]>
Commit | Line | Data |
---|---|---|
064af421 | 1 | /* |
c475ae67 | 2 | * Copyright (c) 2009, 2010 Nicira Networks. |
064af421 | 3 | * |
a14bc59f BP |
4 | * Licensed under the Apache License, Version 2.0 (the "License"); |
5 | * you may not use this file except in compliance with the License. | |
6 | * You may obtain a copy of the License at: | |
064af421 | 7 | * |
a14bc59f BP |
8 | * http://www.apache.org/licenses/LICENSE-2.0 |
9 | * | |
10 | * Unless required by applicable law or agreed to in writing, software | |
11 | * distributed under the License is distributed on an "AS IS" BASIS, | |
12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | |
13 | * See the License for the specific language governing permissions and | |
14 | * limitations under the License. | |
064af421 BP |
15 | */ |
16 | ||
17 | #include <config.h> | |
18 | #include "ofproto.h" | |
19 | #include <errno.h> | |
20 | #include <inttypes.h> | |
21 | #include <net/if.h> | |
22 | #include <netinet/in.h> | |
23 | #include <stdbool.h> | |
24 | #include <stdlib.h> | |
25 | #include "classifier.h" | |
26 | #include "coverage.h" | |
27 | #include "discovery.h" | |
28 | #include "dpif.h" | |
4f2cad2c | 29 | #include "dynamic-string.h" |
064af421 BP |
30 | #include "fail-open.h" |
31 | #include "in-band.h" | |
32 | #include "mac-learning.h" | |
33 | #include "netdev.h" | |
34 | #include "netflow.h" | |
35 | #include "odp-util.h" | |
36 | #include "ofp-print.h" | |
72b06300 | 37 | #include "ofproto-sflow.h" |
064af421 BP |
38 | #include "ofpbuf.h" |
39 | #include "openflow/nicira-ext.h" | |
40 | #include "openflow/openflow.h" | |
064af421 BP |
41 | #include "openvswitch/datapath-protocol.h" |
42 | #include "packets.h" | |
43 | #include "pinsched.h" | |
44 | #include "pktbuf.h" | |
45 | #include "poll-loop.h" | |
46 | #include "port-array.h" | |
47 | #include "rconn.h" | |
48 | #include "shash.h" | |
49 | #include "status.h" | |
50 | #include "stp.h" | |
fe55ad15 | 51 | #include "stream-ssl.h" |
064af421 BP |
52 | #include "svec.h" |
53 | #include "tag.h" | |
54 | #include "timeval.h" | |
4f2cad2c | 55 | #include "unixctl.h" |
064af421 | 56 | #include "vconn.h" |
064af421 BP |
57 | #include "xtoxll.h" |
58 | ||
59 | #define THIS_MODULE VLM_ofproto | |
60 | #include "vlog.h" | |
61 | ||
72b06300 | 62 | #include "sflow_api.h" |
064af421 BP |
63 | |
64 | enum { | |
65 | TABLEID_HASH = 0, | |
66 | TABLEID_CLASSIFIER = 1 | |
67 | }; | |
68 | ||
69 | struct ofport { | |
70 | struct netdev *netdev; | |
71 | struct ofp_phy_port opp; /* In host byte order. */ | |
72 | }; | |
73 | ||
74 | static void ofport_free(struct ofport *); | |
75 | static void hton_ofp_phy_port(struct ofp_phy_port *); | |
76 | ||
77 | static int xlate_actions(const union ofp_action *in, size_t n_in, | |
78 | const flow_t *flow, struct ofproto *ofproto, | |
79 | const struct ofpbuf *packet, | |
80 | struct odp_actions *out, tag_type *tags, | |
6a07af36 | 81 | bool *may_set_up_flow, uint16_t *nf_output_iface); |
064af421 BP |
82 | |
83 | struct rule { | |
84 | struct cls_rule cr; | |
85 | ||
39997502 JP |
86 | uint64_t flow_cookie; /* Controller-issued identifier. |
87 | (Kept in network-byte order.) */ | |
064af421 BP |
88 | uint16_t idle_timeout; /* In seconds from time of last use. */ |
89 | uint16_t hard_timeout; /* In seconds from time of creation. */ | |
ca069229 | 90 | bool send_flow_removed; /* Send a flow removed message? */ |
064af421 BP |
91 | long long int used; /* Last-used time (0 if never used). */ |
92 | long long int created; /* Creation time. */ | |
93 | uint64_t packet_count; /* Number of packets received. */ | |
94 | uint64_t byte_count; /* Number of bytes received. */ | |
95 | uint64_t accounted_bytes; /* Number of bytes passed to account_cb. */ | |
064af421 | 96 | tag_type tags; /* Tags (set only by hooks). */ |
0193b2af | 97 | struct netflow_flow nf_flow; /* Per-flow NetFlow tracking data. */ |
064af421 BP |
98 | |
99 | /* If 'super' is non-NULL, this rule is a subrule, that is, it is an | |
100 | * exact-match rule (having cr.wc.wildcards of 0) generated from the | |
101 | * wildcard rule 'super'. In this case, 'list' is an element of the | |
102 | * super-rule's list. | |
103 | * | |
104 | * If 'super' is NULL, this rule is a super-rule, and 'list' is the head of | |
105 | * a list of subrules. A super-rule with no wildcards (where | |
106 | * cr.wc.wildcards is 0) will never have any subrules. */ | |
107 | struct rule *super; | |
108 | struct list list; | |
109 | ||
110 | /* OpenFlow actions. | |
79eee1eb BP |
111 | * |
112 | * 'n_actions' is the number of elements in the 'actions' array. A single | |
113 | * action may take up more more than one element's worth of space. | |
064af421 BP |
114 | * |
115 | * A subrule has no actions (it uses the super-rule's actions). */ | |
116 | int n_actions; | |
117 | union ofp_action *actions; | |
118 | ||
119 | /* Datapath actions. | |
120 | * | |
121 | * A super-rule with wildcard fields never has ODP actions (since the | |
122 | * datapath only supports exact-match flows). */ | |
123 | bool installed; /* Installed in datapath? */ | |
124 | bool may_install; /* True ordinarily; false if actions must | |
125 | * be reassessed for every packet. */ | |
126 | int n_odp_actions; | |
127 | union odp_action *odp_actions; | |
128 | }; | |
129 | ||
130 | static inline bool | |
131 | rule_is_hidden(const struct rule *rule) | |
132 | { | |
133 | /* Subrules are merely an implementation detail, so hide them from the | |
134 | * controller. */ | |
135 | if (rule->super != NULL) { | |
136 | return true; | |
137 | } | |
138 | ||
8cd4882f | 139 | /* Rules with priority higher than UINT16_MAX are set up by ofproto itself |
064af421 BP |
140 | * (e.g. by in-band control) and are intentionally hidden from the |
141 | * controller. */ | |
142 | if (rule->cr.priority > UINT16_MAX) { | |
143 | return true; | |
144 | } | |
145 | ||
146 | return false; | |
147 | } | |
148 | ||
0193b2af JG |
149 | static struct rule *rule_create(struct ofproto *, struct rule *super, |
150 | const union ofp_action *, size_t n_actions, | |
ca069229 | 151 | uint16_t idle_timeout, uint16_t hard_timeout, |
39997502 | 152 | uint64_t flow_cookie, bool send_flow_removed); |
064af421 BP |
153 | static void rule_free(struct rule *); |
154 | static void rule_destroy(struct ofproto *, struct rule *); | |
155 | static struct rule *rule_from_cls_rule(const struct cls_rule *); | |
156 | static void rule_insert(struct ofproto *, struct rule *, | |
157 | struct ofpbuf *packet, uint16_t in_port); | |
158 | static void rule_remove(struct ofproto *, struct rule *); | |
159 | static bool rule_make_actions(struct ofproto *, struct rule *, | |
160 | const struct ofpbuf *packet); | |
161 | static void rule_install(struct ofproto *, struct rule *, | |
162 | struct rule *displaced_rule); | |
163 | static void rule_uninstall(struct ofproto *, struct rule *); | |
164 | static void rule_post_uninstall(struct ofproto *, struct rule *); | |
ca069229 JP |
165 | static void send_flow_removed(struct ofproto *p, struct rule *rule, |
166 | long long int now, uint8_t reason); | |
064af421 | 167 | |
76ce9432 BP |
168 | /* ofproto supports two kinds of OpenFlow connections: |
169 | * | |
170 | * - "Controller connections": Connections to ordinary OpenFlow controllers. | |
171 | * ofproto maintains persistent connections to these controllers and by | |
172 | * default sends them asynchronous messages such as packet-ins. | |
173 | * | |
174 | * - "Transient connections", e.g. from ovs-ofctl. When these connections | |
175 | * drop, it is the other side's responsibility to reconnect them if | |
176 | * necessary. ofproto does not send them asynchronous messages by default. | |
177 | */ | |
178 | enum ofconn_type { | |
179 | OFCONN_CONTROLLER, /* An OpenFlow controller. */ | |
180 | OFCONN_TRANSIENT /* A transient connection. */ | |
181 | }; | |
064af421 | 182 | |
76ce9432 BP |
183 | /* An OpenFlow connection. */ |
184 | struct ofconn { | |
185 | struct ofproto *ofproto; /* The ofproto that owns this connection. */ | |
186 | struct list node; /* In struct ofproto's "all_conns" list. */ | |
187 | struct rconn *rconn; /* OpenFlow connection. */ | |
188 | enum ofconn_type type; /* Type. */ | |
189 | ||
190 | /* OFPT_PACKET_IN related data. */ | |
191 | struct rconn_packet_counter *packet_in_counter; /* # queued on 'rconn'. */ | |
192 | struct pinsched *schedulers[2]; /* Indexed by reason code; see below. */ | |
193 | struct pktbuf *pktbuf; /* OpenFlow packet buffers. */ | |
194 | int miss_send_len; /* Bytes to send of buffered packets. */ | |
195 | ||
196 | /* Number of OpenFlow messages queued on 'rconn' as replies to OpenFlow | |
197 | * requests, and the maximum number before we stop reading OpenFlow | |
198 | * requests. */ | |
064af421 BP |
199 | #define OFCONN_REPLY_MAX 100 |
200 | struct rconn_packet_counter *reply_counter; | |
76ce9432 BP |
201 | |
202 | /* type == OFCONN_CONTROLLER only. */ | |
9deba63b | 203 | enum nx_role role; /* Role. */ |
76ce9432 BP |
204 | struct hmap_node hmap_node; /* In struct ofproto's "controllers" map. */ |
205 | struct discovery *discovery; /* Controller discovery object, if enabled. */ | |
206 | struct status_category *ss; /* Switch status category. */ | |
d2ede7bc | 207 | enum ofproto_band band; /* In-band or out-of-band? */ |
064af421 BP |
208 | }; |
209 | ||
76ce9432 BP |
210 | /* We use OFPR_NO_MATCH and OFPR_ACTION as indexes into struct ofconn's |
211 | * "schedulers" array. Their values are 0 and 1, and their meanings and values | |
212 | * coincide with _ODPL_MISS_NR and _ODPL_ACTION_NR, so this is convenient. In | |
213 | * case anything ever changes, check their values here. */ | |
214 | #define N_SCHEDULERS 2 | |
215 | BUILD_ASSERT_DECL(OFPR_NO_MATCH == 0); | |
216 | BUILD_ASSERT_DECL(OFPR_NO_MATCH == _ODPL_MISS_NR); | |
217 | BUILD_ASSERT_DECL(OFPR_ACTION == 1); | |
218 | BUILD_ASSERT_DECL(OFPR_ACTION == _ODPL_ACTION_NR); | |
219 | ||
220 | static struct ofconn *ofconn_create(struct ofproto *, struct rconn *, | |
221 | enum ofconn_type); | |
c475ae67 | 222 | static void ofconn_destroy(struct ofconn *); |
064af421 BP |
223 | static void ofconn_run(struct ofconn *, struct ofproto *); |
224 | static void ofconn_wait(struct ofconn *); | |
225 | static void queue_tx(struct ofpbuf *msg, const struct ofconn *ofconn, | |
226 | struct rconn_packet_counter *counter); | |
227 | ||
76ce9432 BP |
228 | static void send_packet_in(struct ofproto *, struct ofpbuf *odp_msg); |
229 | static void do_send_packet_in(struct ofpbuf *odp_msg, void *ofconn); | |
230 | ||
064af421 BP |
231 | struct ofproto { |
232 | /* Settings. */ | |
233 | uint64_t datapath_id; /* Datapath ID. */ | |
234 | uint64_t fallback_dpid; /* Datapath ID if no better choice found. */ | |
5a719c38 JP |
235 | char *mfr_desc; /* Manufacturer. */ |
236 | char *hw_desc; /* Hardware. */ | |
237 | char *sw_desc; /* Software version. */ | |
238 | char *serial_desc; /* Serial number. */ | |
8abc4ed7 | 239 | char *dp_desc; /* Datapath description. */ |
064af421 BP |
240 | |
241 | /* Datapath. */ | |
c228a364 | 242 | struct dpif *dpif; |
e9e28be3 | 243 | struct netdev_monitor *netdev_monitor; |
064af421 BP |
244 | struct port_array ports; /* Index is ODP port nr; ofport->opp.port_no is |
245 | * OFP port nr. */ | |
246 | struct shash port_by_name; | |
247 | uint32_t max_ports; | |
248 | ||
249 | /* Configuration. */ | |
250 | struct switch_status *switch_status; | |
064af421 | 251 | struct fail_open *fail_open; |
064af421 | 252 | struct netflow *netflow; |
72b06300 | 253 | struct ofproto_sflow *sflow; |
064af421 | 254 | |
d2ede7bc BP |
255 | /* In-band control. */ |
256 | struct in_band *in_band; | |
257 | long long int next_in_band_update; | |
917e50e1 BP |
258 | struct sockaddr_in *extra_in_band_remotes; |
259 | size_t n_extra_remotes; | |
260 | ||
064af421 BP |
261 | /* Flow table. */ |
262 | struct classifier cls; | |
263 | bool need_revalidate; | |
264 | long long int next_expiration; | |
265 | struct tag_set revalidate_set; | |
659586ef | 266 | bool tun_id_from_cookie; |
064af421 BP |
267 | |
268 | /* OpenFlow connections. */ | |
76ce9432 BP |
269 | struct hmap controllers; /* Controller "struct ofconn"s. */ |
270 | struct list all_conns; /* Contains "struct ofconn"s. */ | |
064af421 BP |
271 | struct pvconn **listeners; |
272 | size_t n_listeners; | |
273 | struct pvconn **snoops; | |
274 | size_t n_snoops; | |
275 | ||
276 | /* Hooks for ovs-vswitchd. */ | |
277 | const struct ofhooks *ofhooks; | |
278 | void *aux; | |
279 | ||
280 | /* Used by default ofhooks. */ | |
281 | struct mac_learning *ml; | |
282 | }; | |
283 | ||
284 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
285 | ||
286 | static const struct ofhooks default_ofhooks; | |
287 | ||
fa60c019 | 288 | static uint64_t pick_datapath_id(const struct ofproto *); |
064af421 | 289 | static uint64_t pick_fallback_dpid(void); |
76ce9432 | 290 | |
064af421 | 291 | static void update_used(struct ofproto *); |
0193b2af JG |
292 | static void update_stats(struct ofproto *, struct rule *, |
293 | const struct odp_flow_stats *); | |
064af421 | 294 | static void expire_rule(struct cls_rule *, void *ofproto); |
0193b2af | 295 | static void active_timeout(struct ofproto *ofproto, struct rule *rule); |
064af421 BP |
296 | static bool revalidate_rule(struct ofproto *p, struct rule *rule); |
297 | static void revalidate_cb(struct cls_rule *rule_, void *p_); | |
298 | ||
299 | static void handle_odp_msg(struct ofproto *, struct ofpbuf *); | |
300 | ||
301 | static void handle_openflow(struct ofconn *, struct ofproto *, | |
302 | struct ofpbuf *); | |
303 | ||
72b06300 BP |
304 | static void refresh_port_groups(struct ofproto *); |
305 | ||
064af421 BP |
306 | static void update_port(struct ofproto *, const char *devname); |
307 | static int init_ports(struct ofproto *); | |
308 | static void reinit_ports(struct ofproto *); | |
309 | ||
310 | int | |
1a6f1e2a JG |
311 | ofproto_create(const char *datapath, const char *datapath_type, |
312 | const struct ofhooks *ofhooks, void *aux, | |
064af421 BP |
313 | struct ofproto **ofprotop) |
314 | { | |
064af421 BP |
315 | struct odp_stats stats; |
316 | struct ofproto *p; | |
c228a364 | 317 | struct dpif *dpif; |
064af421 BP |
318 | int error; |
319 | ||
320 | *ofprotop = NULL; | |
321 | ||
322 | /* Connect to datapath and start listening for messages. */ | |
1a6f1e2a | 323 | error = dpif_open(datapath, datapath_type, &dpif); |
064af421 BP |
324 | if (error) { |
325 | VLOG_ERR("failed to open datapath %s: %s", datapath, strerror(error)); | |
326 | return error; | |
327 | } | |
c228a364 | 328 | error = dpif_get_dp_stats(dpif, &stats); |
064af421 BP |
329 | if (error) { |
330 | VLOG_ERR("failed to obtain stats for datapath %s: %s", | |
331 | datapath, strerror(error)); | |
c228a364 | 332 | dpif_close(dpif); |
064af421 BP |
333 | return error; |
334 | } | |
72b06300 | 335 | error = dpif_recv_set_mask(dpif, ODPL_MISS | ODPL_ACTION | ODPL_SFLOW); |
064af421 BP |
336 | if (error) { |
337 | VLOG_ERR("failed to listen on datapath %s: %s", | |
338 | datapath, strerror(error)); | |
c228a364 | 339 | dpif_close(dpif); |
064af421 BP |
340 | return error; |
341 | } | |
c228a364 | 342 | dpif_flow_flush(dpif); |
8f24562a | 343 | dpif_recv_purge(dpif); |
064af421 BP |
344 | |
345 | /* Initialize settings. */ | |
ec6fde61 | 346 | p = xzalloc(sizeof *p); |
064af421 | 347 | p->fallback_dpid = pick_fallback_dpid(); |
fa60c019 | 348 | p->datapath_id = p->fallback_dpid; |
5a719c38 JP |
349 | p->mfr_desc = xstrdup(DEFAULT_MFR_DESC); |
350 | p->hw_desc = xstrdup(DEFAULT_HW_DESC); | |
351 | p->sw_desc = xstrdup(DEFAULT_SW_DESC); | |
352 | p->serial_desc = xstrdup(DEFAULT_SERIAL_DESC); | |
23ff2821 | 353 | p->dp_desc = xstrdup(DEFAULT_DP_DESC); |
064af421 BP |
354 | |
355 | /* Initialize datapath. */ | |
356 | p->dpif = dpif; | |
8b61709d | 357 | p->netdev_monitor = netdev_monitor_create(); |
064af421 BP |
358 | port_array_init(&p->ports); |
359 | shash_init(&p->port_by_name); | |
360 | p->max_ports = stats.max_ports; | |
361 | ||
362 | /* Initialize submodules. */ | |
363 | p->switch_status = switch_status_create(p); | |
364 | p->in_band = NULL; | |
064af421 | 365 | p->fail_open = NULL; |
064af421 | 366 | p->netflow = NULL; |
72b06300 | 367 | p->sflow = NULL; |
064af421 BP |
368 | |
369 | /* Initialize flow table. */ | |
370 | classifier_init(&p->cls); | |
371 | p->need_revalidate = false; | |
372 | p->next_expiration = time_msec() + 1000; | |
373 | tag_set_init(&p->revalidate_set); | |
374 | ||
375 | /* Initialize OpenFlow connections. */ | |
376 | list_init(&p->all_conns); | |
76ce9432 | 377 | hmap_init(&p->controllers); |
064af421 BP |
378 | p->listeners = NULL; |
379 | p->n_listeners = 0; | |
380 | p->snoops = NULL; | |
381 | p->n_snoops = 0; | |
382 | ||
383 | /* Initialize hooks. */ | |
384 | if (ofhooks) { | |
385 | p->ofhooks = ofhooks; | |
386 | p->aux = aux; | |
387 | p->ml = NULL; | |
388 | } else { | |
389 | p->ofhooks = &default_ofhooks; | |
390 | p->aux = p; | |
391 | p->ml = mac_learning_create(); | |
392 | } | |
393 | ||
fa60c019 BP |
394 | /* Pick final datapath ID. */ |
395 | p->datapath_id = pick_datapath_id(p); | |
b123cc3c | 396 | VLOG_INFO("using datapath ID %016"PRIx64, p->datapath_id); |
fa60c019 | 397 | |
064af421 BP |
398 | *ofprotop = p; |
399 | return 0; | |
400 | } | |
401 | ||
402 | void | |
403 | ofproto_set_datapath_id(struct ofproto *p, uint64_t datapath_id) | |
404 | { | |
405 | uint64_t old_dpid = p->datapath_id; | |
fa60c019 | 406 | p->datapath_id = datapath_id ? datapath_id : pick_datapath_id(p); |
064af421 | 407 | if (p->datapath_id != old_dpid) { |
76ce9432 BP |
408 | struct ofconn *ofconn; |
409 | ||
b123cc3c | 410 | VLOG_INFO("datapath ID changed to %016"PRIx64, p->datapath_id); |
76ce9432 BP |
411 | |
412 | /* Force all active connections to reconnect, since there is no way to | |
413 | * notify a controller that the datapath ID has changed. */ | |
414 | LIST_FOR_EACH (ofconn, struct ofconn, node, &p->all_conns) { | |
415 | rconn_reconnect(ofconn->rconn); | |
416 | } | |
064af421 BP |
417 | } |
418 | } | |
419 | ||
76ce9432 BP |
420 | static bool |
421 | is_discovery_controller(const struct ofproto_controller *c) | |
422 | { | |
423 | return !strcmp(c->target, "discover"); | |
424 | } | |
425 | ||
426 | static bool | |
427 | is_in_band_controller(const struct ofproto_controller *c) | |
428 | { | |
429 | return is_discovery_controller(c) || c->band == OFPROTO_IN_BAND; | |
430 | } | |
431 | ||
432 | /* Creates a new controller in 'ofproto'. Some of the settings are initially | |
433 | * drawn from 'c', but update_controller() needs to be called later to finish | |
434 | * the new ofconn's configuration. */ | |
435 | static void | |
436 | add_controller(struct ofproto *ofproto, const struct ofproto_controller *c) | |
437 | { | |
438 | struct discovery *discovery; | |
439 | struct ofconn *ofconn; | |
440 | ||
441 | if (is_discovery_controller(c)) { | |
442 | int error = discovery_create(c->accept_re, c->update_resolv_conf, | |
443 | ofproto->dpif, ofproto->switch_status, | |
444 | &discovery); | |
445 | if (error) { | |
446 | return; | |
447 | } | |
448 | } else { | |
449 | discovery = NULL; | |
450 | } | |
451 | ||
452 | ofconn = ofconn_create(ofproto, rconn_create(5, 8), OFCONN_CONTROLLER); | |
453 | ofconn->pktbuf = pktbuf_create(); | |
454 | ofconn->miss_send_len = OFP_DEFAULT_MISS_SEND_LEN; | |
455 | if (discovery) { | |
456 | ofconn->discovery = discovery; | |
457 | } else { | |
458 | rconn_connect(ofconn->rconn, c->target); | |
459 | } | |
460 | hmap_insert(&ofproto->controllers, &ofconn->hmap_node, | |
461 | hash_string(c->target, 0)); | |
462 | } | |
463 | ||
464 | /* Reconfigures 'ofconn' to match 'c'. This function cannot update an ofconn's | |
465 | * target or turn discovery on or off (these are done by creating new ofconns | |
