summaryrefslogtreecommitdiff
path: root/src/common/network.c
diff options
context:
space:
mode:
authorblacksirius <blacksirius@54d463be-8e91-2dee-dedb-b68131a5f0ec>2012-06-12 00:59:55 +0000
committerblacksirius <blacksirius@54d463be-8e91-2dee-dedb-b68131a5f0ec>2012-06-12 00:59:55 +0000
commit49586af4b75b96010380576fe9ce0a9b67a83ac2 (patch)
treeafebe591c3a1a372f66422f495f5e1fd0538032e /src/common/network.c
parent0598cc569db02ee93d7fc0470defecb64e995f5c (diff)
downloadhercules-49586af4b75b96010380576fe9ce0a9b67a83ac2.tar.gz
hercules-49586af4b75b96010380576fe9ce0a9b67a83ac2.tar.bz2
hercules-49586af4b75b96010380576fe9ce0a9b67a83ac2.tar.xz
hercules-49586af4b75b96010380576fe9ce0a9b67a83ac2.zip
added new networking subsystem (early stage - files are not compiled yet during normal build!)
Note The files i added / modifications i did, are not affecting a normal build nothing gets changed yet! Linux 2.5+ only yet. git-svn-id: https://rathena.svn.sourceforge.net/svnroot/rathena/trunk@16271 54d463be-8e91-2dee-dedb-b68131a5f0ec
Diffstat (limited to 'src/common/network.c')
-rw-r--r--src/common/network.c1062
1 files changed, 1062 insertions, 0 deletions
diff --git a/src/common/network.c b/src/common/network.c
new file mode 100644
index 000000000..2574f3cc1
--- /dev/null
+++ b/src/common/network.c
@@ -0,0 +1,1062 @@
+ //
+// Network Subsystem (previously known as socket system)
+//
+// Author: Florian Wilkemeyer <fw@f-ws.de>
+//
+// Copyright (c) rAthena Project (www.rathena.org) - Licensed under GNU GPL
+// For more information, see LICENCE in the main folder
+//
+//
+//#ifdef HAVE_ACCETP4
+#define _GNU_SOURCE
+//#endif
+
+#include <stdio.h>
+#include <string.h>
+#include <unistd.h>
+#include <errno.h>
+
+#include <sys/types.h>
+#include <sys/fcntl.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <netinet/tcp.h>
+#include <arpa/inet.h>
+
+
+#include "../common/cbasetypes.h"
+#include "../common/showmsg.h"
+#include "../common/timer.h"
+#include "../common/evdp.h"
+#include "../common/netbuffer.h"
+
+#include "../common/network.h"
+
+#define ENABLE_IPV6
+#define HAVE_ACCEPT4
+#define EVENTS_PER_CYCLE 10
+#define PARANOID_CHECKS
+
+// Local Vars (settings..)
+static int l_ListenBacklog = 64;
+
+//
+// Global Session Array (previously exported as session[]
+//
+SESSION g_Session[MAXCONN];
+
+
+//
+static bool onSend(int32 fd);
+
+
+#define _network_free_netbuf_async( buf ) add_timer( 0, _network_async_free_netbuf_proc, 0, (intptr_t) buf)
+static int _network_async_free_netbuf_proc(int tid, unsigned int tick, int id, intptr_t data){
+ // netbuf is in data
+ netbuffer_put( (netbuf)data );
+
+ return 0;
+}//end: _network_async_free_netbuf_proc()
+
+
+
+void network_init(){
+ SESSION *s;
+ int32 i;
+
+ memset(g_Session, 0x00, (sizeof(SESSION) * MAXCONN) );
+
+ for(i = 0; i < MAXCONN; i++){
+ s = &g_Session[i];
+
+ s->type = NST_FREE;
+ s->disconnect_in_progress = false;
+
+ }
+
+ // Initialize the correspondig event dispatcher
+ evdp_init();
+
+ //
+ add_timer_func_list(_network_async_free_netbuf_proc, "_network_async_free_netbuf_proc");
+
+}//end: network_init()
+
+
+void network_final(){
+
+ // @TODO:
+ // .. disconnect and cleanup everything!
