_execute(ofproto, facet, packet); | |
2127 | facet_install(ofproto, facet, true); | |
2128 | return; | |
2129 | } | |
2130 | ||
2131 | /* We can't account anything to a facet. If we were to try, then that | |
2132 | * facet would have a non-matching rule, busting our invariants. */ | |
f29152ca | 2133 | action_xlate_ctx_init(&ctx, ofproto, &flow, packet); |
cdee00fd | 2134 | odp_actions = xlate_actions(&ctx, rule->actions, rule->n_actions); |
bcf84111 | 2135 | size = packet->size; |
856081f6 | 2136 | if (execute_odp_actions(ofproto, &flow, odp_actions->data, |
cdee00fd | 2137 | odp_actions->size, packet)) { |
064af421 | 2138 | rule->used = time_msec(); |
bcf84111 BP |
2139 | rule->packet_count++; |
2140 | rule->byte_count += size; | |
878ae780 | 2141 | flow_push_stats(ofproto, rule, &flow, 1, size, rule->used); |
064af421 | 2142 | } |
cdee00fd | 2143 | ofpbuf_delete(odp_actions); |
064af421 BP |
2144 | } |
2145 | ||
afe75089 | 2146 | /* Inserts 'rule' into 'p''s flow table. */ |
064af421 | 2147 | static void |
afe75089 | 2148 | rule_insert(struct ofproto *p, struct rule *rule) |
064af421 BP |
2149 | { |
2150 | struct rule *displaced_rule; | |
2151 | ||
064af421 | 2152 | displaced_rule = rule_from_cls_rule(classifier_insert(&p->cls, &rule->cr)); |
bcf84111 BP |
2153 | if (displaced_rule) { |
2154 | rule_destroy(p, displaced_rule); | |
064af421 | 2155 | } |
bcf84111 | 2156 | p->need_revalidate = true; |
064af421 BP |
2157 | } |
2158 | ||
bcf84111 BP |
2159 | /* Creates and returns a new facet within 'ofproto' owned by 'rule', given a |
2160 | * 'flow' and an example 'packet' within that flow. | |
2161 | * | |
2162 | * The caller must already have determined that no facet with an identical | |
2163 | * 'flow' exists in 'ofproto' and that 'flow' is the best match for 'rule' in | |
2164 | * 'ofproto''s classifier table. */ | |
2165 | static struct facet * | |
2166 | facet_create(struct ofproto *ofproto, struct rule *rule, | |
2167 | const struct flow *flow, const struct ofpbuf *packet) | |
064af421 | 2168 | { |
bcf84111 | 2169 | struct facet *facet; |
fbb2ea0b | 2170 | |
bcf84111 BP |
2171 | facet = xzalloc(sizeof *facet); |
2172 | facet->used = time_msec(); | |
2173 | hmap_insert(&ofproto->facets, &facet->hmap_node, flow_hash(flow, 0)); | |
2174 | list_push_back(&rule->facets, &facet->list_node); | |
2175 | facet->rule = rule; | |
2176 | facet->flow = *flow; | |
2177 | netflow_flow_init(&facet->nf_flow); | |
2178 | netflow_flow_update_time(ofproto->netflow, &facet->nf_flow, facet->used); | |
2179 | ||
2180 | facet_make_actions(ofproto, facet, packet); | |
064af421 | 2181 | |
bcf84111 BP |
2182 | return facet; |
2183 | } | |
2184 | ||
2185 | static void | |
2186 | facet_free(struct facet *facet) | |
2187 | { | |
2188 | free(facet->actions); | |
2189 | free(facet); | |
064af421 BP |
2190 | } |
2191 | ||
431d4707 | 2192 | /* Remove 'rule' from 'ofproto' and free up the associated memory: |
431d4707 BP |
2193 | * |
2194 | * - Removes 'rule' from the classifier. | |
2195 | * | |
bcf84111 BP |
2196 | * - If 'rule' has facets, revalidates them (and possibly uninstalls and |
2197 | * destroys them), via rule_destroy(). | |
431d4707 | 2198 | */ |
064af421 BP |
2199 | static void |
2200 | rule_remove(struct ofproto *ofproto, struct rule *rule) | |
2201 | { | |
bcf84111 BP |
2202 | COVERAGE_INC(ofproto_del_rule); |
2203 | ofproto->need_revalidate = true; | |
064af421 BP |
2204 | classifier_remove(&ofproto->cls, &rule->cr); |
2205 | rule_destroy(ofproto, rule); | |
2206 | } | |
2207 | ||
bcf84111 BP |
2208 | /* Remove 'facet' from 'ofproto' and free up the associated memory: |
2209 | * | |
2210 | * - If 'facet' was installed in the datapath, uninstalls it and updates its | |
2211 | * rule's statistics, via facet_uninstall(). | |
2212 | * | |
2213 | * - Removes 'facet' from its rule and from ofproto->facets. | |
2214 | */ | |
2215 | static void | |
2216 | facet_remove(struct ofproto *ofproto, struct facet *facet) | |
2217 | { | |
2218 | facet_uninstall(ofproto, facet); | |
d530fcd2 | 2219 | facet_flush_stats(ofproto, facet); |
bcf84111 BP |
2220 | hmap_remove(&ofproto->facets, &facet->hmap_node); |
2221 | list_remove(&facet->list_node); | |
2222 | facet_free(facet); | |
2223 | } | |
2224 | ||
7f7ae89d BP |
2225 | /* Composes the ODP actions for 'facet' based on its rule's actions. */ |
2226 | static void | |
bcf84111 BP |
2227 | facet_make_actions(struct ofproto *p, struct facet *facet, |
2228 | const struct ofpbuf *packet) | |
064af421 | 2229 | { |
bcf84111 | 2230 | const struct rule *rule = facet->rule; |
cdee00fd | 2231 | struct ofpbuf *odp_actions; |
f29152ca | 2232 | struct action_xlate_ctx ctx; |
064af421 | 2233 | |
f29152ca | 2234 | action_xlate_ctx_init(&ctx, p, &facet->flow, packet); |
cdee00fd | 2235 | odp_actions = xlate_actions(&ctx, rule->actions, rule->n_actions); |
19739aee JP |
2236 | facet->tags = ctx.tags; |
2237 | facet->may_install = ctx.may_set_up_flow; | |
2238 | facet->nf_flow.output_iface = ctx.nf_output_iface; | |
064af421 | 2239 | |
cdee00fd BP |
2240 | if (facet->actions_len != odp_actions->size |
2241 | || memcmp(facet->actions, odp_actions->data, odp_actions->size)) { | |
7f7ae89d | 2242 | free(facet->actions); |
cdee00fd BP |
2243 | facet->actions_len = odp_actions->size; |
2244 | facet->actions = xmemdup(odp_actions->data, odp_actions->size); | |
064af421 | 2245 | } |
cdee00fd BP |
2246 | |
2247 | ofpbuf_delete(odp_actions); | |
064af421 BP |
2248 | } |
2249 | ||
2250 | static int | |
ba25b8f4 | 2251 | facet_put__(struct ofproto *ofproto, struct facet *facet, |
c62b0064 BP |
2252 | const struct nlattr *actions, size_t actions_len, |
2253 | struct dpif_flow_stats *stats) | |
064af421 | 2254 | { |
36956a7d | 2255 | uint32_t keybuf[ODPUTIL_FLOW_KEY_U32S]; |
c62b0064 | 2256 | enum dpif_flow_put_flags flags; |
36956a7d BP |
2257 | struct ofpbuf key; |
2258 | ||
c62b0064 BP |
2259 | flags = DPIF_FP_CREATE | DPIF_FP_MODIFY; |
2260 | if (stats) { | |
2261 | flags |= DPIF_FP_ZERO_STATS; | |
3394b5b6 EJ |
2262 | facet->dp_packet_count = 0; |
2263 | facet->dp_byte_count = 0; | |
c62b0064 BP |
2264 | } |
2265 | ||
36956a7d BP |
2266 | ofpbuf_use_stack(&key, keybuf, sizeof keybuf); |
2267 | odp_flow_key_from_flow(&key, &facet->flow); | |
2268 | assert(key.base == keybuf); | |
2269 | ||
feebdea2 | 2270 | return dpif_flow_put(ofproto->dpif, flags, key.data, key.size, |
c62b0064 | 2271 | actions, actions_len, stats); |
064af421 BP |
2272 | } |
2273 | ||
bcf84111 BP |
2274 | /* If 'facet' is installable, inserts or re-inserts it into 'p''s datapath. If |
2275 | * 'zero_stats' is true, clears any existing statistics from the datapath for | |
2276 | * 'facet'. */ | |
064af421 | 2277 | static void |
bcf84111 | 2278 | facet_install(struct ofproto *p, struct facet *facet, bool zero_stats) |
064af421 | 2279 | { |
c62b0064 BP |
2280 | struct dpif_flow_stats stats; |
2281 | ||
2282 | if (facet->may_install | |
2283 | && !facet_put__(p, facet, facet->actions, facet->actions_len, | |
2284 | zero_stats ? &stats : NULL)) { | |
2285 | facet->installed = true; | |
064af421 BP |
2286 | } |
2287 | } | |
2288 | ||
bcf84111 BP |
2289 | /* Ensures that the bytes in 'facet', plus 'extra_bytes', have been passed up |
2290 | * to the accounting hook function in the ofhooks structure. */ | |
064af421 | 2291 | static void |
bcf84111 BP |
2292 | facet_account(struct ofproto *ofproto, |
2293 | struct facet *facet, uint64_t extra_bytes) | |
064af421 | 2294 | { |
bcf84111 | 2295 | uint64_t total_bytes = facet->byte_count + extra_bytes; |
064af421 BP |
2296 | |
2297 | if (ofproto->ofhooks->account_flow_cb | |
bcf84111 | 2298 | && total_bytes > facet->accounted_bytes) |
064af421 BP |
2299 | { |
2300 | ofproto->ofhooks->account_flow_cb( | |
cdee00fd | 2301 | &facet->flow, facet->tags, facet->actions, facet->actions_len, |
bcf84111 BP |
2302 | total_bytes - facet->accounted_bytes, ofproto->aux); |
2303 | facet->accounted_bytes = total_bytes; | |
064af421 BP |
2304 | } |
2305 | } | |
2306 | ||
d530fcd2 | 2307 | /* If 'rule' is installed in the datapath, uninstalls it. */ |
064af421 | 2308 | static void |
bcf84111 | 2309 | facet_uninstall(struct ofproto *p, struct facet *facet) |
064af421 | 2310 | { |
bcf84111 | 2311 | if (facet->installed) { |
36956a7d | 2312 | uint32_t keybuf[ODPUTIL_FLOW_KEY_U32S]; |
c97fb132 | 2313 | struct dpif_flow_stats stats; |
36956a7d | 2314 | struct ofpbuf key; |
064af421 | 2315 | |
36956a7d BP |
2316 | ofpbuf_use_stack(&key, keybuf, sizeof keybuf); |
2317 | odp_flow_key_from_flow(&key, &facet->flow); | |
2318 | assert(key.base == keybuf); | |
2319 | ||
feebdea2 BP |
2320 | if (!dpif_flow_del(p->dpif, key.data, key.size, &stats)) { |
2321 | facet_update_stats(p, facet, &stats); | |
064af421 | 2322 | } |
bcf84111 | 2323 | facet->installed = false; |
3394b5b6 EJ |
2324 | facet->dp_packet_count = 0; |
2325 | facet->dp_byte_count = 0; | |
a2c6a63c EJ |
2326 | } else { |
2327 | assert(facet->dp_packet_count == 0); | |
2328 | assert(facet->dp_byte_count == 0); | |
064af421 BP |
2329 | } |
2330 | } | |
2331 | ||
bcf84111 BP |
2332 | /* Returns true if the only action for 'facet' is to send to the controller. |
2333 | * (We don't report NetFlow expiration messages for such facets because they | |
2334 | * are just part of the control logic for the network, not real traffic). */ | |
0193b2af | 2335 | static bool |
bcf84111 | 2336 | facet_is_controller_flow(struct facet *facet) |
0193b2af | 2337 | { |
bcf84111 BP |
2338 | return (facet |
2339 | && facet->rule->n_actions == 1 | |
2340 | && action_outputs_to_port(&facet->rule->actions[0], | |
c1c9c9c4 | 2341 | htons(OFPP_CONTROLLER))); |
0193b2af JG |
2342 | } |
2343 | ||
bcf84111 | 2344 | /* Folds all of 'facet''s statistics into its rule. Also updates the |
3394b5b6 EJ |
2345 | * accounting ofhook and emits a NetFlow expiration if appropriate. All of |
2346 | * 'facet''s statistics in the datapath should have been zeroed and folded into | |
2347 | * its packet and byte counts before this function is called. */ | |
064af421 | 2348 | static void |
d530fcd2 | 2349 | facet_flush_stats(struct ofproto *ofproto, struct facet *facet) |
064af421 | 2350 | { |
3394b5b6 EJ |
2351 | assert(!facet->dp_byte_count); |
2352 | assert(!facet->dp_packet_count); | |
2353 | ||
878ae780 | 2354 | facet_push_stats(ofproto, facet); |
bcf84111 | 2355 | facet_account(ofproto, facet, 0); |
064af421 | 2356 | |
bcf84111 | 2357 | if (ofproto->netflow && !facet_is_controller_flow(facet)) { |
064af421 | 2358 | struct ofexpired expired; |
bcf84111 BP |
2359 | expired.flow = facet->flow; |
2360 | expired.packet_count = facet->packet_count; | |
2361 | expired.byte_count = facet->byte_count; | |
2362 | expired.used = facet->used; | |
2363 | netflow_expire(ofproto->netflow, &facet->nf_flow, &expired); | |
2364 | } | |
2365 | ||
2366 | facet->rule->packet_count += facet->packet_count; | |
2367 | facet->rule->byte_count += facet->byte_count; | |
2368 | ||
2369 | /* Reset counters to prevent double counting if 'facet' ever gets | |
2370 | * reinstalled. */ | |
2371 | facet->packet_count = 0; | |
2372 | facet->byte_count = 0; | |
878ae780 EJ |
2373 | facet->rs_packet_count = 0; |
2374 | facet->rs_byte_count = 0; | |
bcf84111 BP |
2375 | facet->accounted_bytes = 0; |
2376 | ||
2377 | netflow_flow_clear(&facet->nf_flow); | |
2378 | } | |
2379 | ||
2380 | /* Searches 'ofproto''s table of facets for one exactly equal to 'flow'. | |
2381 | * Returns it if found, otherwise a null pointer. | |
2382 | * | |
2383 | * The returned facet might need revalidation; use facet_lookup_valid() | |
2384 | * instead if that is important. */ | |
2385 | static struct facet * | |
2386 | facet_find(struct ofproto *ofproto, const struct flow *flow) | |
2387 | { | |
2388 | struct facet *facet; | |
2389 | ||
2390 | HMAP_FOR_EACH_WITH_HASH (facet, hmap_node, flow_hash(flow, 0), | |
2391 | &ofproto->facets) { | |
2392 | if (flow_equal(flow, &facet->flow)) { | |
2393 | return facet; | |
2394 | } | |
2395 | } | |
2396 | ||
2397 | return NULL; | |
2398 | } | |
2399 | ||
2400 | /* Searches 'ofproto''s table of facets for one exactly equal to 'flow'. | |
2401 | * Returns it if found, otherwise a null pointer. | |
2402 | * | |
2403 | * The returned facet is guaranteed to be valid. */ | |
2404 | static struct facet * | |
2405 | facet_lookup_valid(struct ofproto *ofproto, const struct flow *flow) | |
2406 | { | |
2407 | struct facet *facet = facet_find(ofproto, flow); | |
2408 | ||
2409 | /* The facet we found might not be valid, since we could be in need of | |
2410 | * revalidation. If it is not valid, don't return it. */ | |
2411 | if (facet | |
2412 | && ofproto->need_revalidate | |
2413 | && !facet_revalidate(ofproto, facet)) { | |
2414 | COVERAGE_INC(ofproto_invalidated); | |
2415 | return NULL; | |
064af421 | 2416 | } |
064af421 | 2417 | |
bcf84111 BP |
2418 | return facet; |
2419 | } | |
0193b2af | 2420 | |
bcf84111 BP |
2421 | /* Re-searches 'ofproto''s classifier for a rule matching 'facet': |
2422 | * | |
2423 | * - If the rule found is different from 'facet''s current rule, moves | |
2424 | * 'facet' to the new rule and recompiles its actions. | |
2425 | * | |
2426 | * - If the rule found is the same as 'facet''s current rule, leaves 'facet' | |
2427 | * where it is and recompiles its actions anyway. | |
2428 | * | |
2429 | * - If there is none, destroys 'facet'. | |
2430 | * | |
d530fcd2 | 2431 | * Returns true if 'facet' still exists, false if it has been destroyed. */ |
bcf84111 BP |
2432 | static bool |
2433 | facet_revalidate(struct ofproto *ofproto, struct facet *facet) | |
2434 | { | |
f29152ca | 2435 | struct action_xlate_ctx ctx; |
cdee00fd | 2436 | struct ofpbuf *odp_actions; |
d530fcd2 | 2437 | struct rule *new_rule; |
d530fcd2 | 2438 | bool actions_changed; |
bcf84111 BP |
2439 | |
2440 | COVERAGE_INC(facet_revalidate); | |
d530fcd2 BP |
2441 | |
2442 | /* Determine the new rule. */ | |
2443 | new_rule = rule_lookup(ofproto, &facet->flow); | |
2444 | if (!new_rule) { | |
2445 | /* No new rule, so delete the facet. */ | |
bcf84111 BP |
2446 | facet_remove(ofproto, facet); |
2447 | return false; | |
2448 | } | |
2449 | ||
d530fcd2 BP |
2450 | /* Calculate new ODP actions. |
2451 | * | |
cdee00fd BP |
2452 | * We do not modify any 'facet' state yet, because we might need to, e.g., |
2453 | * emit a NetFlow expiration and, if so, we need to have the old state | |
2454 | * around to properly compose it. */ | |
f29152ca | 2455 | action_xlate_ctx_init(&ctx, ofproto, &facet->flow, NULL); |
cdee00fd BP |
2456 | odp_actions = xlate_actions(&ctx, new_rule->actions, new_rule->n_actions); |
2457 | actions_changed = (facet->actions_len != odp_actions->size | |
2458 | || memcmp(facet->actions, odp_actions->data, | |
2459 | facet->actions_len)); | |
d530fcd2 BP |
2460 | |
2461 | /* If the ODP actions changed or the installability changed, then we need | |
2462 | * to talk to the datapath. */ | |
e4add896 BP |
2463 | if (actions_changed || ctx.may_set_up_flow != facet->installed) { |
2464 | if (ctx.may_set_up_flow) { | |
c97fb132 | 2465 | struct dpif_flow_stats stats; |
d530fcd2 | 2466 | |
c62b0064 BP |
2467 | facet_put__(ofproto, facet, |
2468 | odp_actions->data, odp_actions->size, &stats); | |
feebdea2 | 2469 | facet_update_stats(ofproto, facet, &stats); |
d530fcd2 BP |
2470 | } else { |
2471 | facet_uninstall(ofproto, facet); | |
2472 | } | |
2473 | ||
2474 | /* The datapath flow is gone or has zeroed stats, so push stats out of | |
2475 | * 'facet' into 'rule'. */ | |
2476 | facet_flush_stats(ofproto, facet); | |
2477 | } | |
2478 | ||
2479 | /* Update 'facet' now that we've taken care of all the old state. */ | |
f29152ca BP |
2480 | facet->tags = ctx.tags; |
2481 | facet->nf_flow.output_iface = ctx.nf_output_iface; | |
2482 | facet->may_install = ctx.may_set_up_flow; | |
d530fcd2 BP |
2483 | if (actions_changed) { |
2484 | free(facet->actions); | |
cdee00fd BP |
2485 | facet->actions_len = odp_actions->size; |
2486 | facet->actions = xmemdup(odp_actions->data, odp_actions->size); | |
d530fcd2 BP |
2487 | } |
2488 | if (facet->rule != new_rule) { | |
bcf84111 BP |