466 | * and deleting old ones), but it can update the rest of an ofconn's | |
467 | * settings. */ | |
468 | static void | |
469 | update_controller(struct ofconn *ofconn, const struct ofproto_controller *c) | |
064af421 | 470 | { |
76ce9432 BP |
471 | struct ofproto *ofproto = ofconn->ofproto; |
472 | int probe_interval; | |
473 | int i; | |
79c9f2ee | 474 | |
d2ede7bc BP |
475 | ofconn->band = (is_in_band_controller(c) |
476 | ? OFPROTO_IN_BAND : OFPROTO_OUT_OF_BAND); | |
477 | ||
76ce9432 | 478 | rconn_set_max_backoff(ofconn->rconn, c->max_backoff); |
79c9f2ee | 479 | |
76ce9432 BP |
480 | probe_interval = c->probe_interval ? MAX(c->probe_interval, 5) : 0; |
481 | rconn_set_probe_interval(ofconn->rconn, probe_interval); | |
79c9f2ee | 482 | |
76ce9432 BP |
483 | if (ofconn->discovery) { |
484 | discovery_set_update_resolv_conf(ofconn->discovery, | |
485 | c->update_resolv_conf); | |
486 | discovery_set_accept_controller_re(ofconn->discovery, c->accept_re); | |
487 | } | |
79c9f2ee | 488 | |
76ce9432 BP |
489 | for (i = 0; i < N_SCHEDULERS; i++) { |
490 | struct pinsched **s = &ofconn->schedulers[i]; | |
79c9f2ee | 491 | |
76ce9432 BP |
492 | if (c->rate_limit > 0) { |
493 | if (!*s) { | |
494 | *s = pinsched_create(c->rate_limit, c->burst_limit, | |
495 | ofproto->switch_status); | |
79c9f2ee | 496 | } else { |
76ce9432 | 497 | pinsched_set_limits(*s, c->rate_limit, c->burst_limit); |
79c9f2ee | 498 | } |
76ce9432 BP |
499 | } else { |
500 | pinsched_destroy(*s); | |
501 | *s = NULL; | |
79c9f2ee | 502 | } |
76ce9432 BP |
503 | } |
504 | } | |
79c9f2ee | 505 | |
76ce9432 BP |
506 | static const char * |
507 | ofconn_get_target(const struct ofconn *ofconn) | |
508 | { | |
509 | return ofconn->discovery ? "discover" : rconn_get_name(ofconn->rconn); | |
510 | } | |
511 | ||
512 | static struct ofconn * | |
513 | find_controller_by_target(struct ofproto *ofproto, const char *target) | |
514 | { | |
515 | struct ofconn *ofconn; | |
516 | ||
517 | HMAP_FOR_EACH_WITH_HASH (ofconn, struct ofconn, hmap_node, | |
518 | hash_string(target, 0), &ofproto->controllers) { | |
519 | if (!strcmp(ofconn_get_target(ofconn), target)) { | |
520 | return ofconn; | |
79c9f2ee | 521 | } |
064af421 | 522 | } |
76ce9432 BP |
523 | return NULL; |
524 | } | |
064af421 | 525 | |
d2ede7bc BP |
526 | static void |
527 | update_in_band_remotes(struct ofproto *ofproto) | |
528 | { | |
529 | const struct ofconn *ofconn; | |
530 | struct sockaddr_in *addrs; | |
917e50e1 | 531 | size_t max_addrs, n_addrs; |
d2ede7bc | 532 | bool discovery; |
917e50e1 | 533 | size_t i; |
d2ede7bc | 534 | |
917e50e1 BP |
535 | /* Allocate enough memory for as many remotes as we could possibly have. */ |
536 | max_addrs = ofproto->n_extra_remotes + hmap_count(&ofproto->controllers); | |
537 | addrs = xmalloc(max_addrs * sizeof *addrs); | |
d2ede7bc BP |
538 | n_addrs = 0; |
539 | ||
540 | /* Add all the remotes. */ | |
541 | discovery = false; | |
542 | HMAP_FOR_EACH (ofconn, struct ofconn, hmap_node, &ofproto->controllers) { | |
543 | struct sockaddr_in *sin = &addrs[n_addrs]; | |
544 | ||
545 | sin->sin_addr.s_addr = rconn_get_remote_ip(ofconn->rconn); | |
546 | if (sin->sin_addr.s_addr) { | |
547 | sin->sin_port = rconn_get_remote_port(ofconn->rconn); | |
548 | n_addrs++; | |
549 | } | |
550 | if (ofconn->discovery) { | |
551 | discovery = true; | |
552 | } | |
553 | } | |
917e50e1 BP |
554 | for (i = 0; i < ofproto->n_extra_remotes; i++) { |
555 | addrs[n_addrs++] = ofproto->extra_in_band_remotes[i]; | |
556 | } | |
d2ede7bc BP |
557 | |
558 | /* Create or update or destroy in-band. | |
559 | * | |
560 | * Ordinarily we only enable in-band if there's at least one remote | |
561 | * address, but discovery needs the in-band rules for DHCP to be installed | |
562 | * even before we know any remote addresses. */ | |
563 | if (n_addrs || discovery) { | |
564 | if (!ofproto->in_band) { | |
565 | in_band_create(ofproto, ofproto->dpif, ofproto->switch_status, | |
566 | &ofproto->in_band); | |
567 | } | |
568 | in_band_set_remotes(ofproto->in_band, addrs, n_addrs); | |
569 | ofproto->next_in_band_update = time_msec() + 1000; | |
570 | } else { | |
571 | in_band_destroy(ofproto->in_band); | |
572 | ofproto->in_band = NULL; | |
573 | } | |
574 | ||
575 | /* Clean up. */ | |
576 | free(addrs); | |
577 | } | |
578 | ||
76ce9432 BP |
579 | void |
580 | ofproto_set_controllers(struct ofproto *p, | |
581 | const struct ofproto_controller *controllers, | |
582 | size_t n_controllers) | |
583 | { | |
584 | struct shash new_controllers; | |
76ce9432 BP |
585 | enum ofproto_fail_mode fail_mode; |
586 | struct ofconn *ofconn, *next; | |
587 | bool ss_exists; | |
76ce9432 | 588 | size_t i; |
79c9f2ee | 589 | |
76ce9432 BP |
590 | shash_init(&new_controllers); |
591 | for (i = 0; i < n_controllers; i++) { | |
592 | const struct ofproto_controller *c = &controllers[i]; | |
593 | ||
594 | shash_add_once(&new_controllers, c->target, &controllers[i]); | |
595 | if (!find_controller_by_target(p, c->target)) { | |
596 | add_controller(p, c); | |
597 | } | |
598 | } | |
599 | ||
76ce9432 BP |
600 | fail_mode = OFPROTO_FAIL_STANDALONE; |
601 | ss_exists = false; | |
602 | HMAP_FOR_EACH_SAFE (ofconn, next, struct ofconn, hmap_node, | |
603 | &p->controllers) { | |
604 | struct ofproto_controller *c; | |
605 | ||
606 | c = shash_find_data(&new_controllers, ofconn_get_target(ofconn)); | |
607 | if (!c) { | |
608 | ofconn_destroy(ofconn); | |
79c9f2ee | 609 | } else { |
76ce9432 | 610 | update_controller(ofconn, c); |
76ce9432 BP |
611 | if (ofconn->ss) { |
612 | ss_exists = true; | |
613 | } | |
76ce9432 BP |
614 | if (c->fail == OFPROTO_FAIL_SECURE) { |
615 | fail_mode = OFPROTO_FAIL_SECURE; | |
616 | } | |
617 | } | |
618 | } | |
619 | shash_destroy(&new_controllers); | |
620 | ||
d2ede7bc | 621 | update_in_band_remotes(p); |
76ce9432 BP |
622 | |
623 | if (!hmap_is_empty(&p->controllers) | |
624 | && fail_mode == OFPROTO_FAIL_STANDALONE) { | |
625 | struct rconn **rconns; | |
626 | size_t n; | |
79c9f2ee | 627 | |
79c9f2ee | 628 | if (!p->fail_open) { |
76ce9432 BP |
629 | p->fail_open = fail_open_create(p, p->switch_status); |
630 | } | |
631 | ||
632 | n = 0; | |
633 | rconns = xmalloc(hmap_count(&p->controllers) * sizeof *rconns); | |
634 | HMAP_FOR_EACH (ofconn, struct ofconn, hmap_node, &p->controllers) { | |
635 | rconns[n++] = ofconn->rconn; | |
79c9f2ee | 636 | } |
76ce9432 BP |
637 | |
638 | fail_open_set_controllers(p->fail_open, rconns, n); | |
639 | /* p->fail_open takes ownership of 'rconns'. */ | |
79c9f2ee BP |
640 | } else { |
641 | fail_open_destroy(p->fail_open); | |
642 | p->fail_open = NULL; | |
643 | } | |
644 | ||
76ce9432 BP |
645 | if (!hmap_is_empty(&p->controllers) && !ss_exists) { |
646 | ofconn = CONTAINER_OF(hmap_first(&p->controllers), | |
647 | struct ofconn, hmap_node); | |
648 | ofconn->ss = switch_status_register(p->switch_status, "remote", | |
649 | rconn_status_cb, ofconn->rconn); | |
79c9f2ee | 650 | } |
064af421 BP |
651 | } |
652 | ||
917e50e1 BP |
653 | static bool |
654 | any_extras_changed(const struct ofproto *ofproto, | |
655 | const struct sockaddr_in *extras, size_t n) | |
656 | { | |
657 | size_t i; | |
658 | ||
659 | if (n != ofproto->n_extra_remotes) { | |
660 | return true; | |
661 | } | |
662 | ||
663 | for (i = 0; i < n; i++) { | |
664 | const struct sockaddr_in *old = &ofproto->extra_in_band_remotes[i]; | |
665 | const struct sockaddr_in *new = &extras[i]; | |
666 | ||
667 | if (old->sin_addr.s_addr != new->sin_addr.s_addr || | |
668 | old->sin_port != new->sin_port) { | |
669 | return true; | |
670 | } | |
671 | } | |
672 | ||
673 | return false; | |
674 | } | |
675 | ||
676 | /* Sets the 'n' TCP port addresses in 'extras' as ones to which 'ofproto''s | |
677 | * in-band control should guarantee access, in the same way that in-band | |
678 | * control guarantees access to OpenFlow controllers. */ | |
679 | void | |
680 | ofproto_set_extra_in_band_remotes(struct ofproto *ofproto, | |
681 | const struct sockaddr_in *extras, size_t n) | |
682 | { | |
683 | if (!any_extras_changed(ofproto, extras, n)) { | |
684 | return; | |
685 | } | |
686 | ||
687 | free(ofproto->extra_in_band_remotes); | |
688 | ofproto->n_extra_remotes = n; | |
689 | ofproto->extra_in_band_remotes = xmemdup(extras, n * sizeof *extras); | |
690 | ||
691 | update_in_band_remotes(ofproto); | |
692 | } | |
693 | ||
064af421 BP |
694 | void |
695 | ofproto_set_desc(struct ofproto *p, | |
5a719c38 JP |
696 | const char *mfr_desc, const char *hw_desc, |
697 | const char *sw_desc, const char *serial_desc, | |
8abc4ed7 | 698 | const char *dp_desc) |
064af421 | 699 | { |
5a719c38 JP |
700 | struct ofp_desc_stats *ods; |
701 | ||
702 | if (mfr_desc) { | |
703 | if (strlen(mfr_desc) >= sizeof ods->mfr_desc) { | |
704 | VLOG_WARN("truncating mfr_desc, must be less than %zu characters", | |
705 | sizeof ods->mfr_desc); | |
706 | } | |
707 | free(p->mfr_desc); | |
708 | p->mfr_desc = xstrdup(mfr_desc); | |
064af421 | 709 | } |
5a719c38 JP |
710 | if (hw_desc) { |
711 | if (strlen(hw_desc) >= sizeof ods->hw_desc) { | |
712 | VLOG_WARN("truncating hw_desc, must be less than %zu characters", | |
713 | sizeof ods->hw_desc); | |
714 | } | |
715 | free(p->hw_desc); | |
716 | p->hw_desc = xstrdup(hw_desc); | |
064af421 | 717 | } |
5a719c38 JP |
718 | if (sw_desc) { |
719 | if (strlen(sw_desc) >= sizeof ods->sw_desc) { | |
720 | VLOG_WARN("truncating sw_desc, must be less than %zu characters", | |
721 | sizeof ods->sw_desc); | |
722 | } | |
723 | free(p->sw_desc); | |
724 | p->sw_desc = xstrdup(sw_desc); | |
725 | } | |
726 | if (serial_desc) { | |
727 | if (strlen(serial_desc) >= sizeof ods->serial_num) { | |
728 | VLOG_WARN("truncating serial_desc, must be less than %zu " | |
729 | "characters", | |
730 | sizeof ods->serial_num); | |
731 | } | |
732 | free(p->serial_desc); | |
733 | p->serial_desc = xstrdup(serial_desc); | |
064af421 | 734 | } |
8abc4ed7 | 735 | if (dp_desc) { |
5a719c38 JP |
736 | if (strlen(dp_desc) >= sizeof ods->dp_desc) { |
737 | VLOG_WARN("truncating dp_desc, must be less than %zu characters", | |
738 | sizeof ods->dp_desc); | |
739 | } | |
8abc4ed7 JP |
740 | free(p->dp_desc); |
741 | p->dp_desc = xstrdup(dp_desc); | |
742 | } | |
064af421 BP |
743 | } |
744 | ||
064af421 BP |
745 | static int |
746 | set_pvconns(struct pvconn ***pvconnsp, size_t *n_pvconnsp, | |
747 | const struct svec *svec) | |
748 | { | |
749 | struct pvconn **pvconns = *pvconnsp; | |
750 | size_t n_pvconns = *n_pvconnsp; | |
751 | int retval = 0; | |
752 | size_t i; | |
753 | ||
754 | for (i = 0; i < n_pvconns; i++) { | |
755 | pvconn_close(pvconns[i]); | |
756 | } | |
757 | free(pvconns); | |
758 | ||
759 | pvconns = xmalloc(svec->n * sizeof *pvconns); | |
760 | n_pvconns = 0; | |
761 | for (i = 0; i < svec->n; i++) { | |
762 | const char *name = svec->names[i]; | |
763 | struct pvconn *pvconn; | |
764 | int error; | |
765 | ||
766 | error = pvconn_open(name, &pvconn); | |
767 | if (!error) { | |
768 | pvconns[n_pvconns++] = pvconn; | |
769 | } else { | |
770 | VLOG_ERR("failed to listen on %s: %s", name, strerror(error)); | |
771 | if (!retval) { | |
772 | retval = error; | |
773 | } | |
774 | } | |
775 | } | |
776 | ||
777 | *pvconnsp = pvconns; | |
778 | *n_pvconnsp = n_pvconns; | |
779 | ||
780 | return retval; | |
781 | } | |
782 | ||
783 | int | |
784 | ofproto_set_listeners(struct ofproto *ofproto, const struct svec *listeners) | |
785 | { | |
786 | return set_pvconns(&ofproto->listeners, &ofproto->n_listeners, listeners); | |
787 | } | |
788 | ||
789 | int | |
790 | ofproto_set_snoops(struct ofproto *ofproto, const struct svec *snoops) | |
791 | { | |
792 | return set_pvconns(&ofproto->snoops, &ofproto->n_snoops, snoops); | |
793 | } | |
794 | ||
795 | int | |
0193b2af JG |
796 | ofproto_set_netflow(struct ofproto *ofproto, |
797 | const struct netflow_options *nf_options) | |
064af421 | 798 | { |
76343538 | 799 | if (nf_options && nf_options->collectors.n) { |
064af421 BP |
800 | if (!ofproto->netflow) { |
801 | ofproto->netflow = netflow_create(); | |
802 | } | |
0193b2af | 803 | return netflow_set_options(ofproto->netflow, nf_options); |
064af421 BP |
804 | } else { |
805 | netflow_destroy(ofproto->netflow); | |
806 | ofproto->netflow = NULL; | |
807 | return 0; | |
808 | } | |
809 | } | |
810 | ||
72b06300 BP |
811 | void |
812 | ofproto_set_sflow(struct ofproto *ofproto, | |
813 | const struct ofproto_sflow_options *oso) | |
814 | { | |
815 | struct ofproto_sflow *os = ofproto->sflow; | |
816 | if (oso) { | |
817 | if (!os) { | |
818 | struct ofport *ofport; | |
819 | unsigned int odp_port; | |
820 | ||
821 | os = ofproto->sflow = ofproto_sflow_create(ofproto->dpif); | |
822 | refresh_port_groups(ofproto); | |
823 | PORT_ARRAY_FOR_EACH (ofport, &ofproto->ports, odp_port) { | |
824 | ofproto_sflow_add_port(os, odp_port, | |
825 | netdev_get_name(ofport->netdev)); | |
826 | } | |
827 | } | |
828 | ofproto_sflow_set_options(os, oso); | |
829 | } else { | |
830 | ofproto_sflow_destroy(os); | |
831 | ofproto->sflow = NULL; | |
832 | } | |
833 | } | |
834 | ||
064af421 | 835 | int |
67a4917b | 836 | ofproto_set_stp(struct ofproto *ofproto OVS_UNUSED, bool enable_stp) |
064af421 BP |
837 | { |
838 | /* XXX */ | |
839 | if (enable_stp) { | |
840 | VLOG_WARN("STP is not yet implemented"); | |
841 | return EINVAL; | |
842 | } else { | |
843 | return 0; | |
844 | } | |
845 | } | |
846 | ||
064af421 BP |
847 | uint64_t |
848 | ofproto_get_datapath_id(const struct ofproto *ofproto) | |
849 | { | |
850 | return ofproto->datapath_id; | |
851 | } | |
852 | ||
76ce9432 BP |
853 | bool |
854 | ofproto_has_controller(const struct ofproto *ofproto) | |
064af421 | 855 | { |
76ce9432 | 856 | return !hmap_is_empty(&ofproto->controllers); |
064af421 BP |
857 | } |
858 | ||
859 | void | |
860 | ofproto_get_listeners(const struct ofproto *ofproto, struct svec *listeners) | |
861 | { | |
862 | size_t i; | |
863 | ||
864 | for (i = 0; i < ofproto->n_listeners; i++) { | |
865 | svec_add(listeners, pvconn_get_name(ofproto->listeners[i])); | |
866 | } | |
867 | } | |
868 | ||
869 | void | |
870 | ofproto_get_snoops(const struct ofproto *ofproto, struct svec *snoops) | |
871 | { | |
872 | size_t i; | |
873 | ||
874 | for (i = 0; i < ofproto->n_snoops; i++) { | |
875 | svec_add(snoops, pvconn_get_name(ofproto->snoops[i])); | |
876 | } | |
877 | } | |
878 | ||
879 | void | |
880 | ofproto_destroy(struct ofproto *p) | |
881 | { | |
882 | struct ofconn *ofconn, *next_ofconn; | |
883 | struct ofport *ofport; | |
884 | unsigned int port_no; | |
885 | size_t i; | |
886 | ||
887 | if (!p) { | |
888 | return; | |
889 | } | |
890 | ||
f7de2cdf | 891 | /* Destroy fail-open and in-band early, since they touch the classifier. */ |
79c9f2ee BP |
892 | fail_open_destroy(p->fail_open); |
893 | p->fail_open = NULL; | |
894 | ||
895 | in_band_destroy(p->in_band); | |
896 | p->in_band = NULL; | |
917e50e1 | 897 | free(p->extra_in_band_remotes); |
2f6d3445 | 898 | |
064af421 BP |
899 | ofproto_flush_flows(p); |
900 | classifier_destroy(&p->cls); | |
901 | ||
902 | LIST_FOR_EACH_SAFE (ofconn, next_ofconn, struct ofconn, node, | |
903 | &p->all_conns) { | |
c475ae67 | 904 | ofconn_destroy(ofconn); |
064af421 | 905 | } |
76ce9432 | 906 | hmap_destroy(&p->controllers); |
064af421 | 907 | |
c228a364 | 908 | dpif_close(p->dpif); |
e9e28be3 | 909 | netdev_monitor_destroy(p->netdev_monitor); |
064af421 BP |
910 | PORT_ARRAY_FOR_EACH (ofport, &p->ports, port_no) { |
911 | ofport_free(ofport); | |
912 | } | |
913 | shash_destroy(&p->port_by_name); | |
914 | ||
915 | switch_status_destroy(p->switch_status); | |
064af421 | 916 | netflow_destroy(p->netflow); |
72b06300 | 917 | ofproto_sflow_destroy(p->sflow); |
064af421 | 918 | |
064af421 BP |
919 | for (i = 0; i < p->n_listeners; i++) { |
920 | pvconn_close(p->listeners[i]); | |
921 | } | |
922 | free(p->listeners); | |
923 | ||
924 | for (i = 0; i < p->n_snoops; i++) { | |
925 | pvconn_close(p->snoops[i]); | |
926 | } | |
927 | free(p->snoops); | |
928 | ||
929 | mac_learning_destroy(p->ml); | |
930 | ||
5a719c38 JP |
931 | free(p->mfr_desc); |
932 | free(p->hw_desc); | |
933 | free(p->sw_desc); | |
934 | free(p->serial_desc); | |
cb871ae0 JP |
935 | free(p->dp_desc); |
936 | ||
3b917492 BP |
937 | port_array_destroy(&p->ports); |
938 | ||
064af421 BP |
939 | free(p); |
940 | } | |
941 | ||
942 | int | |
943 | ofproto_run(struct ofproto *p) | |
944 | { | |
945 | int error = ofproto_run1(p); | |
946 | if (!error) { | |
947 | error = ofproto_run2(p, false); | |
948 | } | |
949 | return error; | |
950 | } | |
951 | ||
e9e28be3 BP |
952 | static void |
953 | process_port_change(struct ofproto *ofproto, int error, char *devname) | |
954 | { | |
955 | if (error == ENOBUFS) { | |
956 | reinit_ports(ofproto); | |
957 | } else if (!error) { | |
958 | update_port(ofproto, devname); | |
959 | free(devname); | |
960 | } | |
961 | } | |
962 | ||
76ce9432 BP |
963 | /* One of ofproto's "snoop" pvconns has accepted a new connection on 'vconn'. |
964 | * Connects this vconn to a controller. */ | |
965 | static void | |
966 | add_snooper(struct ofproto *ofproto, struct vconn *vconn) | |
967 | { | |
968 | struct ofconn *ofconn; | |
969 | ||
970 | /* Arbitrarily pick the first controller in the list for monitoring. We | |
971 | * could do something smarter or more flexible later, if it ever proves | |
972 | * useful. */ | |
973 | LIST_FOR_EACH (ofconn, struct ofconn, node, &ofproto->all_conns) { | |
974 | if (ofconn->type == OFCONN_CONTROLLER) { | |
975 | rconn_add_monitor(ofconn->rconn, vconn); | |
976 | return; | |
977 | } | |
978 | ||
979 | } | |
980 | VLOG_INFO_RL(&rl, "no controller connection to monitor"); | |
981 | vconn_close(vconn); | |
982 | } | |
983 | ||
064af421 BP |
984 | int |
985 | ofproto_run1(struct ofproto *p) | |
986 | { | |
987 | struct ofconn *ofconn, *next_ofconn; | |
988 | char *devname; | |
989 | int error; | |
990 | int i; | |
991 | ||
149f577a JG |
992 | if (shash_is_empty(&p->port_by_name)) { |
993 | init_ports(p); | |
994 | } | |
995 | ||
064af421 BP |
996 | for (i = 0; i < 50; i++) { |
997 | struct ofpbuf *buf; | |
998 | int error; | |
999 | ||
c228a364 | 1000 | error = dpif_recv(p->dpif, &buf); |
064af421 BP |
1001 | if (error) { |
1002 | if (error == ENODEV) { | |
1003 | /* Someone destroyed the datapath behind our back. The caller | |
1004 | * better destroy us and give up, because we're just going to | |
1005 | * spin from here on out. */ | |
39a559f2 BP |
1006 | static struct vlog_rate_limit rl2 = VLOG_RATE_LIMIT_INIT(1, 5); |
1007 | VLOG_ERR_RL(&rl2, "%s: datapath was destroyed externally", | |
c228a364 | 1008 | dpif_name(p->dpif)); |
064af421 BP |
1009 | return ENODEV; |
1010 | } | |
1011 | break; | |
1012 | } | |
1013 | ||
1014 | handle_odp_msg(p, buf); | |
1015 | } | |
1016 | ||
e9e28be3 BP |
1017 | while ((error = dpif_port_poll(p->dpif, &devname)) != EAGAIN) { |
1018 | process_port_change(p, error, devname); | |
1019 | } | |
1020 | while ((error = netdev_monitor_poll(p->netdev_monitor, | |
1021 | &devname)) != EAGAIN) { | |
1022 | process_port_change(p, error, devname); | |
064af421 BP |
1023 | } |
1024 | ||
1025 | if (p->in_band) { | |
d2ede7bc BP |
1026 | if (time_msec() >= p->next_in_band_update) { |
1027 | update_in_band_remotes(p); | |
1028 | } | |
064af421 BP |
1029 | in_band_run(p->in_band); |
1030 | } | |
064af421 BP |
1031 | |
1032 | LIST_FOR_EACH_SAFE (ofconn, next_ofconn, struct ofconn, node, | |
1033 | &p->all_conns) { | |
1034 | ofconn_run(ofconn, p); | |
1035 | } | |
1036 | ||
7778bd15 BP |
1037 | /* Fail-open maintenance. Do this after processing the ofconns since |
1038 | * fail-open checks the status of the controller rconn. */ | |
1039 | if (p->fail_open) { | |
1040 | fail_open_run(p->fail_open); | |
1041 | } | |
1042 | ||
064af421 BP |
