2 net.c -- most of the network code
3 Copyright (C) 2014 Guus Sliepen <guus@meshlink.io>
5 This program is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation; either version 2 of the License, or
8 (at your option) any later version.
10 This program is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 GNU General Public License for more details.
15 You should have received a copy of the GNU General Public License along
16 with this program; if not, write to the Free Software Foundation, Inc.,
17 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 #include "connection.h"
27 #include "meshlink_internal.h"
35 Terminate a connection:
37 - Remove the edge representing this connection
39 - Check if we need to retry making an outgoing connection
41 void terminate_connection(meshlink_handle_t *mesh, connection_t *c, bool report) {
42 logger(DEBUG_CONNECTIONS, LOG_NOTICE, "Closing connection with %s (%s)", c->name, c->hostname);
44 c->status.active = false;
46 if(c->node && c->node->connection == c)
47 c->node->connection = NULL;
51 send_del_edge(mesh, mesh->everyone, c->edge);
53 edge_del(mesh, c->edge);
56 /* Run MST and SSSP algorithms */
60 /* If the node is not reachable anymore but we remember it had an edge to us, clean it up */
62 if(report && !c->node->status.reachable) {
64 e = lookup_edge(c->node, mesh->self);
66 send_del_edge(mesh, mesh->everyone, e);
72 outgoing_t *outgoing = c->outgoing;
73 connection_del(mesh, c);
75 /* Check if this was our outgoing connection */
78 do_outgoing_connection(mesh, outgoing);
81 /* Clean up dead proxy processes */
83 while(waitpid(-1, NULL, WNOHANG) > 0);
88 Check if the other end is active.
89 If we have sent packets, but didn't receive any,
90 then possibly the other end is dead. We send a
91 PING request over the meta connection. If the other
92 end does not reply in time, we consider them dead
93 and close the connection.
95 static void timeout_handler(event_loop_t *loop, void *data) {
96 meshlink_handle_t *mesh = loop->data;
98 for list_each(connection_t, c, mesh->connections) {
99 if(c->last_ping_time + mesh->pingtimeout <= mesh->loop.now.tv_sec) {
100 if(c->status.active) {
101 if(c->status.pinged) {
102 logger(DEBUG_CONNECTIONS, LOG_INFO, "%s (%s) didn't respond to PING in %ld seconds", c->name, c->hostname, (long)mesh->loop.now.tv_sec - c->last_ping_time);
103 } else if(c->last_ping_time + mesh->pinginterval <= mesh->loop.now.tv_sec) {
110 if(c->status.connecting)
111 logger(DEBUG_CONNECTIONS, LOG_WARNING, "Timeout while connecting to %s (%s)", c->name, c->hostname);
113 logger(DEBUG_CONNECTIONS, LOG_WARNING, "Timeout from %s (%s) during authentication", c->name, c->hostname);
115 terminate_connection(mesh, c, c->status.active);
119 timeout_set(&mesh->loop, data, &(struct timeval){mesh->pingtimeout, rand() % 100000});
122 /// Utility function to establish connections based on condition check
123 /** The function iterates over all nodes, but skips those that do
124 * not pass the condition check.
126 * The condition check function is passed
127 * a pointer to a random number r between 0 and rand_modulo, a pointer to the
128 * current node index i, and the node pointer n. This function should return true
129 * if a connection attempt to the node should be made.
131 * @param mesh A pointer to the mesh structure
132 * @param rand_modulo Random index is selected between 0 and rand_modulo
133 * @cond_check A function pointer. This function should return true
134 * if a connection attempt to the node should be made
136 static void cond_add_connection(meshlink_handle_t *mesh, int rand_modulo, bool (*cond_check)(int*, int*, node_t*)) {
137 int r = rand() % rand_modulo;
140 for splay_each(node_t, n, mesh->nodes) {
141 /* skip nodes that do not pass condition check */
142 if(!(*cond_check)(&i, &r, n))
145 /* check if there is already a connection attempt to this node */
147 for list_each(outgoing_t, outgoing, mesh->outgoings) {
148 if(!strcmp(outgoing->name, n->name)) {
155 //TODO: if the node is blacklisted the connection will not happen, but
156 //the user will read this debug message "Autoconnecting to %s" that is misleading
157 logger(DEBUG_CONNECTIONS, LOG_INFO, "Autoconnecting to %s", n->name);
158 outgoing_t *outgoing = xzalloc(sizeof *outgoing);
159 outgoing->name = xstrdup(n->name);
160 list_insert_tail(mesh->outgoings, outgoing);
161 setup_outgoing_connection(mesh, outgoing);
167 static bool found_random_node(int *i, int *r, node_t *n) {
177 static bool found_random_unreachable_node(int *i, int *r, node_t *n) {
178 if(n->status.reachable)
190 static void periodic_handler(event_loop_t *loop, void *data) {
191 meshlink_handle_t *mesh = loop->data;
193 /* Check if there are too many contradicting ADD_EDGE and DEL_EDGE messages.
