// Called from the pingInteral timeout.
static void pingCallback(void* vic)
{
struct Context* ic = Identity_cast((struct Context*) vic);
uint64_t now = Time_currentTimeMilliseconds(ic->eventBase);
ic->pingCount++;
// scan for endpoints have not sent anything recently.
for (uint32_t i = 0; i < ic->peerMap.count; i++) {
struct IFCPeer* ep = ic->peerMap.values[i];
if (now > ep->timeOfLastMessage + ic->pingAfterMilliseconds) {
#ifdef Log_DEBUG
uint8_t key[56];
Base32_encode(key, 56, CryptoAuth_getHerPublicKey(ep->cryptoAuthIf), 32);
#endif
if (ep->transient && now > ep->timeOfLastMessage + ic->forgetAfterMilliseconds) {
Log_debug(ic->logger, "Unresponsive peer [%s.k] has not responded in [%u] "
"seconds, dropping connection",
key, ic->forgetAfterMilliseconds / 1024);
Allocator_free(ep->external->allocator);
} else if (now > ep->timeOfLastMessage + ic->unresponsiveAfterMilliseconds) {
// Lets skip 87% of pings when they're really down.
if (ic->pingCount % 8) {
continue;
}
ep->state = InterfaceController_PeerState_UNRESPONSIVE;
uint32_t lag = ((now - ep->timeOfLastMessage) / 1024);
Log_debug(ic->logger, "Pinging unresponsive peer [%s.k] lag [%u]", key, lag);
} else {
uint32_t lag = ((now - ep->timeOfLastMessage) / 1024);
Log_debug(ic->logger, "Pinging lazy peer [%s] lag [%u]", key, lag);
}
struct SwitchPinger_Ping* ping =
SwitchPinger_newPing(ep->switchLabel,
String_CONST(""),
ic->timeoutMilliseconds,
onPingResponse,
ic->switchPinger);
ping->onResponseContext = ep;
SwitchPinger_sendPing(ping);
}
}
}
// Called from the pingInteral timeout.
static void pingCallback(void* vic)
{
struct Context* ic = Identity_cast((struct Context*) vic);
uint64_t now = Time_currentTimeMilliseconds(ic->eventBase);
ic->pingCount++;
// scan for endpoints have not sent anything recently.
for (uint32_t i = 0; i < ic->peerMap.count; i++) {
struct IFCPeer* ep = ic->peerMap.values[i];
// This is here because of a pathological state where the connection is in ESTABLISHED
// state but the *direct peer* has somehow been dropped from the routing table.
// TODO: understand the cause of this issue rather than checking for it once per second.
struct Node* peerNode = RouterModule_getNode(ep->switchLabel, ic->routerModule);
if (now > ep->timeOfLastMessage + ic->pingAfterMilliseconds || !peerNode) {
#ifdef Log_DEBUG
uint8_t key[56];
Base32_encode(key, 56, CryptoAuth_getHerPublicKey(ep->cryptoAuthIf), 32);
#endif
if (ep->isIncomingConnection
&& now > ep->timeOfLastMessage + ic->forgetAfterMilliseconds)
{
Log_debug(ic->logger, "Unresponsive peer [%s.k] has not responded in [%u] "
"seconds, dropping connection",
key, ic->forgetAfterMilliseconds / 1024);
Allocator_free(ep->external->allocator);
return;
}
bool unresponsive = (now > ep->timeOfLastMessage + ic->unresponsiveAfterMilliseconds);
uint32_t lag = ~0u;
if (unresponsive) {
// flush the peer from the table...
RouterModule_brokenPath(ep->switchLabel, ic->routerModule);
// Lets skip 87% of pings when they're really down.
if (ic->pingCount % 8) {
continue;
}
ep->state = InterfaceController_PeerState_UNRESPONSIVE;
lag = ((now - ep->timeOfLastMessage) / 1024);
} else {
lag = ((now - ep->timeOfLastMessage) / 1024);
}
struct SwitchPinger_Ping* ping =
SwitchPinger_newPing(ep->switchLabel,
String_CONST(""),
ic->timeoutMilliseconds,
onPingResponse,
ic->allocator,
ic->switchPinger);
if (!ping) {
Log_debug(ic->logger,
"Failed to ping %s peer [%s.k] lag [%u], out of ping slots.",
(unresponsive ? "unresponsive" : "lazy"), key, lag);
return;
}
ping->onResponseContext = ep;
SwitchPinger_sendPing(ping);
Log_debug(ic->logger,
"Pinging %s peer [%s.k] lag [%u]",
(unresponsive ? "unresponsive" : "lazy"), key, lag);
}
}
}