1043 | for (i = 0; i < p->n_listeners; i++) { |
1044 | struct vconn *vconn; | |
1045 | int retval; | |
1046 | ||
1047 | retval = pvconn_accept(p->listeners[i], OFP_VERSION, &vconn); | |
1048 | if (!retval) { | |
76ce9432 BP |
1049 | ofconn_create(p, rconn_new_from_vconn("passive", vconn), |
1050 | OFCONN_TRANSIENT); | |
064af421 BP |
1051 | } else if (retval != EAGAIN) { |
1052 | VLOG_WARN_RL(&rl, "accept failed (%s)", strerror(retval)); | |
1053 | } | |
1054 | } | |
1055 | ||
1056 | for (i = 0; i < p->n_snoops; i++) { | |
1057 | struct vconn *vconn; | |
1058 | int retval; | |
1059 | ||
1060 | retval = pvconn_accept(p->snoops[i], OFP_VERSION, &vconn); | |
1061 | if (!retval) { | |
76ce9432 | 1062 | add_snooper(p, vconn); |
064af421 BP |
1063 | } else if (retval != EAGAIN) { |
1064 | VLOG_WARN_RL(&rl, "accept failed (%s)", strerror(retval)); | |
1065 | } | |
1066 | } | |
1067 | ||
1068 | if (time_msec() >= p->next_expiration) { | |
1069 | COVERAGE_INC(ofproto_expiration); | |
1070 | p->next_expiration = time_msec() + 1000; | |
1071 | update_used(p); | |
1072 | ||
1073 | classifier_for_each(&p->cls, CLS_INC_ALL, expire_rule, p); | |
1074 | ||
1075 | /* Let the hook know that we're at a stable point: all outstanding data | |
1076 | * in existing flows has been accounted to the account_cb. Thus, the | |
1077 | * hook can now reasonably do operations that depend on having accurate | |
1078 | * flow volume accounting (currently, that's just bond rebalancing). */ | |
1079 | if (p->ofhooks->account_checkpoint_cb) { | |
1080 | p->ofhooks->account_checkpoint_cb(p->aux); | |
1081 | } | |
1082 | } | |
1083 | ||
1084 | if (p->netflow) { | |
1085 | netflow_run(p->netflow); | |
1086 | } | |
72b06300 BP |
1087 | if (p->sflow) { |
1088 | ofproto_sflow_run(p->sflow); | |
1089 | } | |
064af421 BP |
1090 | |
1091 | return 0; | |
1092 | } | |
1093 | ||
1094 | struct revalidate_cbdata { | |
1095 | struct ofproto *ofproto; | |
1096 | bool revalidate_all; /* Revalidate all exact-match rules? */ | |
1097 | bool revalidate_subrules; /* Revalidate all exact-match subrules? */ | |
1098 | struct tag_set revalidate_set; /* Set of tags to revalidate. */ | |
1099 | }; | |
1100 | ||
1101 | int | |
1102 | ofproto_run2(struct ofproto *p, bool revalidate_all) | |
1103 | { | |
1104 | if (p->need_revalidate || revalidate_all | |
1105 | || !tag_set_is_empty(&p->revalidate_set)) { | |
1106 | struct revalidate_cbdata cbdata; | |
1107 | cbdata.ofproto = p; | |
1108 | cbdata.revalidate_all = revalidate_all; | |
1109 | cbdata.revalidate_subrules = p->need_revalidate; | |
1110 | cbdata.revalidate_set = p->revalidate_set; | |
1111 | tag_set_init(&p->revalidate_set); | |
1112 | COVERAGE_INC(ofproto_revalidate); | |
1113 | classifier_for_each(&p->cls, CLS_INC_EXACT, revalidate_cb, &cbdata); | |
1114 | p->need_revalidate = false; | |
1115 | } | |
1116 | ||
1117 | return 0; | |
1118 | } | |
1119 | ||
1120 | void | |
1121 | ofproto_wait(struct ofproto *p) | |
1122 | { | |
1123 | struct ofconn *ofconn; | |
1124 | size_t i; | |
1125 | ||
c228a364 | 1126 | dpif_recv_wait(p->dpif); |
e9e28be3 BP |
1127 | dpif_port_poll_wait(p->dpif); |
1128 | netdev_monitor_poll_wait(p->netdev_monitor); | |
064af421 BP |
1129 | LIST_FOR_EACH (ofconn, struct ofconn, node, &p->all_conns) { |
1130 | ofconn_wait(ofconn); | |
1131 | } | |
1132 | if (p->in_band) { | |
d2ede7bc | 1133 | poll_timer_wait(p->next_in_band_update - time_msec()); |
064af421 BP |
1134 | in_band_wait(p->in_band); |
1135 | } | |
064af421 BP |
1136 | if (p->fail_open) { |
1137 | fail_open_wait(p->fail_open); | |
1138 | } | |
72b06300 BP |
1139 | if (p->sflow) { |
1140 | ofproto_sflow_wait(p->sflow); | |
1141 | } | |
064af421 BP |
1142 | if (!tag_set_is_empty(&p->revalidate_set)) { |
1143 | poll_immediate_wake(); | |
1144 | } | |
1145 | if (p->need_revalidate) { | |
1146 | /* Shouldn't happen, but if it does just go around again. */ | |
1147 | VLOG_DBG_RL(&rl, "need revalidate in ofproto_wait_cb()"); | |
1148 | poll_immediate_wake(); | |
1149 | } else if (p->next_expiration != LLONG_MAX) { | |
1150 | poll_timer_wait(p->next_expiration - time_msec()); | |
1151 | } | |
1152 | for (i = 0; i < p->n_listeners; i++) { | |
1153 | pvconn_wait(p->listeners[i]); | |
1154 | } | |
1155 | for (i = 0; i < p->n_snoops; i++) { | |
1156 | pvconn_wait(p->snoops[i]); | |
1157 | } | |
1158 | } | |
1159 | ||
1160 | void | |
1161 | ofproto_revalidate(struct ofproto *ofproto, tag_type tag) | |
1162 | { | |
1163 | tag_set_add(&ofproto->revalidate_set, tag); | |
1164 | } | |
1165 | ||
1166 | struct tag_set * | |
1167 | ofproto_get_revalidate_set(struct ofproto *ofproto) | |
1168 | { | |
1169 | return &ofproto->revalidate_set; | |
1170 | } | |
1171 | ||
1172 | bool | |
1173 | ofproto_is_alive(const struct ofproto *p) | |
1174 | { | |
76ce9432 | 1175 | return !hmap_is_empty(&p->controllers); |
064af421 BP |
1176 | } |
1177 | ||
1178 | int | |
1179 | ofproto_send_packet(struct ofproto *p, const flow_t *flow, | |
1180 | const union ofp_action *actions, size_t n_actions, | |
1181 | const struct ofpbuf *packet) | |
1182 | { | |
1183 | struct odp_actions odp_actions; | |
1184 | int error; | |
1185 | ||
1186 | error = xlate_actions(actions, n_actions, flow, p, packet, &odp_actions, | |
6a07af36 | 1187 | NULL, NULL, NULL); |
064af421 BP |
1188 | if (error) { |
1189 | return error; | |
1190 | } | |
1191 | ||
1192 | /* XXX Should we translate the dpif_execute() errno value into an OpenFlow | |
1193 | * error code? */ | |
c228a364 | 1194 | dpif_execute(p->dpif, flow->in_port, odp_actions.actions, |
064af421 BP |
1195 | odp_actions.n_actions, packet); |
1196 | return 0; | |
1197 | } | |
1198 | ||
1199 | void | |
1200 | ofproto_add_flow(struct ofproto *p, | |
1201 | const flow_t *flow, uint32_t wildcards, unsigned int priority, | |
1202 | const union ofp_action *actions, size_t n_actions, | |
1203 | int idle_timeout) | |
1204 | { | |
1205 | struct rule *rule; | |
0193b2af | 1206 | rule = rule_create(p, NULL, actions, n_actions, |
ca069229 | 1207 | idle_timeout >= 0 ? idle_timeout : 5 /* XXX */, |
39997502 | 1208 | 0, 0, false); |
659586ef | 1209 | cls_rule_from_flow(flow, wildcards, priority, &rule->cr); |
064af421 BP |
1210 | rule_insert(p, rule, NULL, 0); |
1211 | } | |
1212 | ||
1213 | void | |
1214 | ofproto_delete_flow(struct ofproto *ofproto, const flow_t *flow, | |
1215 | uint32_t wildcards, unsigned int priority) | |
1216 | { | |
1217 | struct rule *rule; | |
1218 | ||
1219 | rule = rule_from_cls_rule(classifier_find_rule_exactly(&ofproto->cls, | |
1220 | flow, wildcards, | |
1221 | priority)); | |
1222 | if (rule) { | |
1223 | rule_remove(ofproto, rule); | |
1224 | } | |
1225 | } | |
1226 | ||
1227 | static void | |
1228 | destroy_rule(struct cls_rule *rule_, void *ofproto_) | |
1229 | { | |
1230 | struct rule *rule = rule_from_cls_rule(rule_); | |
1231 | struct ofproto *ofproto = ofproto_; | |
1232 | ||
1233 | /* Mark the flow as not installed, even though it might really be | |
1234 | * installed, so that rule_remove() doesn't bother trying to uninstall it. | |
1235 | * There is no point in uninstalling it individually since we are about to | |
1236 | * blow away all the flows with dpif_flow_flush(). */ | |
1237 | rule->installed = false; | |
1238 | ||
1239 | rule_remove(ofproto, rule); | |
1240 | } | |
1241 | ||
1242 | void | |
1243 | ofproto_flush_flows(struct ofproto *ofproto) | |
1244 | { | |
1245 | COVERAGE_INC(ofproto_flush); | |
1246 | classifier_for_each(&ofproto->cls, CLS_INC_ALL, destroy_rule, ofproto); | |
c228a364 | 1247 | dpif_flow_flush(ofproto->dpif); |
064af421 BP |
1248 | if (ofproto->in_band) { |
1249 | in_band_flushed(ofproto->in_band); | |
1250 | } | |
1251 | if (ofproto->fail_open) { | |
1252 | fail_open_flushed(ofproto->fail_open); | |
1253 | } | |
1254 | } | |
1255 | \f | |
1256 | static void | |
1257 | reinit_ports(struct ofproto *p) | |
1258 | { | |
1259 | struct svec devnames; | |
1260 | struct ofport *ofport; | |
1261 | unsigned int port_no; | |
1262 | struct odp_port *odp_ports; | |
1263 | size_t n_odp_ports; | |
1264 | size_t i; | |
1265 | ||
1266 | svec_init(&devnames); | |
1267 | PORT_ARRAY_FOR_EACH (ofport, &p->ports, port_no) { | |
1268 | svec_add (&devnames, (char *) ofport->opp.name); | |
1269 | } | |
c228a364 | 1270 | dpif_port_list(p->dpif, &odp_ports, &n_odp_ports); |
064af421 BP |
1271 | for (i = 0; i < n_odp_ports; i++) { |
1272 | svec_add (&devnames, odp_ports[i].devname); | |
1273 | } | |
1274 | free(odp_ports); | |
1275 | ||
1276 | svec_sort_unique(&devnames); | |
1277 | for (i = 0; i < devnames.n; i++) { | |
1278 | update_port(p, devnames.names[i]); | |
1279 | } | |
1280 | svec_destroy(&devnames); | |
1281 | } | |
1282 | ||
72b06300 | 1283 | static size_t |
064af421 BP |
1284 | refresh_port_group(struct ofproto *p, unsigned int group) |
1285 | { | |
1286 | uint16_t *ports; | |
1287 | size_t n_ports; | |
1288 | struct ofport *port; | |
1289 | unsigned int port_no; | |
1290 | ||
1291 | assert(group == DP_GROUP_ALL || group == DP_GROUP_FLOOD); | |
1292 | ||
1293 | ports = xmalloc(port_array_count(&p->ports) * sizeof *ports); | |
1294 | n_ports = 0; | |
1295 | PORT_ARRAY_FOR_EACH (port, &p->ports, port_no) { | |
1296 | if (group == DP_GROUP_ALL || !(port->opp.config & OFPPC_NO_FLOOD)) { | |
1297 | ports[n_ports++] = port_no; | |
1298 | } | |
1299 | } | |
c228a364 | 1300 | dpif_port_group_set(p->dpif, group, ports, n_ports); |
064af421 | 1301 | free(ports); |
72b06300 BP |
1302 | |
1303 | return n_ports; | |
064af421 BP |
1304 | } |
1305 | ||
1306 | static void | |
1307 | refresh_port_groups(struct ofproto *p) | |
1308 | { | |
72b06300 BP |
1309 | size_t n_flood = refresh_port_group(p, DP_GROUP_FLOOD); |
1310 | size_t n_all = refresh_port_group(p, DP_GROUP_ALL); | |
1311 | if (p->sflow) { | |
1312 | ofproto_sflow_set_group_sizes(p->sflow, n_flood, n_all); | |
1313 | } | |
064af421 BP |
1314 | } |
1315 | ||
1316 | static struct ofport * | |
1317 | make_ofport(const struct odp_port *odp_port) | |
1318 | { | |
149f577a | 1319 | struct netdev_options netdev_options; |
064af421 BP |
1320 | enum netdev_flags flags; |
1321 | struct ofport *ofport; | |
1322 | struct netdev *netdev; | |
1323 | bool carrier; | |
1324 | int error; | |
1325 | ||
149f577a JG |
1326 | memset(&netdev_options, 0, sizeof netdev_options); |
1327 | netdev_options.name = odp_port->devname; | |
1328 | netdev_options.ethertype = NETDEV_ETH_TYPE_NONE; | |
1329 | netdev_options.may_open = true; | |
1330 | ||
1331 | error = netdev_open(&netdev_options, &netdev); | |
064af421 BP |
1332 | if (error) { |
1333 | VLOG_WARN_RL(&rl, "ignoring port %s (%"PRIu16") because netdev %s " | |
1334 | "cannot be opened (%s)", | |
1335 | odp_port->devname, odp_port->port, | |
1336 | odp_port->devname, strerror(error)); | |
1337 | return NULL; | |
1338 | } | |
1339 | ||
1340 | ofport = xmalloc(sizeof *ofport); | |
1341 | ofport->netdev = netdev; | |
1342 | ofport->opp.port_no = odp_port_to_ofp_port(odp_port->port); | |
80992a35 | 1343 | netdev_get_etheraddr(netdev, ofport->opp.hw_addr); |
064af421 BP |
1344 | memcpy(ofport->opp.name, odp_port->devname, |
1345 | MIN(sizeof ofport->opp.name, sizeof odp_port->devname)); | |
1346 | ofport->opp.name[sizeof ofport->opp.name - 1] = '\0'; | |
1347 | ||
1348 | netdev_get_flags(netdev, &flags); | |
1349 | ofport->opp.config = flags & NETDEV_UP ? 0 : OFPPC_PORT_DOWN; | |
1350 | ||
1351 | netdev_get_carrier(netdev, &carrier); | |
1352 | ofport->opp.state = carrier ? 0 : OFPPS_LINK_DOWN; | |
1353 | ||
1354 | netdev_get_features(netdev, | |
1355 | &ofport->opp.curr, &ofport->opp.advertised, | |
1356 | &ofport->opp.supported, &ofport->opp.peer); | |
1357 | return ofport; | |
1358 | } | |
1359 | ||
1360 | static bool | |
1361 | ofport_conflicts(const struct ofproto *p, const struct odp_port *odp_port) | |
1362 | { | |
1363 | if (port_array_get(&p->ports, odp_port->port)) { | |
1364 | VLOG_WARN_RL(&rl, "ignoring duplicate port %"PRIu16" in datapath", | |
1365 | odp_port->port); | |
1366 | return true; | |
1367 | } else if (shash_find(&p->port_by_name, odp_port->devname)) { | |
1368 | VLOG_WARN_RL(&rl, "ignoring duplicate device %s in datapath", | |
1369 | odp_port->devname); | |
1370 | return true; | |
1371 | } else { | |
1372 | return false; | |
1373 | } | |
1374 | } | |
1375 | ||
1376 | static int | |
1377 | ofport_equal(const struct ofport *a_, const struct ofport *b_) | |
1378 | { | |
1379 | const struct ofp_phy_port *a = &a_->opp; | |
1380 | const struct ofp_phy_port *b = &b_->opp; | |
1381 | ||
1382 | BUILD_ASSERT_DECL(sizeof *a == 48); /* Detect ofp_phy_port changes. */ | |
1383 | return (a->port_no == b->port_no | |
1384 | && !memcmp(a->hw_addr, b->hw_addr, sizeof a->hw_addr) | |
1385 | && !strcmp((char *) a->name, (char *) b->name) | |
1386 | && a->state == b->state | |
1387 | && a->config == b->config | |
1388 | && a->curr == b->curr | |
1389 | && a->advertised == b->advertised | |
1390 | && a->supported == b->supported | |
1391 | && a->peer == b->peer); | |
1392 | } | |
1393 | ||
1394 | static void | |
1395 | send_port_status(struct ofproto *p, const struct ofport *ofport, | |
1396 | uint8_t reason) | |
1397 | { | |
1398 | /* XXX Should limit the number of queued port status change messages. */ | |
1399 | struct ofconn *ofconn; | |
1400 | LIST_FOR_EACH (ofconn, struct ofconn, node, &p->all_conns) { | |
1401 | struct ofp_port_status *ops; | |
1402 | struct ofpbuf *b; | |
1403 | ||
9deba63b BP |
1404 | if (ofconn->role == NX_ROLE_SLAVE) { |
1405 | continue; | |
1406 | } | |
1407 | ||
064af421 BP |
1408 | ops = make_openflow_xid(sizeof *ops, OFPT_PORT_STATUS, 0, &b); |
1409 | ops->reason = reason; | |
1410 | ops->desc = ofport->opp; | |
1411 | hton_ofp_phy_port(&ops->desc); | |
1412 | queue_tx(b, ofconn, NULL); | |
1413 | } | |
1414 | if (p->ofhooks->port_changed_cb) { | |
1415 | p->ofhooks->port_changed_cb(reason, &ofport->opp, p->aux); | |
1416 | } | |
1417 | } | |
1418 | ||
1419 | static void | |
1420 | ofport_install(struct ofproto *p, struct ofport *ofport) | |
1421 | { | |
72b06300 BP |
1422 | uint16_t odp_port = ofp_port_to_odp_port(ofport->opp.port_no); |
1423 | const char *netdev_name = (const char *) ofport->opp.name; | |
1424 | ||
e9e28be3 | 1425 | netdev_monitor_add(p->netdev_monitor, ofport->netdev); |
72b06300 BP |
1426 | port_array_set(&p->ports, odp_port, ofport); |
1427 | shash_add(&p->port_by_name, netdev_name, ofport); | |
1428 | if (p->sflow) { | |
1429 | ofproto_sflow_add_port(p->sflow, odp_port, netdev_name); | |
1430 | } | |
064af421 BP |
1431 | } |
1432 | ||
1433 | static void | |
1434 | ofport_remove(struct ofproto *p, struct ofport *ofport) | |
1435 | { | |
72b06300 BP |
1436 | uint16_t odp_port = ofp_port_to_odp_port(ofport->opp.port_no); |
1437 | ||
e9e28be3 | 1438 | netdev_monitor_remove(p->netdev_monitor, ofport->netdev); |
72b06300 | 1439 | port_array_set(&p->ports, odp_port, NULL); |
064af421 BP |
1440 | shash_delete(&p->port_by_name, |
1441 | shash_find(&p->port_by_name, (char *) ofport->opp.name)); | |
72b06300 BP |
1442 | if (p->sflow) { |
1443 | ofproto_sflow_del_port(p->sflow, odp_port); | |
1444 | } | |
064af421 BP |
1445 | } |
1446 | ||
1447 | static void | |
1448 | ofport_free(struct ofport *ofport) | |
1449 | { | |
1450 | if (ofport) { | |
1451 | netdev_close(ofport->netdev); | |
1452 | free(ofport); | |
1453 | } | |
1454 | } | |
1455 | ||
1456 | static void | |
1457 | update_port(struct ofproto *p, const char *devname) | |
1458 | { | |
1459 | struct odp_port odp_port; | |
c874dc6d BP |
1460 | struct ofport *old_ofport; |
1461 | struct ofport *new_ofport; | |
064af421 BP |
1462 | int error; |
1463 | ||
1464 | COVERAGE_INC(ofproto_update_port); | |
c874dc6d BP |
1465 | |
1466 | /* Query the datapath for port information. */ | |
c228a364 | 1467 | error = dpif_port_query_by_name(p->dpif, devname, &odp_port); |
064af421 | 1468 | |
c874dc6d BP |
1469 | /* Find the old ofport. */ |
1470 | old_ofport = shash_find_data(&p->port_by_name, devname); | |
1471 | if (!error) { | |
1472 | if (!old_ofport) { | |
1473 | /* There's no port named 'devname' but there might be a port with | |
1474 | * the same port number. This could happen if a port is deleted | |
1475 | * and then a new one added in its place very quickly, or if a port | |
1476 | * is renamed. In the former case we want to send an OFPPR_DELETE | |
1477 | * and an OFPPR_ADD, and in the latter case we want to send a | |
1478 | * single OFPPR_MODIFY. We can distinguish the cases by comparing | |
1479 | * the old port's ifindex against the new port, or perhaps less | |
1480 | * reliably but more portably by comparing the old port's MAC | |
1481 | * against the new port's MAC. However, this code isn't that smart | |
1482 | * and always sends an OFPPR_MODIFY (XXX). */ | |
1483 | old_ofport = port_array_get(&p->ports, odp_port.port); | |
064af421 | 1484 | } |
c874dc6d | 1485 | } else if (error != ENOENT && error != ENODEV) { |
064af421 BP |
1486 | VLOG_WARN_RL(&rl, "dpif_port_query_by_name returned unexpected error " |
1487 | "%s", strerror(error)); | |
1488 | return; | |
1489 | } | |
c874dc6d BP |
1490 | |
1491 | /* Create a new ofport. */ | |
1492 | new_ofport = !error ? make_ofport(&odp_port) : NULL; | |
1493 | ||
1494 | /* Eliminate a few pathological cases. */ | |
1495 | if (!old_ofport && !new_ofport) { | |
1496 | return; | |
1497 | } else if (old_ofport && new_ofport) { | |
1498 | /* Most of the 'config' bits are OpenFlow soft state, but | |
1499 | * OFPPC_PORT_DOWN is maintained the kernel. So transfer the OpenFlow | |
1500 | * bits from old_ofport. (make_ofport() only sets OFPPC_PORT_DOWN and | |
1501 | * leaves the other bits 0.) */ | |
1502 | new_ofport->opp.config |= old_ofport->opp.config & ~OFPPC_PORT_DOWN; | |
1503 | ||
1504 | if (ofport_equal(old_ofport, new_ofport)) { | |
1505 | /* False alarm--no change. */ | |
1506 | ofport_free(new_ofport); | |
1507 | return; | |
1508 | } | |
1509 | } | |
1510 | ||
1511 | /* Now deal with the normal cases. */ | |
1512 | if (old_ofport) { | |
1513 | ofport_remove(p, old_ofport); | |
1514 | } | |
1515 | if (new_ofport) { | |
1516 | ofport_install(p, new_ofport); | |
1517 | } | |
1518 | send_port_status(p, new_ofport ? new_ofport : old_ofport, | |
1519 | (!old_ofport ? OFPPR_ADD | |
1520 | : !new_ofport ? OFPPR_DELETE | |
1521 | : OFPPR_MODIFY)); | |
1522 | ofport_free(old_ofport); | |
1523 | ||
1524 | /* Update port groups. */ | |
064af421 BP |
1525 | refresh_port_groups(p); |
1526 | } | |
1527 | ||
1528 | static int | |
1529 | init_ports(struct ofproto *p) | |
1530 | { | |
1531 | struct odp_port *ports; | |
1532 | size_t n_ports; | |
1533 | size_t i; | |
1534 | int error; | |
1535 | ||
c228a364 | 1536 | error = dpif_port_list(p->dpif, &ports, &n_ports); |
064af421 BP |
1537 | if (error) { |
1538 | return error; | |
1539 | } | |
1540 | ||
1541 | for (i = 0; i < n_ports; i++) { | |
1542 | const struct odp_port *odp_port = &ports[i]; | |
1543 | if (!ofport_conflicts(p, odp_port)) { | |
1544 | struct ofport *ofport = make_ofport(odp_port); | |
1545 | if (ofport) { | |
1546 | ofport_install(p, ofport); | |
1547 | } | |
1548 | } | |
1549 | } | |
1550 | free(ports); | |
1551 | refresh_port_groups(p); | |
1552 | return 0; | |
1553 | } | |
1554 | \f | |
1555 | static struct ofconn * | |
76ce9432 | 1556 | ofconn_create(struct ofproto *p, struct rconn *rconn, enum ofconn_type type) |
064af421 | 1557 | { |
76ce9432 BP |
1558 | struct ofconn *ofconn = xzalloc(sizeof *ofconn); |
1559 | ofconn->ofproto = p; | |
064af421 BP |
1560 | list_push_back(&p->all_conns, &ofconn->node); |
1561 | ofconn->rconn = rconn; | |
76ce9432 | 1562 | ofconn->type = type; |
9deba63b | 1563 | ofconn->role = NX_ROLE_OTHER; |
76ce9432 | 1564 | ofconn->packet_in_counter = rconn_packet_counter_create (); |
064af421 | 1565 | ofconn->pktbuf = NULL; |
064af421 | 1566 | ofconn->miss_send_len = 0; |
064af421 BP |