2489 | COVERAGE_INC(facet_changed_rule); |
2490 | list_remove(&facet->list_node); | |
d530fcd2 BP |
2491 | list_push_back(&new_rule->facets, &facet->list_node); |
2492 | facet->rule = new_rule; | |
2493 | facet->used = new_rule->created; | |
5e1b3214 | 2494 | facet->rs_used = facet->used; |
0c0afbec | 2495 | } |
bcf84111 | 2496 | |
cdfcd496 BP |
2497 | ofpbuf_delete(odp_actions); |
2498 | ||
bcf84111 | 2499 | return true; |
064af421 BP |
2500 | } |
2501 | \f | |
2502 | static void | |
2503 | queue_tx(struct ofpbuf *msg, const struct ofconn *ofconn, | |
2504 | struct rconn_packet_counter *counter) | |
2505 | { | |
2506 | update_openflow_length(msg); | |
2507 | if (rconn_send(ofconn->rconn, msg, counter)) { | |
2508 | ofpbuf_delete(msg); | |
2509 | } | |
2510 | } | |
2511 | ||
064af421 BP |
2512 | static void |
2513 | send_error_oh(const struct ofconn *ofconn, const struct ofp_header *oh, | |
2514 | int error) | |
2515 | { | |
dc4762ed | 2516 | struct ofpbuf *buf = ofputil_encode_error_msg(error, oh); |
26c112c2 BP |
2517 | if (buf) { |
2518 | COVERAGE_INC(ofproto_error); | |
2519 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
2520 | } | |
064af421 BP |
2521 | } |
2522 | ||
2523 | static void | |
2524 | hton_ofp_phy_port(struct ofp_phy_port *opp) | |
2525 | { | |
2526 | opp->port_no = htons(opp->port_no); | |
2527 | opp->config = htonl(opp->config); | |
2528 | opp->state = htonl(opp->state); | |
2529 | opp->curr = htonl(opp->curr); | |
2530 | opp->advertised = htonl(opp->advertised); | |
2531 | opp->supported = htonl(opp->supported); | |
2532 | opp->peer = htonl(opp->peer); | |
2533 | } | |
2534 | ||
2535 | static int | |
d1e2cf21 | 2536 | handle_echo_request(struct ofconn *ofconn, const struct ofp_header *oh) |
064af421 | 2537 | { |
d1e2cf21 | 2538 | queue_tx(make_echo_reply(oh), ofconn, ofconn->reply_counter); |
064af421 BP |
2539 | return 0; |
2540 | } | |
2541 | ||
2542 | static int | |
d1e2cf21 | 2543 | handle_features_request(struct ofconn *ofconn, const struct ofp_header *oh) |
064af421 BP |
2544 | { |
2545 | struct ofp_switch_features *osf; | |
2546 | struct ofpbuf *buf; | |
064af421 BP |
2547 | struct ofport *port; |
2548 | ||
2549 | osf = make_openflow_xid(sizeof *osf, OFPT_FEATURES_REPLY, oh->xid, &buf); | |
3269c562 | 2550 | osf->datapath_id = htonll(ofconn->ofproto->datapath_id); |
064af421 BP |
2551 | osf->n_buffers = htonl(pktbuf_capacity()); |
2552 | osf->n_tables = 2; | |
2553 | osf->capabilities = htonl(OFPC_FLOW_STATS | OFPC_TABLE_STATS | | |
0254ae23 | 2554 | OFPC_PORT_STATS | OFPC_ARP_MATCH_IP); |
064af421 BP |
2555 | osf->actions = htonl((1u << OFPAT_OUTPUT) | |
2556 | (1u << OFPAT_SET_VLAN_VID) | | |
2557 | (1u << OFPAT_SET_VLAN_PCP) | | |
2558 | (1u << OFPAT_STRIP_VLAN) | | |
2559 | (1u << OFPAT_SET_DL_SRC) | | |
2560 | (1u << OFPAT_SET_DL_DST) | | |
2561 | (1u << OFPAT_SET_NW_SRC) | | |
2562 | (1u << OFPAT_SET_NW_DST) | | |
959a2ecd | 2563 | (1u << OFPAT_SET_NW_TOS) | |
064af421 | 2564 | (1u << OFPAT_SET_TP_SRC) | |
c1c9c9c4 BP |
2565 | (1u << OFPAT_SET_TP_DST) | |
2566 | (1u << OFPAT_ENQUEUE)); | |
064af421 | 2567 | |
3269c562 | 2568 | HMAP_FOR_EACH (port, hmap_node, &ofconn->ofproto->ports) { |
064af421 BP |
2569 | hton_ofp_phy_port(ofpbuf_put(buf, &port->opp, sizeof port->opp)); |
2570 | } | |
2571 | ||
2572 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
2573 | return 0; | |
2574 | } | |
2575 | ||
2576 | static int | |
d1e2cf21 | 2577 | handle_get_config_request(struct ofconn *ofconn, const struct ofp_header *oh) |
064af421 BP |
2578 | { |
2579 | struct ofpbuf *buf; | |
2580 | struct ofp_switch_config *osc; | |
2581 | uint16_t flags; | |
2582 | bool drop_frags; | |
2583 | ||
2584 | /* Figure out flags. */ | |
3269c562 | 2585 | dpif_get_drop_frags(ofconn->ofproto->dpif, &drop_frags); |
064af421 | 2586 | flags = drop_frags ? OFPC_FRAG_DROP : OFPC_FRAG_NORMAL; |
064af421 BP |
2587 | |
2588 | /* Send reply. */ | |
2589 | osc = make_openflow_xid(sizeof *osc, OFPT_GET_CONFIG_REPLY, oh->xid, &buf); | |
2590 | osc->flags = htons(flags); | |
2591 | osc->miss_send_len = htons(ofconn->miss_send_len); | |
2592 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
2593 | ||
2594 | return 0; | |
2595 | } | |
2596 | ||
2597 | static int | |
d1e2cf21 | 2598 | handle_set_config(struct ofconn *ofconn, const struct ofp_switch_config *osc) |
064af421 | 2599 | { |
d1e2cf21 | 2600 | uint16_t flags = ntohs(osc->flags); |
064af421 | 2601 | |
5899143f | 2602 | if (ofconn->type == OFCONN_PRIMARY && ofconn->role != NX_ROLE_SLAVE) { |
064af421 BP |
2603 | switch (flags & OFPC_FRAG_MASK) { |
2604 | case OFPC_FRAG_NORMAL: | |
3269c562 | 2605 | dpif_set_drop_frags(ofconn->ofproto->dpif, false); |
064af421 BP |
2606 | break; |
2607 | case OFPC_FRAG_DROP: | |
3269c562 | 2608 | dpif_set_drop_frags(ofconn->ofproto->dpif, true); |
064af421 BP |
2609 | break; |
2610 | default: | |
2611 | VLOG_WARN_RL(&rl, "requested bad fragment mode (flags=%"PRIx16")", | |
2612 | osc->flags); | |
2613 | break; | |
2614 | } | |
2615 | } | |
2616 | ||
064af421 BP |
2617 | ofconn->miss_send_len = ntohs(osc->miss_send_len); |
2618 | ||
2619 | return 0; | |
2620 | } | |
2621 | ||
064af421 BP |
2622 | static void do_xlate_actions(const union ofp_action *in, size_t n_in, |
2623 | struct action_xlate_ctx *ctx); | |
2624 | ||
2625 | static void | |
2626 | add_output_action(struct action_xlate_ctx *ctx, uint16_t port) | |
2627 | { | |
ca0f572c | 2628 | const struct ofport *ofport = get_port(ctx->ofproto, port); |
6cfaf517 BP |
2629 | |
2630 | if (ofport) { | |
2631 | if (ofport->opp.config & OFPPC_NO_FWD) { | |
2632 | /* Forwarding disabled on port. */ | |
2633 | return; | |
2634 | } | |
2635 | } else { | |
2636 | /* | |
2637 | * We don't have an ofport record for this port, but it doesn't hurt to | |
2638 | * allow forwarding to it anyhow. Maybe such a port will appear later | |
2639 | * and we're pre-populating the flow table. | |
2640 | */ | |
064af421 | 2641 | } |
6cfaf517 | 2642 | |
7aec165d | 2643 | nl_msg_put_u32(ctx->odp_actions, ODP_ACTION_ATTR_OUTPUT, port); |
6a07af36 | 2644 | ctx->nf_output_iface = port; |
064af421 BP |
2645 | } |
2646 | ||
2647 | static struct rule * | |
bcf84111 | 2648 | rule_lookup(struct ofproto *ofproto, const struct flow *flow) |
064af421 | 2649 | { |
3c4486a5 | 2650 | return rule_from_cls_rule(classifier_lookup(&ofproto->cls, flow)); |
064af421 BP |
2651 | } |
2652 | ||
2653 | static void | |
2654 | xlate_table_action(struct action_xlate_ctx *ctx, uint16_t in_port) | |
2655 | { | |
5f8bfd69 | 2656 | if (ctx->recurse < MAX_RESUBMIT_RECURSION) { |
2c5d1389 | 2657 | uint16_t old_in_port; |
064af421 | 2658 | struct rule *rule; |
064af421 | 2659 | |
2c5d1389 BP |
2660 | /* Look up a flow with 'in_port' as the input port. Then restore the |
2661 | * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will | |
2662 | * have surprising behavior). */ | |
2663 | old_in_port = ctx->flow.in_port; | |
e18fe8a2 | 2664 | ctx->flow.in_port = in_port; |
bcf84111 | 2665 | rule = rule_lookup(ctx->ofproto, &ctx->flow); |
2c5d1389 BP |
2666 | ctx->flow.in_port = old_in_port; |
2667 | ||
7aa697dd BP |
2668 | if (ctx->resubmit_hook) { |
2669 | ctx->resubmit_hook(ctx, rule); | |
2670 | } | |
2671 | ||
064af421 | 2672 | if (rule) { |
064af421 BP |
2673 | ctx->recurse++; |
2674 | do_xlate_actions(rule->actions, rule->n_actions, ctx); | |
2675 | ctx->recurse--; | |
2676 | } | |
5f8bfd69 | 2677 | } else { |
db5ce514 | 2678 | static struct vlog_rate_limit recurse_rl = VLOG_RATE_LIMIT_INIT(1, 1); |
5f8bfd69 BP |
2679 | |
2680 | VLOG_ERR_RL(&recurse_rl, "NXAST_RESUBMIT recursed over %d times", | |
2681 | MAX_RESUBMIT_RECURSION); | |
064af421 BP |
2682 | } |
2683 | } | |
2684 | ||
f1588b1f BP |
2685 | static void |
2686 | flood_packets(struct ofproto *ofproto, uint16_t odp_in_port, uint32_t mask, | |
cdee00fd | 2687 | uint16_t *nf_output_iface, struct ofpbuf *odp_actions) |
f1588b1f BP |
2688 | { |
2689 | struct ofport *ofport; | |
2690 | ||
2691 | HMAP_FOR_EACH (ofport, hmap_node, &ofproto->ports) { | |
2692 | uint16_t odp_port = ofport->odp_port; | |
2693 | if (odp_port != odp_in_port && !(ofport->opp.config & mask)) { | |
7aec165d | 2694 | nl_msg_put_u32(odp_actions, ODP_ACTION_ATTR_OUTPUT, odp_port); |
f1588b1f BP |
2695 | } |
2696 | } | |
2697 | *nf_output_iface = NF_OUT_FLOOD; | |
2698 | } | |
2699 | ||
064af421 | 2700 | static void |
aae51f53 BP |
2701 | xlate_output_action__(struct action_xlate_ctx *ctx, |
2702 | uint16_t port, uint16_t max_len) | |
064af421 BP |
2703 | { |
2704 | uint16_t odp_port; | |
6a07af36 JG |
2705 | uint16_t prev_nf_output_iface = ctx->nf_output_iface; |
2706 | ||
2707 | ctx->nf_output_iface = NF_OUT_DROP; | |
064af421 | 2708 | |
aae51f53 | 2709 | switch (port) { |
064af421 | 2710 | case OFPP_IN_PORT: |
e18fe8a2 | 2711 | add_output_action(ctx, ctx->flow.in_port); |
064af421 BP |
2712 | break; |
2713 | case OFPP_TABLE: | |
e18fe8a2 | 2714 | xlate_table_action(ctx, ctx->flow.in_port); |
064af421 BP |
2715 | break; |
2716 | case OFPP_NORMAL: | |
e18fe8a2 | 2717 | if (!ctx->ofproto->ofhooks->normal_cb(&ctx->flow, ctx->packet, |
cdee00fd | 2718 | ctx->odp_actions, &ctx->tags, |
6a07af36 | 2719 | &ctx->nf_output_iface, |
064af421 BP |
2720 | ctx->ofproto->aux)) { |
2721 | COVERAGE_INC(ofproto_uninstallable); | |
d6fbec6d | 2722 | ctx->may_set_up_flow = false; |
064af421 BP |
2723 | } |
2724 | break; | |
2725 | case OFPP_FLOOD: | |
f1588b1f | 2726 | flood_packets(ctx->ofproto, ctx->flow.in_port, OFPPC_NO_FLOOD, |
cdee00fd | 2727 | &ctx->nf_output_iface, ctx->odp_actions); |
9628cd42 | 2728 | break; |
064af421 | 2729 | case OFPP_ALL: |
f1588b1f | 2730 | flood_packets(ctx->ofproto, ctx->flow.in_port, 0, |
cdee00fd | 2731 | &ctx->nf_output_iface, ctx->odp_actions); |
064af421 BP |
2732 | break; |
2733 | case OFPP_CONTROLLER: | |
7aec165d | 2734 | nl_msg_put_u64(ctx->odp_actions, ODP_ACTION_ATTR_CONTROLLER, max_len); |
064af421 BP |
2735 | break; |
2736 | case OFPP_LOCAL: | |
2737 | add_output_action(ctx, ODPP_LOCAL); | |
2738 | break; | |
2739 | default: | |
aae51f53 | 2740 | odp_port = ofp_port_to_odp_port(port); |
e18fe8a2 | 2741 | if (odp_port != ctx->flow.in_port) { |
064af421 BP |
2742 | add_output_action(ctx, odp_port); |
2743 | } | |
2744 | break; | |
2745 | } | |
6a07af36 JG |
2746 | |
2747 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
2748 | ctx->nf_output_iface = NF_OUT_FLOOD; | |
2749 | } else if (ctx->nf_output_iface == NF_OUT_DROP) { | |
2750 | ctx->nf_output_iface = prev_nf_output_iface; | |
2751 | } else if (prev_nf_output_iface != NF_OUT_DROP && | |
2752 | ctx->nf_output_iface != NF_OUT_FLOOD) { | |
2753 | ctx->nf_output_iface = NF_OUT_MULTI; | |
2754 | } | |
064af421 BP |
2755 | } |
2756 | ||
aae51f53 BP |
2757 | static void |
2758 | xlate_output_action(struct action_xlate_ctx *ctx, | |
2759 | const struct ofp_action_output *oao) | |
2760 | { | |
2761 | xlate_output_action__(ctx, ntohs(oao->port), ntohs(oao->max_len)); | |
2762 | } | |
2763 | ||
c1c9c9c4 BP |
2764 | /* If the final ODP action in 'ctx' is "pop priority", drop it, as an |
2765 | * optimization, because we're going to add another action that sets the | |
2766 | * priority immediately after, or because there are no actions following the | |
2767 | * pop. */ | |
2768 | static void | |
2769 | remove_pop_action(struct action_xlate_ctx *ctx) | |
2770 | { | |
cdee00fd BP |
2771 | if (ctx->odp_actions->size == ctx->last_pop_priority) { |
2772 | ctx->odp_actions->size -= NLA_ALIGN(NLA_HDRLEN); | |
2773 | ctx->last_pop_priority = -1; | |
2774 | } | |
2775 | } | |
2776 | ||
2777 | static void | |
2778 | add_pop_action(struct action_xlate_ctx *ctx) | |
2779 | { | |
2780 | if (ctx->odp_actions->size != ctx->last_pop_priority) { | |
7aec165d | 2781 | nl_msg_put_flag(ctx->odp_actions, ODP_ACTION_ATTR_POP_PRIORITY); |
cdee00fd | 2782 | ctx->last_pop_priority = ctx->odp_actions->size; |
c1c9c9c4 BP |
2783 | } |
2784 | } | |
2785 | ||
2786 | static void | |
2787 | xlate_enqueue_action(struct action_xlate_ctx *ctx, | |
2788 | const struct ofp_action_enqueue *oae) | |
2789 | { | |
2790 | uint16_t ofp_port, odp_port; | |
aae51f53 BP |
2791 | uint32_t priority; |
2792 | int error; | |
2793 | ||
2794 | error = dpif_queue_to_priority(ctx->ofproto->dpif, ntohl(oae->queue_id), | |
2795 | &priority); | |
2796 | if (error) { | |
2797 | /* Fall back to ordinary output action. */ | |
2798 | xlate_output_action__(ctx, ntohs(oae->port), 0); | |
2799 | return; | |
2800 | } | |
c1c9c9c4 BP |
2801 | |
2802 | /* Figure out ODP output port. */ | |
2803 | ofp_port = ntohs(oae->port); | |
2804 | if (ofp_port != OFPP_IN_PORT) { | |
2805 | odp_port = ofp_port_to_odp_port(ofp_port); | |
2806 | } else { | |
2807 | odp_port = ctx->flow.in_port; | |
2808 | } | |
2809 | ||
2810 | /* Add ODP actions. */ | |
2811 | remove_pop_action(ctx); | |
7aec165d | 2812 | nl_msg_put_u32(ctx->odp_actions, ODP_ACTION_ATTR_SET_PRIORITY, priority); |
c1c9c9c4 | 2813 | add_output_action(ctx, odp_port); |
cdee00fd | 2814 | add_pop_action(ctx); |
c1c9c9c4 BP |
2815 | |
2816 | /* Update NetFlow output port. */ | |
2817 | if (ctx->nf_output_iface == NF_OUT_DROP) { | |
2818 | ctx->nf_output_iface = odp_port; | |
2819 | } else if (ctx->nf_output_iface != NF_OUT_FLOOD) { | |
2820 | ctx->nf_output_iface = NF_OUT_MULTI; | |
2821 | } | |
2822 | } | |
2823 | ||
eedc0097 JP |
2824 | static void |
2825 | xlate_set_queue_action(struct action_xlate_ctx *ctx, | |
2826 | const struct nx_action_set_queue *nasq) | |
2827 | { | |
2828 | uint32_t priority; | |
2829 | int error; | |
2830 | ||
2831 | error = dpif_queue_to_priority(ctx->ofproto->dpif, ntohl(nasq->queue_id), | |
2832 | &priority); | |
2833 | if (error) { | |
2834 | /* Couldn't translate queue to a priority, so ignore. A warning | |
2835 | * has already been logged. */ | |
2836 | return; | |
2837 | } | |
2838 | ||
2839 | remove_pop_action(ctx); | |
7aec165d | 2840 | nl_msg_put_u32(ctx->odp_actions, ODP_ACTION_ATTR_SET_PRIORITY, priority); |
eedc0097 JP |
2841 | } |
2842 | ||
350a665f BP |
2843 | static void |
2844 | xlate_set_dl_tci(struct action_xlate_ctx *ctx) | |
2845 | { | |
66642cb4 BP |
2846 | ovs_be16 tci = ctx->flow.vlan_tci; |
2847 | if (!(tci & htons(VLAN_CFI))) { | |
7aec165d | 2848 | nl_msg_put_flag(ctx->odp_actions, ODP_ACTION_ATTR_STRIP_VLAN); |
350a665f | 2849 | } else { |
7aec165d | 2850 | nl_msg_put_be16(ctx->odp_actions, ODP_ACTION_ATTR_SET_DL_TCI, |
cdee00fd | 2851 | tci & ~htons(VLAN_CFI)); |
350a665f BP |
2852 | } |
2853 | } | |
2854 | ||
7b064a79 BP |
2855 | struct xlate_reg_state { |
2856 | ovs_be16 vlan_tci; | |
2857 | ovs_be64 tun_id; | |
2858 | }; | |
2859 | ||
b6c9e612 | 2860 | static void |
7b064a79 BP |
2861 | save_reg_state(const struct action_xlate_ctx *ctx, |
2862 | struct xlate_reg_state *state) | |
b6c9e612 | 2863 | { |
7b064a79 BP |
2864 | state->vlan_tci = ctx->flow.vlan_tci; |
2865 | state->tun_id = ctx->flow.tun_id; | |
2866 | } | |
b6c9e612 | 2867 | |
7b064a79 BP |
2868 | static void |
2869 | update_reg_state(struct action_xlate_ctx *ctx, | |
2870 | const struct xlate_reg_state *state) | |
2871 | { | |
2872 | if (ctx->flow.vlan_tci != state->vlan_tci) { | |
b6c9e612 BP |
2873 | xlate_set_dl_tci(ctx); |
2874 | } | |
7b064a79 | 2875 | if (ctx->flow.tun_id != state->tun_id) { |
7aec165d BP |
2876 | nl_msg_put_be64(ctx->odp_actions, |
2877 | ODP_ACTION_ATTR_SET_TUNNEL, ctx->flow.tun_id); | |
926947e6 | 2878 | } |
b6c9e612 BP |
2879 | } |
2880 | ||
064af421 BP |
2881 | static void |
2882 | xlate_nicira_action(struct action_xlate_ctx *ctx, | |
2883 | const struct nx_action_header *nah) | |
2884 | { | |
2885 | const struct nx_action_resubmit *nar; | |
659586ef | 2886 | const struct nx_action_set_tunnel *nast; |
eedc0097 | 2887 | const struct nx_action_set_queue *nasq; |
53ddd40a | 2888 | const struct nx_action_multipath *nam; |
e41a9130 | 2889 | enum nx_action_subtype subtype = ntohs(nah->subtype); |
7b064a79 | 2890 | struct xlate_reg_state state; |
b9298d3f | 2891 | ovs_be64 tun_id; |
064af421 BP |
2892 | |
2893 | assert(nah->vendor == htonl(NX_VENDOR_ID)); | |
2894 | switch (subtype) { | |
2895 | case NXAST_RESUBMIT: | |
2896 | nar = (const struct nx_action_resubmit *) nah; | |
