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"
36 static const int min(int a, int b) {
41 Terminate a connection:
43 - Remove the edge representing this connection
45 - Check if we need to retry making an outgoing connection
47 void terminate_connection(meshlink_handle_t *mesh, connection_t *c, bool report) {
48 logger(mesh, MESHLINK_INFO, "Closing connection with %s (%s)", c->name, c->hostname);
50 c->status.active = false;
52 if(c->node && c->node->connection == c)
53 c->node->connection = NULL;
57 send_del_edge(mesh, mesh->everyone, c->edge);
59 edge_del(mesh, c->edge);
62 /* Run MST and SSSP algorithms */
66 /* If the node is not reachable anymore but we remember it had an edge to us, clean it up */
68 if(report && !c->node->status.reachable) {
70 e = lookup_edge(c->node, mesh->self);
72 send_del_edge(mesh, mesh->everyone, e);
78 outgoing_t *outgoing = c->outgoing;
79 connection_del(mesh, c);
81 /* Check if this was our outgoing connection */
84 do_outgoing_connection(mesh, outgoing);
87 /* Clean up dead proxy processes */
89 while(waitpid(-1, NULL, WNOHANG) > 0);
94 Check if the other end is active.
95 If we have sent packets, but didn't receive any,
96 then possibly the other end is dead. We send a
97 PING request over the meta connection. If the other
98 end does not reply in time, we consider them dead
99 and close the connection.
101 static void timeout_handler(event_loop_t *loop, void *data) {
102 meshlink_handle_t *mesh = loop->data;
103 logger(mesh, MESHLINK_DEBUG, "timeout_handler()");
105 for list_each(connection_t, c, mesh->connections) {
106 // Also make sure that if outstanding key requests for the UDP counterpart of a connection has timed out, we restart it.
108 if(c->node->status.waitingforkey && c->node->last_req_key + mesh->pingtimeout <= mesh->loop.now.tv_sec)
109 send_req_key(mesh, c->node);
111 if(c->last_ping_time + mesh->pingtimeout <= mesh->loop.now.tv_sec) {
112 if(c->status.active) {
113 if(c->status.pinged) {
114 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);
115 } else if(c->last_ping_time + mesh->pinginterval <= mesh->loop.now.tv_sec) {
122 if(c->status.connecting)
123 logger(mesh, MESHLINK_WARNING, "Timeout while connecting to %s (%s)", c->name, c->hostname);
125 logger(mesh, MESHLINK_WARNING, "Timeout from %s (%s) during authentication", c->name, c->hostname);
127 terminate_connection(mesh, c, c->status.active);
131 timeout_set(&mesh->loop, data, &(struct timeval){mesh->pingtimeout, rand() % 100000});
134 // devclass asc, last_connect_try desc
135 static int node_compare_devclass_asc_last_connect_try_desc(const void *a, const void *b)
137 const node_t *na = a, *nb = b;
139 if(na->devclass < nb->devclass)
142 if(na->devclass > nb->devclass)
145 if(na->last_connect_try == nb->last_connect_try)
148 if(nb->last_connect_try == 0 || na->last_connect_try < nb->last_connect_try)
151 if(na->last_connect_try == 0 || na->last_connect_try > nb->last_connect_try)
157 // last_connect_try desc
158 static int node_compare_last_connect_try_desc(const void *a, const void *b)
160 const node_t *na = a, *nb = b;
162 if(na->last_connect_try == nb->last_connect_try)
165 if(nb->last_connect_try == 0 || na->last_connect_try < nb->last_connect_try)
168 if(na->last_connect_try == 0 || na->last_connect_try > nb->last_connect_try)
175 static int node_compare_devclass_desc(const void *a, const void *b)
177 const node_t *na = a, *nb = b;
179 if(na->devclass < nb->devclass)
182 if(na->devclass > nb->devclass)
196 // find the best one for initial connect
201 where dclass <= my.dclass and !connection and (timestamp - last_retry) > retry_timeout
202 order by dclass asc, last_connection desc
208 // find better nodes to connect to: in case we have less than min connections within [BACKBONE, i] and there are nodes which we are not connected to within the range
212 for i = BACKBONE to my.dclass
213 j += count(from connections where node.dclass = i)
217 where dclass = i and !connection and (timestamp - last_retry) > retry_timeout
218 order by last_connection desc
230 where dclass <= my.dclass and !reachable and (timestamp - last_retry) > retry_timeout
231 order by dclass asc, last_connection desc
243 // disconnect outgoing connections in case we have more than min connections within [BACKBONE, i] and there are nodes which we are connected to within the range [i, PORTABLE]
247 for i = BACKBONE to my.dclass
248 j += count(from connections where node.dclass = i)
252 where dclass >= i and outgoing_connection
260 // disconnect connections in case we have more than enough connections
265 where outgoing_connection
277 next (timeout, autoconnect)
285 static void periodic_handler(event_loop_t *loop, void *data) {
286 meshlink_handle_t *mesh = loop->data;
288 /* Check if there are too many contradicting ADD_EDGE and DEL_EDGE messages.