+
+ evdp_final();
+
+}//end: network_final()
+
+
+void network_do(){
+ struct EVDP_EVENT l_events[EVENTS_PER_CYCLE];
+ register struct EVDP_EVENT *ev;
+ register int n, nfds;
+ register SESSION *s;
+
+ nfds = evdp_wait( l_events, EVENTS_PER_CYCLE, 1000); // @TODO: timer_getnext()
+
+ for(n = 0; n < nfds; n++){
+ ev = &l_events[n];
+ s = &g_Session[ ev->fd ];
+
+ if(ev->events & EVDP_EVENT_HUP){
+ network_disconnect( ev->fd );
+ continue; // no further event processing.
+ }// endif vent is HUP (disconnect)
+
+
+ if(ev->events & EVDP_EVENT_IN){
+
+ if(s->onRecv != NULL){
+ if( false == s->onRecv(ev->fd) ){
+ network_disconnect(ev->fd);
+ continue; // ..
+ }
+ }else{
+ ShowError("network_do: fd #%u has no onRecv proc set. - disconnecting\n", ev->fd);
+ network_disconnect(ev->fd);
+ continue;
+ }
+
+ }// endif event is IN (recv)
+
+
+ if(ev->events & EVDP_EVENT_OUT){
+ if(s->onSend != NULL){
+ if( false == s->onSend(ev->fd) ){
+ network_disconnect(ev->fd);
+ continue;
+ }
+ }else{
+ ShowError("network_do: fd #%u has no onSend proc set. - disconnecting\n", ev->fd);
+ network_disconnect(ev->fd);
+ continue;
+ }
+ }// endif event is OUT (send)
+
+ }//endfor
+
+}//end: network_do()
+
+
+static bool _setnonblock(int32 fd){
+ int flags = fcntl(fd, F_GETFL, 0);
+ if(fcntl(fd, F_SETFL, flags | O_NONBLOCK) != 0)
+ return false;
+
+ return true;
+}//end: _setnonblock()
+
+
+static bool _network_accept(int32 fd){
+ SESSION *listener = &g_Session[fd];
+ SESSION *s;
+ union{
+ struct sockaddr_in v4;
+#ifdef ENABLE_IPV6
+ struct sockaddr_in6 v6;
+#endif
+ } _addr;
+ int newfd;
+ socklen_t addrlen;
+ struct sockaddr *addr;
+
+ // Accept until OS returns - nothing to accept anymore
+ // - this is required due to our EVDP abstraction. (which handles on listening sockets similar to epoll's EPOLLET flag.)
+ while(1){
+#ifdef ENABLE_IPV6
+ if(listener->v6 == true){
+ addrlen = sizeof(_addr.v6);
+ addr = (struct sockaddr*)&_addr.v6;
+ }else{
+#endif
+ addrlen = sizeof(_addr.v4);
+ addr = (struct sockaddr*)&_addr.v4;
+#ifdef ENABLE_IPV6
+ }
+#endif
+
+#ifdef HAVE_ACCEPT4
+ newfd = accept4(fd, addr, &addrlen, SOCK_NONBLOCK);
+#else
+ newfd = accept(fd, addr, &addrlen);
+#endif
+
+ if(newfd == -1){
+ if(errno == EAGAIN || errno == EWOULDBLOCK)
+ break; // this is fully valid & whished., se explaination on top of while(1)
+
+ // Otherwis .. we have serious problems :( seems tahat our listner has gone away..
+ // @TODO handle this ..
+ ShowError("_network_accept: accept() returned error. closing listener. (errno: %u / %s)\n", errno, strerror(errno));
+
+ return false; // will call disconnect after return.
+ //break;
+ }
+
+#ifndef HAVE_ACCEPT4 // no accept4 means, we have to set nonblock by ourself. ..
+ if(_setnonblock(newfd) == false){
+ ShowError("_network_accept: failed to set newly accepted connection nonblocking (errno: %u / %s). - disconnecting.\n", errno, strerror(errno));
+ close(newfd);
+ continue;
+ }
+#endif
+
+ // Check connection limits.
+ if(newfd >= MAXCONN){
+ ShowError("_network_accept: failed to accept connection - MAXCONN (%u) exceeded.\n", MAXCONN);
+ close(newfd);
+ continue; // we have to loop over the events (and disconnect them too ..) but otherwise we would leak event notifications.
+ }
+
+
+ // Create new Session.
+ s = &g_Session[newfd];
+ s->type = NST_CLIENT;
+
+ // The new connection inherits listenr's handlers.
+ s->onDisconnect = listener->onDisconnect;
+ s->onConnect = listener->onConnect; // maybe useless but .. fear the future .. :~
+
+ // Register the new connection @ EVDP
+ if( evdp_addclient(newfd, &s->evdp_data) == false){
+ ShowError("_network_accept: failed to accept connection - event subsystem returned an error.\n");
+ close(newfd);
+ s->type = NST_FREE;
+ }
+
+ // Call the onConnect handler on the listener.