2897 | xlate_table_action(ctx, ofp_port_to_odp_port(ntohs(nar->in_port))); | |
2898 | break; | |
2899 | ||
659586ef JG |
2900 | case NXAST_SET_TUNNEL: |
2901 | nast = (const struct nx_action_set_tunnel *) nah; | |
b9298d3f | 2902 | tun_id = htonll(ntohl(nast->tun_id)); |
7aec165d | 2903 | nl_msg_put_be64(ctx->odp_actions, ODP_ACTION_ATTR_SET_TUNNEL, tun_id); |
b9298d3f | 2904 | ctx->flow.tun_id = tun_id; |
659586ef JG |
2905 | break; |
2906 | ||
401eeb92 BP |
2907 | case NXAST_DROP_SPOOFED_ARP: |
2908 | if (ctx->flow.dl_type == htons(ETH_TYPE_ARP)) { | |
7aec165d BP |
2909 | nl_msg_put_flag(ctx->odp_actions, |
2910 | ODP_ACTION_ATTR_DROP_SPOOFED_ARP); | |
401eeb92 BP |
2911 | } |
2912 | break; | |
2913 | ||
eedc0097 JP |
2914 | case NXAST_SET_QUEUE: |
2915 | nasq = (const struct nx_action_set_queue *) nah; | |
2916 | xlate_set_queue_action(ctx, nasq); | |
2917 | break; | |
2918 | ||
2919 | case NXAST_POP_QUEUE: | |
cdee00fd | 2920 | add_pop_action(ctx); |
eedc0097 JP |
2921 | break; |
2922 | ||
b6c9e612 | 2923 | case NXAST_REG_MOVE: |
7b064a79 BP |
2924 | save_reg_state(ctx, &state); |
2925 | nxm_execute_reg_move((const struct nx_action_reg_move *) nah, | |
2926 | &ctx->flow); | |
2927 | update_reg_state(ctx, &state); | |
b6c9e612 BP |
2928 | break; |
2929 | ||
2930 | case NXAST_REG_LOAD: | |
7b064a79 | 2931 | save_reg_state(ctx, &state); |
b6c9e612 BP |
2932 | nxm_execute_reg_load((const struct nx_action_reg_load *) nah, |
2933 | &ctx->flow); | |
7b064a79 BP |
2934 | update_reg_state(ctx, &state); |
2935 | break; | |
96fc46e8 BP |
2936 | |
2937 | case NXAST_NOTE: | |
2938 | /* Nothing to do. */ | |
b6c9e612 BP |
2939 | break; |
2940 | ||
b9298d3f BP |
2941 | case NXAST_SET_TUNNEL64: |
2942 | tun_id = ((const struct nx_action_set_tunnel64 *) nah)->tun_id; | |
7aec165d | 2943 | nl_msg_put_be64(ctx->odp_actions, ODP_ACTION_ATTR_SET_TUNNEL, tun_id); |
b9298d3f BP |
2944 | ctx->flow.tun_id = tun_id; |
2945 | break; | |
2946 | ||
53ddd40a BP |
2947 | case NXAST_MULTIPATH: |
2948 | nam = (const struct nx_action_multipath *) nah; | |
2949 | multipath_execute(nam, &ctx->flow); | |
2950 | break; | |
2951 | ||
999f0d45 | 2952 | /* If you add a new action here that modifies flow data, don't forget to |
c1c9c9c4 | 2953 | * update the flow key in ctx->flow at the same time. */ |
999f0d45 | 2954 | |
e41a9130 | 2955 | case NXAST_SNAT__OBSOLETE: |
064af421 | 2956 | default: |
e41a9130 | 2957 | VLOG_DBG_RL(&rl, "unknown Nicira action type %d", (int) subtype); |
064af421 BP |
2958 | break; |
2959 | } | |
2960 | } | |
2961 | ||
2962 | static void | |
2963 | do_xlate_actions(const union ofp_action *in, size_t n_in, | |
2964 | struct action_xlate_ctx *ctx) | |
2965 | { | |
2966 | struct actions_iterator iter; | |
2967 | const union ofp_action *ia; | |
2968 | const struct ofport *port; | |
2969 | ||
ca0f572c | 2970 | port = get_port(ctx->ofproto, ctx->flow.in_port); |
064af421 | 2971 | if (port && port->opp.config & (OFPPC_NO_RECV | OFPPC_NO_RECV_STP) && |
ba186119 | 2972 | port->opp.config & (eth_addr_equals(ctx->flow.dl_dst, eth_addr_stp) |
064af421 BP |
2973 | ? OFPPC_NO_RECV_STP : OFPPC_NO_RECV)) { |
2974 | /* Drop this flow. */ | |
2975 | return; | |
2976 | } | |
2977 | ||
2978 | for (ia = actions_first(&iter, in, n_in); ia; ia = actions_next(&iter)) { | |
e41a9130 | 2979 | enum ofp_action_type type = ntohs(ia->type); |
cdee00fd | 2980 | const struct ofp_action_dl_addr *oada; |
064af421 BP |
2981 | |
2982 | switch (type) { | |
2983 | case OFPAT_OUTPUT: | |
2984 | xlate_output_action(ctx, &ia->output); | |
2985 | break; | |
2986 | ||
2987 | case OFPAT_SET_VLAN_VID: | |
66642cb4 BP |
2988 | ctx->flow.vlan_tci &= ~htons(VLAN_VID_MASK); |
2989 | ctx->flow.vlan_tci |= ia->vlan_vid.vlan_vid | htons(VLAN_CFI); | |
350a665f | 2990 | xlate_set_dl_tci(ctx); |
064af421 BP |
2991 | break; |
2992 | ||
2993 | case OFPAT_SET_VLAN_PCP: | |
66642cb4 BP |
2994 | ctx->flow.vlan_tci &= ~htons(VLAN_PCP_MASK); |
2995 | ctx->flow.vlan_tci |= htons( | |
2996 | (ia->vlan_pcp.vlan_pcp << VLAN_PCP_SHIFT) | VLAN_CFI); | |
350a665f | 2997 | xlate_set_dl_tci(ctx); |
064af421 BP |
2998 | break; |
2999 | ||
3000 | case OFPAT_STRIP_VLAN: | |
66642cb4 | 3001 | ctx->flow.vlan_tci = htons(0); |
350a665f | 3002 | xlate_set_dl_tci(ctx); |
064af421 BP |
3003 | break; |
3004 | ||
3005 | case OFPAT_SET_DL_SRC: | |
cdee00fd | 3006 | oada = ((struct ofp_action_dl_addr *) ia); |
7aec165d | 3007 | nl_msg_put_unspec(ctx->odp_actions, ODP_ACTION_ATTR_SET_DL_SRC, |
cdee00fd BP |
3008 | oada->dl_addr, ETH_ADDR_LEN); |
3009 | memcpy(ctx->flow.dl_src, oada->dl_addr, ETH_ADDR_LEN); | |
064af421 BP |
3010 | break; |
3011 | ||
3012 | case OFPAT_SET_DL_DST: | |
cdee00fd | 3013 | oada = ((struct ofp_action_dl_addr *) ia); |
7aec165d | 3014 | nl_msg_put_unspec(ctx->odp_actions, ODP_ACTION_ATTR_SET_DL_DST, |
cdee00fd BP |
3015 | oada->dl_addr, ETH_ADDR_LEN); |
3016 | memcpy(ctx->flow.dl_dst, oada->dl_addr, ETH_ADDR_LEN); | |
064af421 BP |
3017 | break; |
3018 | ||
3019 | case OFPAT_SET_NW_SRC: | |
7aec165d | 3020 | nl_msg_put_be32(ctx->odp_actions, ODP_ACTION_ATTR_SET_NW_SRC, |
cdee00fd BP |
3021 | ia->nw_addr.nw_addr); |
3022 | ctx->flow.nw_src = ia->nw_addr.nw_addr; | |
064af421 BP |
3023 | break; |
3024 | ||
2d70a31a | 3025 | case OFPAT_SET_NW_DST: |
7aec165d | 3026 | nl_msg_put_be32(ctx->odp_actions, ODP_ACTION_ATTR_SET_NW_DST, |
cdee00fd BP |
3027 | ia->nw_addr.nw_addr); |
3028 | ctx->flow.nw_dst = ia->nw_addr.nw_addr; | |
2d38e234 | 3029 | break; |
959a2ecd JP |
3030 | |
3031 | case OFPAT_SET_NW_TOS: | |
7aec165d | 3032 | nl_msg_put_u8(ctx->odp_actions, ODP_ACTION_ATTR_SET_NW_TOS, |
cdee00fd BP |
3033 | ia->nw_tos.nw_tos); |
3034 | ctx->flow.nw_tos = ia->nw_tos.nw_tos; | |
2d70a31a JP |
3035 | break; |
3036 | ||
064af421 | 3037 | case OFPAT_SET_TP_SRC: |
7aec165d | 3038 | nl_msg_put_be16(ctx->odp_actions, ODP_ACTION_ATTR_SET_TP_SRC, |
cdee00fd BP |
3039 | ia->tp_port.tp_port); |
3040 | ctx->flow.tp_src = ia->tp_port.tp_port; | |
064af421 BP |
3041 | break; |
3042 | ||
2d70a31a | 3043 | case OFPAT_SET_TP_DST: |
7aec165d | 3044 | nl_msg_put_be16(ctx->odp_actions, ODP_ACTION_ATTR_SET_TP_DST, |
cdee00fd BP |
3045 | ia->tp_port.tp_port); |
3046 | ctx->flow.tp_dst = ia->tp_port.tp_port; | |
2d70a31a JP |
3047 | break; |
3048 | ||
064af421 BP |
3049 | case OFPAT_VENDOR: |
3050 | xlate_nicira_action(ctx, (const struct nx_action_header *) ia); | |
3051 | break; | |
3052 | ||
c1c9c9c4 BP |
3053 | case OFPAT_ENQUEUE: |
3054 | xlate_enqueue_action(ctx, (const struct ofp_action_enqueue *) ia); | |
3055 | break; | |
3056 | ||
064af421 | 3057 | default: |
e41a9130 | 3058 | VLOG_DBG_RL(&rl, "unknown action type %d", (int) type); |
064af421 BP |
3059 | break; |
3060 | } | |
3061 | } | |
3062 | } | |
3063 | ||
f29152ca BP |
3064 | static void |
3065 | action_xlate_ctx_init(struct action_xlate_ctx *ctx, | |
3066 | struct ofproto *ofproto, const struct flow *flow, | |
3067 | const struct ofpbuf *packet) | |
064af421 | 3068 | { |
f29152ca BP |
3069 | ctx->ofproto = ofproto; |
3070 | ctx->flow = *flow; | |
3071 | ctx->packet = packet; | |
7aa697dd | 3072 | ctx->resubmit_hook = NULL; |
ebe482fd | 3073 | ctx->check_special = true; |
f29152ca | 3074 | } |
1eb0942d | 3075 | |
cdee00fd | 3076 | static struct ofpbuf * |
f29152ca BP |
3077 | xlate_actions(struct action_xlate_ctx *ctx, |
3078 | const union ofp_action *in, size_t n_in) | |
3079 | { | |
064af421 | 3080 | COVERAGE_INC(ofproto_ofp2odp); |
cdee00fd BP |
3081 | |
3082 | ctx->odp_actions = ofpbuf_new(512); | |
f29152ca BP |
3083 | ctx->tags = 0; |
3084 | ctx->may_set_up_flow = true; | |
3085 | ctx->nf_output_iface = NF_OUT_DROP; | |
3086 | ctx->recurse = 0; | |
cdee00fd | 3087 | ctx->last_pop_priority = -1; |
ebe482fd EJ |
3088 | |
3089 | if (!ctx->check_special | |
7a7ae75a BP |
3090 | || !ctx->ofproto->ofhooks->special_cb |
3091 | || ctx->ofproto->ofhooks->special_cb(&ctx->flow, ctx->packet, | |
3092 | ctx->ofproto->aux)) { | |
ebe482fd EJ |
3093 | do_xlate_actions(in, n_in, ctx); |
3094 | } else { | |
3095 | ctx->may_set_up_flow = false; | |
3096 | } | |
3097 | ||
f29152ca | 3098 | remove_pop_action(ctx); |
0ad9b732 | 3099 | |
d6fbec6d | 3100 | /* Check with in-band control to see if we're allowed to set up this |
0ad9b732 | 3101 | * flow. */ |
cdee00fd BP |
3102 | if (!in_band_rule_check(ctx->ofproto->in_band, &ctx->flow, |
3103 | ctx->odp_actions->data, ctx->odp_actions->size)) { | |
f29152ca | 3104 | ctx->may_set_up_flow = false; |
0ad9b732 JP |
3105 | } |
3106 | ||
cdee00fd | 3107 | return ctx->odp_actions; |
064af421 BP |
3108 | } |
3109 | ||
9deba63b BP |
3110 | /* Checks whether 'ofconn' is a slave controller. If so, returns an OpenFlow |
3111 | * error message code (composed with ofp_mkerr()) for the caller to propagate | |
3112 | * upward. Otherwise, returns 0. | |
3113 | * | |
2228b50d | 3114 | * The log message mentions 'msg_type'. */ |
9deba63b | 3115 | static int |
2228b50d | 3116 | reject_slave_controller(struct ofconn *ofconn, const const char *msg_type) |
9deba63b | 3117 | { |
5899143f | 3118 | if (ofconn->type == OFCONN_PRIMARY && ofconn->role == NX_ROLE_SLAVE) { |
9deba63b | 3119 | static struct vlog_rate_limit perm_rl = VLOG_RATE_LIMIT_INIT(1, 5); |
9deba63b | 3120 | VLOG_WARN_RL(&perm_rl, "rejecting %s message from slave controller", |
2228b50d | 3121 | msg_type); |
9deba63b BP |
3122 | |
3123 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_EPERM); | |
3124 | } else { | |
3125 | return 0; | |
3126 | } | |
3127 | } | |
3128 | ||
064af421 | 3129 | static int |
d1e2cf21 | 3130 | handle_packet_out(struct ofconn *ofconn, const struct ofp_header *oh) |
064af421 | 3131 | { |
3269c562 | 3132 | struct ofproto *p = ofconn->ofproto; |
064af421 BP |
3133 | struct ofp_packet_out *opo; |
3134 | struct ofpbuf payload, *buffer; | |
ac51afaf | 3135 | union ofp_action *ofp_actions; |
f29152ca | 3136 | struct action_xlate_ctx ctx; |
cdee00fd | 3137 | struct ofpbuf *odp_actions; |
ac51afaf | 3138 | struct ofpbuf request; |
ae412e7d | 3139 | struct flow flow; |
ac51afaf | 3140 | size_t n_ofp_actions; |
064af421 | 3141 | uint16_t in_port; |
064af421 BP |
3142 | int error; |
3143 | ||
ac51afaf BP |
3144 | COVERAGE_INC(ofproto_packet_out); |
3145 | ||
2228b50d | 3146 | error = reject_slave_controller(ofconn, "OFPT_PACKET_OUT"); |
9deba63b BP |
3147 | if (error) { |
3148 | return error; | |
3149 | } | |
3150 | ||
ac51afaf | 3151 | /* Get ofp_packet_out. */ |
0bc9407d | 3152 | ofpbuf_use_const(&request, oh, ntohs(oh->length)); |
bbc32a88 | 3153 | opo = ofpbuf_pull(&request, offsetof(struct ofp_packet_out, actions)); |
ac51afaf BP |
3154 | |
3155 | /* Get actions. */ | |
3156 | error = ofputil_pull_actions(&request, ntohs(opo->actions_len), | |
3157 | &ofp_actions, &n_ofp_actions); | |
064af421 BP |
3158 | if (error) { |
3159 | return error; | |
3160 | } | |
064af421 | 3161 | |
ac51afaf | 3162 | /* Get payload. */ |
064af421 BP |
3163 | if (opo->buffer_id != htonl(UINT32_MAX)) { |
3164 | error = pktbuf_retrieve(ofconn->pktbuf, ntohl(opo->buffer_id), | |
3165 | &buffer, &in_port); | |
7778bd15 | 3166 | if (error || !buffer) { |
064af421 BP |
3167 | return error; |
3168 | } | |
3169 | payload = *buffer; | |
3170 | } else { | |
ac51afaf | 3171 | payload = request; |
064af421 BP |
3172 | buffer = NULL; |
3173 | } | |
3174 | ||
ac51afaf BP |
3175 | /* Extract flow, check actions. */ |
3176 | flow_extract(&payload, 0, ofp_port_to_odp_port(ntohs(opo->in_port)), | |
3177 | &flow); | |
f1defbf9 | 3178 | error = validate_actions(ofp_actions, n_ofp_actions, &flow, p->max_ports); |
ac51afaf BP |
3179 | if (error) { |
3180 | goto exit; | |
3181 | } | |
3182 | ||
3183 | /* Send. */ | |
f29152ca | 3184 | action_xlate_ctx_init(&ctx, p, &flow, &payload); |
cdee00fd BP |
3185 | odp_actions = xlate_actions(&ctx, ofp_actions, n_ofp_actions); |
3186 | dpif_execute(p->dpif, odp_actions->data, odp_actions->size, &payload); | |
3187 | ofpbuf_delete(odp_actions); | |
064af421 | 3188 | |
ac51afaf BP |
3189 | exit: |
3190 | ofpbuf_delete(buffer); | |
3191 | return 0; | |
064af421 BP |
3192 | } |
3193 | ||
3194 | static void | |
3195 | update_port_config(struct ofproto *p, struct ofport *port, | |
3196 | uint32_t config, uint32_t mask) | |
3197 | { | |
3198 | mask &= config ^ port->opp.config; | |
3199 | if (mask & OFPPC_PORT_DOWN) { | |
3200 | if (config & OFPPC_PORT_DOWN) { | |
3201 | netdev_turn_flags_off(port->netdev, NETDEV_UP, true); | |
3202 | } else { | |
3203 | netdev_turn_flags_on(port->netdev, NETDEV_UP, true); | |
3204 | } | |
3205 | } | |
f1588b1f BP |
3206 | #define REVALIDATE_BITS (OFPPC_NO_RECV | OFPPC_NO_RECV_STP | \ |
3207 | OFPPC_NO_FWD | OFPPC_NO_FLOOD) | |
064af421 BP |
3208 | if (mask & REVALIDATE_BITS) { |
3209 | COVERAGE_INC(ofproto_costly_flags); | |
3210 | port->opp.config ^= mask & REVALIDATE_BITS; | |
3211 | p->need_revalidate = true; | |
3212 | } | |
3213 | #undef REVALIDATE_BITS | |
064af421 BP |
3214 | if (mask & OFPPC_NO_PACKET_IN) { |
3215 | port->opp.config ^= OFPPC_NO_PACKET_IN; | |
3216 | } | |
3217 | } | |
3218 | ||
3219 | static int | |
d1e2cf21 | 3220 | handle_port_mod(struct ofconn *ofconn, const struct ofp_header *oh) |
064af421 | 3221 | { |
3269c562 | 3222 | struct ofproto *p = ofconn->ofproto; |
d1e2cf21 | 3223 | const struct ofp_port_mod *opm = (const struct ofp_port_mod *) oh; |
064af421 BP |
3224 | struct ofport *port; |
3225 | int error; | |
3226 | ||
2228b50d | 3227 | error = reject_slave_controller(ofconn, "OFPT_PORT_MOD"); |
9deba63b BP |
3228 | if (error) { |
3229 | return error; | |
3230 | } | |
064af421 | 3231 | |
ca0f572c | 3232 | port = get_port(p, ofp_port_to_odp_port(ntohs(opm->port_no))); |
064af421 BP |
3233 | if (!port) { |
3234 | return ofp_mkerr(OFPET_PORT_MOD_FAILED, OFPPMFC_BAD_PORT); | |
3235 | } else if (memcmp(port->opp.hw_addr, opm->hw_addr, OFP_ETH_ALEN)) { | |
3236 | return ofp_mkerr(OFPET_PORT_MOD_FAILED, OFPPMFC_BAD_HW_ADDR); | |
3237 | } else { | |
3238 | update_port_config(p, port, ntohl(opm->config), ntohl(opm->mask)); | |
3239 | if (opm->advertise) { | |
3240 | netdev_set_advertisements(port->netdev, ntohl(opm->advertise)); | |
3241 | } | |
3242 | } | |
3243 | return 0; | |
3244 | } | |
3245 | ||
3246 | static struct ofpbuf * | |
06a5e131 | 3247 | make_ofp_stats_reply(ovs_be32 xid, ovs_be16 type, size_t body_len) |
064af421 BP |
3248 | { |
3249 | struct ofp_stats_reply *osr; | |
3250 | struct ofpbuf *msg; | |
3251 | ||
3252 | msg = ofpbuf_new(MIN(sizeof *osr + body_len, UINT16_MAX)); | |
3253 | osr = put_openflow_xid(sizeof *osr, OFPT_STATS_REPLY, xid, msg); | |
3254 | osr->type = type; | |
3255 | osr->flags = htons(0); | |
3256 | return msg; | |
3257 | } | |
3258 | ||
3259 | static struct ofpbuf * | |
d1e2cf21 | 3260 | start_ofp_stats_reply(const struct ofp_header *request, size_t body_len) |
064af421 | 3261 | { |
d1e2cf21 BP |
3262 | const struct ofp_stats_request *osr |
3263 | = (const struct ofp_stats_request *) request; | |
3264 | return make_ofp_stats_reply(osr->header.xid, osr->type, body_len); | |
064af421 BP |
3265 | } |
3266 | ||
3267 | static void * | |
06a5e131 BP |
3268 | append_ofp_stats_reply(size_t nbytes, struct ofconn *ofconn, |
3269 | struct ofpbuf **msgp) | |
064af421 BP |
3270 | { |
3271 | struct ofpbuf *msg = *msgp; | |
3272 | assert(nbytes <= UINT16_MAX - sizeof(struct ofp_stats_reply)); | |
3273 | if (nbytes + msg->size > UINT16_MAX) { | |
3274 | struct ofp_stats_reply *reply = msg->data; | |
3275 | reply->flags = htons(OFPSF_REPLY_MORE); | |
06a5e131 | 3276 | *msgp = make_ofp_stats_reply(reply->header.xid, reply->type, nbytes); |
064af421 BP |
3277 | queue_tx(msg, ofconn, ofconn->reply_counter); |
3278 | } | |
3279 | return ofpbuf_put_uninit(*msgp, nbytes); | |
3280 | } | |
3281 | ||
09246b99 BP |
3282 | static struct ofpbuf * |
3283 | make_nxstats_reply(ovs_be32 xid, ovs_be32 subtype, size_t body_len) | |
3284 | { | |
3285 | struct nicira_stats_msg *nsm; | |
3286 | struct ofpbuf *msg; | |
3287 | ||
3288 | msg = ofpbuf_new(MIN(sizeof *nsm + body_len, UINT16_MAX)); | |
3289 | nsm = put_openflow_xid(sizeof *nsm, OFPT_STATS_REPLY, xid, msg); | |
3290 | nsm->type = htons(OFPST_VENDOR); | |