194 This usually only happens when another node has the same Name as this node.
195 If so, sleep for a short while to prevent a storm of contradicting messages.
198 if(mesh->contradicting_del_edge > 100 && mesh->contradicting_add_edge > 100) {
199 logger(DEBUG_ALWAYS, LOG_WARNING, "Possible node with same Name as us! Sleeping %d seconds.", mesh->sleeptime);
200 usleep(mesh->sleeptime * 1000000LL);
201 mesh->sleeptime *= 2;
202 if(mesh->sleeptime < 0)
203 mesh->sleeptime = 3600;
205 mesh->sleeptime /= 2;
206 if(mesh->sleeptime < 10)
207 mesh->sleeptime = 10;
210 mesh->contradicting_add_edge = 0;
211 mesh->contradicting_del_edge = 0;
213 /* If AutoConnect is set, check if we need to make or break connections. */
215 if(autoconnect && mesh->nodes->count > 1) {
216 /* Count number of active connections */
218 for list_each(connection_t, c, mesh->connections) {
223 /* Count number of unreachable nodes */
224 int num_unreachable = 0;
225 for splay_each(node_t, n, mesh->nodes) {
226 if(!n->status.reachable)
230 if(nc < autoconnect) {
231 /* Not enough active connections, try to add one.
232 Choose a random node, if we don't have a connection to it,
233 and we are not already trying to make one, create an
234 outgoing connection to this node.
236 cond_add_connection(mesh, mesh->nodes->count, &found_random_node);
237 } else if(num_unreachable > 0) {
238 /* Min number of connections established. Now try
239 to connect to some unreachable nodes to attempt
240 to heal possible partitions.
242 cond_add_connection(mesh, num_unreachable, &found_random_unreachable_node);
245 if(nc > autoconnect) {
246 /* Too many active connections, try to remove one.
247 Choose a random outgoing connection to a node
248 that has at least one other connection.
253 for list_each(connection_t, c, mesh->connections) {
254 if(!c->status.active)
260 if(!c->outgoing || !c->node || c->node->edge_tree->count < 2)
263 logger(DEBUG_CONNECTIONS, LOG_INFO, "Autodisconnecting from %s", c->name);
264 list_delete(mesh->outgoings, c->outgoing);
266 terminate_connection(mesh, c, c->status.active);
271 if(nc >= autoconnect) {
272 /* If we have enough active connections,
273 remove any pending outgoing connections.
274 Do not remove pending connections to unreachable
277 node_t *o_node = NULL;
278 for list_each(outgoing_t, o, mesh->outgoings) {
279 o_node = lookup_node(mesh, o->name);
280 /* o_node is NULL if it is not part of the graph yet */
281 if(!o_node || !o_node->status.reachable)
285 for list_each(connection_t, c, mesh->connections) {
286 if(c->outgoing == o) {
292 logger(DEBUG_CONNECTIONS, LOG_INFO, "Cancelled outgoing connection to %s", o->name);
293 /* The node variable is leaked in from using the list_each macro.
294 The o variable could be used, but using node directly
297 list_delete_node(mesh->outgoings, node);
303 timeout_set(&mesh->loop, data, &(struct timeval){5, rand() % 100000});
306 void handle_meta_connection_data(meshlink_handle_t *mesh, connection_t *c) {
307 if (!receive_meta(mesh, c)) {
308 terminate_connection(mesh, c, c->status.active);
313 void retry(meshlink_handle_t *mesh) {
314 /* Reset the reconnection timers for all outgoing connections */
315 for list_each(outgoing_t, outgoing, mesh->outgoings) {
316 outgoing->timeout = 0;
318 timeout_set(&mesh->loop, &outgoing->ev, &(struct timeval){0, 0});
321 /* Check for outgoing connections that are in progress, and reset their ping timers */
322 for list_each(connection_t, c, mesh->connections) {
323 if(c->outgoing && !c->node)
324 c->last_ping_time = 0;
327 /* Kick the ping timeout handler */
328 timeout_set(&mesh->loop, &mesh->pingtimer, &(struct timeval){0, 0});
332 this is where it all happens...
334 int main_loop(meshlink_handle_t *mesh) {
335 timeout_add(&mesh->loop, &mesh->pingtimer, timeout_handler, &mesh->pingtimer, &(struct timeval){mesh->pingtimeout, rand() % 100000});
336 timeout_add(&mesh->loop, &mesh->periodictimer, periodic_handler, &mesh->periodictimer, &(struct timeval){mesh->pingtimeout, rand() % 100000});
339 mesh->datafromapp.signum = 0;
340 signal_add(&(mesh->loop),&(mesh->datafromapp), (signal_cb_t)meshlink_send_from_queue,mesh, mesh->datafromapp.signum);
342 if(!event_loop_run(&mesh->loop)) {
343 logger(DEBUG_ALWAYS, LOG_ERR, "Error while waiting for input: %s", strerror(errno));
347 timeout_del(&mesh->loop, &mesh->periodictimer);
348 timeout_del(&mesh->loop, &mesh->pingtimer);