+ if( listener->onConnect(newfd) == false ){
+ // Resfused by onConnect handler..
+ evdp_remove(newfd, &s->evdp_data);
+
+ close(newfd);
+ s->type = NST_FREE;
+
+ s->data = NULL; // be on the safe side ~ !
+ continue;
+ }
+
+
+ }
+
+ return true;
+}//end: _network_accept()
+
+
+void network_disconnect(int32 fd){
+ SESSION *s = &g_Session[fd];
+ netbuf b, bn;
+
+ // Prevent recursive calls
+ // by wrong implemented on disconnect handlers.. and such..
+ if(s->disconnect_in_progress == true)
+ return;
+
+ s->disconnect_in_progress = true;
+
+
+ // Disconnect Todo:
+ // - Call onDisconnect Handler
+ // - Release all Assigned buffers.
+ // - remove from event system (notifications)
+ // - cleanup session structure
+ // - close connection.
+ //
+
+ if(s->onDisconnect != NULL &&
+ s->type != NST_LISTENER){
+
+ s->onDisconnect( fd );
+ }
+
+ // Read Buffer
+ if(s->read.buf != NULL){
+ netbuffer_put(s->read.buf);
+ s->read.buf = NULL;
+ }
+
+ // Write Buffer(s)
+ b = s->write.buf;
+ while(1){
+ if(b == NULL) break;
+
+ bn = b->next;
+
+ netbuffer_put(b);
+
+ b = bn;
+ }
+ s->write.buf = NULL;
+ s->write.buf_last = NULL;
+
+ s->write.n_outstanding = 0;
+ s->write.max_outstanding = 0;
+
+
+ // Remove from event system.
+ evdp_remove(fd, &s->evdp_data);
+
+ // Cleanup Session Structure.
+ s->type = NST_FREE;
+ s->data = NULL; // no application level data assigned
+ s->disconnect_in_progress = false;
+
+
+ // Close connection
+ close(fd);
+
+}//end: network_disconnect()
+
+
+int32 network_addlistener(bool v6, const char *addr, uint16 port){
+ SESSION *s;
+ int optval, fd;
+
+#if !defined(ENABLE_IPV6)
+ if(v6 == true){
+ ShowError("network_addlistener(%c, '%s', %u): this release has no IPV6 support.\n", (v6==true?'t':'f'), addr, port);
+ return -1;
+ }
+#endif
+
+
+#ifdef ENABLE_IPV6
+ if(v6 == true)
+ fd = socket(AF_INET6, SOCK_STREAM, 0);
+ else
+#endif
+ fd = socket(AF_INET, SOCK_STREAM, 0);
+
+ // Error?
+ if(fd == -1){
+ ShowError("network_addlistener(%c, '%s', %u): socket() failed (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+ return -1;
+ }
+
+ // Too many connections?
+ if(fd >= MAXCONN){
+ ShowError("network_addlistener(%c, '%s', %u): cannot create listener, exceeds more than supported connections (%u).\n", (v6==true?'t':'f'), addr, port, MAXCONN);
+ close(fd);
+ return -1;
+ }
+
+
+ s = &g_Session[fd];
+ if(s->type != NST_FREE){ // additional checks.. :)
+ ShowError("network_addlistener(%c, '%s', %u): failed, got fd #%u which is already in use in local session table?!\n", (v6==true?'t':'f'), addr, port, fd);
+ close(fd);
+ return -1;
+ }
+
+
+ // Fill ip addr structs
+#ifdef ENABLE_IPV6
+ if(v6 == true){
+ memset(&s->addr.v6, 0x00, sizeof(s->addr.v6));
+ s->addr.v6.sin6_family = AF_INET6;
+ s->addr.v6.sin6_port = htons(port);
+ if(inet_pton(AF_INET6, addr, &s->addr.v6.sin6_addr) != 1){
+ ShowError("network_addlistener(%c, '%s', %u): failed to parse the given IPV6 address.\n", (v6==true?'t':'f'), addr, port);
+ close(fd);
+ return -1;
+ }
+
+ }else{
+#endif
+ memset(&s->addr.v4, 0x00, sizeof(s->addr.v4));
+ s->addr.v4.sin_family = AF_INET;
+ s->addr.v4.sin_port = htons(port);
+ s->addr.v4.sin_addr.s_addr = inet_addr(addr);
+#ifdef ENABLE_IPV6
+ }
+#endif
+
+
+ // if OS has support for SO_REUSEADDR, apply the flag
+ // so the address could be used when there're still time_wait sockets outstanding from previous application run.