3291 | nsm->flags = htons(0); | |
3292 | nsm->vendor = htonl(NX_VENDOR_ID); | |
d5f2379b | 3293 | nsm->subtype = subtype; |
09246b99 BP |
3294 | return msg; |
3295 | } | |
3296 | ||
3297 | static struct ofpbuf * | |
3298 | start_nxstats_reply(const struct nicira_stats_msg *request, size_t body_len) | |
3299 | { | |
3300 | return make_nxstats_reply(request->header.xid, request->subtype, body_len); | |
3301 | } | |
3302 | ||
3303 | static void | |
3304 | append_nxstats_reply(size_t nbytes, struct ofconn *ofconn, | |
3305 | struct ofpbuf **msgp) | |
3306 | { | |
3307 | struct ofpbuf *msg = *msgp; | |
3308 | assert(nbytes <= UINT16_MAX - sizeof(struct nicira_stats_msg)); | |
3309 | if (nbytes + msg->size > UINT16_MAX) { | |
3310 | struct nicira_stats_msg *reply = msg->data; | |
3311 | reply->flags = htons(OFPSF_REPLY_MORE); | |
3312 | *msgp = make_nxstats_reply(reply->header.xid, reply->subtype, nbytes); | |
3313 | queue_tx(msg, ofconn, ofconn->reply_counter); | |
3314 | } | |
3315 | ofpbuf_prealloc_tailroom(*msgp, nbytes); | |
3316 | } | |
3317 | ||
064af421 | 3318 | static int |
3269c562 | 3319 | handle_desc_stats_request(struct ofconn *ofconn, |
d1e2cf21 | 3320 | const struct ofp_header *request) |
064af421 | 3321 | { |
3269c562 | 3322 | struct ofproto *p = ofconn->ofproto; |
064af421 BP |
3323 | struct ofp_desc_stats *ods; |
3324 | struct ofpbuf *msg; | |
3325 | ||
06a5e131 BP |
3326 | msg = start_ofp_stats_reply(request, sizeof *ods); |
3327 | ods = append_ofp_stats_reply(sizeof *ods, ofconn, &msg); | |
5a719c38 JP |
3328 | memset(ods, 0, sizeof *ods); |
3329 | ovs_strlcpy(ods->mfr_desc, p->mfr_desc, sizeof ods->mfr_desc); | |
3330 | ovs_strlcpy(ods->hw_desc, p->hw_desc, sizeof ods->hw_desc); | |
3331 | ovs_strlcpy(ods->sw_desc, p->sw_desc, sizeof ods->sw_desc); | |
3332 | ovs_strlcpy(ods->serial_num, p->serial_desc, sizeof ods->serial_num); | |
3333 | ovs_strlcpy(ods->dp_desc, p->dp_desc, sizeof ods->dp_desc); | |
064af421 BP |
3334 | queue_tx(msg, ofconn, ofconn->reply_counter); |
3335 | ||
3336 | return 0; | |
3337 | } | |
3338 | ||
064af421 | 3339 | static int |
3269c562 | 3340 | handle_table_stats_request(struct ofconn *ofconn, |
d1e2cf21 | 3341 | const struct ofp_header *request) |
064af421 | 3342 | { |
3269c562 | 3343 | struct ofproto *p = ofconn->ofproto; |
064af421 BP |
3344 | struct ofp_table_stats *ots; |
3345 | struct ofpbuf *msg; | |
064af421 | 3346 | |
06a5e131 | 3347 | msg = start_ofp_stats_reply(request, sizeof *ots * 2); |
064af421 | 3348 | |
064af421 | 3349 | /* Classifier table. */ |
06a5e131 | 3350 | ots = append_ofp_stats_reply(sizeof *ots, ofconn, &msg); |
064af421 | 3351 | memset(ots, 0, sizeof *ots); |
064af421 | 3352 | strcpy(ots->name, "classifier"); |
b70eac89 | 3353 | ots->wildcards = (ofconn->flow_format == NXFF_OPENFLOW10 |
f9bfea14 | 3354 | ? htonl(OFPFW_ALL) : htonl(OVSFW_ALL)); |
ad828225 | 3355 | ots->max_entries = htonl(1024 * 1024); /* An arbitrary big number. */ |
bcf84111 | 3356 | ots->active_count = htonl(classifier_count(&p->cls)); |
c4617b3c BP |
3357 | put_32aligned_be64(&ots->lookup_count, htonll(0)); /* XXX */ |
3358 | put_32aligned_be64(&ots->matched_count, htonll(0)); /* XXX */ | |
064af421 BP |
3359 | |
3360 | queue_tx(msg, ofconn, ofconn->reply_counter); | |
3361 | return 0; | |
3362 | } | |
3363 | ||
abaad8cf | 3364 | static void |
ca0f572c | 3365 | append_port_stat(struct ofport *port, struct ofconn *ofconn, |
a4948b95 | 3366 | struct ofpbuf **msgp) |
abaad8cf JP |
3367 | { |
3368 | struct netdev_stats stats; | |
3369 | struct ofp_port_stats *ops; | |
3370 | ||
d295e8e9 JP |
3371 | /* Intentionally ignore return value, since errors will set |
3372 | * 'stats' to all-1s, which is correct for OpenFlow, and | |
abaad8cf JP |
3373 | * netdev_get_stats() will log errors. */ |
3374 | netdev_get_stats(port->netdev, &stats); | |
3375 | ||
06a5e131 | 3376 | ops = append_ofp_stats_reply(sizeof *ops, ofconn, msgp); |
ca0f572c | 3377 | ops->port_no = htons(port->opp.port_no); |
abaad8cf | 3378 | memset(ops->pad, 0, sizeof ops->pad); |
c4617b3c BP |
3379 | put_32aligned_be64(&ops->rx_packets, htonll(stats.rx_packets)); |
3380 | put_32aligned_be64(&ops->tx_packets, htonll(stats.tx_packets)); | |
3381 | put_32aligned_be64(&ops->rx_bytes, htonll(stats.rx_bytes)); | |
3382 | put_32aligned_be64(&ops->tx_bytes, htonll(stats.tx_bytes)); | |
3383 | put_32aligned_be64(&ops->rx_dropped, htonll(stats.rx_dropped)); | |
3384 | put_32aligned_be64(&ops->tx_dropped, htonll(stats.tx_dropped)); | |
3385 | put_32aligned_be64(&ops->rx_errors, htonll(stats.rx_errors)); | |
3386 | put_32aligned_be64(&ops->tx_errors, htonll(stats.tx_errors)); | |
3387 | put_32aligned_be64(&ops->rx_frame_err, htonll(stats.rx_frame_errors)); | |
3388 | put_32aligned_be64(&ops->rx_over_err, htonll(stats.rx_over_errors)); | |
3389 | put_32aligned_be64(&ops->rx_crc_err, htonll(stats.rx_crc_errors)); | |
3390 | put_32aligned_be64(&ops->collisions, htonll(stats.collisions)); | |
abaad8cf JP |
3391 | } |
3392 | ||
064af421 | 3393 | static int |
d1e2cf21 | 3394 | handle_port_stats_request(struct ofconn *ofconn, const struct ofp_header *oh) |
064af421 | 3395 | { |
3269c562 | 3396 | struct ofproto *p = ofconn->ofproto; |
d1e2cf21 | 3397 | const struct ofp_port_stats_request *psr = ofputil_stats_body(oh); |
064af421 BP |
3398 | struct ofp_port_stats *ops; |
3399 | struct ofpbuf *msg; | |
3400 | struct ofport *port; | |
064af421 | 3401 | |
d1e2cf21 | 3402 | msg = start_ofp_stats_reply(oh, sizeof *ops * 16); |
abaad8cf | 3403 | if (psr->port_no != htons(OFPP_NONE)) { |
ca0f572c | 3404 | port = get_port(p, ofp_port_to_odp_port(ntohs(psr->port_no))); |
abaad8cf | 3405 | if (port) { |
ca0f572c | 3406 | append_port_stat(port, ofconn, &msg); |
abaad8cf JP |
3407 | } |
3408 | } else { | |
4e8e4213 | 3409 | HMAP_FOR_EACH (port, hmap_node, &p->ports) { |
ca0f572c | 3410 | append_port_stat(port, ofconn, &msg); |
abaad8cf | 3411 | } |
064af421 BP |
3412 | } |
3413 | ||
3414 | queue_tx(msg, ofconn, ofconn->reply_counter); | |
3415 | return 0; | |
3416 | } | |
3417 | ||
c6ebb8fb BP |
3418 | static void |
3419 | calc_flow_duration(long long int start, ovs_be32 *sec, ovs_be32 *nsec) | |
3420 | { | |
3421 | long long int msecs = time_msec() - start; | |
3422 | *sec = htonl(msecs / 1000); | |
3423 | *nsec = htonl((msecs % 1000) * (1000 * 1000)); | |
3424 | } | |
3425 | ||
064af421 | 3426 | static void |
5ecc9d81 BP |
3427 | put_ofp_flow_stats(struct ofconn *ofconn, struct rule *rule, |
3428 | ovs_be16 out_port, struct ofpbuf **replyp) | |
064af421 | 3429 | { |
064af421 BP |
3430 | struct ofp_flow_stats *ofs; |
3431 | uint64_t packet_count, byte_count; | |
c4617b3c | 3432 | ovs_be64 cookie; |
064af421 BP |
3433 | size_t act_len, len; |
3434 | ||
5ecc9d81 | 3435 | if (rule_is_hidden(rule) || !rule_has_out_port(rule, out_port)) { |
064af421 BP |
3436 | return; |
3437 | } | |
3438 | ||
3439 | act_len = sizeof *rule->actions * rule->n_actions; | |
3440 | len = offsetof(struct ofp_flow_stats, actions) + act_len; | |
3441 | ||
3394b5b6 | 3442 | rule_get_stats(rule, &packet_count, &byte_count); |
064af421 | 3443 | |
5ecc9d81 | 3444 | ofs = append_ofp_stats_reply(len, ofconn, replyp); |
064af421 | 3445 | ofs->length = htons(len); |
ad828225 | 3446 | ofs->table_id = 0; |
064af421 | 3447 | ofs->pad = 0; |
ff9d3826 | 3448 | ofputil_cls_rule_to_match(&rule->cr, ofconn->flow_format, &ofs->match, |
c4617b3c BP |
3449 | rule->flow_cookie, &cookie); |
3450 | put_32aligned_be64(&ofs->cookie, cookie); | |
c6ebb8fb | 3451 | calc_flow_duration(rule->created, &ofs->duration_sec, &ofs->duration_nsec); |
064af421 BP |
3452 | ofs->priority = htons(rule->cr.priority); |
3453 | ofs->idle_timeout = htons(rule->idle_timeout); | |
3454 | ofs->hard_timeout = htons(rule->hard_timeout); | |
39997502 | 3455 | memset(ofs->pad2, 0, sizeof ofs->pad2); |
c4617b3c BP |
3456 | put_32aligned_be64(&ofs->packet_count, htonll(packet_count)); |
3457 | put_32aligned_be64(&ofs->byte_count, htonll(byte_count)); | |
3dffcf07 BP |
3458 | if (rule->n_actions > 0) { |
3459 | memcpy(ofs->actions, rule->actions, act_len); | |
3460 | } | |
064af421 BP |
3461 | } |
3462 | ||
3c4486a5 BP |
3463 | static bool |
3464 | is_valid_table(uint8_t table_id) | |
064af421 | 3465 | { |
a02e5331 BP |
3466 | if (table_id == 0 || table_id == 0xff) { |
3467 | return true; | |
3468 | } else { | |
3469 | /* It would probably be better to reply with an error but there doesn't | |
3470 | * seem to be any appropriate value, so that might just be | |
3471 | * confusing. */ | |
3472 | VLOG_WARN_RL(&rl, "controller asked for invalid table %"PRIu8, | |
3473 | table_id); | |
3474 | return false; | |
3475 | } | |
064af421 BP |
3476 | } |
3477 | ||
3478 | static int | |
d1e2cf21 | 3479 | handle_flow_stats_request(struct ofconn *ofconn, const struct ofp_header *oh) |
064af421 | 3480 | { |
d1e2cf21 | 3481 | const struct ofp_flow_stats_request *fsr = ofputil_stats_body(oh); |
5ecc9d81 | 3482 | struct ofpbuf *reply; |
064af421 | 3483 | |
064af421 | 3484 | COVERAGE_INC(ofproto_flows_req); |
d1e2cf21 | 3485 | reply = start_ofp_stats_reply(oh, 1024); |
3c4486a5 | 3486 | if (is_valid_table(fsr->table_id)) { |
5ecc9d81 | 3487 | struct cls_cursor cursor; |
3c4486a5 | 3488 | struct cls_rule target; |
5ecc9d81 | 3489 | struct rule *rule; |
3c4486a5 | 3490 | |
d8ae4d67 BP |
3491 | ofputil_cls_rule_from_match(&fsr->match, 0, NXFF_OPENFLOW10, 0, |
3492 | &target); | |
5ecc9d81 BP |
3493 | cls_cursor_init(&cursor, &ofconn->ofproto->cls, &target); |
3494 | CLS_CURSOR_FOR_EACH (rule, cr, &cursor) { | |
3495 | put_ofp_flow_stats(ofconn, rule, fsr->out_port, &reply); | |
3496 | } | |
3c4486a5 | 3497 | } |
5ecc9d81 | 3498 | queue_tx(reply, ofconn, ofconn->reply_counter); |
3c4486a5 | 3499 | |
064af421 BP |
3500 | return 0; |
3501 | } | |
3502 | ||
09246b99 | 3503 | static void |
5ecc9d81 BP |
3504 | put_nx_flow_stats(struct ofconn *ofconn, struct rule *rule, |
3505 | ovs_be16 out_port, struct ofpbuf **replyp) | |
09246b99 | 3506 | { |
09246b99 BP |
3507 | struct nx_flow_stats *nfs; |
3508 | uint64_t packet_count, byte_count; | |
3509 | size_t act_len, start_len; | |
5ecc9d81 | 3510 | struct ofpbuf *reply; |
09246b99 | 3511 | |
5ecc9d81 | 3512 | if (rule_is_hidden(rule) || !rule_has_out_port(rule, out_port)) { |
09246b99 BP |
3513 | return; |
3514 | } | |
3515 | ||
3394b5b6 | 3516 | rule_get_stats(rule, &packet_count, &byte_count); |
09246b99 BP |
3517 | |
3518 | act_len = sizeof *rule->actions * rule->n_actions; | |
3519 | ||
5ecc9d81 | 3520 | append_nxstats_reply(sizeof *nfs + NXM_MAX_LEN + act_len, ofconn, replyp); |
1dfee98d | 3521 | start_len = (*replyp)->size; |
5ecc9d81 BP |
3522 | reply = *replyp; |
3523 | ||
3524 | nfs = ofpbuf_put_uninit(reply, sizeof *nfs); | |
09246b99 BP |
3525 | nfs->table_id = 0; |
3526 | nfs->pad = 0; | |
3527 | calc_flow_duration(rule->created, &nfs->duration_sec, &nfs->duration_nsec); | |
3528 | nfs->cookie = rule->flow_cookie; | |
3529 | nfs->priority = htons(rule->cr.priority); | |
3530 | nfs->idle_timeout = htons(rule->idle_timeout); | |
3531 | nfs->hard_timeout = htons(rule->hard_timeout); | |
5ecc9d81 | 3532 | nfs->match_len = htons(nx_put_match(reply, &rule->cr)); |
09246b99 BP |
3533 | memset(nfs->pad2, 0, sizeof nfs->pad2); |
3534 | nfs->packet_count = htonll(packet_count); | |
3535 | nfs->byte_count = htonll(byte_count); | |
3536 | if (rule->n_actions > 0) { | |
5ecc9d81 | 3537 | ofpbuf_put(reply, rule->actions, act_len); |
09246b99 | 3538 | } |
5ecc9d81 | 3539 | nfs->length = htons(reply->size - start_len); |
09246b99 BP |
3540 | } |
3541 | ||
3542 | static int | |
d1e2cf21 | 3543 | handle_nxst_flow(struct ofconn *ofconn, const struct ofp_header *oh) |
09246b99 BP |
3544 | { |
3545 | struct nx_flow_stats_request *nfsr; | |
09246b99 | 3546 | struct cls_rule target; |
5ecc9d81 | 3547 | struct ofpbuf *reply; |
d1e2cf21 | 3548 | struct ofpbuf b; |
09246b99 BP |
3549 | int error; |
3550 | ||
0bc9407d | 3551 | ofpbuf_use_const(&b, oh, ntohs(oh->length)); |
d1e2cf21 | 3552 | |
09246b99 | 3553 | /* Dissect the message. */ |
bbc32a88 | 3554 | nfsr = ofpbuf_pull(&b, sizeof *nfsr); |
d1e2cf21 | 3555 | error = nx_pull_match(&b, ntohs(nfsr->match_len), 0, &target); |
09246b99 BP |
3556 | if (error) { |
3557 | return error; | |
3558 | } | |
d1e2cf21 BP |
3559 | if (b.size) { |
3560 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); | |
3561 | } | |
09246b99 BP |
3562 | |
3563 | COVERAGE_INC(ofproto_flows_req); | |
5ecc9d81 | 3564 | reply = start_nxstats_reply(&nfsr->nsm, 1024); |
3c4486a5 | 3565 | if (is_valid_table(nfsr->table_id)) { |
5ecc9d81 BP |
3566 | struct cls_cursor cursor; |
3567 | struct rule *rule; | |
3568 | ||
3569 | cls_cursor_init(&cursor, &ofconn->ofproto->cls, &target); | |
3570 | CLS_CURSOR_FOR_EACH (rule, cr, &cursor) { | |
3571 | put_nx_flow_stats(ofconn, rule, nfsr->out_port, &reply); | |
3572 | } | |
3c4486a5 | 3573 | } |
5ecc9d81 BP |
3574 | queue_tx(reply, ofconn, ofconn->reply_counter); |
3575 | ||
09246b99 BP |
3576 | return 0; |
3577 | } | |
3578 | ||
4f2cad2c | 3579 | static void |
3394b5b6 | 3580 | flow_stats_ds(struct rule *rule, struct ds *results) |
4f2cad2c | 3581 | { |
4f2cad2c JP |
3582 | uint64_t packet_count, byte_count; |
3583 | size_t act_len = sizeof *rule->actions * rule->n_actions; | |
3584 | ||
3394b5b6 | 3585 | rule_get_stats(rule, &packet_count, &byte_count); |
4f2cad2c JP |
3586 | |
3587 | ds_put_format(results, "duration=%llds, ", | |
3588 | (time_msec() - rule->created) / 1000); | |
3693a563 | 3589 | ds_put_format(results, "idle=%.3fs, ", (time_msec() - rule->used) / 1000.0); |
52ae00b3 | 3590 | ds_put_format(results, "priority=%u, ", rule->cr.priority); |
4f2cad2c JP |
3591 | ds_put_format(results, "n_packets=%"PRIu64", ", packet_count); |
3592 | ds_put_format(results, "n_bytes=%"PRIu64", ", byte_count); | |
cb833cf6 | 3593 | cls_rule_format(&rule->cr, results); |
a5df0e72 | 3594 | ds_put_char(results, ','); |
3dffcf07 BP |
3595 | if (act_len > 0) { |
3596 | ofp_print_actions(results, &rule->actions->header, act_len); | |
3c8552c1 JP |
3597 | } else { |
3598 | ds_put_cstr(results, "drop"); | |
3dffcf07 | 3599 | } |
4f2cad2c JP |
3600 | ds_put_cstr(results, "\n"); |
3601 | } | |
3602 | ||
d295e8e9 | 3603 | /* Adds a pretty-printed description of all flows to 'results', including |
ee8b231c | 3604 | * hidden flows (e.g., set up by in-band control). */ |
4f2cad2c JP |
3605 | void |
3606 | ofproto_get_all_flows(struct ofproto *p, struct ds *results) | |
3607 | { | |
5ecc9d81 BP |
3608 | struct cls_cursor cursor; |
3609 | struct rule *rule; | |
064af421 | 3610 | |
5ecc9d81 BP |
3611 | cls_cursor_init(&cursor, &p->cls, NULL); |
3612 | CLS_CURSOR_FOR_EACH (rule, cr, &cursor) { | |
3394b5b6 | 3613 | flow_stats_ds(rule, results); |
064af421 | 3614 | } |
064af421 BP |
3615 | } |
3616 | ||
27d34fce BP |
3617 | static void |
3618 | query_aggregate_stats(struct ofproto *ofproto, struct cls_rule *target, | |
734bbeb4 | 3619 | ovs_be16 out_port, uint8_t table_id, |
27d34fce BP |
3620 | struct ofp_aggregate_stats_reply *oasr) |
3621 | { | |
5ecc9d81 BP |
3622 | uint64_t total_packets = 0; |
3623 | uint64_t total_bytes = 0; | |
3624 | int n_flows = 0; | |
27d34fce BP |
3625 | |
3626 | COVERAGE_INC(ofproto_agg_request); | |
5ecc9d81 | 3627 | |
3c4486a5 | 3628 | if (is_valid_table(table_id)) { |
5ecc9d81 BP |
3629 | struct cls_cursor cursor; |
3630 | struct rule *rule; | |
3c4486a5 | 3631 | |
5ecc9d81 BP |
3632 | cls_cursor_init(&cursor, &ofproto->cls, target); |
3633 | CLS_CURSOR_FOR_EACH (rule, cr, &cursor) { | |
3634 | if (!rule_is_hidden(rule) && rule_has_out_port(rule, out_port)) { | |
3635 | uint64_t packet_count; | |
3636 | uint64_t byte_count; | |
3637 | ||
3394b5b6 | 3638 | rule_get_stats(rule, &packet_count, &byte_count); |
5ecc9d81 BP |
3639 | |
3640 | total_packets += packet_count; | |
3641 | total_bytes += byte_count; | |
3642 | n_flows++; | |
3643 | } | |
3644 | } | |
3c4486a5 | 3645 | } |
27d34fce | 3646 | |
5ecc9d81 | 3647 | oasr->flow_count = htonl(n_flows); |
c4617b3c BP |
3648 | put_32aligned_be64(&oasr->packet_count, htonll(total_packets)); |
3649 | put_32aligned_be64(&oasr->byte_count, htonll(total_bytes)); | |