1567 | ofconn->reply_counter = rconn_packet_counter_create (); |
1568 | return ofconn; | |
1569 | } | |
1570 | ||
1571 | static void | |
c475ae67 | 1572 | ofconn_destroy(struct ofconn *ofconn) |
064af421 | 1573 | { |
76ce9432 BP |
1574 | if (ofconn->type == OFCONN_CONTROLLER) { |
1575 | hmap_remove(&ofconn->ofproto->controllers, &ofconn->hmap_node); | |
1576 | } | |
1577 | discovery_destroy(ofconn->discovery); | |
1578 | ||
064af421 | 1579 | list_remove(&ofconn->node); |
76ce9432 | 1580 | switch_status_unregister(ofconn->ss); |
064af421 BP |
1581 | rconn_destroy(ofconn->rconn); |
1582 | rconn_packet_counter_destroy(ofconn->packet_in_counter); | |
1583 | rconn_packet_counter_destroy(ofconn->reply_counter); | |
1584 | pktbuf_destroy(ofconn->pktbuf); | |
1585 | free(ofconn); | |
1586 | } | |
1587 | ||
1588 | static void | |
1589 | ofconn_run(struct ofconn *ofconn, struct ofproto *p) | |
1590 | { | |
1591 | int iteration; | |
76ce9432 BP |
1592 | size_t i; |
1593 | ||
1594 | if (ofconn->discovery) { | |
1595 | char *controller_name; | |
1596 | if (rconn_is_connectivity_questionable(ofconn->rconn)) { | |
1597 | discovery_question_connectivity(ofconn->discovery); | |
1598 | } | |
1599 | if (discovery_run(ofconn->discovery, &controller_name)) { | |
1600 | if (controller_name) { | |
1601 | rconn_connect(ofconn->rconn, controller_name); | |
1602 | } else { | |
1603 | rconn_disconnect(ofconn->rconn); | |
1604 | } | |
1605 | } | |
1606 | } | |
1607 | ||
1608 | for (i = 0; i < N_SCHEDULERS; i++) { | |
1609 | pinsched_run(ofconn->schedulers[i], do_send_packet_in, ofconn); | |
1610 | } | |
064af421 BP |
1611 | |
1612 | rconn_run(ofconn->rconn); | |
1613 | ||
1614 | if (rconn_packet_counter_read (ofconn->reply_counter) < OFCONN_REPLY_MAX) { | |
1615 | /* Limit the number of iterations to prevent other tasks from | |
1616 | * starving. */ | |
1617 | for (iteration = 0; iteration < 50; iteration++) { | |
1618 | struct ofpbuf *of_msg = rconn_recv(ofconn->rconn); | |
1619 | if (!of_msg) { | |
1620 | break; | |
1621 | } | |
7778bd15 BP |
1622 | if (p->fail_open) { |
1623 | fail_open_maybe_recover(p->fail_open); | |
1624 | } | |
064af421 BP |
1625 | handle_openflow(ofconn, p, of_msg); |
1626 | ofpbuf_delete(of_msg); | |
1627 | } | |
1628 | } | |
1629 | ||
76ce9432 | 1630 | if (!ofconn->discovery && !rconn_is_alive(ofconn->rconn)) { |
c475ae67 | 1631 | ofconn_destroy(ofconn); |
064af421 BP |
1632 | } |
1633 | } | |
1634 | ||
1635 | static void | |
1636 | ofconn_wait(struct ofconn *ofconn) | |
1637 | { | |
76ce9432 BP |
1638 | int i; |
1639 | ||
1640 | if (ofconn->discovery) { | |
1641 | discovery_wait(ofconn->discovery); | |
1642 | } | |
1643 | for (i = 0; i < N_SCHEDULERS; i++) { | |
1644 | pinsched_wait(ofconn->schedulers[i]); | |
1645 | } | |
064af421 BP |
1646 | rconn_run_wait(ofconn->rconn); |
1647 | if (rconn_packet_counter_read (ofconn->reply_counter) < OFCONN_REPLY_MAX) { | |
1648 | rconn_recv_wait(ofconn->rconn); | |
1649 | } else { | |
1650 | COVERAGE_INC(ofproto_ofconn_stuck); | |
1651 | } | |
1652 | } | |
1653 | \f | |
1654 | /* Caller is responsible for initializing the 'cr' member of the returned | |
1655 | * rule. */ | |
1656 | static struct rule * | |
0193b2af | 1657 | rule_create(struct ofproto *ofproto, struct rule *super, |
064af421 | 1658 | const union ofp_action *actions, size_t n_actions, |
ca069229 | 1659 | uint16_t idle_timeout, uint16_t hard_timeout, |
39997502 | 1660 | uint64_t flow_cookie, bool send_flow_removed) |
064af421 | 1661 | { |
ec6fde61 | 1662 | struct rule *rule = xzalloc(sizeof *rule); |
064af421 BP |
1663 | rule->idle_timeout = idle_timeout; |
1664 | rule->hard_timeout = hard_timeout; | |
39997502 | 1665 | rule->flow_cookie = flow_cookie; |
064af421 | 1666 | rule->used = rule->created = time_msec(); |
ca069229 | 1667 | rule->send_flow_removed = send_flow_removed; |
064af421 BP |
1668 | rule->super = super; |
1669 | if (super) { | |
1670 | list_push_back(&super->list, &rule->list); | |
1671 | } else { | |
1672 | list_init(&rule->list); | |
1673 | } | |
1674 | rule->n_actions = n_actions; | |
1675 | rule->actions = xmemdup(actions, n_actions * sizeof *actions); | |
0193b2af JG |
1676 | netflow_flow_clear(&rule->nf_flow); |
1677 | netflow_flow_update_time(ofproto->netflow, &rule->nf_flow, rule->created); | |
1678 | ||
064af421 BP |
1679 | return rule; |
1680 | } | |
1681 | ||
1682 | static struct rule * | |
1683 | rule_from_cls_rule(const struct cls_rule *cls_rule) | |
1684 | { | |
1685 | return cls_rule ? CONTAINER_OF(cls_rule, struct rule, cr) : NULL; | |
1686 | } | |
1687 | ||
1688 | static void | |
1689 | rule_free(struct rule *rule) | |
1690 | { | |
1691 | free(rule->actions); | |
1692 | free(rule->odp_actions); | |
1693 | free(rule); | |
1694 | } | |
1695 | ||
1696 | /* Destroys 'rule'. If 'rule' is a subrule, also removes it from its | |
1697 | * super-rule's list of subrules. If 'rule' is a super-rule, also iterates | |
1698 | * through all of its subrules and revalidates them, destroying any that no | |
1699 | * longer has a super-rule (which is probably all of them). | |
1700 | * | |
1701 | * Before calling this function, the caller must make have removed 'rule' from | |
1702 | * the classifier. If 'rule' is an exact-match rule, the caller is also | |
1703 | * responsible for ensuring that it has been uninstalled from the datapath. */ | |
1704 | static void | |
1705 | rule_destroy(struct ofproto *ofproto, struct rule *rule) | |
1706 | { | |
1707 | if (!rule->super) { | |
1708 | struct rule *subrule, *next; | |
1709 | LIST_FOR_EACH_SAFE (subrule, next, struct rule, list, &rule->list) { | |
1710 | revalidate_rule(ofproto, subrule); | |
1711 | } | |
1712 | } else { | |
1713 | list_remove(&rule->list); | |
1714 | } | |
1715 | rule_free(rule); | |
1716 | } | |
1717 | ||
1718 | static bool | |
1719 | rule_has_out_port(const struct rule *rule, uint16_t out_port) | |
1720 | { | |
1721 | const union ofp_action *oa; | |
1722 | struct actions_iterator i; | |
1723 | ||
1724 | if (out_port == htons(OFPP_NONE)) { | |
1725 | return true; | |
1726 | } | |
1727 | for (oa = actions_first(&i, rule->actions, rule->n_actions); oa; | |
1728 | oa = actions_next(&i)) { | |
1729 | if (oa->type == htons(OFPAT_OUTPUT) && oa->output.port == out_port) { | |
1730 | return true; | |
1731 | } | |
1732 | } | |
1733 | return false; | |
1734 | } | |
1735 | ||
1736 | /* Executes the actions indicated by 'rule' on 'packet', which is in flow | |
1737 | * 'flow' and is considered to have arrived on ODP port 'in_port'. | |
1738 | * | |
1739 | * The flow that 'packet' actually contains does not need to actually match | |
1740 | * 'rule'; the actions in 'rule' will be applied to it either way. Likewise, | |
1741 | * the packet and byte counters for 'rule' will be credited for the packet sent | |
1742 | * out whether or not the packet actually matches 'rule'. | |
1743 | * | |
1744 | * If 'rule' is an exact-match rule and 'flow' actually equals the rule's flow, | |
1745 | * the caller must already have accurately composed ODP actions for it given | |
1746 | * 'packet' using rule_make_actions(). If 'rule' is a wildcard rule, or if | |
1747 | * 'rule' is an exact-match rule but 'flow' is not the rule's flow, then this | |
1748 | * function will compose a set of ODP actions based on 'rule''s OpenFlow | |
1749 | * actions and apply them to 'packet'. */ | |
1750 | static void | |
1751 | rule_execute(struct ofproto *ofproto, struct rule *rule, | |
1752 | struct ofpbuf *packet, const flow_t *flow) | |
1753 | { | |
1754 | const union odp_action *actions; | |
1755 | size_t n_actions; | |
1756 | struct odp_actions a; | |
1757 | ||
1758 | /* Grab or compose the ODP actions. | |
1759 | * | |
1760 | * The special case for an exact-match 'rule' where 'flow' is not the | |
1761 | * rule's flow is important to avoid, e.g., sending a packet out its input | |
1762 | * port simply because the ODP actions were composed for the wrong | |
1763 | * scenario. */ | |
1764 | if (rule->cr.wc.wildcards || !flow_equal(flow, &rule->cr.flow)) { | |
1765 | struct rule *super = rule->super ? rule->super : rule; | |
1766 | if (xlate_actions(super->actions, super->n_actions, flow, ofproto, | |
6a07af36 | 1767 | packet, &a, NULL, 0, NULL)) { |
064af421 BP |
1768 | return; |
1769 | } | |
1770 | actions = a.actions; | |
1771 | n_actions = a.n_actions; | |
1772 | } else { | |
1773 | actions = rule->odp_actions; | |
1774 | n_actions = rule->n_odp_actions; | |
1775 | } | |
1776 | ||
1777 | /* Execute the ODP actions. */ | |
c228a364 | 1778 | if (!dpif_execute(ofproto->dpif, flow->in_port, |
064af421 BP |
1779 | actions, n_actions, packet)) { |
1780 | struct odp_flow_stats stats; | |
1781 | flow_extract_stats(flow, packet, &stats); | |
0193b2af | 1782 | update_stats(ofproto, rule, &stats); |
064af421 | 1783 | rule->used = time_msec(); |
0193b2af | 1784 | netflow_flow_update_time(ofproto->netflow, &rule->nf_flow, rule->used); |
064af421 BP |
1785 | } |
1786 | } | |
1787 | ||
1788 | static void | |
1789 | rule_insert(struct ofproto *p, struct rule *rule, struct ofpbuf *packet, | |
1790 | uint16_t in_port) | |
1791 | { | |
1792 | struct rule *displaced_rule; | |
1793 | ||
1794 | /* Insert the rule in the classifier. */ | |
1795 | displaced_rule = rule_from_cls_rule(classifier_insert(&p->cls, &rule->cr)); | |
1796 | if (!rule->cr.wc.wildcards) { | |
1797 | rule_make_actions(p, rule, packet); | |
1798 | } | |
1799 | ||
1800 | /* Send the packet and credit it to the rule. */ | |
1801 | if (packet) { | |
1802 | flow_t flow; | |
659586ef | 1803 | flow_extract(packet, 0, in_port, &flow); |
064af421 BP |
1804 | rule_execute(p, rule, packet, &flow); |
1805 | } | |
1806 | ||
1807 | /* Install the rule in the datapath only after sending the packet, to | |
1808 | * avoid packet reordering. */ | |
1809 | if (rule->cr.wc.wildcards) { | |
1810 | COVERAGE_INC(ofproto_add_wc_flow); | |
1811 | p->need_revalidate = true; | |
1812 | } else { | |
1813 | rule_install(p, rule, displaced_rule); | |
1814 | } | |
1815 | ||
1816 | /* Free the rule that was displaced, if any. */ | |
1817 | if (displaced_rule) { | |
1818 | rule_destroy(p, displaced_rule); | |
1819 | } | |
1820 | } | |
1821 | ||
1822 | static struct rule * | |
1823 | rule_create_subrule(struct ofproto *ofproto, struct rule *rule, | |
1824 | const flow_t *flow) | |
1825 | { | |
0193b2af | 1826 | struct rule *subrule = rule_create(ofproto, rule, NULL, 0, |
ca069229 | 1827 | rule->idle_timeout, rule->hard_timeout, |
39997502 | 1828 | 0, false); |
064af421 | 1829 | COVERAGE_INC(ofproto_subrule_create); |
659586ef JG |
1830 | cls_rule_from_flow(flow, 0, (rule->cr.priority <= UINT16_MAX ? UINT16_MAX |
1831 | : rule->cr.priority), &subrule->cr); | |
064af421 BP |
1832 | classifier_insert_exact(&ofproto->cls, &subrule->cr); |
1833 | ||
1834 | return subrule; | |
1835 | } | |
1836 | ||
1837 | static void | |
1838 | rule_remove(struct ofproto *ofproto, struct rule *rule) | |
1839 | { | |
1840 | if (rule->cr.wc.wildcards) { | |
1841 | COVERAGE_INC(ofproto_del_wc_flow); | |
1842 | ofproto->need_revalidate = true; | |
1843 | } else { | |
1844 | rule_uninstall(ofproto, rule); | |
1845 | } | |
1846 | classifier_remove(&ofproto->cls, &rule->cr); | |
1847 | rule_destroy(ofproto, rule); | |
1848 | } | |
1849 | ||
1850 | /* Returns true if the actions changed, false otherwise. */ | |
1851 | static bool | |
1852 | rule_make_actions(struct ofproto *p, struct rule *rule, | |
1853 | const struct ofpbuf *packet) | |
1854 | { | |
1855 | const struct rule *super; | |
1856 | struct odp_actions a; | |
1857 | size_t actions_len; | |
1858 | ||
1859 | assert(!rule->cr.wc.wildcards); | |
1860 | ||
1861 | super = rule->super ? rule->super : rule; | |
1862 | rule->tags = 0; | |
1863 | xlate_actions(super->actions, super->n_actions, &rule->cr.flow, p, | |
6a07af36 | 1864 | packet, &a, &rule->tags, &rule->may_install, |
0193b2af | 1865 | &rule->nf_flow.output_iface); |
064af421 BP |
1866 | |
1867 | actions_len = a.n_actions * sizeof *a.actions; | |
1868 | if (rule->n_odp_actions != a.n_actions | |
1869 | || memcmp(rule->odp_actions, a.actions, actions_len)) { | |
1870 | COVERAGE_INC(ofproto_odp_unchanged); | |
1871 | free(rule->odp_actions); | |
1872 | rule->n_odp_actions = a.n_actions; | |
1873 | rule->odp_actions = xmemdup(a.actions, actions_len); | |
1874 | return true; | |
1875 | } else { | |
1876 | return false; | |
1877 | } | |
1878 | } | |
1879 | ||
1880 | static int | |
1881 | do_put_flow(struct ofproto *ofproto, struct rule *rule, int flags, | |
1882 | struct odp_flow_put *put) | |
1883 | { | |
1884 | memset(&put->flow.stats, 0, sizeof put->flow.stats); | |
1885 | put->flow.key = rule->cr.flow; | |
1886 | put->flow.actions = rule->odp_actions; | |
1887 | put->flow.n_actions = rule->n_odp_actions; | |
ab48643b | 1888 | put->flow.flags = 0; |
064af421 | 1889 | put->flags = flags; |
c228a364 | 1890 | return dpif_flow_put(ofproto->dpif, put); |
064af421 BP |
1891 | } |
1892 | ||
1893 | static void | |
1894 | rule_install(struct ofproto *p, struct rule *rule, struct rule *displaced_rule) | |
1895 | { | |
1896 | assert(!rule->cr.wc.wildcards); | |
1897 | ||
1898 | if (rule->may_install) { | |
1899 | struct odp_flow_put put; | |
1900 | if (!do_put_flow(p, rule, | |
1901 | ODPPF_CREATE | ODPPF_MODIFY | ODPPF_ZERO_STATS, | |
1902 | &put)) { | |
1903 | rule->installed = true; | |
1904 | if (displaced_rule) { | |
14986b31 | 1905 | update_stats(p, displaced_rule, &put.flow.stats); |
064af421 BP |
1906 | rule_post_uninstall(p, displaced_rule); |
1907 | } | |
1908 | } | |
1909 | } else if (displaced_rule) { | |
1910 | rule_uninstall(p, displaced_rule); | |
1911 | } | |
1912 | } | |
1913 | ||
1914 | static void | |
1915 | rule_reinstall(struct ofproto *ofproto, struct rule *rule) | |
1916 | { | |
1917 | if (rule->installed) { | |
1918 | struct odp_flow_put put; | |
1919 | COVERAGE_INC(ofproto_dp_missed); | |
1920 | do_put_flow(ofproto, rule, ODPPF_CREATE | ODPPF_MODIFY, &put); | |
1921 | } else { | |
1922 | rule_install(ofproto, rule, NULL); | |
1923 | } | |
1924 | } | |
1925 | ||
1926 | static void | |
1927 | rule_update_actions(struct ofproto *ofproto, struct rule *rule) | |
1928 | { | |
42c3641c JG |
1929 | bool actions_changed; |
1930 | uint16_t new_out_iface, old_out_iface; | |
1931 | ||
1932 | old_out_iface = rule->nf_flow.output_iface; | |
1933 | actions_changed = rule_make_actions(ofproto, rule, NULL); | |
1934 | ||
064af421 BP |
1935 | if (rule->may_install) { |
1936 | if (rule->installed) { | |
1937 | if (actions_changed) { | |
064af421 | 1938 | struct odp_flow_put put; |
42c3641c JG |
1939 | do_put_flow(ofproto, rule, ODPPF_CREATE | ODPPF_MODIFY |
1940 | | ODPPF_ZERO_STATS, &put); | |
1941 | update_stats(ofproto, rule, &put.flow.stats); | |
1942 | ||
1943 | /* Temporarily set the old output iface so that NetFlow | |
1944 | * messages have the correct output interface for the old | |
1945 | * stats. */ | |
1946 | new_out_iface = rule->nf_flow.output_iface; | |
1947 | rule->nf_flow.output_iface = old_out_iface; | |
1948 | rule_post_uninstall(ofproto, rule); | |
1949 | rule->nf_flow.output_iface = new_out_iface; | |
064af421 BP |
1950 | } |
1951 | } else { | |
1952 | rule_install(ofproto, rule, NULL); | |
1953 | } | |
1954 | } else { | |
1955 | rule_uninstall(ofproto, rule); | |
1956 | } | |
1957 | } | |
1958 | ||
1959 | static void | |
1960 | rule_account(struct ofproto *ofproto, struct rule *rule, uint64_t extra_bytes) | |
1961 | { | |
1962 | uint64_t total_bytes = rule->byte_count + extra_bytes; | |
1963 | ||
1964 | if (ofproto->ofhooks->account_flow_cb | |
1965 | && total_bytes > rule->accounted_bytes) | |
1966 | { | |
1967 | ofproto->ofhooks->account_flow_cb( | |
1968 | &rule->cr.flow, rule->odp_actions, rule->n_odp_actions, | |
1969 | total_bytes - rule->accounted_bytes, ofproto->aux); | |
1970 | rule->accounted_bytes = total_bytes; | |
1971 | } | |
1972 | } | |
1973 | ||
1974 | static void | |
1975 | rule_uninstall(struct ofproto *p, struct rule *rule) | |
1976 | { | |
1977 | assert(!rule->cr.wc.wildcards); | |
1978 | if (rule->installed) { | |
1979 | struct odp_flow odp_flow; | |
1980 | ||
1981 | odp_flow.key = rule->cr.flow; | |
1982 | odp_flow.actions = NULL; | |
1983 | odp_flow.n_actions = 0; | |
ab48643b | 1984 | odp_flow.flags = 0; |
c228a364 | 1985 | if (!dpif_flow_del(p->dpif, &odp_flow)) { |
0193b2af | 1986 | update_stats(p, rule, &odp_flow.stats); |
064af421 BP |
1987 | } |
1988 | rule->installed = false; | |
1989 | ||
1990 | rule_post_uninstall(p, rule); | |
1991 | } | |
1992 | } | |
1993 | ||
0193b2af JG |
1994 | static bool |
1995 | is_controller_rule(struct rule *rule) | |
1996 | { | |
1997 | /* If the only action is send to the controller then don't report | |
1998 | * NetFlow expiration messages since it is just part of the control | |
1999 | * logic for the network and not real traffic. */ | |
2000 | ||
2001 | if (rule && rule->super) { | |
2002 | struct rule *super = rule->super; | |
2003 | ||
2004 | return super->n_actions == 1 && | |
2005 | super->actions[0].type == htons(OFPAT_OUTPUT) && | |
2006 | super->actions[0].output.port == htons(OFPP_CONTROLLER); | |
2007 | } | |
2008 | ||
2009 | return false; | |
2010 | } | |
2011 | ||
064af421 BP |
2012 | static void |
2013 | rule_post_uninstall(struct ofproto *ofproto, struct rule *rule) | |
2014 | { | |
2015 | struct rule *super = rule->super; | |
2016 | ||
2017 | rule_account(ofproto, rule, 0); | |
6a07af36 | 2018 | |
0193b2af | 2019 | if (ofproto->netflow && !is_controller_rule(rule)) { |
064af421 BP |
2020 | struct ofexpired expired; |
2021 | expired.flow = rule->cr.flow; | |
2022 | expired.packet_count = rule->packet_count; | |
2023 | expired.byte_count = rule->byte_count; | |
2024 | expired.used = rule->used; | |
0193b2af | 2025 | netflow_expire(ofproto->netflow, &rule->nf_flow, &expired); |
064af421 BP |
2026 | } |
2027 | if (super) { | |
2028 | super->packet_count += rule->packet_count; | |
2029 | super->byte_count += rule->byte_count; | |
064af421 | 2030 | |
0c0afbec JG |
2031 | /* Reset counters to prevent double counting if the rule ever gets |
2032 | * reinstalled. */ | |
2033 | rule->packet_count = 0; | |
2034 | rule->byte_count = 0; | |
2035 | rule->accounted_bytes = 0; | |
0193b2af JG |
2036 | |
2037 | netflow_flow_clear(&rule->nf_flow); | |
0c0afbec | 2038 | } |
064af421 BP |
2039 | } |
2040 | \f | |
2041 | static void | |
2042 | queue_tx(struct ofpbuf *msg, const struct ofconn *ofconn, | |
2043 | struct rconn_packet_counter *counter) | |
2044 | { | |
2045 | update_openflow_length(msg); | |
2046 | if (rconn_send(ofconn->rconn, msg, counter)) { | |
2047 | ofpbuf_delete(msg); | |
2048 | } | |
2049 | } | |
2050 | ||
2051 | static void | |
2052 | send_error(const struct ofconn *ofconn, const struct ofp_header *oh, | |
2053 | int error, const void *data, size_t len) | |
2054 | { | |
2055 | struct ofpbuf *buf; | |
2056 | struct ofp_error_msg *oem; | |
2057 | ||
2058 | if (!(error >> 16)) { | |
2059 | VLOG_WARN_RL(&rl, "not sending bad error code %d to controller", | |
2060 | error); | |
2061 | return; | |
2062 | } | |
2063 | ||
2064 | COVERAGE_INC(ofproto_error); | |
2065 | oem = make_openflow_xid(len + sizeof *oem, OFPT_ERROR, | |
2066 | oh ? oh->xid : 0, &buf); | |
2067 | oem->type = htons((unsigned int) error >> 16); | |
2068 | oem->code = htons(error & 0xffff); | |
2069 | memcpy(oem->data, data, len); | |
2070 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
2071 | } | |
2072 | ||
2073 | static void | |
2074 | send_error_oh(const struct ofconn *ofconn, const struct ofp_header *oh, | |
2075 | int error) | |
2076 | { | |
2077 | size_t oh_length = ntohs(oh->length); | |
2078 | send_error(ofconn, oh, error, oh, MIN(oh_length, 64)); | |
2079 | } | |
2080 | ||
2081 | static void | |
2082 | hton_ofp_phy_port(struct ofp_phy_port *opp) | |
2083 | { | |
2084 | opp->port_no = htons(opp->port_no); | |
2085 | opp->config = htonl(opp->config); | |
