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"
27 #include "poll-loop.h"
33 #define THIS_MODULE VLM_fail_open
39 * In fail-open mode, the switch detects when the controller cannot be
40 * contacted or when the controller is dropping switch connections because the
41 * switch does not pass its admission control policy. In those situations the
42 * switch sets up flows itself using the "normal" action.
44 * There is a little subtlety to implementation, to properly handle the case
45 * where the controller allows switch connections but drops them a few seconds
46 * later for admission control reasons. Because of this case, we don't want to
47 * just stop setting up flows when we connect to the controller: if we did,
48 * then new flow setup and existing flows would stop during the duration of
49 * connection to the controller, and thus the whole network would go down for
50 * that period of time.
52 * So, instead, we add some special caseswhen we are connected to a controller,
53 * but not yet sure that it has admitted us:
55 * - We set up flows immediately ourselves, but simultaneously send out an
56 * OFPT_PACKET_IN to the controller. We put a special bogus buffer-id in
57 * these OFPT_PACKET_IN messages so that duplicate packets don't get sent
58 * out to the network when the controller replies.
60 * - We also send out OFPT_PACKET_IN messages for totally bogus packets
61 * every so often, in case no real new flows are arriving in the network.
63 * - We don't flush the flow table at the time we connect, because this
64 * could cause network stuttering in a switch with lots of flows or very
65 * high-bandwidth flows by suddenly throwing lots of packets down to
70 struct ofproto *ofproto;
71 struct rconn *controller;
73 int last_disconn_secs;
74 struct status_category *ss_cat;
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 true if 'fo' should be in fail-open mode, otherwise false. */
83 should_fail_open(const struct fail_open *fo)
85 return rconn_failure_duration(fo->controller) >= fo->trigger_duration;
88 /* Returns true if 'fo' is currently in fail-open mode, otherwise false. */
90 fail_open_is_active(const struct fail_open *fo)
92 return fo->last_disconn_secs != 0;
96 send_bogus_packet_in(struct fail_open *fo)
98 uint8_t mac[ETH_ADDR_LEN];
102 /* Compose ofp_packet_in. */
103 ofpbuf_init(&b, 128);
104 eth_addr_nicira_random(mac);
105 compose_benign_packet(&b, "Open vSwitch Controller Probe", 0xa033, mac);
106 opi = make_packet_in(pktbuf_get_null(), OFPP_LOCAL, OFPR_NO_MATCH, &b, 64);
110 rconn_send_with_limit(fo->controller, opi, fo->bogus_packet_counter, 1);
113 /* Enter fail-open mode if we should be in it. Handle reconnecting to a
114 * controller from fail-open mode. */
116 fail_open_run(struct fail_open *fo)
118 /* Enter fail-open mode if 'fo' is not in it but should be. */
119 if (should_fail_open(fo)) {
120 int disconn_secs = rconn_failure_duration(fo->controller);
121 if (!fail_open_is_active(fo)) {
122 VLOG_WARN("Could not connect to controller (or switch failed "
123 "controller's post-connection admission control "
124 "policy) for %d seconds, failing open", disconn_secs);
125 fo->last_disconn_secs = disconn_secs;
127 /* Flush all OpenFlow and datapath flows. We will set up our
128 * fail-open rule from fail_open_flushed() when
129 * ofproto_flush_flows() calls back to us. */
130 ofproto_flush_flows(fo->ofproto);
131 } else if (disconn_secs > fo->last_disconn_secs + 60) {
132 VLOG_INFO("Still in fail-open mode after %d seconds disconnected "
133 "from controller", disconn_secs);
134 fo->last_disconn_secs = disconn_secs;
138 /* Schedule a bogus packet-in if we're connected and in fail-open. */
139 if (fail_open_is_active(fo)) {
140 if (rconn_is_connected(fo->controller)) {
141 bool expired = time_msec() >= fo->next_bogus_packet_in;
143 send_bogus_packet_in(fo);
145 if (expired || fo->next_bogus_packet_in == LLONG_MAX) {
146 fo->next_bogus_packet_in = time_msec() + 2000;
149 fo->next_bogus_packet_in = LLONG_MAX;
155 /* If 'fo' is currently in fail-open mode and its rconn has connected to the
156 * controller, exits fail open mode. */
158 fail_open_maybe_recover(struct fail_open *fo)
160 if (rconn_is_admitted(fo->controller)) {
161 fail_open_recover(fo);
166 fail_open_recover(struct fail_open *fo)
168 if (fail_open_is_active(fo)) {
171 VLOG_WARN("No longer in fail-open mode");
172 fo->last_disconn_secs = 0;
173 fo->next_bogus_packet_in = LLONG_MAX;
175 memset(&flow, 0, sizeof flow);
176 ofproto_delete_flow(fo->ofproto, &flow, OFPFW_ALL, FAIL_OPEN_PRIORITY);
181 fail_open_wait(struct fail_open *fo)
183 if (fo->next_bogus_packet_in != LLONG_MAX) {
184 poll_timer_wait(fo->next_bogus_packet_in - time_msec());
189 fail_open_flushed(struct fail_open *fo)
191 int disconn_secs = rconn_failure_duration(fo->controller);
192 bool open = disconn_secs >= fo->trigger_duration;
194 union ofp_action action;
197 /* Set up a flow that matches every packet and directs them to
199 memset(&action, 0, sizeof action);
200 action.type = htons(OFPAT_OUTPUT);
201 action.output.len = htons(sizeof action);
202 action.output.port = htons(OFPP_NORMAL);
203 memset(&flow, 0, sizeof flow);
204 ofproto_add_flow(fo->ofproto, &flow, OFPFW_ALL, FAIL_OPEN_PRIORITY,
210 fail_open_status_cb(struct status_reply *sr, void *fo_)
212 struct fail_open *fo = fo_;
213 int cur_duration = rconn_failure_duration(fo->controller);
215 status_reply_put(sr, "trigger-duration=%d", fo->trigger_duration);
216 status_reply_put(sr, "current-duration=%d", cur_duration);
217 status_reply_put(sr, "triggered=%s",
218 cur_duration >= fo->trigger_duration ? "true" : "false");
222 fail_open_create(struct ofproto *ofproto,
223 int trigger_duration, struct switch_status *switch_status,
224 struct rconn *controller)
226 struct fail_open *fo = xmalloc(sizeof *fo);
227 fo->ofproto = ofproto;
228 fo->controller = controller;
229 fo->trigger_duration = trigger_duration;
230 fo->last_disconn_secs = 0;
231 fo->ss_cat = switch_status_register(switch_status, "fail-open",
232 fail_open_status_cb, fo);
233 fo->next_bogus_packet_in = LLONG_MAX;
234 fo->bogus_packet_counter = rconn_packet_counter_create();
239 fail_open_set_trigger_duration(struct fail_open *fo, int trigger_duration)
241 fo->trigger_duration = trigger_duration;
245 fail_open_destroy(struct fail_open *fo)
248 fail_open_recover(fo);
249 /* We don't own fo->controller. */
250 switch_status_unregister(fo->ss_cat);
251 rconn_packet_counter_destroy(fo->bogus_packet_counter);