27d34fce BP |
3650 | memset(oasr->pad, 0, sizeof oasr->pad); |
3651 | } | |
3652 | ||
064af421 | 3653 | static int |
3269c562 | 3654 | handle_aggregate_stats_request(struct ofconn *ofconn, |
d1e2cf21 | 3655 | const struct ofp_header *oh) |
064af421 | 3656 | { |
d1e2cf21 | 3657 | const struct ofp_aggregate_stats_request *request = ofputil_stats_body(oh); |
064af421 | 3658 | struct ofp_aggregate_stats_reply *reply; |
064af421 BP |
3659 | struct cls_rule target; |
3660 | struct ofpbuf *msg; | |
3661 | ||
d8ae4d67 BP |
3662 | ofputil_cls_rule_from_match(&request->match, 0, NXFF_OPENFLOW10, 0, |
3663 | &target); | |
064af421 | 3664 | |
d1e2cf21 | 3665 | msg = start_ofp_stats_reply(oh, sizeof *reply); |
06a5e131 | 3666 | reply = append_ofp_stats_reply(sizeof *reply, ofconn, &msg); |
27d34fce BP |
3667 | query_aggregate_stats(ofconn->ofproto, &target, request->out_port, |
3668 | request->table_id, reply); | |
064af421 BP |
3669 | queue_tx(msg, ofconn, ofconn->reply_counter); |
3670 | return 0; | |
3671 | } | |
3672 | ||
09246b99 | 3673 | static int |
d1e2cf21 | 3674 | handle_nxst_aggregate(struct ofconn *ofconn, const struct ofp_header *oh) |
09246b99 BP |
3675 | { |
3676 | struct nx_aggregate_stats_request *request; | |
3677 | struct ofp_aggregate_stats_reply *reply; | |
3678 | struct cls_rule target; | |
d1e2cf21 | 3679 | struct ofpbuf b; |
09246b99 BP |
3680 | struct ofpbuf *buf; |
3681 | int error; | |
3682 | ||
0bc9407d | 3683 | ofpbuf_use_const(&b, oh, ntohs(oh->length)); |
d1e2cf21 | 3684 | |
09246b99 | 3685 | /* Dissect the message. */ |
bbc32a88 | 3686 | request = ofpbuf_pull(&b, sizeof *request); |
d1e2cf21 | 3687 | error = nx_pull_match(&b, ntohs(request->match_len), 0, &target); |
09246b99 BP |
3688 | if (error) { |
3689 | return error; | |
3690 | } | |
d1e2cf21 BP |
3691 | if (b.size) { |
3692 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); | |
3693 | } | |
09246b99 BP |
3694 | |
3695 | /* Reply. */ | |
3696 | COVERAGE_INC(ofproto_flows_req); | |
3697 | buf = start_nxstats_reply(&request->nsm, sizeof *reply); | |
3698 | reply = ofpbuf_put_uninit(buf, sizeof *reply); | |
3699 | query_aggregate_stats(ofconn->ofproto, &target, request->out_port, | |
3700 | request->table_id, reply); | |
3701 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
3702 | ||
3703 | return 0; | |
3704 | } | |
3705 | ||
c1c9c9c4 BP |
3706 | struct queue_stats_cbdata { |
3707 | struct ofconn *ofconn; | |
ca0f572c | 3708 | struct ofport *ofport; |
c1c9c9c4 | 3709 | struct ofpbuf *msg; |
c1c9c9c4 BP |
3710 | }; |
3711 | ||
3712 | static void | |
db9220c3 | 3713 | put_queue_stats(struct queue_stats_cbdata *cbdata, uint32_t queue_id, |
c1c9c9c4 BP |
3714 | const struct netdev_queue_stats *stats) |
3715 | { | |
3716 | struct ofp_queue_stats *reply; | |
3717 | ||
06a5e131 | 3718 | reply = append_ofp_stats_reply(sizeof *reply, cbdata->ofconn, &cbdata->msg); |
ca0f572c | 3719 | reply->port_no = htons(cbdata->ofport->opp.port_no); |
c1c9c9c4 BP |
3720 | memset(reply->pad, 0, sizeof reply->pad); |
3721 | reply->queue_id = htonl(queue_id); | |
c4617b3c BP |
3722 | put_32aligned_be64(&reply->tx_bytes, htonll(stats->tx_bytes)); |
3723 | put_32aligned_be64(&reply->tx_packets, htonll(stats->tx_packets)); | |
3724 | put_32aligned_be64(&reply->tx_errors, htonll(stats->tx_errors)); | |
c1c9c9c4 BP |
3725 | } |
3726 | ||
3727 | static void | |
db9220c3 | 3728 | handle_queue_stats_dump_cb(uint32_t queue_id, |
c1c9c9c4 BP |
3729 | struct netdev_queue_stats *stats, |
3730 | void *cbdata_) | |
3731 | { | |
3732 | struct queue_stats_cbdata *cbdata = cbdata_; | |
3733 | ||
3734 | put_queue_stats(cbdata, queue_id, stats); | |
3735 | } | |
3736 | ||
3737 | static void | |
ca0f572c | 3738 | handle_queue_stats_for_port(struct ofport *port, uint32_t queue_id, |
c1c9c9c4 BP |
3739 | struct queue_stats_cbdata *cbdata) |
3740 | { | |
ca0f572c | 3741 | cbdata->ofport = port; |
c1c9c9c4 BP |
3742 | if (queue_id == OFPQ_ALL) { |
3743 | netdev_dump_queue_stats(port->netdev, | |
3744 | handle_queue_stats_dump_cb, cbdata); | |
3745 | } else { | |
3746 | struct netdev_queue_stats stats; | |
3747 | ||
1ac788f6 BP |
3748 | if (!netdev_get_queue_stats(port->netdev, queue_id, &stats)) { |
3749 | put_queue_stats(cbdata, queue_id, &stats); | |
3750 | } | |
c1c9c9c4 BP |
3751 | } |
3752 | } | |
3753 | ||
3754 | static int | |
d1e2cf21 | 3755 | handle_queue_stats_request(struct ofconn *ofconn, const struct ofp_header *oh) |
c1c9c9c4 | 3756 | { |
3269c562 | 3757 | struct ofproto *ofproto = ofconn->ofproto; |
d1e2cf21 | 3758 | const struct ofp_queue_stats_request *qsr; |
c1c9c9c4 BP |
3759 | struct queue_stats_cbdata cbdata; |
3760 | struct ofport *port; | |
3761 | unsigned int port_no; | |
3762 | uint32_t queue_id; | |
3763 | ||
d1e2cf21 BP |
3764 | qsr = ofputil_stats_body(oh); |
3765 | if (!qsr) { | |
c1c9c9c4 BP |
3766 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); |
3767 | } | |
c1c9c9c4 BP |
3768 | |
3769 | COVERAGE_INC(ofproto_queue_req); | |
3770 | ||
3771 | cbdata.ofconn = ofconn; | |
d1e2cf21 | 3772 | cbdata.msg = start_ofp_stats_reply(oh, 128); |
c1c9c9c4 BP |
3773 | |
3774 | port_no = ntohs(qsr->port_no); | |
3775 | queue_id = ntohl(qsr->queue_id); | |
3776 | if (port_no == OFPP_ALL) { | |
4e8e4213 | 3777 | HMAP_FOR_EACH (port, hmap_node, &ofproto->ports) { |
ca0f572c | 3778 | handle_queue_stats_for_port(port, queue_id, &cbdata); |
c1c9c9c4 BP |
3779 | } |
3780 | } else if (port_no < ofproto->max_ports) { | |
ca0f572c | 3781 | port = get_port(ofproto, ofp_port_to_odp_port(port_no)); |
c1c9c9c4 | 3782 | if (port) { |
ca0f572c | 3783 | handle_queue_stats_for_port(port, queue_id, &cbdata); |
c1c9c9c4 BP |
3784 | } |
3785 | } else { | |
3786 | ofpbuf_delete(cbdata.msg); | |
3787 | return ofp_mkerr(OFPET_QUEUE_OP_FAILED, OFPQOFC_BAD_PORT); | |
3788 | } | |
3789 | queue_tx(cbdata.msg, ofconn, ofconn->reply_counter); | |
3790 | ||
3791 | return 0; | |
3792 | } | |
3793 | ||
878ae780 EJ |
3794 | /* Updates 'facet''s used time. Caller is responsible for calling |
3795 | * facet_push_stats() to update the flows which 'facet' resubmits into. */ | |
064af421 | 3796 | static void |
bcf84111 | 3797 | facet_update_time(struct ofproto *ofproto, struct facet *facet, |
3394b5b6 | 3798 | long long int used) |
064af421 | 3799 | { |
bcf84111 BP |
3800 | if (used > facet->used) { |
3801 | facet->used = used; | |
3802 | if (used > facet->rule->used) { | |
3803 | facet->rule->used = used; | |
4836f9f2 | 3804 | } |
bcf84111 | 3805 | netflow_flow_update_time(ofproto->netflow, &facet->nf_flow, used); |
064af421 BP |
3806 | } |
3807 | } | |
3808 | ||
bcf84111 BP |
3809 | /* Folds the statistics from 'stats' into the counters in 'facet'. |
3810 | * | |
3811 | * Because of the meaning of a facet's counters, it only makes sense to do this | |
3812 | * if 'stats' are not tracked in the datapath, that is, if 'stats' represents a | |
3813 | * packet that was sent by hand or if it represents statistics that have been | |
3814 | * cleared out of the datapath. */ | |
064af421 | 3815 | static void |
bcf84111 | 3816 | facet_update_stats(struct ofproto *ofproto, struct facet *facet, |
c97fb132 | 3817 | const struct dpif_flow_stats *stats) |
064af421 | 3818 | { |
0b13821f | 3819 | if (stats->n_packets || stats->used > facet->used) { |
3394b5b6 | 3820 | facet_update_time(ofproto, facet, stats->used); |
bcf84111 BP |
3821 | facet->packet_count += stats->n_packets; |
3822 | facet->byte_count += stats->n_bytes; | |
878ae780 | 3823 | facet_push_stats(ofproto, facet); |
bcf84111 | 3824 | netflow_flow_update_flags(&facet->nf_flow, stats->tcp_flags); |
064af421 BP |
3825 | } |
3826 | } | |
3827 | ||
878ae780 EJ |
3828 | static void |
3829 | facet_push_stats(struct ofproto *ofproto, struct facet *facet) | |
3830 | { | |
3831 | uint64_t rs_packets, rs_bytes; | |
3832 | ||
3833 | assert(facet->packet_count >= facet->rs_packet_count); | |
3834 | assert(facet->byte_count >= facet->rs_byte_count); | |
3835 | assert(facet->used >= facet->rs_used); | |
3836 | ||
3837 | rs_packets = facet->packet_count - facet->rs_packet_count; | |
3838 | rs_bytes = facet->byte_count - facet->rs_byte_count; | |
3839 | ||
3840 | if (rs_packets || rs_bytes || facet->used > facet->rs_used) { | |
3841 | facet->rs_packet_count = facet->packet_count; | |
3842 | facet->rs_byte_count = facet->byte_count; | |
3843 | facet->rs_used = facet->used; | |
3844 | ||
3845 | flow_push_stats(ofproto, facet->rule, &facet->flow, | |
3846 | rs_packets, rs_bytes, facet->used); | |
3847 | } | |
3848 | } | |
3849 | ||
3850 | struct ofproto_push { | |
3851 | struct action_xlate_ctx ctx; | |
3852 | uint64_t packets; | |
3853 | uint64_t bytes; | |
3854 | long long int used; | |
3855 | }; | |
3856 | ||
3857 | static void | |
3858 | push_resubmit(struct action_xlate_ctx *ctx, struct rule *rule) | |
3859 | { | |
3860 | struct ofproto_push *push = CONTAINER_OF(ctx, struct ofproto_push, ctx); | |
3861 | ||
3862 | if (rule) { | |
3863 | rule->packet_count += push->packets; | |
3864 | rule->byte_count += push->bytes; | |
3865 | rule->used = MAX(push->used, rule->used); | |
3866 | } | |
3867 | } | |
3868 | ||
3869 | /* Pushes flow statistics to the rules which 'flow' resubmits into given | |
3870 | * 'rule''s actions. */ | |
3871 | static void | |
3872 | flow_push_stats(struct ofproto *ofproto, const struct rule *rule, | |
3873 | struct flow *flow, uint64_t packets, uint64_t bytes, | |
3874 | long long int used) | |
3875 | { | |
3876 | struct ofproto_push push; | |
3877 | ||
3878 | push.packets = packets; | |
3879 | push.bytes = bytes; | |
3880 | push.used = used; | |
3881 | ||
3882 | action_xlate_ctx_init(&push.ctx, ofproto, flow, NULL); | |
3883 | push.ctx.resubmit_hook = push_resubmit; | |
3884 | ofpbuf_delete(xlate_actions(&push.ctx, rule->actions, rule->n_actions)); | |
3885 | } | |
3886 | ||
79eee1eb BP |
3887 | /* Implements OFPFC_ADD and the cases for OFPFC_MODIFY and OFPFC_MODIFY_STRICT |
3888 | * in which no matching flow already exists in the flow table. | |
3889 | * | |
3890 | * Adds the flow specified by 'ofm', which is followed by 'n_actions' | |
3269c562 BP |
3891 | * ofp_actions, to ofconn->ofproto's flow table. Returns 0 on success or an |
3892 | * OpenFlow error code as encoded by ofp_mkerr() on failure. | |
79eee1eb BP |
3893 | * |
3894 | * 'ofconn' is used to retrieve the packet buffer specified in ofm->buffer_id, | |
3895 | * if any. */ | |
064af421 | 3896 | static int |
3052b0c5 | 3897 | add_flow(struct ofconn *ofconn, struct flow_mod *fm) |
064af421 | 3898 | { |
3269c562 | 3899 | struct ofproto *p = ofconn->ofproto; |
064af421 BP |
3900 | struct ofpbuf *packet; |
3901 | struct rule *rule; | |
3902 | uint16_t in_port; | |
3903 | int error; | |
3904 | ||
3052b0c5 BP |
3905 | if (fm->flags & OFPFF_CHECK_OVERLAP |
3906 | && classifier_rule_overlaps(&p->cls, &fm->cr)) { | |
3907 | return ofp_mkerr(OFPET_FLOW_MOD_FAILED, OFPFMFC_OVERLAP); | |
49bdc010 JP |
3908 | } |
3909 | ||
064af421 | 3910 | error = 0; |
3052b0c5 BP |
3911 | if (fm->buffer_id != UINT32_MAX) { |
3912 | error = pktbuf_retrieve(ofconn->pktbuf, fm->buffer_id, | |
064af421 | 3913 | &packet, &in_port); |
212fe71c BP |
3914 | } else { |
3915 | packet = NULL; | |
165cd8a3 | 3916 | in_port = UINT16_MAX; |
064af421 BP |
3917 | } |
3918 | ||
bcf84111 BP |
3919 | rule = rule_create(&fm->cr, fm->actions, fm->n_actions, |
3920 | fm->idle_timeout, fm->hard_timeout, fm->cookie, | |
3921 | fm->flags & OFPFF_SEND_FLOW_REM); | |
afe75089 BP |
3922 | rule_insert(p, rule); |
3923 | if (packet) { | |
3924 | rule_execute(p, rule, in_port, packet); | |
3925 | } | |
064af421 BP |
3926 | return error; |
3927 | } | |
3928 | ||
79eee1eb | 3929 | static struct rule * |
3052b0c5 | 3930 | find_flow_strict(struct ofproto *p, const struct flow_mod *fm) |
064af421 | 3931 | { |
3052b0c5 | 3932 | return rule_from_cls_rule(classifier_find_rule_exactly(&p->cls, &fm->cr)); |
79eee1eb | 3933 | } |
064af421 | 3934 | |
79eee1eb | 3935 | static int |
3269c562 | 3936 | send_buffered_packet(struct ofconn *ofconn, |
3052b0c5 | 3937 | struct rule *rule, uint32_t buffer_id) |
79eee1eb BP |
3938 | { |
3939 | struct ofpbuf *packet; | |
3940 | uint16_t in_port; | |
79eee1eb | 3941 | int error; |
064af421 | 3942 | |
3052b0c5 | 3943 | if (buffer_id == UINT32_MAX) { |
79eee1eb | 3944 | return 0; |
064af421 | 3945 | } |
79eee1eb | 3946 | |
3052b0c5 | 3947 | error = pktbuf_retrieve(ofconn->pktbuf, buffer_id, &packet, &in_port); |
79eee1eb BP |
3948 | if (error) { |
3949 | return error; | |
3950 | } | |
3951 | ||
bcf84111 | 3952 | rule_execute(ofconn->ofproto, rule, in_port, packet); |
79eee1eb | 3953 | |
064af421 BP |
3954 | return 0; |
3955 | } | |
79eee1eb BP |
3956 | \f |
3957 | /* OFPFC_MODIFY and OFPFC_MODIFY_STRICT. */ | |
064af421 BP |
3958 | |
3959 | struct modify_flows_cbdata { | |
3960 | struct ofproto *ofproto; | |
3052b0c5 | 3961 | const struct flow_mod *fm; |
79eee1eb | 3962 | struct rule *match; |
064af421 BP |
3963 | }; |
3964 | ||
3052b0c5 BP |
3965 | static int modify_flow(struct ofproto *, const struct flow_mod *, |
3966 | struct rule *); | |
79eee1eb BP |
3967 | |
3968 | /* Implements OFPFC_MODIFY. Returns 0 on success or an OpenFlow error code as | |
3969 | * encoded by ofp_mkerr() on failure. | |
3970 | * | |
3971 | * 'ofconn' is used to retrieve the packet buffer specified in ofm->buffer_id, | |
3972 | * if any. */ | |
3973 | static int | |
3052b0c5 | 3974 | modify_flows_loose(struct ofconn *ofconn, struct flow_mod *fm) |
79eee1eb | 3975 | { |
5ecc9d81 BP |
3976 | struct ofproto *p = ofconn->ofproto; |
3977 | struct rule *match = NULL; | |
3978 | struct cls_cursor cursor; | |
3979 | struct rule *rule; | |
79eee1eb | 3980 | |
5ecc9d81 BP |
3981 | cls_cursor_init(&cursor, &p->cls, &fm->cr); |
3982 | CLS_CURSOR_FOR_EACH (rule, cr, &cursor) { | |
3983 | if (!rule_is_hidden(rule)) { | |
3984 | match = rule; | |
3985 | modify_flow(p, fm, rule); | |
3986 | } | |
3987 | } | |
79eee1eb | 3988 | |
5ecc9d81 | 3989 | if (match) { |
d6302b0f BP |
3990 | /* This credits the packet to whichever flow happened to match last. |
3991 | * That's weird. Maybe we should do a lookup for the flow that | |
3992 | * actually matches the packet? Who knows. */ | |
5ecc9d81 | 3993 | send_buffered_packet(ofconn, match, fm->buffer_id); |
79eee1eb BP |
3994 | return 0; |
3995 | } else { | |
3052b0c5 | 3996 | return add_flow(ofconn, fm); |
79eee1eb BP |
3997 | } |
3998 | } | |
3999 | ||
4000 | /* Implements OFPFC_MODIFY_STRICT. Returns 0 on success or an OpenFlow error | |
4001 | * code as encoded by ofp_mkerr() on failure. | |
4002 | * | |
4003 | * 'ofconn' is used to retrieve the packet buffer specified in ofm->buffer_id, | |
4004 | * if any. */ | |
4005 | static int | |
3052b0c5 | 4006 | modify_flow_strict(struct ofconn *ofconn, struct flow_mod *fm) |
79eee1eb | 4007 | { |
3052b0c5 BP |
4008 | struct ofproto *p = ofconn->ofproto; |
4009 | struct rule *rule = find_flow_strict(p, fm); | |
79eee1eb | 4010 | if (rule && !rule_is_hidden(rule)) { |
3052b0c5 BP |
4011 | modify_flow(p, fm, rule); |
4012 | return send_buffered_packet(ofconn, rule, fm->buffer_id); | |
79eee1eb | 4013 | } else { |
3052b0c5 | 4014 | return add_flow(ofconn, fm); |
79eee1eb BP |
4015 | } |
4016 | } | |
4017 | ||
79eee1eb BP |
4018 | /* Implements core of OFPFC_MODIFY and OFPFC_MODIFY_STRICT where 'rule' has |
4019 | * been identified as a flow in 'p''s flow table to be modified, by changing | |
4020 | * the rule's actions to match those in 'ofm' (which is followed by 'n_actions' | |
4021 | * ofp_action[] structures). */ | |
064af421 | 4022 | static int |
3052b0c5 | 4023 | modify_flow(struct ofproto *p, const struct flow_mod *fm, struct rule *rule) |
064af421 | 4024 | { |
3052b0c5 | 4025 | size_t actions_len = fm->n_actions * sizeof *rule->actions; |
79eee1eb | 4026 | |
3052b0c5 | 4027 | rule->flow_cookie = fm->cookie; |
79eee1eb BP |
4028 | |
4029 | /* If the actions are the same, do nothing. */ | |
3052b0c5 BP |
4030 | if (fm->n_actions == rule->n_actions |
4031 | && (!fm->n_actions | |
4032 | || !memcmp(fm->actions, rule->actions, actions_len))) { | |
79eee1eb BP |
4033 | return 0; |
4034 | } | |
4035 | ||