2086 | opp->state = htonl(opp->state); | |
2087 | opp->curr = htonl(opp->curr); | |
2088 | opp->advertised = htonl(opp->advertised); | |
2089 | opp->supported = htonl(opp->supported); | |
2090 | opp->peer = htonl(opp->peer); | |
2091 | } | |
2092 | ||
2093 | static int | |
2094 | handle_echo_request(struct ofconn *ofconn, struct ofp_header *oh) | |
2095 | { | |
2096 | struct ofp_header *rq = oh; | |
2097 | queue_tx(make_echo_reply(rq), ofconn, ofconn->reply_counter); | |
2098 | return 0; | |
2099 | } | |
2100 | ||
2101 | static int | |
2102 | handle_features_request(struct ofproto *p, struct ofconn *ofconn, | |
2103 | struct ofp_header *oh) | |
2104 | { | |
2105 | struct ofp_switch_features *osf; | |
2106 | struct ofpbuf *buf; | |
2107 | unsigned int port_no; | |
2108 | struct ofport *port; | |
2109 | ||
2110 | osf = make_openflow_xid(sizeof *osf, OFPT_FEATURES_REPLY, oh->xid, &buf); | |
2111 | osf->datapath_id = htonll(p->datapath_id); | |
2112 | osf->n_buffers = htonl(pktbuf_capacity()); | |
2113 | osf->n_tables = 2; | |
2114 | osf->capabilities = htonl(OFPC_FLOW_STATS | OFPC_TABLE_STATS | | |
0254ae23 | 2115 | OFPC_PORT_STATS | OFPC_ARP_MATCH_IP); |
064af421 BP |
2116 | osf->actions = htonl((1u << OFPAT_OUTPUT) | |
2117 | (1u << OFPAT_SET_VLAN_VID) | | |
2118 | (1u << OFPAT_SET_VLAN_PCP) | | |
2119 | (1u << OFPAT_STRIP_VLAN) | | |
2120 | (1u << OFPAT_SET_DL_SRC) | | |
2121 | (1u << OFPAT_SET_DL_DST) | | |
2122 | (1u << OFPAT_SET_NW_SRC) | | |
2123 | (1u << OFPAT_SET_NW_DST) | | |
959a2ecd | 2124 | (1u << OFPAT_SET_NW_TOS) | |
064af421 BP |
2125 | (1u << OFPAT_SET_TP_SRC) | |
2126 | (1u << OFPAT_SET_TP_DST)); | |
2127 | ||
2128 | PORT_ARRAY_FOR_EACH (port, &p->ports, port_no) { | |
2129 | hton_ofp_phy_port(ofpbuf_put(buf, &port->opp, sizeof port->opp)); | |
2130 | } | |
2131 | ||
2132 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
2133 | return 0; | |
2134 | } | |
2135 | ||
2136 | static int | |
2137 | handle_get_config_request(struct ofproto *p, struct ofconn *ofconn, | |
2138 | struct ofp_header *oh) | |
2139 | { | |
2140 | struct ofpbuf *buf; | |
2141 | struct ofp_switch_config *osc; | |
2142 | uint16_t flags; | |
2143 | bool drop_frags; | |
2144 | ||
2145 | /* Figure out flags. */ | |
c228a364 | 2146 | dpif_get_drop_frags(p->dpif, &drop_frags); |
064af421 | 2147 | flags = drop_frags ? OFPC_FRAG_DROP : OFPC_FRAG_NORMAL; |
064af421 BP |
2148 | |
2149 | /* Send reply. */ | |
2150 | osc = make_openflow_xid(sizeof *osc, OFPT_GET_CONFIG_REPLY, oh->xid, &buf); | |
2151 | osc->flags = htons(flags); | |
2152 | osc->miss_send_len = htons(ofconn->miss_send_len); | |
2153 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
2154 | ||
2155 | return 0; | |
2156 | } | |
2157 | ||
2158 | static int | |
2159 | handle_set_config(struct ofproto *p, struct ofconn *ofconn, | |
2160 | struct ofp_switch_config *osc) | |
2161 | { | |
2162 | uint16_t flags; | |
2163 | int error; | |
2164 | ||
2165 | error = check_ofp_message(&osc->header, OFPT_SET_CONFIG, sizeof *osc); | |
2166 | if (error) { | |
2167 | return error; | |
2168 | } | |
2169 | flags = ntohs(osc->flags); | |
2170 | ||
9deba63b | 2171 | if (ofconn->type == OFCONN_CONTROLLER && ofconn->role != NX_ROLE_SLAVE) { |
064af421 BP |
2172 | switch (flags & OFPC_FRAG_MASK) { |
2173 | case OFPC_FRAG_NORMAL: | |
c228a364 | 2174 | dpif_set_drop_frags(p->dpif, false); |
064af421 BP |
2175 | break; |
2176 | case OFPC_FRAG_DROP: | |
c228a364 | 2177 | dpif_set_drop_frags(p->dpif, true); |
064af421 BP |
2178 | break; |
2179 | default: | |
2180 | VLOG_WARN_RL(&rl, "requested bad fragment mode (flags=%"PRIx16")", | |
2181 | osc->flags); | |
2182 | break; | |
2183 | } | |
2184 | } | |
2185 | ||
064af421 BP |
2186 | ofconn->miss_send_len = ntohs(osc->miss_send_len); |
2187 | ||
2188 | return 0; | |
2189 | } | |
2190 | ||
2191 | static void | |
6a07af36 JG |
2192 | add_output_group_action(struct odp_actions *actions, uint16_t group, |
2193 | uint16_t *nf_output_iface) | |
064af421 BP |
2194 | { |
2195 | odp_actions_add(actions, ODPAT_OUTPUT_GROUP)->output_group.group = group; | |
6a07af36 JG |
2196 | |
2197 | if (group == DP_GROUP_ALL || group == DP_GROUP_FLOOD) { | |
2198 | *nf_output_iface = NF_OUT_FLOOD; | |
2199 | } | |
064af421 BP |
2200 | } |
2201 | ||
2202 | static void | |
2203 | add_controller_action(struct odp_actions *actions, | |
2204 | const struct ofp_action_output *oao) | |
2205 | { | |
2206 | union odp_action *a = odp_actions_add(actions, ODPAT_CONTROLLER); | |
2207 | a->controller.arg = oao->max_len ? ntohs(oao->max_len) : UINT32_MAX; | |
2208 | } | |
2209 | ||
2210 | struct action_xlate_ctx { | |
2211 | /* Input. */ | |
e18fe8a2 | 2212 | flow_t flow; /* Flow to which these actions correspond. */ |
064af421 BP |
2213 | int recurse; /* Recursion level, via xlate_table_action. */ |
2214 | struct ofproto *ofproto; | |
2215 | const struct ofpbuf *packet; /* The packet corresponding to 'flow', or a | |
2216 | * null pointer if we are revalidating | |
2217 | * without a packet to refer to. */ | |
2218 | ||
2219 | /* Output. */ | |
2220 | struct odp_actions *out; /* Datapath actions. */ | |
2221 | tag_type *tags; /* Tags associated with OFPP_NORMAL actions. */ | |
d6fbec6d | 2222 | bool may_set_up_flow; /* True ordinarily; false if the actions must |
064af421 | 2223 | * be reassessed for every packet. */ |
6a07af36 | 2224 | uint16_t nf_output_iface; /* Output interface index for NetFlow. */ |
064af421 BP |
2225 | }; |
2226 | ||
2227 | static void do_xlate_actions(const union ofp_action *in, size_t n_in, | |
2228 | struct action_xlate_ctx *ctx); | |
2229 | ||
2230 | static void | |
2231 | add_output_action(struct action_xlate_ctx *ctx, uint16_t port) | |
2232 | { | |
2233 | const struct ofport *ofport = port_array_get(&ctx->ofproto->ports, port); | |
6cfaf517 BP |
2234 | |
2235 | if (ofport) { | |
2236 | if (ofport->opp.config & OFPPC_NO_FWD) { | |
2237 | /* Forwarding disabled on port. */ | |
2238 | return; | |
2239 | } | |
2240 | } else { | |
2241 | /* | |
2242 | * We don't have an ofport record for this port, but it doesn't hurt to | |
2243 | * allow forwarding to it anyhow. Maybe such a port will appear later | |
2244 | * and we're pre-populating the flow table. | |
2245 | */ | |
064af421 | 2246 | } |
6cfaf517 BP |
2247 | |
2248 | odp_actions_add(ctx->out, ODPAT_OUTPUT)->output.port = port; | |
6a07af36 | 2249 | ctx->nf_output_iface = port; |
064af421 BP |
2250 | } |
2251 | ||
2252 | static struct rule * | |
2253 | lookup_valid_rule(struct ofproto *ofproto, const flow_t *flow) | |
2254 | { | |
2255 | struct rule *rule; | |
2256 | rule = rule_from_cls_rule(classifier_lookup(&ofproto->cls, flow)); | |
2257 | ||
2258 | /* The rule we found might not be valid, since we could be in need of | |
2259 | * revalidation. If it is not valid, don't return it. */ | |
2260 | if (rule | |
2261 | && rule->super | |
2262 | && ofproto->need_revalidate | |
2263 | && !revalidate_rule(ofproto, rule)) { | |
2264 | COVERAGE_INC(ofproto_invalidated); | |
2265 | return NULL; | |
2266 | } | |
2267 | ||
2268 | return rule; | |
2269 | } | |
2270 | ||
2271 | static void | |
2272 | xlate_table_action(struct action_xlate_ctx *ctx, uint16_t in_port) | |
2273 | { | |
2274 | if (!ctx->recurse) { | |
2c5d1389 | 2275 | uint16_t old_in_port; |
064af421 | 2276 | struct rule *rule; |
064af421 | 2277 | |
2c5d1389 BP |
2278 | /* Look up a flow with 'in_port' as the input port. Then restore the |
2279 | * original input port (otherwise OFPP_NORMAL and OFPP_IN_PORT will | |
2280 | * have surprising behavior). */ | |
2281 | old_in_port = ctx->flow.in_port; | |
e18fe8a2 BP |
2282 | ctx->flow.in_port = in_port; |
2283 | rule = lookup_valid_rule(ctx->ofproto, &ctx->flow); | |
2c5d1389 BP |
2284 | ctx->flow.in_port = old_in_port; |
2285 | ||
064af421 BP |
2286 | if (rule) { |
2287 | if (rule->super) { | |
2288 | rule = rule->super; | |
2289 | } | |
2290 | ||
2291 | ctx->recurse++; | |
2292 | do_xlate_actions(rule->actions, rule->n_actions, ctx); | |
2293 | ctx->recurse--; | |
2294 | } | |
2295 | } | |
2296 | } | |
2297 | ||
2298 | static void | |
2299 | xlate_output_action(struct action_xlate_ctx *ctx, | |
2300 | const struct ofp_action_output *oao) | |
2301 | { | |
2302 | uint16_t odp_port; | |
6a07af36 JG |
2303 | uint16_t prev_nf_output_iface = ctx->nf_output_iface; |
2304 | ||
2305 | ctx->nf_output_iface = NF_OUT_DROP; | |
064af421 BP |
2306 | |
2307 | switch (ntohs(oao->port)) { | |
2308 | case OFPP_IN_PORT: | |
e18fe8a2 | 2309 | add_output_action(ctx, ctx->flow.in_port); |
064af421 BP |
2310 | break; |
2311 | case OFPP_TABLE: | |
e18fe8a2 | 2312 | xlate_table_action(ctx, ctx->flow.in_port); |
064af421 BP |
2313 | break; |
2314 | case OFPP_NORMAL: | |
e18fe8a2 | 2315 | if (!ctx->ofproto->ofhooks->normal_cb(&ctx->flow, ctx->packet, |
064af421 | 2316 | ctx->out, ctx->tags, |
6a07af36 | 2317 | &ctx->nf_output_iface, |
064af421 BP |
2318 | ctx->ofproto->aux)) { |
2319 | COVERAGE_INC(ofproto_uninstallable); | |
d6fbec6d | 2320 | ctx->may_set_up_flow = false; |
064af421 BP |
2321 | } |
2322 | break; | |
2323 | case OFPP_FLOOD: | |
6a07af36 JG |
2324 | add_output_group_action(ctx->out, DP_GROUP_FLOOD, |
2325 | &ctx->nf_output_iface); | |
064af421 BP |
2326 | break; |
2327 | case OFPP_ALL: | |
6a07af36 | 2328 | add_output_group_action(ctx->out, DP_GROUP_ALL, &ctx->nf_output_iface); |
064af421 BP |
2329 | break; |
2330 | case OFPP_CONTROLLER: | |
2331 | add_controller_action(ctx->out, oao); | |
2332 | break; | |
2333 | case OFPP_LOCAL: | |
2334 | add_output_action(ctx, ODPP_LOCAL); | |
2335 | break; | |
2336 | default: | |
2337 | odp_port = ofp_port_to_odp_port(ntohs(oao->port)); | |
e18fe8a2 | 2338 | if (odp_port != ctx->flow.in_port) { |
064af421 BP |
2339 | add_output_action(ctx, odp_port); |
2340 | } | |
2341 | break; | |
2342 | } | |
6a07af36 JG |
2343 | |
2344 | if (prev_nf_output_iface == NF_OUT_FLOOD) { | |
2345 | ctx->nf_output_iface = NF_OUT_FLOOD; | |
2346 | } else if (ctx->nf_output_iface == NF_OUT_DROP) { | |
2347 | ctx->nf_output_iface = prev_nf_output_iface; | |
2348 | } else if (prev_nf_output_iface != NF_OUT_DROP && | |
2349 | ctx->nf_output_iface != NF_OUT_FLOOD) { | |
2350 | ctx->nf_output_iface = NF_OUT_MULTI; | |
2351 | } | |
064af421 BP |
2352 | } |
2353 | ||
2354 | static void | |
2355 | xlate_nicira_action(struct action_xlate_ctx *ctx, | |
2356 | const struct nx_action_header *nah) | |
2357 | { | |
2358 | const struct nx_action_resubmit *nar; | |
659586ef JG |
2359 | const struct nx_action_set_tunnel *nast; |
2360 | union odp_action *oa; | |
064af421 BP |
2361 | int subtype = ntohs(nah->subtype); |
2362 | ||
2363 | assert(nah->vendor == htonl(NX_VENDOR_ID)); | |
2364 | switch (subtype) { | |
2365 | case NXAST_RESUBMIT: | |
2366 | nar = (const struct nx_action_resubmit *) nah; | |
2367 | xlate_table_action(ctx, ofp_port_to_odp_port(ntohs(nar->in_port))); | |
2368 | break; | |
2369 | ||
659586ef JG |
2370 | case NXAST_SET_TUNNEL: |
2371 | nast = (const struct nx_action_set_tunnel *) nah; | |
2372 | oa = odp_actions_add(ctx->out, ODPAT_SET_TUNNEL); | |
2373 | ctx->flow.tun_id = oa->tunnel.tun_id = nast->tun_id; | |
2374 | break; | |
2375 | ||
999f0d45 BP |
2376 | /* If you add a new action here that modifies flow data, don't forget to |
2377 | * update the flow key in ctx->flow in the same key. */ | |
2378 | ||
064af421 BP |
2379 | default: |
2380 | VLOG_DBG_RL(&rl, "unknown Nicira action type %"PRIu16, subtype); | |
2381 | break; | |
2382 | } | |
2383 | } | |
2384 | ||
2385 | static void | |
2386 | do_xlate_actions(const union ofp_action *in, size_t n_in, | |
2387 | struct action_xlate_ctx *ctx) | |
2388 | { | |
2389 | struct actions_iterator iter; | |
2390 | const union ofp_action *ia; | |
2391 | const struct ofport *port; | |
2392 | ||
e18fe8a2 | 2393 | port = port_array_get(&ctx->ofproto->ports, ctx->flow.in_port); |
064af421 | 2394 | if (port && port->opp.config & (OFPPC_NO_RECV | OFPPC_NO_RECV_STP) && |
e18fe8a2 | 2395 | port->opp.config & (eth_addr_equals(ctx->flow.dl_dst, stp_eth_addr) |
064af421 BP |
2396 | ? OFPPC_NO_RECV_STP : OFPPC_NO_RECV)) { |
2397 | /* Drop this flow. */ | |
2398 | return; | |
2399 | } | |
2400 | ||
2401 | for (ia = actions_first(&iter, in, n_in); ia; ia = actions_next(&iter)) { | |
2402 | uint16_t type = ntohs(ia->type); | |
2403 | union odp_action *oa; | |
2404 | ||
2405 | switch (type) { | |
2406 | case OFPAT_OUTPUT: | |
2407 | xlate_output_action(ctx, &ia->output); | |
2408 | break; | |
2409 | ||
2410 | case OFPAT_SET_VLAN_VID: | |
2411 | oa = odp_actions_add(ctx->out, ODPAT_SET_VLAN_VID); | |
999f0d45 | 2412 | ctx->flow.dl_vlan = oa->vlan_vid.vlan_vid = ia->vlan_vid.vlan_vid; |
064af421 BP |
2413 | break; |
2414 | ||
2415 | case OFPAT_SET_VLAN_PCP: | |
2416 | oa = odp_actions_add(ctx->out, ODPAT_SET_VLAN_PCP); | |
999f0d45 | 2417 | ctx->flow.dl_vlan_pcp = oa->vlan_pcp.vlan_pcp = ia->vlan_pcp.vlan_pcp; |
064af421 BP |
2418 | break; |
2419 | ||
2420 | case OFPAT_STRIP_VLAN: | |
2421 | odp_actions_add(ctx->out, ODPAT_STRIP_VLAN); | |
999f0d45 BP |
2422 | ctx->flow.dl_vlan = OFP_VLAN_NONE; |
2423 | ctx->flow.dl_vlan_pcp = 0; | |
064af421 BP |
2424 | break; |
2425 | ||
2426 | case OFPAT_SET_DL_SRC: | |
2427 | oa = odp_actions_add(ctx->out, ODPAT_SET_DL_SRC); | |
2428 | memcpy(oa->dl_addr.dl_addr, | |
2429 | ((struct ofp_action_dl_addr *) ia)->dl_addr, ETH_ADDR_LEN); | |
999f0d45 BP |
2430 | memcpy(ctx->flow.dl_src, |
2431 | ((struct ofp_action_dl_addr *) ia)->dl_addr, ETH_ADDR_LEN); | |
064af421 BP |
2432 | break; |
2433 | ||
2434 | case OFPAT_SET_DL_DST: | |
2435 | oa = odp_actions_add(ctx->out, ODPAT_SET_DL_DST); | |
2436 | memcpy(oa->dl_addr.dl_addr, | |
2437 | ((struct ofp_action_dl_addr *) ia)->dl_addr, ETH_ADDR_LEN); | |
999f0d45 BP |
2438 | memcpy(ctx->flow.dl_dst, |
2439 | ((struct ofp_action_dl_addr *) ia)->dl_addr, ETH_ADDR_LEN); | |
064af421 BP |
2440 | break; |
2441 | ||
2442 | case OFPAT_SET_NW_SRC: | |
2443 | oa = odp_actions_add(ctx->out, ODPAT_SET_NW_SRC); | |
999f0d45 | 2444 | ctx->flow.nw_src = oa->nw_addr.nw_addr = ia->nw_addr.nw_addr; |
064af421 BP |
2445 | break; |
2446 | ||
2d70a31a JP |
2447 | case OFPAT_SET_NW_DST: |
2448 | oa = odp_actions_add(ctx->out, ODPAT_SET_NW_DST); | |
999f0d45 | 2449 | ctx->flow.nw_dst = oa->nw_addr.nw_addr = ia->nw_addr.nw_addr; |
2d38e234 | 2450 | break; |
959a2ecd JP |
2451 | |
2452 | case OFPAT_SET_NW_TOS: | |
2453 | oa = odp_actions_add(ctx->out, ODPAT_SET_NW_TOS); | |
999f0d45 | 2454 | ctx->flow.nw_tos = oa->nw_tos.nw_tos = ia->nw_tos.nw_tos; |
2d70a31a JP |
2455 | break; |
2456 | ||
064af421 BP |
2457 | case OFPAT_SET_TP_SRC: |
2458 | oa = odp_actions_add(ctx->out, ODPAT_SET_TP_SRC); | |
999f0d45 | 2459 | ctx->flow.tp_src = oa->tp_port.tp_port = ia->tp_port.tp_port; |
064af421 BP |
2460 | break; |
2461 | ||
2d70a31a JP |
2462 | case OFPAT_SET_TP_DST: |
2463 | oa = odp_actions_add(ctx->out, ODPAT_SET_TP_DST); | |
999f0d45 | 2464 | ctx->flow.tp_dst = oa->tp_port.tp_port = ia->tp_port.tp_port; |
2d70a31a JP |
2465 | break; |
2466 | ||
064af421 BP |
2467 | case OFPAT_VENDOR: |
2468 | xlate_nicira_action(ctx, (const struct nx_action_header *) ia); | |
2469 | break; | |
2470 | ||
2471 | default: | |
2472 | VLOG_DBG_RL(&rl, "unknown action type %"PRIu16, type); | |
2473 | break; | |
2474 | } | |
2475 | } | |
2476 | } | |
2477 | ||
2478 | static int | |
2479 | xlate_actions(const union ofp_action *in, size_t n_in, | |
2480 | const flow_t *flow, struct ofproto *ofproto, | |
2481 | const struct ofpbuf *packet, | |
6a07af36 JG |
2482 | struct odp_actions *out, tag_type *tags, bool *may_set_up_flow, |
2483 | uint16_t *nf_output_iface) | |
064af421 BP |
2484 | { |
2485 | tag_type no_tags = 0; | |
2486 | struct action_xlate_ctx ctx; | |
2487 | COVERAGE_INC(ofproto_ofp2odp); | |
2488 | odp_actions_init(out); | |
e18fe8a2 | 2489 | ctx.flow = *flow; |
064af421 BP |
2490 | ctx.recurse = 0; |
2491 | ctx.ofproto = ofproto; | |
2492 | ctx.packet = packet; | |
2493 | ctx.out = out; | |
2494 | ctx.tags = tags ? tags : &no_tags; | |
d6fbec6d | 2495 | ctx.may_set_up_flow = true; |
6a07af36 | 2496 | ctx.nf_output_iface = NF_OUT_DROP; |
064af421 | 2497 | do_xlate_actions(in, n_in, &ctx); |
0ad9b732 | 2498 | |
d6fbec6d | 2499 | /* Check with in-band control to see if we're allowed to set up this |
0ad9b732 JP |
2500 | * flow. */ |
2501 | if (!in_band_rule_check(ofproto->in_band, flow, out)) { | |
d6fbec6d | 2502 | ctx.may_set_up_flow = false; |
0ad9b732 JP |
2503 | } |
2504 | ||
d6fbec6d BP |
2505 | if (may_set_up_flow) { |
2506 | *may_set_up_flow = ctx.may_set_up_flow; | |
064af421 | 2507 | } |
6a07af36 JG |
2508 | if (nf_output_iface) { |
2509 | *nf_output_iface = ctx.nf_output_iface; | |
064af421 BP |
2510 | } |
2511 | if (odp_actions_overflow(out)) { | |
2512 | odp_actions_init(out); | |
2513 | return ofp_mkerr(OFPET_BAD_ACTION, OFPBAC_TOO_MANY); | |
2514 | } | |
2515 | return 0; | |
2516 | } | |
2517 | ||
9deba63b BP |
2518 | /* Checks whether 'ofconn' is a slave controller. If so, returns an OpenFlow |
2519 | * error message code (composed with ofp_mkerr()) for the caller to propagate | |
2520 | * upward. Otherwise, returns 0. | |
2521 | * | |
2522 | * 'oh' is used to make log messages more informative. */ | |
2523 | static int | |
2524 | reject_slave_controller(struct ofconn *ofconn, const struct ofp_header *oh) | |
2525 | { | |
2526 | if (ofconn->type == OFCONN_CONTROLLER && ofconn->role == NX_ROLE_SLAVE) { | |
2527 | static struct vlog_rate_limit perm_rl = VLOG_RATE_LIMIT_INIT(1, 5); | |
2528 | char *type_name; | |
2529 | ||
2530 | type_name = ofp_message_type_to_string(oh->type); | |
2531 | VLOG_WARN_RL(&perm_rl, "rejecting %s message from slave controller", | |
2532 | type_name); | |
2533 | free(type_name); | |
2534 | ||
2535 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_EPERM); | |
2536 | } else { | |
2537 | return 0; | |
2538 | } | |
2539 | } | |
2540 | ||
064af421 BP |
2541 | static int |
2542 | handle_packet_out(struct ofproto *p, struct ofconn *ofconn, | |
2543 | struct ofp_header *oh) | |
2544 | { | |
2545 | struct ofp_packet_out *opo; | |
2546 | struct ofpbuf payload, *buffer; | |
2547 | struct odp_actions actions; | |
2548 | int n_actions; | |
2549 | uint16_t in_port; | |
2550 | flow_t flow; | |
2551 | int error; | |
2552 | ||
9deba63b BP |
2553 | error = reject_slave_controller(ofconn, oh); |
2554 | if (error) { | |
2555 | return error; | |
2556 | } | |
2557 | ||
064af421 BP |
2558 | error = check_ofp_packet_out(oh, &payload, &n_actions, p->max_ports); |
2559 | if (error) { | |
2560 | return error; | |
2561 | } | |
2562 | opo = (struct ofp_packet_out *) oh; | |
2563 | ||
2564 | COVERAGE_INC(ofproto_packet_out); | |
2565 | if (opo->buffer_id != htonl(UINT32_MAX)) { | |
2566 | error = pktbuf_retrieve(ofconn->pktbuf, ntohl(opo->buffer_id), | |
2567 | &buffer, &in_port); | |
7778bd15 | 2568 | if (error || !buffer) { |
064af421 BP |
2569 | return error; |
2570 | } | |
2571 | payload = *buffer; | |
2572 | } else { | |
2573 | buffer = NULL; | |
2574 | } | |
2575 | ||
659586ef | 2576 | flow_extract(&payload, 0, ofp_port_to_odp_port(ntohs(opo->in_port)), &flow); |
064af421 | 2577 | error = xlate_actions((const union ofp_action *) opo->actions, n_actions, |
6a07af36 | 2578 | &flow, p, &payload, &actions, NULL, NULL, NULL); |
064af421 BP |
2579 | if (error) { |
2580 | return error; | |
2581 | } | |
2582 | ||
c228a364 | 2583 | dpif_execute(p->dpif, flow.in_port, actions.actions, actions.n_actions, |
064af421 BP |
2584 | &payload); |
2585 | ofpbuf_delete(buffer); | |
2586 | ||
2587 | return 0; | |
2588 | } | |
2589 | ||
2590 | static void | |
2591 | update_port_config(struct ofproto *p, struct ofport *port, | |
2592 | uint32_t config, uint32_t mask) | |
2593 | { | |
2594 | mask &= config ^ port->opp.config; | |
2595 | if (mask & OFPPC_PORT_DOWN) { | |
2596 | if (config & OFPPC_PORT_DOWN) { | |
