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"
34 static const int min(int a, int b) {
39 Terminate a connection:
41 - Remove the edge representing this connection
43 - Check if we need to retry making an outgoing connection
45 void terminate_connection(meshlink_handle_t *mesh, connection_t *c, bool report) {
46 logger(mesh, MESHLINK_INFO, "Closing connection with %s (%s)", c->name, c->hostname);
48 c->status.active = false;
50 if(c->node && c->node->connection == c)
51 c->node->connection = NULL;
55 send_del_edge(mesh, mesh->everyone, c->edge);
57 edge_del(mesh, c->edge);
60 /* Run MST and SSSP algorithms */
64 /* If the node is not reachable anymore but we remember it had an edge to us, clean it up */
66 if(report && !c->node->status.reachable) {
68 e = lookup_edge(c->node, mesh->self);
70 send_del_edge(mesh, mesh->everyone, e);
76 outgoing_t *outgoing = c->outgoing;
77 connection_del(mesh, c);
79 /* Check if this was our outgoing connection */
82 do_outgoing_connection(mesh, outgoing);
85 /* Clean up dead proxy processes */
87 while(waitpid(-1, NULL, WNOHANG) > 0);
92 Check if the other end is active.
93 If we have sent packets, but didn't receive any,
94 then possibly the other end is dead. We send a
95 PING request over the meta connection. If the other
96 end does not reply in time, we consider them dead
97 and close the connection.
99 static void timeout_handler(event_loop_t *loop, void *data) {
100 meshlink_handle_t *mesh = loop->data;
102 for list_each(connection_t, c, mesh->connections) {
103 if(c->last_ping_time + mesh->pingtimeout <= mesh->loop.now.tv_sec) {
104 if(c->status.active) {
105 if(c->status.pinged) {
106 logger(mesh, MESHLINK_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);
107 } else if(c->last_ping_time + mesh->pinginterval <= mesh->loop.now.tv_sec) {
114 if(c->status.connecting)
115 logger(mesh, MESHLINK_WARNING, "Timeout while connecting to %s (%s)", c->name, c->hostname);
117 logger(mesh, MESHLINK_WARNING, "Timeout from %s (%s) during authentication", c->name, c->hostname);
119 terminate_connection(mesh, c, c->status.active);
123 timeout_set(&mesh->loop, data, &(struct timeval){mesh->pingtimeout, rand() % 100000});
126 /// Utility function to establish connections based on condition check
127 /** The function iterates over all nodes, but skips those that do
128 * not pass the condition check.
130 * The condition check function is passed
131 * a pointer to a random number r between 0 and rand_modulo, a pointer to the
132 * current node index i, and the node pointer n. This function should return true
133 * if a connection attempt to the node should be made.
135 * @param mesh A pointer to the mesh structure
136 * @param rand_modulo Random index is selected between 0 and rand_modulo
137 * @cond_check A function pointer. This function should return true
138 * if a connection attempt to the node should be made
140 static void cond_add_connection(meshlink_handle_t *mesh, int rand_modulo, bool (*cond_check)(int*, int*, node_t*)) {
141 int r = rand() % rand_modulo;
144 for splay_each(node_t, n, mesh->nodes) {
145 /* skip nodes that do not pass condition check */
146 if(!(*cond_check)(&i, &r, n))
149 /* check if there is already a connection attempt to this node */
151 for list_each(outgoing_t, outgoing, mesh->outgoings) {
152 if(!strcmp(outgoing->name, n->name)) {
159 //TODO: if the node is blacklisted the connection will not happen, but
160 //the user will read this debug message "Autoconnecting to %s" that is misleading
161 logger(mesh, MESHLINK_INFO, "Autoconnecting to %s", n->name);
162 outgoing_t *outgoing = xzalloc(sizeof *outgoing);
163 outgoing->mesh = mesh;
164 outgoing->name = xstrdup(n->name);
165 list_insert_tail(mesh->outgoings, outgoing);
166 setup_outgoing_connection(mesh, outgoing);
172 static bool found_random_node(int *i, int *r, node_t *n) {
182 static bool found_random_unreachable_node(int *i, int *r, node_t *n) {
183 if(n->status.reachable)
195 static void periodic_handler(event_loop_t *loop, void *data) {
196 meshlink_handle_t *mesh = loop->data;
198 /* Check if there are too many contradicting ADD_EDGE and DEL_EDGE messages.