4036 | /* Replace actions. */ | |
4037 | free(rule->actions); | |
3052b0c5 BP |
4038 | rule->actions = fm->n_actions ? xmemdup(fm->actions, actions_len) : NULL; |
4039 | rule->n_actions = fm->n_actions; | |
79eee1eb | 4040 | |
bcf84111 | 4041 | p->need_revalidate = true; |
79eee1eb BP |
4042 | |
4043 | return 0; | |
4044 | } | |
4045 | \f | |
4046 | /* OFPFC_DELETE implementation. */ | |
4047 | ||
8054fc48 | 4048 | static void delete_flow(struct ofproto *, struct rule *, ovs_be16 out_port); |
79eee1eb BP |
4049 | |
4050 | /* Implements OFPFC_DELETE. */ | |
4051 | static void | |
3052b0c5 | 4052 | delete_flows_loose(struct ofproto *p, const struct flow_mod *fm) |
79eee1eb | 4053 | { |
5ecc9d81 BP |
4054 | struct rule *rule, *next_rule; |
4055 | struct cls_cursor cursor; | |
064af421 | 4056 | |
5ecc9d81 BP |
4057 | cls_cursor_init(&cursor, &p->cls, &fm->cr); |
4058 | CLS_CURSOR_FOR_EACH_SAFE (rule, next_rule, cr, &cursor) { | |
4059 | delete_flow(p, rule, htons(fm->out_port)); | |
4060 | } | |
064af421 BP |
4061 | } |
4062 | ||
79eee1eb BP |
4063 | /* Implements OFPFC_DELETE_STRICT. */ |
4064 | static void | |
3052b0c5 | 4065 | delete_flow_strict(struct ofproto *p, struct flow_mod *fm) |
79eee1eb | 4066 | { |
3052b0c5 | 4067 | struct rule *rule = find_flow_strict(p, fm); |
79eee1eb | 4068 | if (rule) { |
3052b0c5 | 4069 | delete_flow(p, rule, htons(fm->out_port)); |
79eee1eb BP |
4070 | } |
4071 | } | |
4072 | ||
79eee1eb BP |
4073 | /* Implements core of OFPFC_DELETE and OFPFC_DELETE_STRICT where 'rule' has |
4074 | * been identified as a flow to delete from 'p''s flow table, by deleting the | |
4075 | * flow and sending out a OFPT_FLOW_REMOVED message to any interested | |
4076 | * controller. | |
4077 | * | |
4078 | * Will not delete 'rule' if it is hidden. Will delete 'rule' only if | |
4079 | * 'out_port' is htons(OFPP_NONE) or if 'rule' actually outputs to the | |
4080 | * specified 'out_port'. */ | |
4081 | static void | |
8054fc48 | 4082 | delete_flow(struct ofproto *p, struct rule *rule, ovs_be16 out_port) |
79eee1eb BP |
4083 | { |
4084 | if (rule_is_hidden(rule)) { | |
4085 | return; | |
4086 | } | |
4087 | ||
4088 | if (out_port != htons(OFPP_NONE) && !rule_has_out_port(rule, out_port)) { | |
4089 | return; | |
4090 | } | |
4091 | ||
bcf84111 | 4092 | rule_send_removed(p, rule, OFPRR_DELETE); |
79eee1eb BP |
4093 | rule_remove(p, rule); |
4094 | } | |
4095 | \f | |
064af421 | 4096 | static int |
2e4f5fcf | 4097 | handle_flow_mod(struct ofconn *ofconn, const struct ofp_header *oh) |
064af421 | 4098 | { |
3052b0c5 | 4099 | struct ofproto *p = ofconn->ofproto; |
2e4f5fcf | 4100 | struct flow_mod fm; |
064af421 BP |
4101 | int error; |
4102 | ||
3052b0c5 | 4103 | error = reject_slave_controller(ofconn, "flow_mod"); |
9deba63b BP |
4104 | if (error) { |
4105 | return error; | |
4106 | } | |
3052b0c5 | 4107 | |
2e4f5fcf | 4108 | error = ofputil_decode_flow_mod(&fm, oh, ofconn->flow_format); |
064af421 BP |
4109 | if (error) { |
4110 | return error; | |
4111 | } | |
4112 | ||
2e4f5fcf BP |
4113 | /* We do not support the emergency flow cache. It will hopefully get |
4114 | * dropped from OpenFlow in the near future. */ | |
4115 | if (fm.flags & OFPFF_EMERG) { | |
49bdc010 JP |
4116 | /* There isn't a good fit for an error code, so just state that the |
4117 | * flow table is full. */ | |
4118 | return ofp_mkerr(OFPET_FLOW_MOD_FAILED, OFPFMFC_ALL_TABLES_FULL); | |
4119 | } | |
4120 | ||
2e4f5fcf BP |
4121 | error = validate_actions(fm.actions, fm.n_actions, |
4122 | &fm.cr.flow, p->max_ports); | |
4123 | if (error) { | |
4124 | return error; | |
4125 | } | |
4126 | ||
4127 | switch (fm.command) { | |
3052b0c5 | 4128 | case OFPFC_ADD: |
2e4f5fcf | 4129 | return add_flow(ofconn, &fm); |
3052b0c5 BP |
4130 | |
4131 | case OFPFC_MODIFY: | |
2e4f5fcf | 4132 | return modify_flows_loose(ofconn, &fm); |
3052b0c5 BP |
4133 | |
4134 | case OFPFC_MODIFY_STRICT: | |
2e4f5fcf | 4135 | return modify_flow_strict(ofconn, &fm); |
3052b0c5 BP |
4136 | |
4137 | case OFPFC_DELETE: | |
2e4f5fcf | 4138 | delete_flows_loose(p, &fm); |
3052b0c5 BP |
4139 | return 0; |
4140 | ||
4141 | case OFPFC_DELETE_STRICT: | |
2e4f5fcf | 4142 | delete_flow_strict(p, &fm); |
3052b0c5 BP |
4143 | return 0; |
4144 | ||
4145 | default: | |
4146 | return ofp_mkerr(OFPET_FLOW_MOD_FAILED, OFPFMFC_BAD_COMMAND); | |
4147 | } | |
4148 | } | |
4149 | ||
659586ef | 4150 | static int |
d1e2cf21 | 4151 | handle_tun_id_from_cookie(struct ofconn *ofconn, const struct ofp_header *oh) |
659586ef | 4152 | { |
d1e2cf21 BP |
4153 | const struct nxt_tun_id_cookie *msg |
4154 | = (const struct nxt_tun_id_cookie *) oh; | |
659586ef | 4155 | |
b70eac89 | 4156 | ofconn->flow_format = msg->set ? NXFF_TUN_ID_FROM_COOKIE : NXFF_OPENFLOW10; |
659586ef JG |
4157 | return 0; |
4158 | } | |
4159 | ||
9deba63b | 4160 | static int |
d1e2cf21 | 4161 | handle_role_request(struct ofconn *ofconn, const struct ofp_header *oh) |
9deba63b | 4162 | { |
d1e2cf21 | 4163 | struct nx_role_request *nrr = (struct nx_role_request *) oh; |
9deba63b BP |
4164 | struct nx_role_request *reply; |
4165 | struct ofpbuf *buf; | |
4166 | uint32_t role; | |
4167 | ||
5899143f | 4168 | if (ofconn->type != OFCONN_PRIMARY) { |
9deba63b BP |
4169 | VLOG_WARN_RL(&rl, "ignoring role request on non-controller " |
4170 | "connection"); | |
4171 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_EPERM); | |
4172 | } | |
4173 | ||
4174 | role = ntohl(nrr->role); | |
4175 | if (role != NX_ROLE_OTHER && role != NX_ROLE_MASTER | |
4176 | && role != NX_ROLE_SLAVE) { | |
4177 | VLOG_WARN_RL(&rl, "received request for unknown role %"PRIu32, role); | |
4178 | ||
4179 | /* There's no good error code for this. */ | |
4180 | return ofp_mkerr(OFPET_BAD_REQUEST, -1); | |
4181 | } | |
4182 | ||
4183 | if (role == NX_ROLE_MASTER) { | |
4184 | struct ofconn *other; | |
4185 | ||
3269c562 | 4186 | HMAP_FOR_EACH (other, hmap_node, &ofconn->ofproto->controllers) { |
9deba63b BP |
4187 | if (other->role == NX_ROLE_MASTER) { |
4188 | other->role = NX_ROLE_SLAVE; | |
4189 | } | |
4190 | } | |
4191 | } | |
4192 | ofconn->role = role; | |
4193 | ||
d1e2cf21 | 4194 | reply = make_nxmsg_xid(sizeof *reply, NXT_ROLE_REPLY, oh->xid, &buf); |
9deba63b BP |
4195 | reply->role = htonl(role); |
4196 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
4197 | ||
4198 | return 0; | |
4199 | } | |
4200 | ||
09246b99 | 4201 | static int |
d1e2cf21 | 4202 | handle_nxt_set_flow_format(struct ofconn *ofconn, const struct ofp_header *oh) |
09246b99 | 4203 | { |
d1e2cf21 BP |
4204 | const struct nxt_set_flow_format *msg |
4205 | = (const struct nxt_set_flow_format *) oh; | |
09246b99 | 4206 | uint32_t format; |
09246b99 BP |
4207 | |
4208 | format = ntohl(msg->format); | |
4209 | if (format == NXFF_OPENFLOW10 | |
4210 | || format == NXFF_TUN_ID_FROM_COOKIE | |
4211 | || format == NXFF_NXM) { | |
4212 | ofconn->flow_format = format; | |
4213 | return 0; | |
4214 | } else { | |
4215 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_EPERM); | |
4216 | } | |
4217 | } | |
4218 | ||
064af421 | 4219 | static int |
d1e2cf21 | 4220 | handle_barrier_request(struct ofconn *ofconn, const struct ofp_header *oh) |
246e61ea JP |
4221 | { |
4222 | struct ofp_header *ob; | |
4223 | struct ofpbuf *buf; | |
4224 | ||
4225 | /* Currently, everything executes synchronously, so we can just | |
4226 | * immediately send the barrier reply. */ | |
4227 | ob = make_openflow_xid(sizeof *ob, OFPT_BARRIER_REPLY, oh->xid, &buf); | |
4228 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
4229 | return 0; | |
4230 | } | |
4231 | ||
d1e2cf21 BP |
4232 | static int |
4233 | handle_openflow__(struct ofconn *ofconn, const struct ofpbuf *msg) | |
064af421 | 4234 | { |
d1e2cf21 BP |
4235 | const struct ofp_header *oh = msg->data; |
4236 | const struct ofputil_msg_type *type; | |
064af421 BP |
4237 | int error; |
4238 | ||
d1e2cf21 BP |
4239 | error = ofputil_decode_msg_type(oh, &type); |
4240 | if (error) { | |
4241 | return error; | |
4242 | } | |
064af421 | 4243 | |
d1e2cf21 BP |
4244 | switch (ofputil_msg_type_code(type)) { |
4245 | /* OpenFlow requests. */ | |
4246 | case OFPUTIL_OFPT_ECHO_REQUEST: | |
4247 | return handle_echo_request(ofconn, oh); | |
064af421 | 4248 | |
d1e2cf21 BP |
4249 | case OFPUTIL_OFPT_FEATURES_REQUEST: |
4250 | return handle_features_request(ofconn, oh); | |
064af421 | 4251 | |
d1e2cf21 BP |
4252 | case OFPUTIL_OFPT_GET_CONFIG_REQUEST: |
4253 | return handle_get_config_request(ofconn, oh); | |
064af421 | 4254 | |
d1e2cf21 BP |
4255 | case OFPUTIL_OFPT_SET_CONFIG: |
4256 | return handle_set_config(ofconn, msg->data); | |
064af421 | 4257 | |
d1e2cf21 BP |
4258 | case OFPUTIL_OFPT_PACKET_OUT: |
4259 | return handle_packet_out(ofconn, oh); | |
064af421 | 4260 | |
d1e2cf21 BP |
4261 | case OFPUTIL_OFPT_PORT_MOD: |
4262 | return handle_port_mod(ofconn, oh); | |
064af421 | 4263 | |
d1e2cf21 | 4264 | case OFPUTIL_OFPT_FLOW_MOD: |
2e4f5fcf | 4265 | return handle_flow_mod(ofconn, oh); |
064af421 | 4266 | |
d1e2cf21 BP |
4267 | case OFPUTIL_OFPT_BARRIER_REQUEST: |
4268 | return handle_barrier_request(ofconn, oh); | |
064af421 | 4269 | |
d1e2cf21 BP |
4270 | /* OpenFlow replies. */ |
4271 | case OFPUTIL_OFPT_ECHO_REPLY: | |
4272 | return 0; | |
246e61ea | 4273 | |
d1e2cf21 | 4274 | /* Nicira extension requests. */ |
d1e2cf21 BP |
4275 | case OFPUTIL_NXT_TUN_ID_FROM_COOKIE: |
4276 | return handle_tun_id_from_cookie(ofconn, oh); | |
4277 | ||
4278 | case OFPUTIL_NXT_ROLE_REQUEST: | |
4279 | return handle_role_request(ofconn, oh); | |
4280 | ||
4281 | case OFPUTIL_NXT_SET_FLOW_FORMAT: | |
4282 | return handle_nxt_set_flow_format(ofconn, oh); | |
4283 | ||
4284 | case OFPUTIL_NXT_FLOW_MOD: | |
2e4f5fcf | 4285 | return handle_flow_mod(ofconn, oh); |
d1e2cf21 BP |
4286 | |
4287 | /* OpenFlow statistics requests. */ | |
4288 | case OFPUTIL_OFPST_DESC_REQUEST: | |
4289 | return handle_desc_stats_request(ofconn, oh); | |
4290 | ||
4291 | case OFPUTIL_OFPST_FLOW_REQUEST: | |
4292 | return handle_flow_stats_request(ofconn, oh); | |
4293 | ||
4294 | case OFPUTIL_OFPST_AGGREGATE_REQUEST: | |
4295 | return handle_aggregate_stats_request(ofconn, oh); | |
4296 | ||
4297 | case OFPUTIL_OFPST_TABLE_REQUEST: | |
4298 | return handle_table_stats_request(ofconn, oh); | |
4299 | ||
4300 | case OFPUTIL_OFPST_PORT_REQUEST: | |
4301 | return handle_port_stats_request(ofconn, oh); | |
4302 | ||
4303 | case OFPUTIL_OFPST_QUEUE_REQUEST: | |
4304 | return handle_queue_stats_request(ofconn, oh); | |
4305 | ||
4306 | /* Nicira extension statistics requests. */ | |
4307 | case OFPUTIL_NXST_FLOW_REQUEST: | |
4308 | return handle_nxst_flow(ofconn, oh); | |
4309 | ||
4310 | case OFPUTIL_NXST_AGGREGATE_REQUEST: | |
4311 | return handle_nxst_aggregate(ofconn, oh); | |
4312 | ||
4313 | case OFPUTIL_INVALID: | |
4314 | case OFPUTIL_OFPT_HELLO: | |
4315 | case OFPUTIL_OFPT_ERROR: | |
4316 | case OFPUTIL_OFPT_FEATURES_REPLY: | |
4317 | case OFPUTIL_OFPT_GET_CONFIG_REPLY: | |
4318 | case OFPUTIL_OFPT_PACKET_IN: | |
4319 | case OFPUTIL_OFPT_FLOW_REMOVED: | |
4320 | case OFPUTIL_OFPT_PORT_STATUS: | |
4321 | case OFPUTIL_OFPT_BARRIER_REPLY: | |
4322 | case OFPUTIL_OFPT_QUEUE_GET_CONFIG_REQUEST: | |
4323 | case OFPUTIL_OFPT_QUEUE_GET_CONFIG_REPLY: | |
4324 | case OFPUTIL_OFPST_DESC_REPLY: | |
4325 | case OFPUTIL_OFPST_FLOW_REPLY: | |
4326 | case OFPUTIL_OFPST_QUEUE_REPLY: | |
4327 | case OFPUTIL_OFPST_PORT_REPLY: | |
4328 | case OFPUTIL_OFPST_TABLE_REPLY: | |
4329 | case OFPUTIL_OFPST_AGGREGATE_REPLY: | |
d1e2cf21 BP |
4330 | case OFPUTIL_NXT_ROLE_REPLY: |
4331 | case OFPUTIL_NXT_FLOW_REMOVED: | |
4332 | case OFPUTIL_NXST_FLOW_REPLY: | |
4333 | case OFPUTIL_NXST_AGGREGATE_REPLY: | |
064af421 BP |
4334 | default: |
4335 | if (VLOG_IS_WARN_ENABLED()) { | |
4336 | char *s = ofp_to_string(oh, ntohs(oh->length), 2); | |
4337 | VLOG_DBG_RL(&rl, "OpenFlow message ignored: %s", s); | |
4338 | free(s); | |
4339 | } | |
d1e2cf21 BP |
4340 | if (oh->type == OFPT_STATS_REQUEST || oh->type == OFPT_STATS_REPLY) { |
4341 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_STAT); | |
4342 | } else { | |
4343 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_TYPE); | |
4344 | } | |
064af421 | 4345 | } |
d1e2cf21 | 4346 | } |
064af421 | 4347 | |
d1e2cf21 BP |
4348 | static void |
4349 | handle_openflow(struct ofconn *ofconn, struct ofpbuf *ofp_msg) | |
4350 | { | |
4351 | int error = handle_openflow__(ofconn, ofp_msg); | |
064af421 BP |
4352 | if (error) { |
4353 | send_error_oh(ofconn, ofp_msg->data, error); | |
4354 | } | |
d1e2cf21 | 4355 | COVERAGE_INC(ofproto_recv_openflow); |
064af421 BP |
4356 | } |
4357 | \f | |
4358 | static void | |
856081f6 | 4359 | handle_miss_upcall(struct ofproto *p, struct dpif_upcall *upcall) |
064af421 | 4360 | { |
bcf84111 | 4361 | struct facet *facet; |
ae412e7d | 4362 | struct flow flow; |
064af421 | 4363 | |
856081f6 BP |
4364 | /* Obtain in_port and tun_id, at least. */ |
4365 | odp_flow_key_to_flow(upcall->key, upcall->key_len, &flow); | |
064af421 | 4366 | |
856081f6 BP |
4367 | /* Set header pointers in 'flow'. */ |
4368 | flow_extract(upcall->packet, flow.tun_id, flow.in_port, &flow); | |
4617e2c1 | 4369 | |
ebe482fd EJ |
4370 | if (p->ofhooks->special_cb |
4371 | && !p->ofhooks->special_cb(&flow, upcall->packet, p->aux)) { | |
4372 | ofpbuf_delete(upcall->packet); | |
4373 | return; | |
4374 | } | |
4375 | ||
0ad9b732 JP |
4376 | /* Check with in-band control to see if this packet should be sent |
4377 | * to the local port regardless of the flow table. */ | |
856081f6 | 4378 | if (in_band_msg_in_hook(p->in_band, &flow, upcall->packet)) { |
cdee00fd | 4379 | struct ofpbuf odp_actions; |
0ad9b732 | 4380 | |
cdee00fd | 4381 | ofpbuf_init(&odp_actions, 32); |
7aec165d | 4382 | nl_msg_put_u32(&odp_actions, ODP_ACTION_ATTR_OUTPUT, ODPP_LOCAL); |
856081f6 BP |
4383 | dpif_execute(p->dpif, odp_actions.data, odp_actions.size, |
4384 | upcall->packet); | |
cdee00fd | 4385 | ofpbuf_uninit(&odp_actions); |
0ad9b732 JP |
4386 | } |
4387 | ||
bcf84111 BP |
4388 | facet = facet_lookup_valid(p, &flow); |
4389 | if (!facet) { | |
4390 | struct rule *rule = rule_lookup(p, &flow); | |
4391 | if (!rule) { | |
4392 | /* Don't send a packet-in if OFPPC_NO_PACKET_IN asserted. */ | |
856081f6 | 4393 | struct ofport *port = get_port(p, flow.in_port); |
bcf84111 BP |
4394 | if (port) { |
4395 | if (port->opp.config & OFPPC_NO_PACKET_IN) { | |
4396 | COVERAGE_INC(ofproto_no_packet_in); | |
4397 | /* XXX install 'drop' flow entry */ | |
856081f6 | 4398 | ofpbuf_delete(upcall->packet); |
bcf84111 BP |
4399 | return; |
4400 | } | |
4401 | } else { | |
4402 | VLOG_WARN_RL(&rl, "packet-in on unknown port %"PRIu16, | |
856081f6 | 4403 | flow.in_port); |
064af421 | 4404 | } |
064af421 | 4405 | |
bcf84111 | 4406 | COVERAGE_INC(ofproto_packet_in); |
856081f6 | 4407 | send_packet_in(p, upcall, &flow, false); |
bcf84111 | 4408 | return; |
064af421 | 4409 | } |
bcf84111 | 4410 | |
856081f6 | 4411 | facet = facet_create(p, rule, &flow, upcall->packet); |
bcf84111 BP |
4412 | } else if (!facet->may_install) { |
4413 | /* The facet is not installable, that is, we need to process every | |
4414 | * packet, so process the current packet's actions into 'facet'. */ | |
856081f6 | 4415 | facet_make_actions(p, facet, upcall->packet); |
064af421 BP |
4416 | } |
4417 | ||
bcf84111 | 4418 | if (facet->rule->cr.priority == FAIL_OPEN_PRIORITY) { |
7778bd15 BP |
4419 | /* |
4420 | * Extra-special case for fail-open mode. | |
4421 | * | |
4422 | * We are in fail-open mode and the packet matched the fail-open rule, | |
4423 | * but we are connected to a controller too. We should send the packet | |
4424 | * up to the controller in the hope that it will try to set up a flow | |
4425 | * and thereby allow us to exit fail-open. | |
4426 | * | |
4427 | * See the top-level comment in fail-open.c for more information. | |
4428 | */ | |
856081f6 | 4429 | send_packet_in(p, upcall, &flow, true); |
7778bd15 | 4430 | } |
750638bb | 4431 | |
856081f6 | 4432 | facet_execute(p, facet, upcall->packet); |
bcf84111 | 4433 | facet_install(p, facet, false); |
064af421 | 4434 | } |