2597 | netdev_turn_flags_off(port->netdev, NETDEV_UP, true); | |
2598 | } else { | |
2599 | netdev_turn_flags_on(port->netdev, NETDEV_UP, true); | |
2600 | } | |
2601 | } | |
2602 | #define REVALIDATE_BITS (OFPPC_NO_RECV | OFPPC_NO_RECV_STP | OFPPC_NO_FWD) | |
2603 | if (mask & REVALIDATE_BITS) { | |
2604 | COVERAGE_INC(ofproto_costly_flags); | |
2605 | port->opp.config ^= mask & REVALIDATE_BITS; | |
2606 | p->need_revalidate = true; | |
2607 | } | |
2608 | #undef REVALIDATE_BITS | |
2609 | if (mask & OFPPC_NO_FLOOD) { | |
2610 | port->opp.config ^= OFPPC_NO_FLOOD; | |
72b06300 | 2611 | refresh_port_groups(p); |
064af421 BP |
2612 | } |
2613 | if (mask & OFPPC_NO_PACKET_IN) { | |
2614 | port->opp.config ^= OFPPC_NO_PACKET_IN; | |
2615 | } | |
2616 | } | |
2617 | ||
2618 | static int | |
9deba63b BP |
2619 | handle_port_mod(struct ofproto *p, struct ofconn *ofconn, |
2620 | struct ofp_header *oh) | |
064af421 BP |
2621 | { |
2622 | const struct ofp_port_mod *opm; | |
2623 | struct ofport *port; | |
2624 | int error; | |
2625 | ||
9deba63b BP |
2626 | error = reject_slave_controller(ofconn, oh); |
2627 | if (error) { | |
2628 | return error; | |
2629 | } | |
064af421 BP |
2630 | error = check_ofp_message(oh, OFPT_PORT_MOD, sizeof *opm); |
2631 | if (error) { | |
2632 | return error; | |
2633 | } | |
2634 | opm = (struct ofp_port_mod *) oh; | |
2635 | ||
2636 | port = port_array_get(&p->ports, | |
2637 | ofp_port_to_odp_port(ntohs(opm->port_no))); | |
2638 | if (!port) { | |
2639 | return ofp_mkerr(OFPET_PORT_MOD_FAILED, OFPPMFC_BAD_PORT); | |
2640 | } else if (memcmp(port->opp.hw_addr, opm->hw_addr, OFP_ETH_ALEN)) { | |
2641 | return ofp_mkerr(OFPET_PORT_MOD_FAILED, OFPPMFC_BAD_HW_ADDR); | |
2642 | } else { | |
2643 | update_port_config(p, port, ntohl(opm->config), ntohl(opm->mask)); | |
2644 | if (opm->advertise) { | |
2645 | netdev_set_advertisements(port->netdev, ntohl(opm->advertise)); | |
2646 | } | |
2647 | } | |
2648 | return 0; | |
2649 | } | |
2650 | ||
2651 | static struct ofpbuf * | |
2652 | make_stats_reply(uint32_t xid, uint16_t type, size_t body_len) | |
2653 | { | |
2654 | struct ofp_stats_reply *osr; | |
2655 | struct ofpbuf *msg; | |
2656 | ||
2657 | msg = ofpbuf_new(MIN(sizeof *osr + body_len, UINT16_MAX)); | |
2658 | osr = put_openflow_xid(sizeof *osr, OFPT_STATS_REPLY, xid, msg); | |
2659 | osr->type = type; | |
2660 | osr->flags = htons(0); | |
2661 | return msg; | |
2662 | } | |
2663 | ||
2664 | static struct ofpbuf * | |
2665 | start_stats_reply(const struct ofp_stats_request *request, size_t body_len) | |
2666 | { | |
2667 | return make_stats_reply(request->header.xid, request->type, body_len); | |
2668 | } | |
2669 | ||
2670 | static void * | |
2671 | append_stats_reply(size_t nbytes, struct ofconn *ofconn, struct ofpbuf **msgp) | |
2672 | { | |
2673 | struct ofpbuf *msg = *msgp; | |
2674 | assert(nbytes <= UINT16_MAX - sizeof(struct ofp_stats_reply)); | |
2675 | if (nbytes + msg->size > UINT16_MAX) { | |
2676 | struct ofp_stats_reply *reply = msg->data; | |
2677 | reply->flags = htons(OFPSF_REPLY_MORE); | |
2678 | *msgp = make_stats_reply(reply->header.xid, reply->type, nbytes); | |
2679 | queue_tx(msg, ofconn, ofconn->reply_counter); | |
2680 | } | |
2681 | return ofpbuf_put_uninit(*msgp, nbytes); | |
2682 | } | |
2683 | ||
2684 | static int | |
2685 | handle_desc_stats_request(struct ofproto *p, struct ofconn *ofconn, | |
2686 | struct ofp_stats_request *request) | |
2687 | { | |
2688 | struct ofp_desc_stats *ods; | |
2689 | struct ofpbuf *msg; | |
2690 | ||
2691 | msg = start_stats_reply(request, sizeof *ods); | |
2692 | ods = append_stats_reply(sizeof *ods, ofconn, &msg); | |
5a719c38 JP |
2693 | memset(ods, 0, sizeof *ods); |
2694 | ovs_strlcpy(ods->mfr_desc, p->mfr_desc, sizeof ods->mfr_desc); | |
2695 | ovs_strlcpy(ods->hw_desc, p->hw_desc, sizeof ods->hw_desc); | |
2696 | ovs_strlcpy(ods->sw_desc, p->sw_desc, sizeof ods->sw_desc); | |
2697 | ovs_strlcpy(ods->serial_num, p->serial_desc, sizeof ods->serial_num); | |
2698 | ovs_strlcpy(ods->dp_desc, p->dp_desc, sizeof ods->dp_desc); | |
064af421 BP |
2699 | queue_tx(msg, ofconn, ofconn->reply_counter); |
2700 | ||
2701 | return 0; | |
2702 | } | |
2703 | ||
2704 | static void | |
2705 | count_subrules(struct cls_rule *cls_rule, void *n_subrules_) | |
2706 | { | |
2707 | struct rule *rule = rule_from_cls_rule(cls_rule); | |
2708 | int *n_subrules = n_subrules_; | |
2709 | ||
2710 | if (rule->super) { | |
2711 | (*n_subrules)++; | |
2712 | } | |
2713 | } | |
2714 | ||
2715 | static int | |
2716 | handle_table_stats_request(struct ofproto *p, struct ofconn *ofconn, | |
2717 | struct ofp_stats_request *request) | |
2718 | { | |
2719 | struct ofp_table_stats *ots; | |
2720 | struct ofpbuf *msg; | |
2721 | struct odp_stats dpstats; | |
2722 | int n_exact, n_subrules, n_wild; | |
2723 | ||
2724 | msg = start_stats_reply(request, sizeof *ots * 2); | |
2725 | ||
2726 | /* Count rules of various kinds. */ | |
2727 | n_subrules = 0; | |
2728 | classifier_for_each(&p->cls, CLS_INC_EXACT, count_subrules, &n_subrules); | |
2729 | n_exact = classifier_count_exact(&p->cls) - n_subrules; | |
2730 | n_wild = classifier_count(&p->cls) - classifier_count_exact(&p->cls); | |
2731 | ||
2732 | /* Hash table. */ | |
c228a364 | 2733 | dpif_get_dp_stats(p->dpif, &dpstats); |
064af421 BP |
2734 | ots = append_stats_reply(sizeof *ots, ofconn, &msg); |
2735 | memset(ots, 0, sizeof *ots); | |
2736 | ots->table_id = TABLEID_HASH; | |
2737 | strcpy(ots->name, "hash"); | |
2738 | ots->wildcards = htonl(0); | |
2739 | ots->max_entries = htonl(dpstats.max_capacity); | |
2740 | ots->active_count = htonl(n_exact); | |
2741 | ots->lookup_count = htonll(dpstats.n_frags + dpstats.n_hit + | |
2742 | dpstats.n_missed); | |
2743 | ots->matched_count = htonll(dpstats.n_hit); /* XXX */ | |
2744 | ||
2745 | /* Classifier table. */ | |
2746 | ots = append_stats_reply(sizeof *ots, ofconn, &msg); | |
2747 | memset(ots, 0, sizeof *ots); | |
2748 | ots->table_id = TABLEID_CLASSIFIER; | |
2749 | strcpy(ots->name, "classifier"); | |
659586ef JG |
2750 | ots->wildcards = p->tun_id_from_cookie ? htonl(OVSFW_ALL) |
2751 | : htonl(OFPFW_ALL); | |
064af421 BP |
2752 | ots->max_entries = htonl(65536); |
2753 | ots->active_count = htonl(n_wild); | |
2754 | ots->lookup_count = htonll(0); /* XXX */ | |
2755 | ots->matched_count = htonll(0); /* XXX */ | |
2756 | ||
2757 | queue_tx(msg, ofconn, ofconn->reply_counter); | |
2758 | return 0; | |
2759 | } | |
2760 | ||
abaad8cf JP |
2761 | static void |
2762 | append_port_stat(struct ofport *port, uint16_t port_no, struct ofconn *ofconn, | |
2763 | struct ofpbuf *msg) | |
2764 | { | |
2765 | struct netdev_stats stats; | |
2766 | struct ofp_port_stats *ops; | |
2767 | ||
2768 | /* Intentionally ignore return value, since errors will set | |
2769 | * 'stats' to all-1s, which is correct for OpenFlow, and | |
2770 | * netdev_get_stats() will log errors. */ | |
2771 | netdev_get_stats(port->netdev, &stats); | |
2772 | ||
2773 | ops = append_stats_reply(sizeof *ops, ofconn, &msg); | |
2774 | ops->port_no = htons(odp_port_to_ofp_port(port_no)); | |
2775 | memset(ops->pad, 0, sizeof ops->pad); | |
2776 | ops->rx_packets = htonll(stats.rx_packets); | |
2777 | ops->tx_packets = htonll(stats.tx_packets); | |
2778 | ops->rx_bytes = htonll(stats.rx_bytes); | |
2779 | ops->tx_bytes = htonll(stats.tx_bytes); | |
2780 | ops->rx_dropped = htonll(stats.rx_dropped); | |
2781 | ops->tx_dropped = htonll(stats.tx_dropped); | |
2782 | ops->rx_errors = htonll(stats.rx_errors); | |
2783 | ops->tx_errors = htonll(stats.tx_errors); | |
2784 | ops->rx_frame_err = htonll(stats.rx_frame_errors); | |
2785 | ops->rx_over_err = htonll(stats.rx_over_errors); | |
2786 | ops->rx_crc_err = htonll(stats.rx_crc_errors); | |
2787 | ops->collisions = htonll(stats.collisions); | |
2788 | } | |
2789 | ||
064af421 BP |
2790 | static int |
2791 | handle_port_stats_request(struct ofproto *p, struct ofconn *ofconn, | |
abaad8cf JP |
2792 | struct ofp_stats_request *osr, |
2793 | size_t arg_size) | |
064af421 | 2794 | { |
abaad8cf | 2795 | struct ofp_port_stats_request *psr; |
064af421 BP |
2796 | struct ofp_port_stats *ops; |
2797 | struct ofpbuf *msg; | |
2798 | struct ofport *port; | |
2799 | unsigned int port_no; | |
2800 | ||
abaad8cf JP |
2801 | if (arg_size != sizeof *psr) { |
2802 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); | |
2803 | } | |
2804 | psr = (struct ofp_port_stats_request *) osr->body; | |
2805 | ||
2806 | msg = start_stats_reply(osr, sizeof *ops * 16); | |
2807 | if (psr->port_no != htons(OFPP_NONE)) { | |
2808 | port = port_array_get(&p->ports, | |
2809 | ofp_port_to_odp_port(ntohs(psr->port_no))); | |
2810 | if (port) { | |
2811 | append_port_stat(port, ntohs(psr->port_no), ofconn, msg); | |
2812 | } | |
2813 | } else { | |
2814 | PORT_ARRAY_FOR_EACH (port, &p->ports, port_no) { | |
2815 | append_port_stat(port, port_no, ofconn, msg); | |
2816 | } | |
064af421 BP |
2817 | } |
2818 | ||
2819 | queue_tx(msg, ofconn, ofconn->reply_counter); | |
2820 | return 0; | |
2821 | } | |
2822 | ||
2823 | struct flow_stats_cbdata { | |
2824 | struct ofproto *ofproto; | |
2825 | struct ofconn *ofconn; | |
2826 | uint16_t out_port; | |
2827 | struct ofpbuf *msg; | |
2828 | }; | |
2829 | ||
01149cfd BP |
2830 | /* Obtains statistic counters for 'rule' within 'p' and stores them into |
2831 | * '*packet_countp' and '*byte_countp'. If 'rule' is a wildcarded rule, the | |
2832 | * returned statistic include statistics for all of 'rule''s subrules. */ | |
064af421 BP |
2833 | static void |
2834 | query_stats(struct ofproto *p, struct rule *rule, | |
2835 | uint64_t *packet_countp, uint64_t *byte_countp) | |
2836 | { | |
2837 | uint64_t packet_count, byte_count; | |
2838 | struct rule *subrule; | |
2839 | struct odp_flow *odp_flows; | |
2840 | size_t n_odp_flows; | |
2841 | ||
01149cfd BP |
2842 | /* Start from historical data for 'rule' itself that are no longer tracked |
2843 | * by the datapath. This counts, for example, subrules that have | |
2844 | * expired. */ | |
b3137fe8 JG |
2845 | packet_count = rule->packet_count; |
2846 | byte_count = rule->byte_count; | |
2847 | ||
01149cfd BP |
2848 | /* Prepare to ask the datapath for statistics on 'rule', or if it is |
2849 | * wildcarded then on all of its subrules. | |
2850 | * | |
2851 | * Also, add any statistics that are not tracked by the datapath for each | |
2852 | * subrule. This includes, for example, statistics for packets that were | |
2853 | * executed "by hand" by ofproto via dpif_execute() but must be accounted | |
2854 | * to a flow. */ | |
064af421 | 2855 | n_odp_flows = rule->cr.wc.wildcards ? list_size(&rule->list) : 1; |
ec6fde61 | 2856 | odp_flows = xzalloc(n_odp_flows * sizeof *odp_flows); |
064af421 BP |
2857 | if (rule->cr.wc.wildcards) { |
2858 | size_t i = 0; | |
2859 | LIST_FOR_EACH (subrule, struct rule, list, &rule->list) { | |
2860 | odp_flows[i++].key = subrule->cr.flow; | |
b3137fe8 JG |
2861 | packet_count += subrule->packet_count; |
2862 | byte_count += subrule->byte_count; | |
064af421 BP |
2863 | } |
2864 | } else { | |
2865 | odp_flows[0].key = rule->cr.flow; | |
2866 | } | |
2867 | ||
28998b22 | 2868 | /* Fetch up-to-date statistics from the datapath and add them in. */ |
c228a364 | 2869 | if (!dpif_flow_get_multiple(p->dpif, odp_flows, n_odp_flows)) { |
064af421 BP |
2870 | size_t i; |
2871 | for (i = 0; i < n_odp_flows; i++) { | |
2872 | struct odp_flow *odp_flow = &odp_flows[i]; | |
2873 | packet_count += odp_flow->stats.n_packets; | |
2874 | byte_count += odp_flow->stats.n_bytes; | |
2875 | } | |
2876 | } | |
2877 | free(odp_flows); | |
2878 | ||
01149cfd | 2879 | /* Return the stats to the caller. */ |
064af421 BP |
2880 | *packet_countp = packet_count; |
2881 | *byte_countp = byte_count; | |
2882 | } | |
2883 | ||
2884 | static void | |
2885 | flow_stats_cb(struct cls_rule *rule_, void *cbdata_) | |
2886 | { | |
2887 | struct rule *rule = rule_from_cls_rule(rule_); | |
2888 | struct flow_stats_cbdata *cbdata = cbdata_; | |
2889 | struct ofp_flow_stats *ofs; | |
2890 | uint64_t packet_count, byte_count; | |
2891 | size_t act_len, len; | |
26c3f94a JP |
2892 | long long int tdiff = time_msec() - rule->created; |
2893 | uint32_t sec = tdiff / 1000; | |
2894 | uint32_t msec = tdiff - (sec * 1000); | |
064af421 BP |
2895 | |
2896 | if (rule_is_hidden(rule) || !rule_has_out_port(rule, cbdata->out_port)) { | |
2897 | return; | |
2898 | } | |
2899 | ||
2900 | act_len = sizeof *rule->actions * rule->n_actions; | |
2901 | len = offsetof(struct ofp_flow_stats, actions) + act_len; | |
2902 | ||
2903 | query_stats(cbdata->ofproto, rule, &packet_count, &byte_count); | |
2904 | ||
2905 | ofs = append_stats_reply(len, cbdata->ofconn, &cbdata->msg); | |
2906 | ofs->length = htons(len); | |
2907 | ofs->table_id = rule->cr.wc.wildcards ? TABLEID_CLASSIFIER : TABLEID_HASH; | |
2908 | ofs->pad = 0; | |
659586ef JG |
2909 | flow_to_match(&rule->cr.flow, rule->cr.wc.wildcards, |
2910 | cbdata->ofproto->tun_id_from_cookie, &ofs->match); | |
26c3f94a JP |
2911 | ofs->duration_sec = htonl(sec); |
2912 | ofs->duration_nsec = htonl(msec * 1000000); | |
39997502 | 2913 | ofs->cookie = rule->flow_cookie; |
064af421 BP |
2914 | ofs->priority = htons(rule->cr.priority); |
2915 | ofs->idle_timeout = htons(rule->idle_timeout); | |
2916 | ofs->hard_timeout = htons(rule->hard_timeout); | |
39997502 | 2917 | memset(ofs->pad2, 0, sizeof ofs->pad2); |
064af421 BP |
2918 | ofs->packet_count = htonll(packet_count); |
2919 | ofs->byte_count = htonll(byte_count); | |
2920 | memcpy(ofs->actions, rule->actions, act_len); | |
2921 | } | |
2922 | ||
2923 | static int | |
2924 | table_id_to_include(uint8_t table_id) | |
2925 | { | |
2926 | return (table_id == TABLEID_HASH ? CLS_INC_EXACT | |
2927 | : table_id == TABLEID_CLASSIFIER ? CLS_INC_WILD | |
2928 | : table_id == 0xff ? CLS_INC_ALL | |
2929 | : 0); | |
2930 | } | |
2931 | ||
2932 | static int | |
2933 | handle_flow_stats_request(struct ofproto *p, struct ofconn *ofconn, | |
2934 | const struct ofp_stats_request *osr, | |
2935 | size_t arg_size) | |
2936 | { | |
2937 | struct ofp_flow_stats_request *fsr; | |
2938 | struct flow_stats_cbdata cbdata; | |
2939 | struct cls_rule target; | |
2940 | ||
2941 | if (arg_size != sizeof *fsr) { | |
49bdc010 | 2942 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); |
064af421 BP |
2943 | } |
2944 | fsr = (struct ofp_flow_stats_request *) osr->body; | |
2945 | ||
2946 | COVERAGE_INC(ofproto_flows_req); | |
2947 | cbdata.ofproto = p; | |
2948 | cbdata.ofconn = ofconn; | |
2949 | cbdata.out_port = fsr->out_port; | |
2950 | cbdata.msg = start_stats_reply(osr, 1024); | |
659586ef | 2951 | cls_rule_from_match(&fsr->match, 0, false, 0, &target); |
064af421 BP |
2952 | classifier_for_each_match(&p->cls, &target, |
2953 | table_id_to_include(fsr->table_id), | |
2954 | flow_stats_cb, &cbdata); | |
2955 | queue_tx(cbdata.msg, ofconn, ofconn->reply_counter); | |
2956 | return 0; | |
2957 | } | |
2958 | ||
4f2cad2c JP |
2959 | struct flow_stats_ds_cbdata { |
2960 | struct ofproto *ofproto; | |
2961 | struct ds *results; | |
2962 | }; | |
2963 | ||
2964 | static void | |
2965 | flow_stats_ds_cb(struct cls_rule *rule_, void *cbdata_) | |
2966 | { | |
2967 | struct rule *rule = rule_from_cls_rule(rule_); | |
2968 | struct flow_stats_ds_cbdata *cbdata = cbdata_; | |
2969 | struct ds *results = cbdata->results; | |
2970 | struct ofp_match match; | |
2971 | uint64_t packet_count, byte_count; | |
2972 | size_t act_len = sizeof *rule->actions * rule->n_actions; | |
2973 | ||
2974 | /* Don't report on subrules. */ | |
2975 | if (rule->super != NULL) { | |
2976 | return; | |
2977 | } | |
2978 | ||
2979 | query_stats(cbdata->ofproto, rule, &packet_count, &byte_count); | |
659586ef JG |
2980 | flow_to_match(&rule->cr.flow, rule->cr.wc.wildcards, |
2981 | cbdata->ofproto->tun_id_from_cookie, &match); | |
4f2cad2c JP |
2982 | |
2983 | ds_put_format(results, "duration=%llds, ", | |
2984 | (time_msec() - rule->created) / 1000); | |
52ae00b3 | 2985 | ds_put_format(results, "priority=%u, ", rule->cr.priority); |
4f2cad2c JP |
2986 | ds_put_format(results, "n_packets=%"PRIu64", ", packet_count); |
2987 | ds_put_format(results, "n_bytes=%"PRIu64", ", byte_count); | |
2988 | ofp_print_match(results, &match, true); | |
2989 | ofp_print_actions(results, &rule->actions->header, act_len); | |
2990 | ds_put_cstr(results, "\n"); | |
2991 | } | |
2992 | ||
2993 | /* Adds a pretty-printed description of all flows to 'results', including | |
2994 | * those marked hidden by secchan (e.g., by in-band control). */ | |
2995 | void | |
2996 | ofproto_get_all_flows(struct ofproto *p, struct ds *results) | |
2997 | { | |
2998 | struct ofp_match match; | |
2999 | struct cls_rule target; | |
3000 | struct flow_stats_ds_cbdata cbdata; | |
3001 | ||
3002 | memset(&match, 0, sizeof match); | |
659586ef | 3003 | match.wildcards = htonl(OVSFW_ALL); |
4f2cad2c JP |
3004 | |
3005 | cbdata.ofproto = p; | |
3006 | cbdata.results = results; | |
3007 | ||
659586ef | 3008 | cls_rule_from_match(&match, 0, false, 0, &target); |
4f2cad2c JP |
3009 | classifier_for_each_match(&p->cls, &target, CLS_INC_ALL, |
3010 | flow_stats_ds_cb, &cbdata); | |
3011 | } | |
3012 | ||
064af421 BP |
3013 | struct aggregate_stats_cbdata { |
3014 | struct ofproto *ofproto; | |
3015 | uint16_t out_port; | |
3016 | uint64_t packet_count; | |
3017 | uint64_t byte_count; | |
3018 | uint32_t n_flows; | |
3019 | }; | |
3020 | ||
3021 | static void | |
3022 | aggregate_stats_cb(struct cls_rule *rule_, void *cbdata_) | |
3023 | { | |
3024 | struct rule *rule = rule_from_cls_rule(rule_); | |
3025 | struct aggregate_stats_cbdata *cbdata = cbdata_; | |
3026 | uint64_t packet_count, byte_count; | |
3027 | ||
3028 | if (rule_is_hidden(rule) || !rule_has_out_port(rule, cbdata->out_port)) { | |
3029 | return; | |
3030 | } | |
3031 | ||
3032 | query_stats(cbdata->ofproto, rule, &packet_count, &byte_count); | |
3033 | ||
3034 | cbdata->packet_count += packet_count; | |
3035 | cbdata->byte_count += byte_count; | |
3036 | cbdata->n_flows++; | |
3037 | } | |
3038 | ||
3039 | static int | |
3040 | handle_aggregate_stats_request(struct ofproto *p, struct ofconn *ofconn, | |
3041 | const struct ofp_stats_request *osr, | |
3042 | size_t arg_size) | |
3043 | { | |
3044 | struct ofp_aggregate_stats_request *asr; | |
3045 | struct ofp_aggregate_stats_reply *reply; | |
3046 | struct aggregate_stats_cbdata cbdata; | |
3047 | struct cls_rule target; | |
3048 | struct ofpbuf *msg; | |
3049 | ||
3050 | if (arg_size != sizeof *asr) { | |
49bdc010 | 3051 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); |
064af421 BP |
3052 | } |
3053 | asr = (struct ofp_aggregate_stats_request *) osr->body; | |
3054 | ||
3055 | COVERAGE_INC(ofproto_agg_request); | |
3056 | cbdata.ofproto = p; | |
3057 | cbdata.out_port = asr->out_port; | |
3058 | cbdata.packet_count = 0; | |
3059 | cbdata.byte_count = 0; | |
3060 | cbdata.n_flows = 0; | |
659586ef | 3061 | cls_rule_from_match(&asr->match, 0, false, 0, &target); |
064af421 BP |
3062 | classifier_for_each_match(&p->cls, &target, |
3063 | table_id_to_include(asr->table_id), | |
3064 | aggregate_stats_cb, &cbdata); | |
3065 | ||
3066 | msg = start_stats_reply(osr, sizeof *reply); | |
3067 | reply = append_stats_reply(sizeof *reply, ofconn, &msg); | |
3068 | reply->flow_count = htonl(cbdata.n_flows); | |
3069 | reply->packet_count = htonll(cbdata.packet_count); | |
3070 | reply->byte_count = htonll(cbdata.byte_count); | |
3071 | queue_tx(msg, ofconn, ofconn->reply_counter); | |
3072 | return 0; | |
3073 | } | |
3074 | ||
3075 | static int | |
3076 | handle_stats_request(struct ofproto *p, struct ofconn *ofconn, | |
3077 | struct ofp_header *oh) | |
3078 | { | |
3079 | struct ofp_stats_request *osr; | |
3080 | size_t arg_size; | |
3081 | int error; | |