289 This usually only happens when another node has the same Name as this node.
290 If so, sleep for a short while to prevent a storm of contradicting messages.
293 if(mesh->contradicting_del_edge > 100 && mesh->contradicting_add_edge > 100) {
294 logger(mesh, MESHLINK_WARNING, "Possible node with same Name as us! Sleeping %d seconds.", mesh->sleeptime);
295 usleep(mesh->sleeptime * 1000000LL);
296 mesh->sleeptime *= 2;
297 if(mesh->sleeptime < 0)
298 mesh->sleeptime = 3600;
300 mesh->sleeptime /= 2;
301 if(mesh->sleeptime < 10)
302 mesh->sleeptime = 10;
305 mesh->contradicting_add_edge = 0;
306 mesh->contradicting_del_edge = 0;
310 /* Check if we need to make or break connections. */
312 if(mesh->nodes->count > 1) {
314 logger(mesh, MESHLINK_INFO, "--- autoconnect begin ---");
317 int retry_timeout = min(mesh->nodes->count * 5, 60);
319 // connect disconnect nodes
321 node_t* connect_to = NULL;
322 node_t* disconnect_from = NULL;
327 int cur_connects = 0;
329 for list_each(connection_t, c, mesh->connections)
331 if(!c->status.remove_unused)
337 logger(mesh, MESHLINK_INFO, "* cur_connects = %d", cur_connects);
340 // get min_connects and max_connects
342 assert(mesh->devclass >= 0 && mesh->devclass <= _DEV_CLASS_MAX);
344 int min_connects = dev_class_traits[mesh->devclass].min_connects;
345 int max_connects = dev_class_traits[mesh->devclass].max_connects;
347 logger(mesh, MESHLINK_INFO, "* min_connects = %d", min_connects);
348 logger(mesh, MESHLINK_INFO, "* max_connects = %d", max_connects);
351 // find the best one for initial connect
353 if(cur_connects < min_connects)
355 splay_tree_t *nodes = splay_alloc_tree(node_compare_devclass_asc_last_connect_try_desc, NULL);
357 for splay_each(node_t, n, mesh->nodes)
359 if(n->devclass <= mesh->devclass && !n->connection && (n->last_connect_try == 0 || (time(NULL) - n->last_connect_try) > retry_timeout))
360 { splay_insert(nodes, n); }
365 logger(mesh, MESHLINK_INFO, "* found best one for initial connect");
368 connect_to = (node_t*)nodes->head->data;
371 splay_free_tree(nodes);
375 // find better nodes to connect to
377 if(!connect_to && min_connects <= cur_connects < max_connects)
379 unsigned int connects = 0;
381 for(int devclass = 0; devclass <= mesh->devclass; ++devclass)
383 for list_each(connection_t, c, mesh->connections)
385 if(!c->status.remove_unused && c->node && c->node->devclass == devclass)
389 if( connects < min_connects )
391 splay_tree_t *nodes = splay_alloc_tree(node_compare_last_connect_try_desc, NULL);
393 for splay_each(node_t, n, mesh->nodes)
395 if(n->devclass == devclass && !n->connection && (n->last_connect_try == 0 || (time(NULL) - n->last_connect_try) > retry_timeout))
396 { splay_insert(nodes, n); }
401 logger(mesh, MESHLINK_INFO, "* found better node");
402 connect_to = (node_t*)nodes->head->data;
404 splay_free_tree(nodes);
408 splay_free_tree(nodes);
418 if(!connect_to && min_connects <= cur_connects < max_connects)
420 splay_tree_t *nodes = splay_alloc_tree(node_compare_devclass_asc_last_connect_try_desc, NULL);
422 for splay_each(node_t, n, mesh->nodes)
424 if(n->devclass <= mesh->devclass && !n->status.reachable && (n->last_connect_try == 0 || (time(NULL) - n->last_connect_try) > retry_timeout))
425 { splay_insert(nodes, n); }
430 logger(mesh, MESHLINK_INFO, "* try to heal partition");
431 connect_to = (node_t*)nodes->head->data;
434 splay_free_tree(nodes);
440 if(connect_to && !connect_to->connection)
442 /* check if there is already a connection attempt to this node */
444 for list_each(outgoing_t, outgoing, mesh->outgoings) {
445 if(!strcmp(outgoing->name, connect_to->name)) {