72b06300 BP |
4435 | |
4436 | static void | |
856081f6 | 4437 | handle_upcall(struct ofproto *p, struct dpif_upcall *upcall) |
72b06300 | 4438 | { |
856081f6 | 4439 | struct flow flow; |
72b06300 | 4440 | |
856081f6 | 4441 | switch (upcall->type) { |
82272ede | 4442 | case DPIF_UC_ACTION: |
72b06300 | 4443 | COVERAGE_INC(ofproto_ctlr_action); |
856081f6 BP |
4444 | odp_flow_key_to_flow(upcall->key, upcall->key_len, &flow); |
4445 | send_packet_in(p, upcall, &flow, false); | |
72b06300 BP |
4446 | break; |
4447 | ||
82272ede | 4448 | case DPIF_UC_SAMPLE: |
72b06300 | 4449 | if (p->sflow) { |
856081f6 BP |
4450 | odp_flow_key_to_flow(upcall->key, upcall->key_len, &flow); |
4451 | ofproto_sflow_received(p->sflow, upcall, &flow); | |
72b06300 | 4452 | } |
856081f6 | 4453 | ofpbuf_delete(upcall->packet); |
72b06300 BP |
4454 | break; |
4455 | ||
82272ede | 4456 | case DPIF_UC_MISS: |
856081f6 | 4457 | handle_miss_upcall(p, upcall); |
72b06300 BP |
4458 | break; |
4459 | ||
982b8810 | 4460 | case DPIF_N_UC_TYPES: |
72b06300 | 4461 | default: |
856081f6 | 4462 | VLOG_WARN_RL(&rl, "upcall has unexpected type %"PRIu32, upcall->type); |
72b06300 BP |
4463 | break; |
4464 | } | |
4465 | } | |
064af421 | 4466 | \f |
4a4cdb3b BP |
4467 | /* Flow expiration. */ |
4468 | ||
0de7a4b4 | 4469 | static int ofproto_dp_max_idle(const struct ofproto *); |
3394b5b6 | 4470 | static void ofproto_update_stats(struct ofproto *); |
5ecc9d81 | 4471 | static void rule_expire(struct ofproto *, struct rule *); |
bcf84111 | 4472 | static void ofproto_expire_facets(struct ofproto *, int dp_max_idle); |
4a4cdb3b BP |
4473 | |
4474 | /* This function is called periodically by ofproto_run(). Its job is to | |
4475 | * collect updates for the flows that have been installed into the datapath, | |
4476 | * most importantly when they last were used, and then use that information to | |
0de7a4b4 BP |
4477 | * expire flows that have not been used recently. |
4478 | * | |
4479 | * Returns the number of milliseconds after which it should be called again. */ | |
4480 | static int | |
4a4cdb3b BP |
4481 | ofproto_expire(struct ofproto *ofproto) |
4482 | { | |
5ecc9d81 BP |
4483 | struct rule *rule, *next_rule; |
4484 | struct cls_cursor cursor; | |
4485 | int dp_max_idle; | |
4a4cdb3b | 4486 | |
3394b5b6 EJ |
4487 | /* Update stats for each flow in the datapath. */ |
4488 | ofproto_update_stats(ofproto); | |
4a4cdb3b | 4489 | |
bcf84111 | 4490 | /* Expire facets that have been idle too long. */ |
5ecc9d81 BP |
4491 | dp_max_idle = ofproto_dp_max_idle(ofproto); |
4492 | ofproto_expire_facets(ofproto, dp_max_idle); | |
bcf84111 BP |
4493 | |
4494 | /* Expire OpenFlow flows whose idle_timeout or hard_timeout has passed. */ | |
5ecc9d81 BP |
4495 | cls_cursor_init(&cursor, &ofproto->cls, NULL); |
4496 | CLS_CURSOR_FOR_EACH_SAFE (rule, next_rule, cr, &cursor) { | |
4497 | rule_expire(ofproto, rule); | |
4498 | } | |
4a4cdb3b BP |
4499 | |
4500 | /* Let the hook know that we're at a stable point: all outstanding data | |
4501 | * in existing flows has been accounted to the account_cb. Thus, the | |
4502 | * hook can now reasonably do operations that depend on having accurate | |
4503 | * flow volume accounting (currently, that's just bond rebalancing). */ | |
4504 | if (ofproto->ofhooks->account_checkpoint_cb) { | |
4505 | ofproto->ofhooks->account_checkpoint_cb(ofproto->aux); | |
4506 | } | |
0de7a4b4 | 4507 | |
5ecc9d81 | 4508 | return MIN(dp_max_idle, 1000); |
4a4cdb3b BP |
4509 | } |
4510 | ||
3394b5b6 | 4511 | /* Update 'packet_count', 'byte_count', and 'used' members of installed facets. |
878ae780 EJ |
4512 | * |
4513 | * This function also pushes statistics updates to rules which each facet | |
4514 | * resubmits into. Generally these statistics will be accurate. However, if a | |
4515 | * facet changes the rule it resubmits into at some time in between | |
4516 | * ofproto_update_stats() runs, it is possible that statistics accrued to the | |
4517 | * old rule will be incorrectly attributed to the new rule. This could be | |
4518 | * avoided by calling ofproto_update_stats() whenever rules are created or | |
4519 | * deleted. However, the performance impact of making so many calls to the | |
4520 | * datapath do not justify the benefit of having perfectly accurate statistics. | |
3394b5b6 | 4521 | */ |
4a4cdb3b | 4522 | static void |
3394b5b6 | 4523 | ofproto_update_stats(struct ofproto *p) |
4a4cdb3b | 4524 | { |
c97fb132 | 4525 | const struct dpif_flow_stats *stats; |
704a1e09 | 4526 | struct dpif_flow_dump dump; |
feebdea2 BP |
4527 | const struct nlattr *key; |
4528 | size_t key_len; | |
4a4cdb3b | 4529 | |
704a1e09 | 4530 | dpif_flow_dump_start(&dump, p->dpif); |
feebdea2 | 4531 | while (dpif_flow_dump_next(&dump, &key, &key_len, NULL, NULL, &stats)) { |
bcf84111 | 4532 | struct facet *facet; |
ae412e7d | 4533 | struct flow flow; |
14608a15 | 4534 | |
feebdea2 | 4535 | if (odp_flow_key_to_flow(key, key_len, &flow)) { |
36956a7d BP |
4536 | struct ds s; |
4537 | ||
4538 | ds_init(&s); | |
feebdea2 | 4539 | odp_flow_key_format(key, key_len, &s); |
36956a7d BP |
4540 | VLOG_WARN_RL(&rl, "failed to convert ODP flow key to flow: %s", |
4541 | ds_cstr(&s)); | |
4542 | ds_destroy(&s); | |
4543 | ||
4544 | continue; | |
4545 | } | |
bcf84111 | 4546 | facet = facet_find(p, &flow); |
4a4cdb3b | 4547 | |
bcf84111 | 4548 | if (facet && facet->installed) { |
3394b5b6 EJ |
4549 | |
4550 | if (stats->n_packets >= facet->dp_packet_count) { | |
4551 | facet->packet_count += stats->n_packets - facet->dp_packet_count; | |
4552 | } else { | |
4553 | VLOG_WARN_RL(&rl, "unexpected packet count from the datapath"); | |
4554 | } | |
4555 | ||
4556 | if (stats->n_bytes >= facet->dp_byte_count) { | |
4557 | facet->byte_count += stats->n_bytes - facet->dp_byte_count; | |
4558 | } else { | |
4559 | VLOG_WARN_RL(&rl, "unexpected byte count from datapath"); | |
4560 | } | |
4561 | ||
4562 | facet->dp_packet_count = stats->n_packets; | |
4563 | facet->dp_byte_count = stats->n_bytes; | |
4564 | ||
4565 | facet_update_time(p, facet, stats->used); | |
feebdea2 | 4566 | facet_account(p, facet, stats->n_bytes); |
878ae780 | 4567 | facet_push_stats(p, facet); |
4a4cdb3b BP |
4568 | } else { |
4569 | /* There's a flow in the datapath that we know nothing about. | |
4570 | * Delete it. */ | |
4571 | COVERAGE_INC(ofproto_unexpected_rule); | |
feebdea2 | 4572 | dpif_flow_del(p->dpif, key, key_len, NULL); |
4a4cdb3b | 4573 | } |
4a4cdb3b | 4574 | } |
704a1e09 | 4575 | dpif_flow_dump_done(&dump); |
4a4cdb3b BP |
4576 | } |
4577 | ||
0de7a4b4 | 4578 | /* Calculates and returns the number of milliseconds of idle time after which |
bcf84111 | 4579 | * facets should expire from the datapath and we should fold their statistics |
0de7a4b4 BP |
4580 | * into their parent rules in userspace. */ |
4581 | static int | |
4582 | ofproto_dp_max_idle(const struct ofproto *ofproto) | |
4583 | { | |
4584 | /* | |
4585 | * Idle time histogram. | |
4586 | * | |
bcf84111 | 4587 | * Most of the time a switch has a relatively small number of facets. When |
0de7a4b4 BP |
4588 | * this is the case we might as well keep statistics for all of them in |
4589 | * userspace and to cache them in the kernel datapath for performance as | |
4590 | * well. | |
4591 | * | |
bcf84111 | 4592 | * As the number of facets increases, the memory required to maintain |
0de7a4b4 | 4593 | * statistics about them in userspace and in the kernel becomes |
bcf84111 | 4594 | * significant. However, with a large number of facets it is likely that |
0de7a4b4 BP |
4595 | * only a few of them are "heavy hitters" that consume a large amount of |
4596 | * bandwidth. At this point, only heavy hitters are worth caching in the | |
bcf84111 | 4597 | * kernel and maintaining in userspaces; other facets we can discard. |
0de7a4b4 BP |
4598 | * |
4599 | * The technique used to compute the idle time is to build a histogram with | |
bcf84111 BP |
4600 | * N_BUCKETS buckets whose width is BUCKET_WIDTH msecs each. Each facet |
4601 | * that is installed in the kernel gets dropped in the appropriate bucket. | |
0de7a4b4 | 4602 | * After the histogram has been built, we compute the cutoff so that only |
bcf84111 BP |
4603 | * the most-recently-used 1% of facets (but at least 1000 flows) are kept |
4604 | * cached. At least the most-recently-used bucket of facets is kept, so | |
4605 | * actually an arbitrary number of facets can be kept in any given | |
0de7a4b4 BP |
4606 | * expiration run (though the next run will delete most of those unless |
4607 | * they receive additional data). | |
4608 | * | |
bcf84111 | 4609 | * This requires a second pass through the facets, in addition to the pass |
3394b5b6 | 4610 | * made by ofproto_update_stats(), because the former function never looks |
bcf84111 | 4611 | * at uninstallable facets. |
0de7a4b4 BP |
4612 | */ |
4613 | enum { BUCKET_WIDTH = ROUND_UP(100, TIME_UPDATE_INTERVAL) }; | |
4614 | enum { N_BUCKETS = 5000 / BUCKET_WIDTH }; | |
4615 | int buckets[N_BUCKETS] = { 0 }; | |
bcf84111 | 4616 | struct facet *facet; |
0de7a4b4 | 4617 | int total, bucket; |
0de7a4b4 BP |
4618 | long long int now; |
4619 | int i; | |
4620 | ||
bcf84111 | 4621 | total = hmap_count(&ofproto->facets); |
0de7a4b4 BP |
4622 | if (total <= 1000) { |
4623 | return N_BUCKETS * BUCKET_WIDTH; | |
4624 | } | |
4625 | ||
4626 | /* Build histogram. */ | |
4627 | now = time_msec(); | |
bcf84111 BP |
4628 | HMAP_FOR_EACH (facet, hmap_node, &ofproto->facets) { |
4629 | long long int idle = now - facet->used; | |
0de7a4b4 BP |
4630 | int bucket = (idle <= 0 ? 0 |
4631 | : idle >= BUCKET_WIDTH * N_BUCKETS ? N_BUCKETS - 1 | |
4632 | : (unsigned int) idle / BUCKET_WIDTH); | |
4633 | buckets[bucket]++; | |
4634 | } | |
4635 | ||
4636 | /* Find the first bucket whose flows should be expired. */ | |
4637 | for (bucket = 0; bucket < N_BUCKETS; bucket++) { | |
4638 | if (buckets[bucket]) { | |
4639 | int subtotal = 0; | |
4640 | do { | |
4641 | subtotal += buckets[bucket++]; | |
4642 | } while (bucket < N_BUCKETS && subtotal < MAX(1000, total / 100)); | |
4643 | break; | |
4644 | } | |
4645 | } | |
4646 | ||
4647 | if (VLOG_IS_DBG_ENABLED()) { | |
4648 | struct ds s; | |
4649 | ||
4650 | ds_init(&s); | |
4651 | ds_put_cstr(&s, "keep"); | |
4652 | for (i = 0; i < N_BUCKETS; i++) { | |
4653 | if (i == bucket) { | |
4654 | ds_put_cstr(&s, ", drop"); | |
4655 | } | |
4656 | if (buckets[i]) { | |
4657 | ds_put_format(&s, " %d:%d", i * BUCKET_WIDTH, buckets[i]); | |
4658 | } | |
4659 | } | |
4660 | VLOG_INFO("%s: %s (msec:count)", | |
4661 | dpif_name(ofproto->dpif), ds_cstr(&s)); | |
4662 | ds_destroy(&s); | |
4663 | } | |
4664 | ||
4665 | return bucket * BUCKET_WIDTH; | |
4666 | } | |
4667 | ||
4a4cdb3b | 4668 | static void |
bcf84111 | 4669 | facet_active_timeout(struct ofproto *ofproto, struct facet *facet) |
4a4cdb3b | 4670 | { |
bcf84111 BP |
4671 | if (ofproto->netflow && !facet_is_controller_flow(facet) && |
4672 | netflow_active_timeout_expired(ofproto->netflow, &facet->nf_flow)) { | |
4a4cdb3b | 4673 | struct ofexpired expired; |
feebdea2 | 4674 | |
bcf84111 | 4675 | if (facet->installed) { |
c97fb132 | 4676 | struct dpif_flow_stats stats; |
c33087b8 BP |
4677 | |
4678 | facet_put__(ofproto, facet, facet->actions, facet->actions_len, | |
4679 | &stats); | |
4680 | facet_update_stats(ofproto, facet, &stats); | |
4a4cdb3b BP |
4681 | } |
4682 | ||
c33087b8 BP |
4683 | expired.flow = facet->flow; |
4684 | expired.packet_count = facet->packet_count; | |
4685 | expired.byte_count = facet->byte_count; | |
4686 | expired.used = facet->used; | |
bcf84111 BP |
4687 | netflow_expire(ofproto->netflow, &facet->nf_flow, &expired); |
4688 | } | |
4689 | } | |
4a4cdb3b | 4690 | |
bcf84111 BP |
4691 | static void |
4692 | ofproto_expire_facets(struct ofproto *ofproto, int dp_max_idle) | |
4693 | { | |
4694 | long long int cutoff = time_msec() - dp_max_idle; | |
4695 | struct facet *facet, *next_facet; | |
4696 | ||
4697 | HMAP_FOR_EACH_SAFE (facet, next_facet, hmap_node, &ofproto->facets) { | |
4698 | facet_active_timeout(ofproto, facet); | |
4699 | if (facet->used < cutoff) { | |
4700 | facet_remove(ofproto, facet); | |
4701 | } | |
4a4cdb3b BP |
4702 | } |
4703 | } | |
4704 | ||
5ecc9d81 BP |
4705 | /* If 'rule' is an OpenFlow rule, that has expired according to OpenFlow rules, |
4706 | * then delete it entirely. */ | |
4a4cdb3b | 4707 | static void |
5ecc9d81 | 4708 | rule_expire(struct ofproto *ofproto, struct rule *rule) |
4a4cdb3b | 4709 | { |
bcf84111 BP |
4710 | struct facet *facet, *next_facet; |
4711 | long long int now; | |
4712 | uint8_t reason; | |
4a4cdb3b | 4713 | |
bcf84111 | 4714 | /* Has 'rule' expired? */ |
4a4cdb3b | 4715 | now = time_msec(); |
bcf84111 BP |
4716 | if (rule->hard_timeout |
4717 | && now > rule->created + rule->hard_timeout * 1000) { | |
4718 | reason = OFPRR_HARD_TIMEOUT; | |
4719 | } else if (rule->idle_timeout && list_is_empty(&rule->facets) | |
4720 | && now >rule->used + rule->idle_timeout * 1000) { | |
4721 | reason = OFPRR_IDLE_TIMEOUT; | |
4a4cdb3b | 4722 | } else { |
bcf84111 | 4723 | return; |
4a4cdb3b | 4724 | } |
064af421 | 4725 | |
bcf84111 | 4726 | COVERAGE_INC(ofproto_expired); |
064af421 | 4727 | |
bcf84111 BP |
4728 | /* Update stats. (This is a no-op if the rule expired due to an idle |
4729 | * timeout, because that only happens when the rule has no facets left.) */ | |
4730 | LIST_FOR_EACH_SAFE (facet, next_facet, list_node, &rule->facets) { | |
5ecc9d81 | 4731 | facet_remove(ofproto, facet); |
064af421 BP |
4732 | } |
4733 | ||
bcf84111 BP |
4734 | /* Get rid of the rule. */ |
4735 | if (!rule_is_hidden(rule)) { | |
5ecc9d81 | 4736 | rule_send_removed(ofproto, rule, reason); |
bcf84111 | 4737 | } |
5ecc9d81 | 4738 | rule_remove(ofproto, rule); |
064af421 | 4739 | } |
bcf84111 | 4740 | \f |
064af421 | 4741 | static struct ofpbuf * |
09246b99 BP |
4742 | compose_ofp_flow_removed(struct ofconn *ofconn, const struct rule *rule, |
4743 | uint8_t reason) | |
064af421 | 4744 | { |
ca069229 | 4745 | struct ofp_flow_removed *ofr; |
064af421 BP |
4746 | struct ofpbuf *buf; |
4747 | ||
0224fbde | 4748 | ofr = make_openflow_xid(sizeof *ofr, OFPT_FLOW_REMOVED, htonl(0), &buf); |
ff9d3826 BP |
4749 | ofputil_cls_rule_to_match(&rule->cr, ofconn->flow_format, &ofr->match, |
4750 | rule->flow_cookie, &ofr->cookie); | |
ca069229 JP |
4751 | ofr->priority = htons(rule->cr.priority); |
4752 | ofr->reason = reason; | |
c6ebb8fb | 4753 | calc_flow_duration(rule->created, &ofr->duration_sec, &ofr->duration_nsec); |
ca069229 JP |
4754 | ofr->idle_timeout = htons(rule->idle_timeout); |
4755 | ofr->packet_count = htonll(rule->packet_count); | |
4756 | ofr->byte_count = htonll(rule->byte_count); | |
064af421 BP |
4757 | |
4758 | return buf; | |
4759 | } | |
4760 | ||
09246b99 BP |
4761 | static struct ofpbuf * |
4762 | compose_nx_flow_removed(const struct rule *rule, uint8_t reason) | |
4763 | { | |
4764 | struct nx_flow_removed *nfr; | |
4765 | struct ofpbuf *buf; | |
4766 | int match_len; | |
4767 | ||
0224fbde | 4768 | make_nxmsg_xid(sizeof *nfr, NXT_FLOW_REMOVED, htonl(0), &buf); |
09246b99 | 4769 | match_len = nx_put_match(buf, &rule->cr); |
0224fbde | 4770 | |
4d0ed519 | 4771 | nfr = buf->data; |
09246b99 BP |
4772 | nfr->cookie = rule->flow_cookie; |
4773 | nfr->priority = htons(rule->cr.priority); | |
4774 | nfr->reason = reason; | |
4775 | calc_flow_duration(rule->created, &nfr->duration_sec, &nfr->duration_nsec); | |
4776 | nfr->idle_timeout = htons(rule->idle_timeout); | |
4777 | nfr->match_len = htons(match_len); | |
4778 | nfr->packet_count = htonll(rule->packet_count); | |
4779 | nfr->byte_count = htonll(rule->byte_count); | |
4780 | ||
4781 | return buf; | |
4782 | } | |
4783 | ||
ca069229 | 4784 | static void |
bcf84111 | 4785 | rule_send_removed(struct ofproto *p, struct rule *rule, uint8_t reason) |
064af421 BP |
4786 | { |
4787 | struct ofconn *ofconn; | |
064af421 | 4788 | |
3b587616 BP |
4789 | if (!rule->send_flow_removed) { |
4790 | return; | |
4791 | } | |
4792 | ||
4e8e4213 | 4793 | LIST_FOR_EACH (ofconn, node, &p->all_conns) { |
7a0efeb5 BP |