3082 | ||
3083 | error = check_ofp_message_array(oh, OFPT_STATS_REQUEST, sizeof *osr, | |
3084 | 1, &arg_size); | |
3085 | if (error) { | |
3086 | return error; | |
3087 | } | |
3088 | osr = (struct ofp_stats_request *) oh; | |
3089 | ||
3090 | switch (ntohs(osr->type)) { | |
3091 | case OFPST_DESC: | |
3092 | return handle_desc_stats_request(p, ofconn, osr); | |
3093 | ||
3094 | case OFPST_FLOW: | |
3095 | return handle_flow_stats_request(p, ofconn, osr, arg_size); | |
3096 | ||
3097 | case OFPST_AGGREGATE: | |
3098 | return handle_aggregate_stats_request(p, ofconn, osr, arg_size); | |
3099 | ||
3100 | case OFPST_TABLE: | |
3101 | return handle_table_stats_request(p, ofconn, osr); | |
3102 | ||
3103 | case OFPST_PORT: | |
abaad8cf | 3104 | return handle_port_stats_request(p, ofconn, osr, arg_size); |
064af421 BP |
3105 | |
3106 | case OFPST_VENDOR: | |
3107 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_VENDOR); | |
3108 | ||
3109 | default: | |
3110 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_STAT); | |
3111 | } | |
3112 | } | |
3113 | ||
3114 | static long long int | |
3115 | msec_from_nsec(uint64_t sec, uint32_t nsec) | |
3116 | { | |
3117 | return !sec ? 0 : sec * 1000 + nsec / 1000000; | |
3118 | } | |
3119 | ||
3120 | static void | |
0193b2af JG |
3121 | update_time(struct ofproto *ofproto, struct rule *rule, |
3122 | const struct odp_flow_stats *stats) | |
064af421 BP |
3123 | { |
3124 | long long int used = msec_from_nsec(stats->used_sec, stats->used_nsec); | |
3125 | if (used > rule->used) { | |
3126 | rule->used = used; | |
4836f9f2 JP |
3127 | if (rule->super && used > rule->super->used) { |
3128 | rule->super->used = used; | |
3129 | } | |
0193b2af | 3130 | netflow_flow_update_time(ofproto->netflow, &rule->nf_flow, used); |
064af421 BP |
3131 | } |
3132 | } | |
3133 | ||
3134 | static void | |
0193b2af JG |
3135 | update_stats(struct ofproto *ofproto, struct rule *rule, |
3136 | const struct odp_flow_stats *stats) | |
064af421 | 3137 | { |
064af421 | 3138 | if (stats->n_packets) { |
0193b2af JG |
3139 | update_time(ofproto, rule, stats); |
3140 | rule->packet_count += stats->n_packets; | |
3141 | rule->byte_count += stats->n_bytes; | |
3142 | netflow_flow_update_flags(&rule->nf_flow, stats->ip_tos, | |
3143 | stats->tcp_flags); | |
064af421 BP |
3144 | } |
3145 | } | |
3146 | ||
79eee1eb BP |
3147 | /* Implements OFPFC_ADD and the cases for OFPFC_MODIFY and OFPFC_MODIFY_STRICT |
3148 | * in which no matching flow already exists in the flow table. | |
3149 | * | |
3150 | * Adds the flow specified by 'ofm', which is followed by 'n_actions' | |
3151 | * ofp_actions, to 'p''s flow table. Returns 0 on success or an OpenFlow error | |
3152 | * code as encoded by ofp_mkerr() on failure. | |
3153 | * | |
3154 | * 'ofconn' is used to retrieve the packet buffer specified in ofm->buffer_id, | |
3155 | * if any. */ | |
064af421 BP |
3156 | static int |
3157 | add_flow(struct ofproto *p, struct ofconn *ofconn, | |
79eee1eb | 3158 | const struct ofp_flow_mod *ofm, size_t n_actions) |
064af421 BP |
3159 | { |
3160 | struct ofpbuf *packet; | |
3161 | struct rule *rule; | |
3162 | uint16_t in_port; | |
3163 | int error; | |
3164 | ||
49bdc010 JP |
3165 | if (ofm->flags & htons(OFPFF_CHECK_OVERLAP)) { |
3166 | flow_t flow; | |
3167 | uint32_t wildcards; | |
3168 | ||
659586ef JG |
3169 | flow_from_match(&ofm->match, p->tun_id_from_cookie, ofm->cookie, |
3170 | &flow, &wildcards); | |
49bdc010 JP |
3171 | if (classifier_rule_overlaps(&p->cls, &flow, wildcards, |
3172 | ntohs(ofm->priority))) { | |
3173 | return ofp_mkerr(OFPET_FLOW_MOD_FAILED, OFPFMFC_OVERLAP); | |
3174 | } | |
3175 | } | |
3176 | ||
0193b2af | 3177 | rule = rule_create(p, NULL, (const union ofp_action *) ofm->actions, |
064af421 | 3178 | n_actions, ntohs(ofm->idle_timeout), |
39997502 | 3179 | ntohs(ofm->hard_timeout), ofm->cookie, |
ca069229 | 3180 | ofm->flags & htons(OFPFF_SEND_FLOW_REM)); |
659586ef JG |
3181 | cls_rule_from_match(&ofm->match, ntohs(ofm->priority), |
3182 | p->tun_id_from_cookie, ofm->cookie, &rule->cr); | |
064af421 | 3183 | |
064af421 BP |
3184 | error = 0; |
3185 | if (ofm->buffer_id != htonl(UINT32_MAX)) { | |
3186 | error = pktbuf_retrieve(ofconn->pktbuf, ntohl(ofm->buffer_id), | |
3187 | &packet, &in_port); | |
212fe71c BP |
3188 | } else { |
3189 | packet = NULL; | |
165cd8a3 | 3190 | in_port = UINT16_MAX; |
064af421 BP |
3191 | } |
3192 | ||
3193 | rule_insert(p, rule, packet, in_port); | |
3194 | ofpbuf_delete(packet); | |
3195 | return error; | |
3196 | } | |
3197 | ||
79eee1eb BP |
3198 | static struct rule * |
3199 | find_flow_strict(struct ofproto *p, const struct ofp_flow_mod *ofm) | |
064af421 | 3200 | { |
064af421 BP |
3201 | uint32_t wildcards; |
3202 | flow_t flow; | |
3203 | ||
659586ef JG |
3204 | flow_from_match(&ofm->match, p->tun_id_from_cookie, ofm->cookie, |
3205 | &flow, &wildcards); | |
79eee1eb | 3206 | return rule_from_cls_rule(classifier_find_rule_exactly( |
064af421 BP |
3207 | &p->cls, &flow, wildcards, |
3208 | ntohs(ofm->priority))); | |
79eee1eb | 3209 | } |
064af421 | 3210 | |
79eee1eb BP |
3211 | static int |
3212 | send_buffered_packet(struct ofproto *ofproto, struct ofconn *ofconn, | |
3213 | struct rule *rule, const struct ofp_flow_mod *ofm) | |
3214 | { | |
3215 | struct ofpbuf *packet; | |
3216 | uint16_t in_port; | |
3217 | flow_t flow; | |
3218 | int error; | |
064af421 | 3219 | |
79eee1eb BP |
3220 | if (ofm->buffer_id == htonl(UINT32_MAX)) { |
3221 | return 0; | |
064af421 | 3222 | } |
79eee1eb BP |
3223 | |
3224 | error = pktbuf_retrieve(ofconn->pktbuf, ntohl(ofm->buffer_id), | |
3225 | &packet, &in_port); | |
3226 | if (error) { | |
3227 | return error; | |
3228 | } | |
3229 | ||
659586ef | 3230 | flow_extract(packet, 0, in_port, &flow); |
79eee1eb BP |
3231 | rule_execute(ofproto, rule, packet, &flow); |
3232 | ofpbuf_delete(packet); | |
3233 | ||
064af421 BP |
3234 | return 0; |
3235 | } | |
79eee1eb BP |
3236 | \f |
3237 | /* OFPFC_MODIFY and OFPFC_MODIFY_STRICT. */ | |
064af421 BP |
3238 | |
3239 | struct modify_flows_cbdata { | |
3240 | struct ofproto *ofproto; | |
3241 | const struct ofp_flow_mod *ofm; | |
064af421 | 3242 | size_t n_actions; |
79eee1eb | 3243 | struct rule *match; |
064af421 BP |
3244 | }; |
3245 | ||
79eee1eb BP |
3246 | static int modify_flow(struct ofproto *, const struct ofp_flow_mod *, |
3247 | size_t n_actions, struct rule *); | |
3248 | static void modify_flows_cb(struct cls_rule *, void *cbdata_); | |
3249 | ||
3250 | /* Implements OFPFC_MODIFY. Returns 0 on success or an OpenFlow error code as | |
3251 | * encoded by ofp_mkerr() on failure. | |
3252 | * | |
3253 | * 'ofconn' is used to retrieve the packet buffer specified in ofm->buffer_id, | |
3254 | * if any. */ | |
3255 | static int | |
3256 | modify_flows_loose(struct ofproto *p, struct ofconn *ofconn, | |
3257 | const struct ofp_flow_mod *ofm, size_t n_actions) | |
3258 | { | |
3259 | struct modify_flows_cbdata cbdata; | |
3260 | struct cls_rule target; | |
3261 | ||
3262 | cbdata.ofproto = p; | |
3263 | cbdata.ofm = ofm; | |
3264 | cbdata.n_actions = n_actions; | |
3265 | cbdata.match = NULL; | |
3266 | ||
659586ef JG |
3267 | cls_rule_from_match(&ofm->match, 0, p->tun_id_from_cookie, ofm->cookie, |
3268 | &target); | |
79eee1eb BP |
3269 | |
3270 | classifier_for_each_match(&p->cls, &target, CLS_INC_ALL, | |
3271 | modify_flows_cb, &cbdata); | |
3272 | if (cbdata.match) { | |
3273 | /* This credits the packet to whichever flow happened to happened to | |
3274 | * match last. That's weird. Maybe we should do a lookup for the | |
3275 | * flow that actually matches the packet? Who knows. */ | |
3276 | send_buffered_packet(p, ofconn, cbdata.match, ofm); | |
3277 | return 0; | |
3278 | } else { | |
3279 | return add_flow(p, ofconn, ofm, n_actions); | |
3280 | } | |
3281 | } | |
3282 | ||
3283 | /* Implements OFPFC_MODIFY_STRICT. Returns 0 on success or an OpenFlow error | |
3284 | * code as encoded by ofp_mkerr() on failure. | |
3285 | * | |
3286 | * 'ofconn' is used to retrieve the packet buffer specified in ofm->buffer_id, | |
3287 | * if any. */ | |
3288 | static int | |
3289 | modify_flow_strict(struct ofproto *p, struct ofconn *ofconn, | |
3290 | struct ofp_flow_mod *ofm, size_t n_actions) | |
3291 | { | |
3292 | struct rule *rule = find_flow_strict(p, ofm); | |
3293 | if (rule && !rule_is_hidden(rule)) { | |
3294 | modify_flow(p, ofm, n_actions, rule); | |
3295 | return send_buffered_packet(p, ofconn, rule, ofm); | |
3296 | } else { | |
3297 | return add_flow(p, ofconn, ofm, n_actions); | |
3298 | } | |
3299 | } | |
3300 | ||
3301 | /* Callback for modify_flows_loose(). */ | |
064af421 BP |
3302 | static void |
3303 | modify_flows_cb(struct cls_rule *rule_, void *cbdata_) | |
3304 | { | |
3305 | struct rule *rule = rule_from_cls_rule(rule_); | |
3306 | struct modify_flows_cbdata *cbdata = cbdata_; | |
3307 | ||
79eee1eb BP |
3308 | if (!rule_is_hidden(rule)) { |
3309 | cbdata->match = rule; | |
3310 | modify_flow(cbdata->ofproto, cbdata->ofm, cbdata->n_actions, rule); | |
064af421 | 3311 | } |
064af421 BP |
3312 | } |
3313 | ||
79eee1eb BP |
3314 | /* Implements core of OFPFC_MODIFY and OFPFC_MODIFY_STRICT where 'rule' has |
3315 | * been identified as a flow in 'p''s flow table to be modified, by changing | |
3316 | * the rule's actions to match those in 'ofm' (which is followed by 'n_actions' | |
3317 | * ofp_action[] structures). */ | |
064af421 | 3318 | static int |
79eee1eb BP |
3319 | modify_flow(struct ofproto *p, const struct ofp_flow_mod *ofm, |
3320 | size_t n_actions, struct rule *rule) | |
064af421 | 3321 | { |
79eee1eb BP |
3322 | size_t actions_len = n_actions * sizeof *rule->actions; |
3323 | ||
3324 | rule->flow_cookie = ofm->cookie; | |
3325 | ||
3326 | /* If the actions are the same, do nothing. */ | |
3327 | if (n_actions == rule->n_actions | |
3328 | && !memcmp(ofm->actions, rule->actions, actions_len)) | |
3329 | { | |
3330 | return 0; | |
3331 | } | |
3332 | ||
3333 | /* Replace actions. */ | |
3334 | free(rule->actions); | |
3335 | rule->actions = xmemdup(ofm->actions, actions_len); | |
3336 | rule->n_actions = n_actions; | |
3337 | ||
3338 | /* Make sure that the datapath gets updated properly. */ | |
3339 | if (rule->cr.wc.wildcards) { | |
3340 | COVERAGE_INC(ofproto_mod_wc_flow); | |
3341 | p->need_revalidate = true; | |
3342 | } else { | |
3343 | rule_update_actions(p, rule); | |
3344 | } | |
3345 | ||
3346 | return 0; | |
3347 | } | |
3348 | \f | |
3349 | /* OFPFC_DELETE implementation. */ | |
3350 | ||
3351 | struct delete_flows_cbdata { | |
3352 | struct ofproto *ofproto; | |
3353 | uint16_t out_port; | |
3354 | }; | |
3355 | ||
3356 | static void delete_flows_cb(struct cls_rule *, void *cbdata_); | |
3357 | static void delete_flow(struct ofproto *, struct rule *, uint16_t out_port); | |
3358 | ||
3359 | /* Implements OFPFC_DELETE. */ | |
3360 | static void | |
3361 | delete_flows_loose(struct ofproto *p, const struct ofp_flow_mod *ofm) | |
3362 | { | |
3363 | struct delete_flows_cbdata cbdata; | |
064af421 BP |
3364 | struct cls_rule target; |
3365 | ||
3366 | cbdata.ofproto = p; | |
79eee1eb | 3367 | cbdata.out_port = ofm->out_port; |
064af421 | 3368 | |
659586ef JG |
3369 | cls_rule_from_match(&ofm->match, 0, p->tun_id_from_cookie, ofm->cookie, |
3370 | &target); | |
064af421 BP |
3371 | |
3372 | classifier_for_each_match(&p->cls, &target, CLS_INC_ALL, | |
79eee1eb | 3373 | delete_flows_cb, &cbdata); |
064af421 BP |
3374 | } |
3375 | ||
79eee1eb BP |
3376 | /* Implements OFPFC_DELETE_STRICT. */ |
3377 | static void | |
3378 | delete_flow_strict(struct ofproto *p, struct ofp_flow_mod *ofm) | |
3379 | { | |
3380 | struct rule *rule = find_flow_strict(p, ofm); | |
3381 | if (rule) { | |
3382 | delete_flow(p, rule, ofm->out_port); | |
3383 | } | |
3384 | } | |
3385 | ||
3386 | /* Callback for delete_flows_loose(). */ | |
3387 | static void | |
3388 | delete_flows_cb(struct cls_rule *rule_, void *cbdata_) | |
3389 | { | |
3390 | struct rule *rule = rule_from_cls_rule(rule_); | |
3391 | struct delete_flows_cbdata *cbdata = cbdata_; | |
3392 | ||
3393 | delete_flow(cbdata->ofproto, rule, cbdata->out_port); | |
3394 | } | |
3395 | ||
3396 | /* Implements core of OFPFC_DELETE and OFPFC_DELETE_STRICT where 'rule' has | |
3397 | * been identified as a flow to delete from 'p''s flow table, by deleting the | |
3398 | * flow and sending out a OFPT_FLOW_REMOVED message to any interested | |
3399 | * controller. | |
3400 | * | |
3401 | * Will not delete 'rule' if it is hidden. Will delete 'rule' only if | |
3402 | * 'out_port' is htons(OFPP_NONE) or if 'rule' actually outputs to the | |
3403 | * specified 'out_port'. */ | |
3404 | static void | |
3405 | delete_flow(struct ofproto *p, struct rule *rule, uint16_t out_port) | |
3406 | { | |
3407 | if (rule_is_hidden(rule)) { | |
3408 | return; | |
3409 | } | |
3410 | ||
3411 | if (out_port != htons(OFPP_NONE) && !rule_has_out_port(rule, out_port)) { | |
3412 | return; | |
3413 | } | |
3414 | ||
3415 | send_flow_removed(p, rule, time_msec(), OFPRR_DELETE); | |
3416 | rule_remove(p, rule); | |
3417 | } | |
3418 | \f | |
064af421 BP |
3419 | static int |
3420 | handle_flow_mod(struct ofproto *p, struct ofconn *ofconn, | |
3421 | struct ofp_flow_mod *ofm) | |
3422 | { | |
3423 | size_t n_actions; | |
3424 | int error; | |
3425 | ||
9deba63b BP |
3426 | error = reject_slave_controller(ofconn, &ofm->header); |
3427 | if (error) { | |
3428 | return error; | |
3429 | } | |
064af421 BP |
3430 | error = check_ofp_message_array(&ofm->header, OFPT_FLOW_MOD, sizeof *ofm, |
3431 | sizeof *ofm->actions, &n_actions); | |
3432 | if (error) { | |
3433 | return error; | |
3434 | } | |
3435 | ||
49bdc010 JP |
3436 | /* We do not support the emergency flow cache. It will hopefully |
3437 | * get dropped from OpenFlow in the near future. */ | |
3438 | if (ofm->flags & htons(OFPFF_EMERG)) { | |
3439 | /* There isn't a good fit for an error code, so just state that the | |
3440 | * flow table is full. */ | |
3441 | return ofp_mkerr(OFPET_FLOW_MOD_FAILED, OFPFMFC_ALL_TABLES_FULL); | |
3442 | } | |
3443 | ||
064af421 BP |
3444 | normalize_match(&ofm->match); |
3445 | if (!ofm->match.wildcards) { | |
3446 | ofm->priority = htons(UINT16_MAX); | |
3447 | } | |
3448 | ||
3449 | error = validate_actions((const union ofp_action *) ofm->actions, | |
3450 | n_actions, p->max_ports); | |
3451 | if (error) { | |
3452 | return error; | |
3453 | } | |
3454 | ||
3455 | switch (ntohs(ofm->command)) { | |
3456 | case OFPFC_ADD: | |
3457 | return add_flow(p, ofconn, ofm, n_actions); | |
3458 | ||
3459 | case OFPFC_MODIFY: | |
79eee1eb | 3460 | return modify_flows_loose(p, ofconn, ofm, n_actions); |
064af421 BP |
3461 | |
3462 | case OFPFC_MODIFY_STRICT: | |
79eee1eb | 3463 | return modify_flow_strict(p, ofconn, ofm, n_actions); |
064af421 BP |
3464 | |
3465 | case OFPFC_DELETE: | |
79eee1eb BP |
3466 | delete_flows_loose(p, ofm); |
3467 | return 0; | |
064af421 BP |
3468 | |
3469 | case OFPFC_DELETE_STRICT: | |
79eee1eb BP |
3470 | delete_flow_strict(p, ofm); |
3471 | return 0; | |
064af421 BP |
3472 | |
3473 | default: | |
3474 | return ofp_mkerr(OFPET_FLOW_MOD_FAILED, OFPFMFC_BAD_COMMAND); | |
3475 | } | |
3476 | } | |
3477 | ||
659586ef JG |
3478 | static int |
3479 | handle_tun_id_from_cookie(struct ofproto *p, struct nxt_tun_id_cookie *msg) | |
3480 | { | |
3481 | int error; | |
3482 | ||
3483 | error = check_ofp_message(&msg->header, OFPT_VENDOR, sizeof *msg); | |
3484 | if (error) { | |
3485 | return error; | |
3486 | } | |
3487 | ||
3488 | p->tun_id_from_cookie = !!msg->set; | |
3489 | return 0; | |
3490 | } | |
3491 | ||
9deba63b BP |
3492 | static int |
3493 | handle_role_request(struct ofproto *ofproto, | |
3494 | struct ofconn *ofconn, struct nicira_header *msg) | |
3495 | { | |
3496 | struct nx_role_request *nrr; | |
3497 | struct nx_role_request *reply; | |
3498 | struct ofpbuf *buf; | |
3499 | uint32_t role; | |
3500 | ||
3501 | if (ntohs(msg->header.length) != sizeof *nrr) { | |
3502 | VLOG_WARN_RL(&rl, "received role request of length %zu (expected %zu)", | |
3503 | ntohs(msg->header.length), sizeof *nrr); | |
3504 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); | |
3505 | } | |
3506 | nrr = (struct nx_role_request *) msg; | |
3507 | ||
3508 | if (ofconn->type != OFCONN_CONTROLLER) { | |
3509 | VLOG_WARN_RL(&rl, "ignoring role request on non-controller " | |
3510 | "connection"); | |
3511 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_EPERM); | |
3512 | } | |
3513 | ||
3514 | role = ntohl(nrr->role); | |
3515 | if (role != NX_ROLE_OTHER && role != NX_ROLE_MASTER | |
3516 | && role != NX_ROLE_SLAVE) { | |
3517 | VLOG_WARN_RL(&rl, "received request for unknown role %"PRIu32, role); | |
3518 | ||
3519 | /* There's no good error code for this. */ | |
3520 | return ofp_mkerr(OFPET_BAD_REQUEST, -1); | |
3521 | } | |
3522 | ||
3523 | if (role == NX_ROLE_MASTER) { | |
3524 | struct ofconn *other; | |
3525 | ||
3526 | HMAP_FOR_EACH (other, struct ofconn, hmap_node, | |
3527 | &ofproto->controllers) { | |
3528 | if (other->role == NX_ROLE_MASTER) { | |
3529 | other->role = NX_ROLE_SLAVE; | |
3530 | } | |
3531 | } | |
3532 | } | |
3533 | ofconn->role = role; | |
3534 | ||
3535 | reply = make_openflow_xid(sizeof *reply, OFPT_VENDOR, msg->header.xid, | |
3536 | &buf); | |
3537 | reply->nxh.vendor = htonl(NX_VENDOR_ID); | |
3538 | reply->nxh.subtype = htonl(NXT_ROLE_REPLY); | |
3539 | reply->role = htonl(role); | |
3540 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
3541 | ||
3542 | return 0; | |
3543 | } | |
3544 | ||
064af421 BP |
3545 | static int |
3546 | handle_vendor(struct ofproto *p, struct ofconn *ofconn, void *msg) | |
3547 | { | |
3548 | struct ofp_vendor_header *ovh = msg; | |
3549 | struct nicira_header *nh; | |
3550 | ||
3551 | if (ntohs(ovh->header.length) < sizeof(struct ofp_vendor_header)) { | |
659586ef JG |
3552 | VLOG_WARN_RL(&rl, "received vendor message of length %zu " |
3553 | "(expected at least %zu)", | |
3554 | ntohs(ovh->header.length), sizeof(struct ofp_vendor_header)); | |
49bdc010 | 3555 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); |
064af421 BP |
3556 | } |
3557 | if (ovh->vendor != htonl(NX_VENDOR_ID)) { | |
3558 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_VENDOR); | |
3559 | } | |
3560 | if (ntohs(ovh->header.length) < sizeof(struct nicira_header)) { | |
659586ef JG |
3561 | VLOG_WARN_RL(&rl, "received Nicira vendor message of length %zu " |
3562 | "(expected at least %zu)", | |
3563 | ntohs(ovh->header.length), sizeof(struct nicira_header)); | |
49bdc010 | 3564 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_LEN); |
064af421 BP |
3565 | } |
3566 | ||
3567 | nh = msg; | |
3568 | switch (ntohl(nh->subtype)) { | |
3569 | case NXT_STATUS_REQUEST: | |
3570 | return switch_status_handle_request(p->switch_status, ofconn->rconn, | |
3571 | msg); | |
659586ef JG |
3572 | |
3573 | case NXT_TUN_ID_FROM_COOKIE: | |
3574 | return handle_tun_id_from_cookie(p, msg); | |
9deba63b BP |
3575 | |
3576 | case NXT_ROLE_REQUEST: | |
3577 | return handle_role_request(p, ofconn, msg); | |
064af421 BP |
3578 | } |
3579 | ||
3580 | return ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_SUBTYPE); | |
3581 | } | |
3582 | ||
246e61ea JP |
3583 | static int |
3584 | handle_barrier_request(struct ofconn *ofconn, struct ofp_header *oh) | |
3585 | { | |
3586 | struct ofp_header *ob; | |
3587 | struct ofpbuf *buf; | |
3588 | ||
3589 | /* Currently, everything executes synchronously, so we can just | |
3590 | * immediately send the barrier reply. */ | |
3591 | ob = make_openflow_xid(sizeof *ob, OFPT_BARRIER_REPLY, oh->xid, &buf); | |
3592 | queue_tx(buf, ofconn, ofconn->reply_counter); | |
3593 | return 0; | |
3594 | } | |
3595 | ||
064af421 BP |
3596 | static void |
3597 | handle_openflow(struct ofconn *ofconn, struct ofproto *p, | |