199 This usually only happens when another node has the same Name as this node.
200 If so, sleep for a short while to prevent a storm of contradicting messages.
203 if(mesh->contradicting_del_edge > 100 && mesh->contradicting_add_edge > 100) {
204 logger(mesh, MESHLINK_WARNING, "Possible node with same Name as us! Sleeping %d seconds.", mesh->sleeptime);
205 usleep(mesh->sleeptime * 1000000LL);
206 mesh->sleeptime *= 2;
207 if(mesh->sleeptime < 0)
208 mesh->sleeptime = 3600;
210 mesh->sleeptime /= 2;
211 if(mesh->sleeptime < 10)
212 mesh->sleeptime = 10;
215 mesh->contradicting_add_edge = 0;
216 mesh->contradicting_del_edge = 0;
220 /* If AutoConnect is set, check if we need to make or break connections. */
222 if(autoconnect && mesh->nodes->count > 1) {
223 /* Count number of active connections */
225 for list_each(connection_t, c, mesh->connections) {
230 /* Count number of unreachable nodes */
231 int num_unreachable = 0;
232 for splay_each(node_t, n, mesh->nodes) {
233 if(!n->status.reachable)
237 if(nc < autoconnect) {
238 /* Not enough active connections, try to add one.
239 Choose a random node, if we don't have a connection to it,
240 and we are not already trying to make one, create an
241 outgoing connection to this node.
243 cond_add_connection(mesh, mesh->nodes->count, &found_random_node);
244 } else if(num_unreachable > 0) {
245 /* Min number of connections established. Now try
246 to connect to some unreachable nodes to attempt
247 to heal possible partitions.
249 cond_add_connection(mesh, num_unreachable, &found_random_unreachable_node);
252 if(nc > autoconnect) {
253 /* Too many active connections, try to remove one.
254 Choose a random outgoing connection to a node
255 that has at least one other connection.
260 for list_each(connection_t, c, mesh->connections) {
261 if(!c->status.active)
267 if(!c->outgoing || !c->node || c->node->edge_tree->count < 2)
270 logger(mesh, MESHLINK_INFO, "Autodisconnecting from %s", c->name);
271 list_delete(mesh->outgoings, c->outgoing);
273 terminate_connection(mesh, c, c->status.active);
278 if(nc >= autoconnect) {
279 /* If we have enough active connections,
280 remove any pending outgoing connections.
281 Do not remove pending connections to unreachable
284 node_t *o_node = NULL;
285 for list_each(outgoing_t, o, mesh->outgoings) {
286 o_node = lookup_node(mesh, o->name);
287 /* o_node is NULL if it is not part of the graph yet */
288 if(!o_node || !o_node->status.reachable)
292 for list_each(connection_t, c, mesh->connections) {
293 if(c->outgoing == o) {
299 logger(mesh, MESHLINK_INFO, "Cancelled outgoing connection to %s", o->name);
300 /* The node variable is leaked in from using the list_each macro.
301 The o variable could be used, but using node directly
304 list_delete_node(mesh->outgoings, node);
309 if (nc + mesh->outgoings->count < min(autoconnect, mesh->nodes->count - 1))
313 timeout_set(&mesh->loop, data, &(struct timeval){timeout, rand() % 100000});
316 void handle_meta_connection_data(meshlink_handle_t *mesh, connection_t *c) {
317 if (!receive_meta(mesh, c)) {
318 terminate_connection(mesh, c, c->status.active);
323 void retry(meshlink_handle_t *mesh) {
324 /* Reset the reconnection timers for all outgoing connections */
325 for list_each(outgoing_t, outgoing, mesh->outgoings) {
326 outgoing->timeout = 0;
328 timeout_set(&mesh->loop, &outgoing->ev, &(struct timeval){0, 0});
331 /* Check for outgoing connections that are in progress, and reset their ping timers */
332 for list_each(connection_t, c, mesh->connections) {
333 if(c->outgoing && !c->node)
334 c->last_ping_time = 0;
337 /* Kick the ping timeout handler */
338 timeout_set(&mesh->loop, &mesh->pingtimer, &(struct timeval){0, 0});
342 this is where it all happens...
344 int main_loop(meshlink_handle_t *mesh) {
345 timeout_add(&mesh->loop, &mesh->pingtimer, timeout_handler, &mesh->pingtimer, &(struct timeval){mesh->pingtimeout, rand() % 100000});
346 timeout_add(&mesh->loop, &mesh->periodictimer, periodic_handler, &mesh->periodictimer, &(struct timeval){0, 0});
349 mesh->datafromapp.signum = 0;
350 signal_add(&(mesh->loop),&(mesh->datafromapp), (signal_cb_t)meshlink_send_from_queue,mesh, mesh->datafromapp.signum);
352 if(!event_loop_run(&mesh->loop)) {
353 logger(mesh, MESHLINK_ERROR, "Error while waiting for input: %s", strerror(errno));
357 timeout_del(&mesh->loop, &mesh->periodictimer);
358 timeout_del(&mesh->loop, &mesh->pingtimer);