4794 | struct ofpbuf *msg; |
4795 | ||
4796 | if (!rconn_is_connected(ofconn->rconn) | |
4797 | || !ofconn_receives_async_msgs(ofconn)) { | |
4798 | continue; | |
064af421 | 4799 | } |
7a0efeb5 | 4800 | |
09246b99 BP |
4801 | msg = (ofconn->flow_format == NXFF_NXM |
4802 | ? compose_nx_flow_removed(rule, reason) | |
4803 | : compose_ofp_flow_removed(ofconn, rule, reason)); | |
6d6c7259 BP |
4804 | |
4805 | /* Account flow expirations under ofconn->reply_counter, the counter | |
4806 | * for replies to OpenFlow requests. That works because preventing | |
4807 | * OpenFlow requests from being processed also prevents new flows from | |
4808 | * being added (and expiring). (It also prevents processing OpenFlow | |
4809 | * requests that would not add new flows, so it is imperfect.) */ | |
7a0efeb5 | 4810 | queue_tx(msg, ofconn, ofconn->reply_counter); |
064af421 BP |
4811 | } |
4812 | } | |
4813 | ||
3394b5b6 EJ |
4814 | /* Obtains statistics for 'rule' and stores them in '*packets' and '*bytes'. |
4815 | * The returned statistics include statistics for all of 'rule''s facets. */ | |
4816 | static void | |
4817 | rule_get_stats(const struct rule *rule, uint64_t *packets, uint64_t *bytes) | |
4818 | { | |
4819 | uint64_t p, b; | |
4820 | struct facet *facet; | |
4821 | ||
4822 | /* Start from historical data for 'rule' itself that are no longer tracked | |
4823 | * in facets. This counts, for example, facets that have expired. */ | |
4824 | p = rule->packet_count; | |
4825 | b = rule->byte_count; | |
4826 | ||
4827 | /* Add any statistics that are tracked by facets. This includes | |
4828 | * statistical data recently updated by ofproto_update_stats() as well as | |
4829 | * stats for packets that were executed "by hand" via dpif_execute(). */ | |
4830 | LIST_FOR_EACH (facet, list_node, &rule->facets) { | |
4831 | p += facet->packet_count; | |
4832 | b += facet->byte_count; | |
4833 | } | |
4834 | ||
4835 | *packets = p; | |
4836 | *bytes = b; | |
4837 | } | |
4838 | ||
856081f6 | 4839 | /* pinsched callback for sending 'ofp_packet_in' on 'ofconn'. */ |
064af421 | 4840 | static void |
856081f6 | 4841 | do_send_packet_in(struct ofpbuf *ofp_packet_in, void *ofconn_) |
064af421 | 4842 | { |
76ce9432 | 4843 | struct ofconn *ofconn = ofconn_; |
43253595 | 4844 | |
856081f6 | 4845 | rconn_send_with_limit(ofconn->rconn, ofp_packet_in, |
43253595 BP |
4846 | ofconn->packet_in_counter, 100); |
4847 | } | |
4848 | ||
856081f6 BP |
4849 | /* Takes 'upcall', whose packet has the flow specified by 'flow', composes an |
4850 | * OpenFlow packet-in message from it, and passes it to 'ofconn''s packet | |
4851 | * scheduler for sending. | |
43253595 | 4852 | * |
856081f6 BP |
4853 | * If 'clone' is true, the caller retains ownership of 'upcall->packet'. |
4854 | * Otherwise, ownership is transferred to this function. */ | |
43253595 | 4855 | static void |
856081f6 BP |
4856 | schedule_packet_in(struct ofconn *ofconn, struct dpif_upcall *upcall, |
4857 | const struct flow *flow, bool clone) | |
43253595 | 4858 | { |
856081f6 | 4859 | enum { OPI_SIZE = offsetof(struct ofp_packet_in, data) }; |
76ce9432 | 4860 | struct ofproto *ofproto = ofconn->ofproto; |
856081f6 BP |
4861 | struct ofp_packet_in *opi; |
4862 | int total_len, send_len; | |
4863 | struct ofpbuf *packet; | |
76ce9432 | 4864 | uint32_t buffer_id; |
82272ede | 4865 | int idx; |
064af421 | 4866 | |
856081f6 | 4867 | /* Get OpenFlow buffer_id. */ |
82272ede | 4868 | if (upcall->type == DPIF_UC_ACTION) { |
76ce9432 | 4869 | buffer_id = UINT32_MAX; |
43253595 BP |
4870 | } else if (ofproto->fail_open && fail_open_is_active(ofproto->fail_open)) { |
4871 | buffer_id = pktbuf_get_null(); | |
89b9612d BP |
4872 | } else if (!ofconn->pktbuf) { |
4873 | buffer_id = UINT32_MAX; | |
76ce9432 | 4874 | } else { |
856081f6 | 4875 | buffer_id = pktbuf_save(ofconn->pktbuf, upcall->packet, flow->in_port); |
76ce9432 | 4876 | } |
372179d4 | 4877 | |
43253595 | 4878 | /* Figure out how much of the packet to send. */ |
856081f6 | 4879 | total_len = send_len = upcall->packet->size; |
43253595 BP |
4880 | if (buffer_id != UINT32_MAX) { |
4881 | send_len = MIN(send_len, ofconn->miss_send_len); | |
4882 | } | |
82272ede | 4883 | if (upcall->type == DPIF_UC_ACTION) { |
856081f6 BP |
4884 | send_len = MIN(send_len, upcall->userdata); |
4885 | } | |
064af421 | 4886 | |
856081f6 | 4887 | /* Copy or steal buffer for OFPT_PACKET_IN. */ |
43253595 | 4888 | if (clone) { |
856081f6 BP |
4889 | packet = ofpbuf_clone_data_with_headroom(upcall->packet->data, |
4890 | send_len, OPI_SIZE); | |
43253595 | 4891 | } else { |
856081f6 BP |
4892 | packet = upcall->packet; |
4893 | packet->size = send_len; | |
43253595 BP |
4894 | } |
4895 | ||
856081f6 BP |
4896 | /* Add OFPT_PACKET_IN. */ |
4897 | opi = ofpbuf_push_zeros(packet, OPI_SIZE); | |
4898 | opi->header.version = OFP_VERSION; | |
4899 | opi->header.type = OFPT_PACKET_IN; | |
4900 | opi->total_len = htons(total_len); | |
4901 | opi->in_port = htons(odp_port_to_ofp_port(flow->in_port)); | |
82272ede | 4902 | opi->reason = upcall->type == DPIF_UC_MISS ? OFPR_NO_MATCH : OFPR_ACTION; |
43253595 BP |
4903 | opi->buffer_id = htonl(buffer_id); |
4904 | update_openflow_length(packet); | |
4905 | ||
4906 | /* Hand over to packet scheduler. It might immediately call into | |
4907 | * do_send_packet_in() or it might buffer it for a while (until a later | |
4908 | * call to pinsched_run()). */ | |
82272ede BP |
4909 | idx = upcall->type == DPIF_UC_MISS ? 0 : 1; |
4910 | pinsched_send(ofconn->schedulers[idx], flow->in_port, | |
43253595 | 4911 | packet, do_send_packet_in, ofconn); |
064af421 BP |
4912 | } |
4913 | ||
82272ede | 4914 | /* Given 'upcall', of type DPIF_UC_ACTION or DPIF_UC_MISS, sends an |
856081f6 BP |
4915 | * OFPT_PACKET_IN message to each OpenFlow controller as necessary according to |
4916 | * their individual configurations. | |
43253595 | 4917 | * |
2c00fecf BP |
4918 | * If 'clone' is true, the caller retains ownership of 'upcall->packet'. |
4919 | * Otherwise, ownership is transferred to this function. */ | |
43253595 | 4920 | static void |
856081f6 BP |
4921 | send_packet_in(struct ofproto *ofproto, struct dpif_upcall *upcall, |
4922 | const struct flow *flow, bool clone) | |
43253595 | 4923 | { |
76ce9432 | 4924 | struct ofconn *ofconn, *prev; |
76ce9432 BP |
4925 | |
4926 | prev = NULL; | |
4e8e4213 | 4927 | LIST_FOR_EACH (ofconn, node, &ofproto->all_conns) { |
c91248b3 | 4928 | if (ofconn_receives_async_msgs(ofconn)) { |
9deba63b | 4929 | if (prev) { |
856081f6 | 4930 | schedule_packet_in(prev, upcall, flow, true); |
9deba63b BP |
4931 | } |
4932 | prev = ofconn; | |
064af421 | 4933 | } |
76ce9432 BP |
4934 | } |
4935 | if (prev) { | |
856081f6 BP |
4936 | schedule_packet_in(prev, upcall, flow, clone); |
4937 | } else if (!clone) { | |
4938 | ofpbuf_delete(upcall->packet); | |
064af421 | 4939 | } |
064af421 BP |
4940 | } |
4941 | ||
4942 | static uint64_t | |
fa60c019 | 4943 | pick_datapath_id(const struct ofproto *ofproto) |
064af421 | 4944 | { |
fa60c019 | 4945 | const struct ofport *port; |
064af421 | 4946 | |
ca0f572c | 4947 | port = get_port(ofproto, ODPP_LOCAL); |
fa60c019 BP |
4948 | if (port) { |
4949 | uint8_t ea[ETH_ADDR_LEN]; | |
4950 | int error; | |
4951 | ||
4952 | error = netdev_get_etheraddr(port->netdev, ea); | |
064af421 BP |
4953 | if (!error) { |
4954 | return eth_addr_to_uint64(ea); | |
4955 | } | |
4956 | VLOG_WARN("could not get MAC address for %s (%s)", | |
fa60c019 | 4957 | netdev_get_name(port->netdev), strerror(error)); |
064af421 | 4958 | } |
fa60c019 | 4959 | return ofproto->fallback_dpid; |
064af421 BP |
4960 | } |
4961 | ||
4962 | static uint64_t | |
4963 | pick_fallback_dpid(void) | |
4964 | { | |
4965 | uint8_t ea[ETH_ADDR_LEN]; | |
70150daf | 4966 | eth_addr_nicira_random(ea); |
064af421 BP |
4967 | return eth_addr_to_uint64(ea); |
4968 | } | |
4969 | \f | |
7aa697dd BP |
4970 | static void |
4971 | ofproto_unixctl_list(struct unixctl_conn *conn, const char *arg OVS_UNUSED, | |
4972 | void *aux OVS_UNUSED) | |
4973 | { | |
4974 | const struct shash_node *node; | |
4975 | struct ds results; | |
4976 | ||
4977 | ds_init(&results); | |
4978 | SHASH_FOR_EACH (node, &all_ofprotos) { | |
4979 | ds_put_format(&results, "%s\n", node->name); | |
4980 | } | |
4981 | unixctl_command_reply(conn, 200, ds_cstr(&results)); | |
4982 | ds_destroy(&results); | |
4983 | } | |
4984 | ||
4985 | struct ofproto_trace { | |
4986 | struct action_xlate_ctx ctx; | |
4987 | struct flow flow; | |
4988 | struct ds *result; | |
4989 | }; | |
4990 | ||
4991 | static void | |
4992 | trace_format_rule(struct ds *result, int level, const struct rule *rule) | |
4993 | { | |
4994 | ds_put_char_multiple(result, '\t', level); | |
4995 | if (!rule) { | |
4996 | ds_put_cstr(result, "No match\n"); | |
4997 | return; | |
4998 | } | |
4999 | ||
5000 | ds_put_format(result, "Rule: cookie=%#"PRIx64" ", | |
5001 | ntohll(rule->flow_cookie)); | |
5002 | cls_rule_format(&rule->cr, result); | |
5003 | ds_put_char(result, '\n'); | |
5004 | ||
5005 | ds_put_char_multiple(result, '\t', level); | |
5006 | ds_put_cstr(result, "OpenFlow "); | |
5007 | ofp_print_actions(result, (const struct ofp_action_header *) rule->actions, | |
5008 | rule->n_actions * sizeof *rule->actions); | |
5009 | ds_put_char(result, '\n'); | |
5010 | } | |
5011 | ||
5012 | static void | |
5013 | trace_format_flow(struct ds *result, int level, const char *title, | |
5014 | struct ofproto_trace *trace) | |
5015 | { | |
5016 | ds_put_char_multiple(result, '\t', level); | |
5017 | ds_put_format(result, "%s: ", title); | |
5018 | if (flow_equal(&trace->ctx.flow, &trace->flow)) { | |
5019 | ds_put_cstr(result, "unchanged"); | |
5020 | } else { | |
5021 | flow_format(result, &trace->ctx.flow); | |
5022 | trace->flow = trace->ctx.flow; | |
5023 | } | |
5024 | ds_put_char(result, '\n'); | |
5025 | } | |
5026 | ||
5027 | static void | |
878ae780 | 5028 | trace_resubmit(struct action_xlate_ctx *ctx, struct rule *rule) |
7aa697dd BP |
5029 | { |
5030 | struct ofproto_trace *trace = CONTAINER_OF(ctx, struct ofproto_trace, ctx); | |
5031 | struct ds *result = trace->result; | |
5032 | ||
5033 | ds_put_char(result, '\n'); | |
5034 | trace_format_flow(result, ctx->recurse + 1, "Resubmitted flow", trace); | |
5035 | trace_format_rule(result, ctx->recurse + 1, rule); | |
5036 | } | |
5037 | ||
5038 | static void | |
5039 | ofproto_unixctl_trace(struct unixctl_conn *conn, const char *args_, | |
5040 | void *aux OVS_UNUSED) | |
5041 | { | |
5042 | char *dpname, *in_port_s, *tun_id_s, *packet_s; | |
5043 | char *args = xstrdup(args_); | |
5044 | char *save_ptr = NULL; | |
5045 | struct ofproto *ofproto; | |
5046 | struct ofpbuf packet; | |
5047 | struct rule *rule; | |
5048 | struct ds result; | |
5049 | struct flow flow; | |
5050 | uint16_t in_port; | |
11e6a15b | 5051 | ovs_be64 tun_id; |
7aa697dd BP |
5052 | char *s; |
5053 | ||
5054 | ofpbuf_init(&packet, strlen(args) / 2); | |
5055 | ds_init(&result); | |
5056 | ||
5057 | dpname = strtok_r(args, " ", &save_ptr); | |
5058 | tun_id_s = strtok_r(NULL, " ", &save_ptr); | |
5059 | in_port_s = strtok_r(NULL, " ", &save_ptr); | |
5060 | packet_s = strtok_r(NULL, "", &save_ptr); /* Get entire rest of line. */ | |
5061 | if (!dpname || !in_port_s || !packet_s) { | |
5062 | unixctl_command_reply(conn, 501, "Bad command syntax"); | |
5063 | goto exit; | |
5064 | } | |
5065 | ||
5066 | ofproto = shash_find_data(&all_ofprotos, dpname); | |
5067 | if (!ofproto) { | |
5068 | unixctl_command_reply(conn, 501, "Unknown ofproto (use ofproto/list " | |
5069 | "for help)"); | |
5070 | goto exit; | |
5071 | } | |
5072 | ||
11e6a15b | 5073 | tun_id = htonll(strtoull(tun_id_s, NULL, 10)); |
7aa697dd BP |
5074 | in_port = ofp_port_to_odp_port(atoi(in_port_s)); |
5075 | ||
5076 | packet_s = ofpbuf_put_hex(&packet, packet_s, NULL); | |
5077 | packet_s += strspn(packet_s, " "); | |
5078 | if (*packet_s != '\0') { | |
5079 | unixctl_command_reply(conn, 501, "Trailing garbage in command"); | |
5080 | goto exit; | |
5081 | } | |
5082 | if (packet.size < ETH_HEADER_LEN) { | |
5083 | unixctl_command_reply(conn, 501, "Packet data too short for Ethernet"); | |
5084 | goto exit; | |
5085 | } | |
5086 | ||
5087 | ds_put_cstr(&result, "Packet: "); | |
5088 | s = ofp_packet_to_string(packet.data, packet.size, packet.size); | |
5089 | ds_put_cstr(&result, s); | |
5090 | free(s); | |
5091 | ||
5092 | flow_extract(&packet, tun_id, in_port, &flow); | |
5093 | ds_put_cstr(&result, "Flow: "); | |
5094 | flow_format(&result, &flow); | |
5095 | ds_put_char(&result, '\n'); | |
5096 | ||
5097 | rule = rule_lookup(ofproto, &flow); | |
5098 | trace_format_rule(&result, 0, rule); | |
5099 | if (rule) { | |
5100 | struct ofproto_trace trace; | |
cdee00fd | 5101 | struct ofpbuf *odp_actions; |
7aa697dd BP |
5102 | |
5103 | trace.result = &result; | |
5104 | trace.flow = flow; | |
5105 | action_xlate_ctx_init(&trace.ctx, ofproto, &flow, &packet); | |
5106 | trace.ctx.resubmit_hook = trace_resubmit; | |
cdee00fd BP |
5107 | odp_actions = xlate_actions(&trace.ctx, |
5108 | rule->actions, rule->n_actions); | |
7aa697dd BP |
5109 | |
5110 | ds_put_char(&result, '\n'); | |
5111 | trace_format_flow(&result, 0, "Final flow", &trace); | |
5112 | ds_put_cstr(&result, "Datapath actions: "); | |
cdee00fd BP |
5113 | format_odp_actions(&result, odp_actions->data, odp_actions->size); |
5114 | ofpbuf_delete(odp_actions); | |
7aa697dd BP |
5115 | } |
5116 | ||
5117 | unixctl_command_reply(conn, 200, ds_cstr(&result)); | |
5118 | ||
5119 | exit: | |
5120 | ds_destroy(&result); | |
5121 | ofpbuf_uninit(&packet); | |
5122 | free(args); | |
5123 | } | |
5124 | ||
5125 | static void | |
5126 | ofproto_unixctl_init(void) | |
5127 | { | |
5128 | static bool registered; | |
5129 | if (registered) { | |
5130 | return; | |
5131 | } | |
5132 | registered = true; | |
5133 | ||
5134 | unixctl_command_register("ofproto/list", ofproto_unixctl_list, NULL); | |
5135 | unixctl_command_register("ofproto/trace", ofproto_unixctl_trace, NULL); | |
5136 | } | |
5137 | \f | |
064af421 | 5138 | static bool |
ae412e7d | 5139 | default_normal_ofhook_cb(const struct flow *flow, const struct ofpbuf *packet, |
cdee00fd | 5140 | struct ofpbuf *odp_actions, tag_type *tags, |
6a07af36 | 5141 | uint16_t *nf_output_iface, void *ofproto_) |
064af421 BP |
5142 | { |
5143 | struct ofproto *ofproto = ofproto_; | |
5144 | int out_port; | |
5145 | ||
5146 | /* Drop frames for reserved multicast addresses. */ | |
5147 | if (eth_addr_is_reserved(flow->dl_dst)) { | |
5148 | return true; | |
5149 | } | |
5150 | ||
5151 | /* Learn source MAC (but don't try to learn from revalidation). */ | |
5152 | if (packet != NULL) { | |
5153 | tag_type rev_tag = mac_learning_learn(ofproto->ml, flow->dl_src, | |
7febb910 JG |
5154 | 0, flow->in_port, |
5155 | GRAT_ARP_LOCK_NONE); | |
064af421 BP |
5156 | if (rev_tag) { |
5157 | /* The log messages here could actually be useful in debugging, | |
5158 | * so keep the rate limit relatively high. */ | |
5159 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
5160 | VLOG_DBG_RL(&rl, "learned that "ETH_ADDR_FMT" is on port %"PRIu16, | |
5161 | ETH_ADDR_ARGS(flow->dl_src), flow->in_port); | |
5162 | ofproto_revalidate(ofproto, rev_tag); | |
5163 | } | |
5164 | } | |
5165 | ||
5166 | /* Determine output port. */ | |
7febb910 JG |
5167 | out_port = mac_learning_lookup_tag(ofproto->ml, flow->dl_dst, 0, tags, |
5168 | NULL); | |
064af421 | 5169 | if (out_port < 0) { |
f1588b1f | 5170 | flood_packets(ofproto, flow->in_port, OFPPC_NO_FLOOD, |
cdee00fd | 5171 | nf_output_iface, odp_actions); |
064af421 | 5172 | } else if (out_port != flow->in_port) { |
7aec165d | 5173 | nl_msg_put_u32(odp_actions, ODP_ACTION_ATTR_OUTPUT, out_port); |
6a07af36 | 5174 | *nf_output_iface = out_port; |
064af421 BP |
5175 | } else { |
5176 | /* Drop. */ | |
5177 | } | |
5178 | ||
5179 | return true; | |
5180 | } | |
5181 | ||
5182 | static const struct ofhooks default_ofhooks = { | |
064af421 BP |
5183 | default_normal_ofhook_cb, |
5184 | NULL, | |
ebe482fd | 5185 | NULL, |
064af421 BP |
5186 | NULL |
5187 | }; |