3598 | struct ofpbuf *ofp_msg) | |
3599 | { | |
3600 | struct ofp_header *oh = ofp_msg->data; | |
3601 | int error; | |
3602 | ||
3603 | COVERAGE_INC(ofproto_recv_openflow); | |
3604 | switch (oh->type) { | |
3605 | case OFPT_ECHO_REQUEST: | |
3606 | error = handle_echo_request(ofconn, oh); | |
3607 | break; | |
3608 | ||
3609 | case OFPT_ECHO_REPLY: | |
3610 | error = 0; | |
3611 | break; | |
3612 | ||
3613 | case OFPT_FEATURES_REQUEST: | |
3614 | error = handle_features_request(p, ofconn, oh); | |
3615 | break; | |
3616 | ||
3617 | case OFPT_GET_CONFIG_REQUEST: | |
3618 | error = handle_get_config_request(p, ofconn, oh); | |
3619 | break; | |
3620 | ||
3621 | case OFPT_SET_CONFIG: | |
3622 | error = handle_set_config(p, ofconn, ofp_msg->data); | |
3623 | break; | |
3624 | ||
3625 | case OFPT_PACKET_OUT: | |
3626 | error = handle_packet_out(p, ofconn, ofp_msg->data); | |
3627 | break; | |
3628 | ||
3629 | case OFPT_PORT_MOD: | |
9deba63b | 3630 | error = handle_port_mod(p, ofconn, oh); |
064af421 BP |
3631 | break; |
3632 | ||
3633 | case OFPT_FLOW_MOD: | |
3634 | error = handle_flow_mod(p, ofconn, ofp_msg->data); | |
3635 | break; | |
3636 | ||
3637 | case OFPT_STATS_REQUEST: | |
3638 | error = handle_stats_request(p, ofconn, oh); | |
3639 | break; | |
3640 | ||
3641 | case OFPT_VENDOR: | |
3642 | error = handle_vendor(p, ofconn, ofp_msg->data); | |
3643 | break; | |
3644 | ||
246e61ea JP |
3645 | case OFPT_BARRIER_REQUEST: |
3646 | error = handle_barrier_request(ofconn, oh); | |
3647 | break; | |
3648 | ||
064af421 BP |
3649 | default: |
3650 | if (VLOG_IS_WARN_ENABLED()) { | |
3651 | char *s = ofp_to_string(oh, ntohs(oh->length), 2); | |
3652 | VLOG_DBG_RL(&rl, "OpenFlow message ignored: %s", s); | |
3653 | free(s); | |
3654 | } | |
3655 | error = ofp_mkerr(OFPET_BAD_REQUEST, OFPBRC_BAD_TYPE); | |
3656 | break; | |
3657 | } | |
3658 | ||
3659 | if (error) { | |
3660 | send_error_oh(ofconn, ofp_msg->data, error); | |
3661 | } | |
3662 | } | |
3663 | \f | |
3664 | static void | |
72b06300 | 3665 | handle_odp_miss_msg(struct ofproto *p, struct ofpbuf *packet) |
064af421 BP |
3666 | { |
3667 | struct odp_msg *msg = packet->data; | |
064af421 BP |
3668 | struct rule *rule; |
3669 | struct ofpbuf payload; | |
3670 | flow_t flow; | |
3671 | ||
064af421 BP |
3672 | payload.data = msg + 1; |
3673 | payload.size = msg->length - sizeof *msg; | |
659586ef | 3674 | flow_extract(&payload, msg->arg, msg->port, &flow); |
064af421 | 3675 | |
0ad9b732 JP |
3676 | /* Check with in-band control to see if this packet should be sent |
3677 | * to the local port regardless of the flow table. */ | |
3678 | if (in_band_msg_in_hook(p->in_band, &flow, &payload)) { | |
3679 | union odp_action action; | |
3680 | ||
3681 | memset(&action, 0, sizeof(action)); | |
3682 | action.output.type = ODPAT_OUTPUT; | |
3683 | action.output.port = ODPP_LOCAL; | |
f1acd62b | 3684 | dpif_execute(p->dpif, flow.in_port, &action, 1, &payload); |
0ad9b732 JP |
3685 | } |
3686 | ||
064af421 BP |
3687 | rule = lookup_valid_rule(p, &flow); |
3688 | if (!rule) { | |
3689 | /* Don't send a packet-in if OFPPC_NO_PACKET_IN asserted. */ | |
3690 | struct ofport *port = port_array_get(&p->ports, msg->port); | |
3691 | if (port) { | |
3692 | if (port->opp.config & OFPPC_NO_PACKET_IN) { | |
3693 | COVERAGE_INC(ofproto_no_packet_in); | |
3694 | /* XXX install 'drop' flow entry */ | |
3695 | ofpbuf_delete(packet); | |
3696 | return; | |
3697 | } | |
3698 | } else { | |
3699 | VLOG_WARN_RL(&rl, "packet-in on unknown port %"PRIu16, msg->port); | |
3700 | } | |
3701 | ||
3702 | COVERAGE_INC(ofproto_packet_in); | |
76ce9432 | 3703 | send_packet_in(p, packet); |
064af421 BP |
3704 | return; |
3705 | } | |
3706 | ||
3707 | if (rule->cr.wc.wildcards) { | |
3708 | rule = rule_create_subrule(p, rule, &flow); | |
3709 | rule_make_actions(p, rule, packet); | |
3710 | } else { | |
3711 | if (!rule->may_install) { | |
3712 | /* The rule is not installable, that is, we need to process every | |
3713 | * packet, so process the current packet and set its actions into | |
3714 | * 'subrule'. */ | |
3715 | rule_make_actions(p, rule, packet); | |
3716 | } else { | |
3717 | /* XXX revalidate rule if it needs it */ | |
3718 | } | |
3719 | } | |
3720 | ||
3721 | rule_execute(p, rule, &payload, &flow); | |
3722 | rule_reinstall(p, rule); | |
7778bd15 | 3723 | |
76ce9432 | 3724 | if (rule->super && rule->super->cr.priority == FAIL_OPEN_PRIORITY) { |
7778bd15 BP |
3725 | /* |
3726 | * Extra-special case for fail-open mode. | |
3727 | * | |
3728 | * We are in fail-open mode and the packet matched the fail-open rule, | |
3729 | * but we are connected to a controller too. We should send the packet | |
3730 | * up to the controller in the hope that it will try to set up a flow | |
3731 | * and thereby allow us to exit fail-open. | |
3732 | * | |
3733 | * See the top-level comment in fail-open.c for more information. | |
3734 | */ | |
76ce9432 | 3735 | send_packet_in(p, packet); |
7778bd15 BP |
3736 | } else { |
3737 | ofpbuf_delete(packet); | |
3738 | } | |
064af421 | 3739 | } |
72b06300 BP |
3740 | |
3741 | static void | |
3742 | handle_odp_msg(struct ofproto *p, struct ofpbuf *packet) | |
3743 | { | |
3744 | struct odp_msg *msg = packet->data; | |
3745 | ||
3746 | switch (msg->type) { | |
3747 | case _ODPL_ACTION_NR: | |
3748 | COVERAGE_INC(ofproto_ctlr_action); | |
76ce9432 | 3749 | send_packet_in(p, packet); |
72b06300 BP |
3750 | break; |
3751 | ||
3752 | case _ODPL_SFLOW_NR: | |
3753 | if (p->sflow) { | |
3754 | ofproto_sflow_received(p->sflow, msg); | |
3755 | } | |
3756 | ofpbuf_delete(packet); | |
3757 | break; | |
3758 | ||
3759 | case _ODPL_MISS_NR: | |
3760 | handle_odp_miss_msg(p, packet); | |
3761 | break; | |
3762 | ||
3763 | default: | |
3764 | VLOG_WARN_RL(&rl, "received ODP message of unexpected type %"PRIu32, | |
3765 | msg->type); | |
3766 | break; | |
3767 | } | |
3768 | } | |
064af421 BP |
3769 | \f |
3770 | static void | |
3771 | revalidate_cb(struct cls_rule *sub_, void *cbdata_) | |
3772 | { | |
3773 | struct rule *sub = rule_from_cls_rule(sub_); | |
3774 | struct revalidate_cbdata *cbdata = cbdata_; | |
3775 | ||
3776 | if (cbdata->revalidate_all | |
3777 | || (cbdata->revalidate_subrules && sub->super) | |
3778 | || (tag_set_intersects(&cbdata->revalidate_set, sub->tags))) { | |
3779 | revalidate_rule(cbdata->ofproto, sub); | |
3780 | } | |
3781 | } | |
3782 | ||
3783 | static bool | |
3784 | revalidate_rule(struct ofproto *p, struct rule *rule) | |
3785 | { | |
3786 | const flow_t *flow = &rule->cr.flow; | |
3787 | ||
3788 | COVERAGE_INC(ofproto_revalidate_rule); | |
3789 | if (rule->super) { | |
3790 | struct rule *super; | |
3791 | super = rule_from_cls_rule(classifier_lookup_wild(&p->cls, flow)); | |
3792 | if (!super) { | |
3793 | rule_remove(p, rule); | |
3794 | return false; | |
3795 | } else if (super != rule->super) { | |
3796 | COVERAGE_INC(ofproto_revalidate_moved); | |
3797 | list_remove(&rule->list); | |
3798 | list_push_back(&super->list, &rule->list); | |
3799 | rule->super = super; | |
3800 | rule->hard_timeout = super->hard_timeout; | |
3801 | rule->idle_timeout = super->idle_timeout; | |
3802 | rule->created = super->created; | |
3803 | rule->used = 0; | |
3804 | } | |
3805 | } | |
3806 | ||
3807 | rule_update_actions(p, rule); | |
3808 | return true; | |
3809 | } | |
3810 | ||
3811 | static struct ofpbuf * | |
659586ef JG |
3812 | compose_flow_removed(struct ofproto *p, const struct rule *rule, |
3813 | long long int now, uint8_t reason) | |
064af421 | 3814 | { |
ca069229 | 3815 | struct ofp_flow_removed *ofr; |
064af421 | 3816 | struct ofpbuf *buf; |
9ca76894 | 3817 | long long int tdiff = now - rule->created; |
26c3f94a JP |
3818 | uint32_t sec = tdiff / 1000; |
3819 | uint32_t msec = tdiff - (sec * 1000); | |
064af421 | 3820 | |
ca069229 | 3821 | ofr = make_openflow(sizeof *ofr, OFPT_FLOW_REMOVED, &buf); |
659586ef JG |
3822 | flow_to_match(&rule->cr.flow, rule->cr.wc.wildcards, p->tun_id_from_cookie, |
3823 | &ofr->match); | |
39997502 | 3824 | ofr->cookie = rule->flow_cookie; |
ca069229 JP |
3825 | ofr->priority = htons(rule->cr.priority); |
3826 | ofr->reason = reason; | |
26c3f94a JP |
3827 | ofr->duration_sec = htonl(sec); |
3828 | ofr->duration_nsec = htonl(msec * 1000000); | |
ca069229 JP |
3829 | ofr->idle_timeout = htons(rule->idle_timeout); |
3830 | ofr->packet_count = htonll(rule->packet_count); | |
3831 | ofr->byte_count = htonll(rule->byte_count); | |
064af421 BP |
3832 | |
3833 | return buf; | |
3834 | } | |
3835 | ||
3836 | static void | |
ca069229 JP |
3837 | uninstall_idle_flow(struct ofproto *ofproto, struct rule *rule) |
3838 | { | |
3839 | assert(rule->installed); | |
3840 | assert(!rule->cr.wc.wildcards); | |
3841 | ||
3842 | if (rule->super) { | |
3843 | rule_remove(ofproto, rule); | |
3844 | } else { | |
3845 | rule_uninstall(ofproto, rule); | |
3846 | } | |
3847 | } | |
9deba63b | 3848 | |
ca069229 JP |
3849 | static void |
3850 | send_flow_removed(struct ofproto *p, struct rule *rule, | |
3851 | long long int now, uint8_t reason) | |
064af421 BP |
3852 | { |
3853 | struct ofconn *ofconn; | |
3854 | struct ofconn *prev; | |
b9b0ce61 | 3855 | struct ofpbuf *buf = NULL; |
064af421 BP |
3856 | |
3857 | /* We limit the maximum number of queued flow expirations it by accounting | |
3858 | * them under the counter for replies. That works because preventing | |
3859 | * OpenFlow requests from being processed also prevents new flows from | |
3860 | * being added (and expiring). (It also prevents processing OpenFlow | |
3861 | * requests that would not add new flows, so it is imperfect.) */ | |
3862 | ||
3863 | prev = NULL; | |
3864 | LIST_FOR_EACH (ofconn, struct ofconn, node, &p->all_conns) { | |
9deba63b BP |
3865 | if (rule->send_flow_removed && rconn_is_connected(ofconn->rconn) |
3866 | && ofconn->role != NX_ROLE_SLAVE) { | |
064af421 | 3867 | if (prev) { |
431d8ad2 | 3868 | queue_tx(ofpbuf_clone(buf), prev, prev->reply_counter); |
064af421 | 3869 | } else { |
659586ef | 3870 | buf = compose_flow_removed(p, rule, now, reason); |
064af421 BP |
3871 | } |
3872 | prev = ofconn; | |
3873 | } | |
3874 | } | |
3875 | if (prev) { | |
431d8ad2 | 3876 | queue_tx(buf, prev, prev->reply_counter); |
064af421 BP |
3877 | } |
3878 | } | |
3879 | ||
064af421 BP |
3880 | |
3881 | static void | |
3882 | expire_rule(struct cls_rule *cls_rule, void *p_) | |
3883 | { | |
3884 | struct ofproto *p = p_; | |
3885 | struct rule *rule = rule_from_cls_rule(cls_rule); | |
3886 | long long int hard_expire, idle_expire, expire, now; | |
3887 | ||
3888 | hard_expire = (rule->hard_timeout | |
3889 | ? rule->created + rule->hard_timeout * 1000 | |
3890 | : LLONG_MAX); | |
3891 | idle_expire = (rule->idle_timeout | |
3892 | && (rule->super || list_is_empty(&rule->list)) | |
3893 | ? rule->used + rule->idle_timeout * 1000 | |
3894 | : LLONG_MAX); | |
3895 | expire = MIN(hard_expire, idle_expire); | |
064af421 BP |
3896 | |
3897 | now = time_msec(); | |
3898 | if (now < expire) { | |
3899 | if (rule->installed && now >= rule->used + 5000) { | |
3900 | uninstall_idle_flow(p, rule); | |
0193b2af JG |
3901 | } else if (!rule->cr.wc.wildcards) { |
3902 | active_timeout(p, rule); | |
064af421 | 3903 | } |
0193b2af | 3904 | |
064af421 BP |
3905 | return; |
3906 | } | |
3907 | ||
3908 | COVERAGE_INC(ofproto_expired); | |
46d6f36f JG |
3909 | |
3910 | /* Update stats. This code will be a no-op if the rule expired | |
3911 | * due to an idle timeout. */ | |
064af421 | 3912 | if (rule->cr.wc.wildcards) { |
064af421 BP |
3913 | struct rule *subrule, *next; |
3914 | LIST_FOR_EACH_SAFE (subrule, next, struct rule, list, &rule->list) { | |
3915 | rule_remove(p, subrule); | |
3916 | } | |
46d6f36f JG |
3917 | } else { |
3918 | rule_uninstall(p, rule); | |
064af421 BP |
3919 | } |
3920 | ||
8fe1a59d | 3921 | if (!rule_is_hidden(rule)) { |
ca069229 JP |
3922 | send_flow_removed(p, rule, now, |
3923 | (now >= hard_expire | |
3924 | ? OFPRR_HARD_TIMEOUT : OFPRR_IDLE_TIMEOUT)); | |
8fe1a59d | 3925 | } |
064af421 BP |
3926 | rule_remove(p, rule); |
3927 | } | |
3928 | ||
0193b2af JG |
3929 | static void |
3930 | active_timeout(struct ofproto *ofproto, struct rule *rule) | |
3931 | { | |
3932 | if (ofproto->netflow && !is_controller_rule(rule) && | |
3933 | netflow_active_timeout_expired(ofproto->netflow, &rule->nf_flow)) { | |
3934 | struct ofexpired expired; | |
3935 | struct odp_flow odp_flow; | |
3936 | ||
3937 | /* Get updated flow stats. */ | |
3938 | memset(&odp_flow, 0, sizeof odp_flow); | |
094e1514 JG |
3939 | if (rule->installed) { |
3940 | odp_flow.key = rule->cr.flow; | |
3941 | odp_flow.flags = ODPFF_ZERO_TCP_FLAGS; | |
d65349ea | 3942 | dpif_flow_get(ofproto->dpif, &odp_flow); |
094e1514 JG |
3943 | |
3944 | if (odp_flow.stats.n_packets) { | |
3945 | update_time(ofproto, rule, &odp_flow.stats); | |
3946 | netflow_flow_update_flags(&rule->nf_flow, odp_flow.stats.ip_tos, | |
3947 | odp_flow.stats.tcp_flags); | |
3948 | } | |
0193b2af JG |
3949 | } |
3950 | ||
3951 | expired.flow = rule->cr.flow; | |
3952 | expired.packet_count = rule->packet_count + | |
3953 | odp_flow.stats.n_packets; | |
3954 | expired.byte_count = rule->byte_count + odp_flow.stats.n_bytes; | |
3955 | expired.used = rule->used; | |
3956 | ||
3957 | netflow_expire(ofproto->netflow, &rule->nf_flow, &expired); | |
3958 | ||
3959 | /* Schedule us to send the accumulated records once we have | |
3960 | * collected all of them. */ | |
3961 | poll_immediate_wake(); | |
3962 | } | |
3963 | } | |
3964 | ||
064af421 BP |
3965 | static void |
3966 | update_used(struct ofproto *p) | |
3967 | { | |
3968 | struct odp_flow *flows; | |
3969 | size_t n_flows; | |
3970 | size_t i; | |
3971 | int error; | |
3972 | ||
c228a364 | 3973 | error = dpif_flow_list_all(p->dpif, &flows, &n_flows); |
064af421 BP |
3974 | if (error) { |
3975 | return; | |
3976 | } | |
3977 | ||
3978 | for (i = 0; i < n_flows; i++) { | |
3979 | struct odp_flow *f = &flows[i]; | |
3980 | struct rule *rule; | |
3981 | ||
3982 | rule = rule_from_cls_rule( | |
3983 | classifier_find_rule_exactly(&p->cls, &f->key, 0, UINT16_MAX)); | |
3984 | if (!rule || !rule->installed) { | |
3985 | COVERAGE_INC(ofproto_unexpected_rule); | |
c228a364 | 3986 | dpif_flow_del(p->dpif, f); |
064af421 BP |
3987 | continue; |
3988 | } | |
3989 | ||
0193b2af | 3990 | update_time(p, rule, &f->stats); |
064af421 BP |
3991 | rule_account(p, rule, f->stats.n_bytes); |
3992 | } | |
3993 | free(flows); | |
3994 | } | |
3995 | ||
3996 | static void | |
76ce9432 | 3997 | do_send_packet_in(struct ofpbuf *packet, void *ofconn_) |
064af421 | 3998 | { |
76ce9432 BP |
3999 | struct ofconn *ofconn = ofconn_; |
4000 | struct ofproto *ofproto = ofconn->ofproto; | |
372179d4 BP |
4001 | struct odp_msg *msg = packet->data; |
4002 | struct ofpbuf payload; | |
4003 | struct ofpbuf *opi; | |
76ce9432 BP |
4004 | uint32_t buffer_id; |
4005 | int send_len; | |
064af421 | 4006 | |
372179d4 | 4007 | /* Extract packet payload from 'msg'. */ |
064af421 BP |
4008 | payload.data = msg + 1; |
4009 | payload.size = msg->length - sizeof *msg; | |
4010 | ||
76ce9432 BP |
4011 | /* Construct packet-in message. */ |
4012 | send_len = INT_MAX; | |
4013 | if (msg->type == _ODPL_ACTION_NR) { | |
4014 | buffer_id = UINT32_MAX; | |
4015 | } else { | |
4016 | if (ofproto->fail_open && fail_open_is_active(ofproto->fail_open)) { | |
4017 | buffer_id = pktbuf_get_null(); | |
4018 | } else { | |
4019 | buffer_id = pktbuf_save(ofconn->pktbuf, &payload, msg->port); | |
4020 | } | |
4021 | if (buffer_id != UINT32_MAX) { | |
4022 | send_len = ofconn->miss_send_len; | |
4023 | } | |
4024 | } | |
4025 | opi = make_packet_in(buffer_id, odp_port_to_ofp_port(msg->port), | |
4026 | msg->type, &payload, send_len); | |
372179d4 BP |
4027 | |
4028 | /* Send. */ | |
4029 | rconn_send_with_limit(ofconn->rconn, opi, ofconn->packet_in_counter, 100); | |
064af421 | 4030 | |
064af421 BP |
4031 | ofpbuf_delete(packet); |
4032 | } | |
4033 | ||
4034 | static void | |
76ce9432 | 4035 | send_packet_in(struct ofproto *ofproto, struct ofpbuf *packet) |
064af421 | 4036 | { |
76ce9432 BP |
4037 | struct odp_msg *msg = packet->data; |
4038 | struct ofconn *ofconn, *prev; | |
064af421 | 4039 | |
76ce9432 BP |
4040 | assert(msg->type == _ODPL_MISS_NR || msg->type == _ODPL_ACTION_NR); |
4041 | ||
4042 | prev = NULL; | |
4043 | LIST_FOR_EACH (ofconn, struct ofconn, node, &ofproto->all_conns) { | |
9deba63b BP |
4044 | if (ofconn->role != NX_ROLE_SLAVE) { |
4045 | if (prev) { | |
4046 | pinsched_send(prev->schedulers[msg->type], msg->port, | |
4047 | ofpbuf_clone(packet), do_send_packet_in, prev); | |
4048 | } | |
4049 | prev = ofconn; | |
064af421 | 4050 | } |
76ce9432 BP |
4051 | } |
4052 | if (prev) { | |
4053 | pinsched_send(prev->schedulers[msg->type], msg->port, | |
4054 | packet, do_send_packet_in, prev); | |
4055 | } else { | |
4056 | ofpbuf_delete(packet); | |
064af421 | 4057 | } |
064af421 BP |
4058 | } |
4059 | ||
4060 | static uint64_t | |
fa60c019 | 4061 | pick_datapath_id(const struct ofproto *ofproto) |
064af421 | 4062 | { |
fa60c019 | 4063 | const struct ofport *port; |
064af421 | 4064 | |
fa60c019 BP |
4065 | port = port_array_get(&ofproto->ports, ODPP_LOCAL); |
4066 | if (port) { | |
4067 | uint8_t ea[ETH_ADDR_LEN]; | |
4068 | int error; | |
4069 | ||
4070 | error = netdev_get_etheraddr(port->netdev, ea); | |
064af421 BP |
4071 | if (!error) { |
4072 | return eth_addr_to_uint64(ea); | |
4073 | } | |
4074 | VLOG_WARN("could not get MAC address for %s (%s)", | |
fa60c019 | 4075 | netdev_get_name(port->netdev), strerror(error)); |
064af421 | 4076 | } |
fa60c019 | 4077 | return ofproto->fallback_dpid; |
064af421 BP |
4078 | } |
4079 | ||
4080 | static uint64_t | |
4081 | pick_fallback_dpid(void) | |
4082 | { | |
4083 | uint8_t ea[ETH_ADDR_LEN]; | |
70150daf | 4084 | eth_addr_nicira_random(ea); |
064af421 BP |
4085 | return eth_addr_to_uint64(ea); |
4086 | } | |
4087 | \f | |
4088 | static bool | |
4089 | default_normal_ofhook_cb(const flow_t *flow, const struct ofpbuf *packet, | |
4090 | struct odp_actions *actions, tag_type *tags, | |
6a07af36 | 4091 | uint16_t *nf_output_iface, void *ofproto_) |
064af421 BP |
4092 | { |
4093 | struct ofproto *ofproto = ofproto_; | |
4094 | int out_port; | |
4095 | ||
4096 | /* Drop frames for reserved multicast addresses. */ | |
4097 | if (eth_addr_is_reserved(flow->dl_dst)) { | |
4098 | return true; | |
4099 | } | |
4100 | ||
4101 | /* Learn source MAC (but don't try to learn from revalidation). */ | |
4102 | if (packet != NULL) { | |
4103 | tag_type rev_tag = mac_learning_learn(ofproto->ml, flow->dl_src, | |
4104 | 0, flow->in_port); | |
4105 | if (rev_tag) { | |
4106 | /* The log messages here could actually be useful in debugging, | |
4107 | * so keep the rate limit relatively high. */ | |
4108 | static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(30, 300); | |
4109 | VLOG_DBG_RL(&rl, "learned that "ETH_ADDR_FMT" is on port %"PRIu16, | |
4110 | ETH_ADDR_ARGS(flow->dl_src), flow->in_port); | |
4111 | ofproto_revalidate(ofproto, rev_tag); | |
4112 | } | |
4113 | } | |
4114 | ||
4115 | /* Determine output port. */ | |
4116 | out_port = mac_learning_lookup_tag(ofproto->ml, flow->dl_dst, 0, tags); | |
4117 | if (out_port < 0) { | |
6a07af36 | 4118 | add_output_group_action(actions, DP_GROUP_FLOOD, nf_output_iface); |
064af421 BP |
4119 | } else if (out_port != flow->in_port) { |
4120 | odp_actions_add(actions, ODPAT_OUTPUT)->output.port = out_port; | |
6a07af36 | 4121 | *nf_output_iface = out_port; |
064af421 BP |
4122 | } else { |
4123 | /* Drop. */ | |
4124 | } | |
4125 | ||
4126 | return true; | |
4127 | } | |
4128 | ||
4129 | static const struct ofhooks default_ofhooks = { | |
4130 | NULL, | |
4131 | default_normal_ofhook_cb, | |
4132 | NULL, | |
4133 | NULL | |
4134 | }; |