2 * Copyright (c) 2008, 2009, 2010 Nicira Networks.
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:
8 * http://www.apache.org/licenses/LICENSE-2.0
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.
18 #include "fail-open.h"
22 #include "mac-learning.h"
26 #include "poll-loop.h"
32 #define THIS_MODULE VLM_fail_open
38 * In fail-open mode, the switch detects when the controller cannot be
39 * contacted or when the controller is dropping switch connections because the
40 * switch does not pass its admission control policy. In those situations the
41 * switch sets up flows itself using the "normal" action.
43 * There is a little subtlety to implementation, to properly handle the case
44 * where the controller allows switch connections but drops them a few seconds
45 * later for admission control reasons. Because of this case, we don't want to
46 * just stop setting up flows when we connect to the controller: if we did,
47 * then new flow setup and existing flows would stop during the duration of
48 * connection to the controller, and thus the whole network would go down for
49 * that period of time.
51 * So, instead, we add some special caseswhen we are connected to a controller,
52 * but not yet sure that it has admitted us:
54 * - We set up flows immediately ourselves, but simultaneously send out an
55 * OFPT_PACKET_IN to the controller. We put a special bogus buffer-id in
56 * these OFPT_PACKET_IN messages so that duplicate packets don't get sent
57 * out to the network when the controller replies.
59 * - We also send out OFPT_PACKET_IN messages for totally bogus packets
60 * every so often, in case no real new flows are arriving in the network.
62 * - We don't flush the flow table at the time we connect, because this
63 * could cause network stuttering in a switch with lots of flows or very
64 * high-bandwidth flows by suddenly throwing lots of packets down to
69 struct ofproto *ofproto;
70 struct rconn *controller;
72 int last_disconn_secs;
73 struct status_category *ss_cat;
74 long long int next_bogus_packet_in;
75 struct rconn_packet_counter *bogus_packet_counter;
78 static void fail_open_recover(struct fail_open *);
80 /* Returns true if 'fo' should be in fail-open mode, otherwise false. */
82 should_fail_open(const struct fail_open *fo)
84 return rconn_failure_duration(fo->controller) >= fo->trigger_duration;
87 /* Returns true if 'fo' is currently in fail-open mode, otherwise false. */
89 fail_open_is_active(const struct fail_open *fo)
91 return fo->last_disconn_secs != 0;
95 send_bogus_packet_in(struct fail_open *fo)
97 uint8_t mac[ETH_ADDR_LEN];
101 /* Compose ofp_packet_in. */
102 ofpbuf_init(&b, 128);
103 eth_addr_nicira_random(mac);
104 compose_benign_packet(&b, "Open vSwitch Controller Probe", 0xa033, mac);
105 opi = make_packet_in(pktbuf_get_null(), OFPP_LOCAL, OFPR_NO_MATCH, &b, 64);
109 rconn_send_with_limit(fo->controller, opi, fo->bogus_packet_counter, 1);
112 /* Enter fail-open mode if we should be in it. Handle reconnecting to a
113 * controller from fail-open mode. */
115 fail_open_run(struct fail_open *fo)
117 /* Enter fail-open mode if 'fo' is not in it but should be. */
118 if (should_fail_open(fo)) {
119 int disconn_secs = rconn_failure_duration(fo->controller);
120 if (!fail_open_is_active(fo)) {
121 VLOG_WARN("Could not connect to controller (or switch failed "
122 "controller's post-connection admission control "
123 "policy) for %d seconds, failing open", disconn_secs);
124 fo->last_disconn_secs = disconn_secs;
126 /* Flush all OpenFlow and datapath flows. We will set up our
127 * fail-open rule from fail_open_flushed() when
128 * ofproto_flush_flows() calls back to us. */
129 ofproto_flush_flows(fo->ofproto);
130 } else if (disconn_secs > fo->last_disconn_secs + 60) {
131 VLOG_INFO("Still in fail-open mode after %d seconds disconnected "
132 "from controller", disconn_secs);
133 fo->last_disconn_secs = disconn_secs;
137 /* Schedule a bogus packet-in if we're connected and in fail-open. */
138 if (fail_open_is_active(fo)) {
139 if (rconn_is_connected(fo->controller)) {
140 bool expired = time_msec() >= fo->next_bogus_packet_in;
142 send_bogus_packet_in(fo);
144 if (expired || fo->next_bogus_packet_in == LLONG_MAX) {
145 fo->next_bogus_packet_in = time_msec() + 2000;
148 fo->next_bogus_packet_in = LLONG_MAX;
154 /* If 'fo' is currently in fail-open mode and its rconn has connected to the
155 * controller, exits fail open mode. */
157 fail_open_maybe_recover(struct fail_open *fo)
159 if (rconn_is_admitted(fo->controller)) {
160 fail_open_recover(fo);
165 fail_open_recover(struct fail_open *fo)
167 if (fail_open_is_active(fo)) {
170 VLOG_WARN("No longer in fail-open mode");
171 fo->last_disconn_secs = 0;
172 fo->next_bogus_packet_in = LLONG_MAX;
174 memset(&flow, 0, sizeof flow);
175 flow.wildcards = OFPFW_ALL;
176 flow.priority = FAIL_OPEN_PRIORITY;
177 ofproto_delete_flow(fo->ofproto, &flow);
182 fail_open_wait(struct fail_open *fo)
184 if (fo->next_bogus_packet_in != LLONG_MAX) {
185 poll_timer_wait(fo->next_bogus_packet_in - time_msec());
190 fail_open_flushed(struct fail_open *fo)
192 int disconn_secs = rconn_failure_duration(fo->controller);
193 bool open = disconn_secs >= fo->trigger_duration;
195 union ofp_action action;
198 /* Set up a flow that matches every packet and directs them to
200 memset(&action, 0, sizeof action);
201 action.type = htons(OFPAT_OUTPUT);
202 action.output.len = htons(sizeof action);
203 action.output.port = htons(OFPP_NORMAL);
204 memset(&flow, 0, sizeof flow);
205 flow.wildcards = OFPFW_ALL;
206 flow.priority = FAIL_OPEN_PRIORITY;
207 ofproto_add_flow(fo->ofproto, &flow, &action, 1, 0);
212 fail_open_status_cb(struct status_reply *sr, void *fo_)
214 struct fail_open *fo = fo_;
215 int cur_duration = rconn_failure_duration(fo->controller);
217 status_reply_put(sr, "trigger-duration=%d", fo->trigger_duration);
218 status_reply_put(sr, "current-duration=%d", cur_duration);
219 status_reply_put(sr, "triggered=%s",
220 cur_duration >= fo->trigger_duration ? "true" : "false");
224 fail_open_create(struct ofproto *ofproto,
225 int trigger_duration, struct switch_status *switch_status,
226 struct rconn *controller)
228 struct fail_open *fo = xmalloc(sizeof *fo);
229 fo->ofproto = ofproto;
230 fo->controller = controller;
231 fo->trigger_duration = trigger_duration;
232 fo->last_disconn_secs = 0;
233 fo->ss_cat = switch_status_register(switch_status, "fail-open",
234 fail_open_status_cb, fo);
235 fo->next_bogus_packet_in = LLONG_MAX;
236 fo->bogus_packet_counter = rconn_packet_counter_create();
241 fail_open_set_trigger_duration(struct fail_open *fo, int trigger_duration)
243 fo->trigger_duration = trigger_duration;
247 fail_open_destroy(struct fail_open *fo)
250 fail_open_recover(fo);
251 /* We don't own fo->controller. */
252 switch_status_unregister(fo->ss_cat);
253 rconn_packet_counter_destroy(fo->bogus_packet_counter);