453 logger(mesh, MESHLINK_INFO, "Autoconnecting to %s", connect_to->name);
454 outgoing_t *outgoing = xzalloc(sizeof(outgoing_t));
455 outgoing->mesh = mesh;
456 outgoing->name = xstrdup(connect_to->name);
457 list_insert_tail(mesh->outgoings, outgoing);
458 setup_outgoing_connection(mesh, outgoing);
463 // disconnect suboptimal outgoing connections
465 if(min_connects < cur_connects <= max_connects)
467 unsigned int connects = 0;
469 for(int devclass = 0; devclass <= mesh->devclass; ++devclass)
471 for list_each(connection_t, c, mesh->connections)
473 if(!c->status.remove_unused && c->node && c->node->devclass == devclass)
477 if( min_connects < connects )
479 splay_tree_t *nodes = splay_alloc_tree(node_compare_devclass_desc, NULL);
481 for list_each(connection_t, c, mesh->connections)
483 if(!c->status.remove_unused && c->outgoing && c->node && c->node->devclass >= devclass)
484 { splay_insert(nodes, c->node); }
489 logger(mesh, MESHLINK_INFO, "* disconnect suboptimal outgoing connection");
490 disconnect_from = (node_t*)nodes->head->data;
493 splay_free_tree(nodes);
500 // disconnect connections (too many connections)
502 if(!disconnect_from && max_connects < cur_connects)
504 splay_tree_t *nodes = splay_alloc_tree(node_compare_devclass_desc, NULL);
506 for list_each(connection_t, c, mesh->connections)
508 if(!c->status.remove_unused && c->node)
509 { splay_insert(nodes, c->node); }
514 logger(mesh, MESHLINK_INFO, "* disconnect connection (too many connections");
517 disconnect_from = (node_t*)nodes->head->data;
520 splay_free_tree(nodes);
524 // perform disconnect
526 if(disconnect_from && disconnect_from->connection)
528 logger(mesh, MESHLINK_INFO, "Autodisconnecting from %s", disconnect_from->connection->name);
529 list_delete(mesh->outgoings, disconnect_from->connection->outgoing);
530 disconnect_from->connection->outgoing = NULL;
531 terminate_connection(mesh, disconnect_from->connection, disconnect_from->connection->status.active);
537 logger(mesh, MESHLINK_INFO, "--- autoconnect end ---");
540 timeout_set(&mesh->loop, data, &(struct timeval){timeout, rand() % 100000});
543 void handle_meta_connection_data(meshlink_handle_t *mesh, connection_t *c) {
544 if (!receive_meta(mesh, c)) {
545 terminate_connection(mesh, c, c->status.active);
550 void retry(meshlink_handle_t *mesh) {
551 /* Reset the reconnection timers for all outgoing connections */
552 for list_each(outgoing_t, outgoing, mesh->outgoings) {
553 outgoing->timeout = 0;
555 timeout_set(&mesh->loop, &outgoing->ev, &(struct timeval){0, 0});
558 /* Check for outgoing connections that are in progress, and reset their ping timers */
559 for list_each(connection_t, c, mesh->connections) {
560 if(c->outgoing && !c->node)
561 c->last_ping_time = 0;
564 /* Kick the ping timeout handler */
565 timeout_set(&mesh->loop, &mesh->pingtimer, &(struct timeval){0, 0});
569 this is where it all happens...
571 int main_loop(meshlink_handle_t *mesh) {
572 timeout_add(&mesh->loop, &mesh->pingtimer, timeout_handler, &mesh->pingtimer, &(struct timeval){mesh->pingtimeout, rand() % 100000});
573 timeout_add(&mesh->loop, &mesh->periodictimer, periodic_handler, &mesh->periodictimer, &(struct timeval){0, 0});
576 mesh->datafromapp.signum = 0;
577 signal_add(&(mesh->loop),&(mesh->datafromapp), (signal_cb_t)meshlink_send_from_queue,mesh, mesh->datafromapp.signum);
579 if(!event_loop_run(&mesh->loop)) {
580 logger(mesh, MESHLINK_ERROR, "Error while waiting for input: %s", strerror(errno));
584 timeout_del(&mesh->loop, &mesh->periodictimer);
585 timeout_del(&mesh->loop, &mesh->pingtimer);