2 * Copyright (c) 2008, 2009, 2010, 2011 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"
21 #include "classifier.h"
24 #include "mac-learning.h"
30 #include "poll-loop.h"
36 VLOG_DEFINE_THIS_MODULE(fail_open);
41 * In fail-open mode, the switch detects when the controller cannot be
42 * contacted or when the controller is dropping switch connections because the
43 * switch does not pass its admission control policy. In those situations the
44 * switch sets up flows itself using the "normal" action.
46 * There is a little subtlety to implementation, to properly handle the case
47 * where the controller allows switch connections but drops them a few seconds
48 * later for admission control reasons. Because of this case, we don't want to
49 * just stop setting up flows when we connect to the controller: if we did,
50 * then new flow setup and existing flows would stop during the duration of
51 * connection to the controller, and thus the whole network would go down for
52 * that period of time.
54 * So, instead, we add some special cases when we are connected to a
55 * controller, but not yet sure that it has admitted us:
57 * - We set up flows immediately ourselves, but simultaneously send out an
58 * OFPT_PACKET_IN to the controller. We put a special bogus buffer-id in
59 * these OFPT_PACKET_IN messages so that duplicate packets don't get sent
60 * out to the network when the controller replies.
62 * - We also send out OFPT_PACKET_IN messages for totally bogus packets
63 * every so often, in case no real new flows are arriving in the network.
65 * - We don't flush the flow table at the time we connect, because this
66 * could cause network stuttering in a switch with lots of flows or very
67 * high-bandwidth flows by suddenly throwing lots of packets down to
72 struct ofproto *ofproto;
73 struct connmgr *connmgr;
74 int last_disconn_secs;
75 long long int next_bogus_packet_in;
76 struct rconn_packet_counter *bogus_packet_counter;
79 static void fail_open_recover(struct fail_open *);
81 /* Returns the number of seconds of disconnection after which fail-open mode
84 trigger_duration(const struct fail_open *fo)
86 if (!connmgr_has_controllers(fo->connmgr)) {
87 /* Shouldn't ever arrive here, but if we do, never fail open. */
90 /* Otherwise, every controller must have a chance to send an
91 * inactivity probe and reconnect before we fail open, so take the
92 * maximum probe interval and multiply by 3:
94 * - The first interval is the idle time before sending an inactivity
97 * - The second interval is the time allowed for a response to the
100 * - The third interval is the time allowed to reconnect after no
101 * response is received.
103 return connmgr_get_max_probe_interval(fo->connmgr) * 3;
107 /* Returns true if 'fo' is currently in fail-open mode, otherwise false. */
109 fail_open_is_active(const struct fail_open *fo)
111 return fo->last_disconn_secs != 0;
115 send_bogus_packet_ins(struct fail_open *fo)
117 uint8_t mac[ETH_ADDR_LEN];
121 /* Compose ofp_packet_in. */
122 ofpbuf_init(&b, 128);
123 eth_addr_nicira_random(mac);
124 compose_benign_packet(&b, "Open vSwitch Controller Probe", 0xa033, mac);
125 opi = make_packet_in(pktbuf_get_null(), OFPP_LOCAL, OFPR_NO_MATCH, &b, 64);
129 connmgr_broadcast(fo->connmgr, opi);
132 /* Enter fail-open mode if we should be in it. */
134 fail_open_run(struct fail_open *fo)
136 int disconn_secs = connmgr_failure_duration(fo->connmgr);
138 /* Enter fail-open mode if 'fo' is not in it but should be. */
139 if (disconn_secs >= trigger_duration(fo)) {
140 if (!fail_open_is_active(fo)) {
141 VLOG_WARN("Could not connect to controller (or switch failed "
142 "controller's post-connection admission control "
143 "policy) for %d seconds, failing open", disconn_secs);
144 fo->last_disconn_secs = disconn_secs;
146 /* Flush all OpenFlow and datapath flows. We will set up our
147 * fail-open rule from fail_open_flushed() when
148 * ofproto_flush_flows() calls back to us. */
149 ofproto_flush_flows(fo->ofproto);
150 } else if (disconn_secs > fo->last_disconn_secs + 60) {
151 VLOG_INFO("Still in fail-open mode after %d seconds disconnected "
152 "from controller", disconn_secs);
153 fo->last_disconn_secs = disconn_secs;
157 /* Schedule a bogus packet-in if we're connected and in fail-open. */
158 if (fail_open_is_active(fo)) {
159 if (connmgr_is_any_controller_connected(fo->connmgr)) {
160 bool expired = time_msec() >= fo->next_bogus_packet_in;
162 send_bogus_packet_ins(fo);
164 if (expired || fo->next_bogus_packet_in == LLONG_MAX) {
165 fo->next_bogus_packet_in = time_msec() + 2000;
168 fo->next_bogus_packet_in = LLONG_MAX;
174 /* If 'fo' is currently in fail-open mode and its rconn has connected to the
175 * controller, exits fail open mode. */
177 fail_open_maybe_recover(struct fail_open *fo)
179 if (fail_open_is_active(fo)
180 && connmgr_is_any_controller_admitted(fo->connmgr)) {
181 fail_open_recover(fo);
186 fail_open_recover(struct fail_open *fo)
188 struct cls_rule rule;
190 VLOG_WARN("No longer in fail-open mode");
191 fo->last_disconn_secs = 0;
192 fo->next_bogus_packet_in = LLONG_MAX;
194 cls_rule_init_catchall(&rule, FAIL_OPEN_PRIORITY);
195 ofproto_delete_flow(fo->ofproto, &rule);
199 fail_open_wait(struct fail_open *fo)
201 if (fo->next_bogus_packet_in != LLONG_MAX) {
202 poll_timer_wait_until(fo->next_bogus_packet_in);
207 fail_open_flushed(struct fail_open *fo)
209 int disconn_secs = connmgr_failure_duration(fo->connmgr);
210 bool open = disconn_secs >= trigger_duration(fo);
212 union ofp_action action;
213 struct cls_rule rule;
215 /* Set up a flow that matches every packet and directs them to
217 memset(&action, 0, sizeof action);
218 action.type = htons(OFPAT_OUTPUT);
219 action.output.len = htons(sizeof action);
220 action.output.port = htons(OFPP_NORMAL);
222 cls_rule_init_catchall(&rule, FAIL_OPEN_PRIORITY);
223 ofproto_add_flow(fo->ofproto, &rule, &action, 1);
227 /* Creates and returns a new struct fail_open for 'ofproto' and 'mgr'. */
229 fail_open_create(struct ofproto *ofproto, struct connmgr *mgr)
231 struct fail_open *fo = xmalloc(sizeof *fo);
232 fo->ofproto = ofproto;
234 fo->last_disconn_secs = 0;
235 fo->next_bogus_packet_in = LLONG_MAX;
236 fo->bogus_packet_counter = rconn_packet_counter_create();
242 fail_open_destroy(struct fail_open *fo)
245 if (fail_open_is_active(fo)) {
246 fail_open_recover(fo);
248 /* We don't own fo->connmgr. */
249 rconn_packet_counter_destroy(fo->bogus_packet_counter);