+#ifdef SO_REUSEADDR
+ optval=1;
+ setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &optval, sizeof(optval));
+#endif
+
+ // Bind
+#ifdef ENABLE_IPV6
+ if(v6 == true){
+ if( bind(fd, (struct sockaddr*)&s->addr.v6, sizeof(s->addr.v6)) == -1) {
+ ShowError("network_addlistener(%c, '%s', %u): bind failed (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+ close(fd);
+ return -1;
+ }
+ }else{
+#endif
+ if( bind(fd, (struct sockaddr*)&s->addr.v4, sizeof(s->addr.v4)) == -1) {
+ ShowError("network_addlistener(%c, '%s', %u): bind failed (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+ close(fd);
+ return -1;
+ }
+#ifdef ENABLE_IPV6
+ }
+#endif
+
+ if( listen(fd, l_ListenBacklog) == -1){
+ ShowError("network_addlistener(%c, '%s', %u): listen failed (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+ close(fd);
+ return -1;
+ }
+
+
+ // Set to nonblock!
+ if(_setnonblock(fd) == false){
+ ShowError("network_addlistener(%c, '%s', %u): cannot set to nonblock (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+ close(fd);
+ return -1;
+ }
+
+
+ // Rgister @ evdp.
+ if( evdp_addlistener(fd, &s->evdp_data) != true){
+ ShowError("network_addlistener(%c, '%s', %u): eventdispatcher subsystem returned an error.\n", (v6==true?'t':'f'), addr, port);
+ close(fd);
+ return -1;
+ }
+
+
+ // Apply flags on Session array for this conneciton.
+ if(v6 == true) s->v6 = true;
+ else s->v6 = false;
+
+ s->type = NST_LISTENER;
+ s->onRecv = _network_accept;
+
+ ShowStatus("Added Listener on '%s':%u\n", addr, port, (v6==true ? "(ipv6)":"(ipv4)") );
+
+ return fd;
+}//end: network_addlistener()
+
+
+static bool _network_connect_establishedHandler(int32 fd){
+ register SESSION *s = &g_Session[fd];
+ int val;
+ socklen_t val_len;
+
+ if(s->type == NST_FREE)
+ return true; // due to multiple non coalesced event notifications
+ // this can happen .. when a previous handled event has already disconnected the connection
+ // within the same cycle..
+
+ val = -1;
+ val_len = sizeof(val);
+ getsockopt(fd, SOL_SOCKET, SO_ERROR, &val, &val_len);
+
+ if(val != 0){
+ // :( .. cleanup session..
+ s->type = NST_FREE;
+ s->onSend = NULL;
+ s->onConnect = NULL;
+ s->onDisconnect = NULL;
+
+ evdp_remove(fd, &s->evdp_data);
+ close(fd);
+
+ return true; // we CANT return false,
+ // becuase the normal disconnect procedure would execute the ondisconnect handler, which we dont want .. in this case.
+ }else{
+ // ok
+ if(s->onConnect(fd) == false) {
+ // onConnect handler has refused the connection ..
+ // cleanup .. and ok
+ s->type = NST_FREE;
+ s->onSend = NULL;
+ s->onConnect = NULL;
+ s->onDisconnect = NULL;
+
+ evdp_remove(fd, &s->evdp_data);
+ close(fd);
+
+ return true; // we dnot want the ondisconnect handler to be executed, so its okay to handle this by ourself.
+ }
+
+ // connection established !
+ //
+ if( evdp_outgoingconnection_established(fd, &s->evdp_data) == false ){
+ return false; // we want the normal disconnect procedure.. with call to ondisconnect handler.
+ }
+
+ s->onSend = NULL;
+
+ ShowStatus("#%u connection successfull!\n", fd);
+ }
+
+ return true;
+}//end: _network_connect_establishedHandler()
+
+
+int32 network_connect(bool v6,
+ const char *addr,
+ uint16 port,
+ const char *from_addr,
+ uint16 from_port,
+ bool (*onConnectionEstablishedHandler)(int32 fd),
+ void (*onConnectionLooseHandler)(int32 fd)
+){
+ register SESSION *s;
+ int32 fd, optval, ret;
+ struct sockaddr_in ip4;
+#ifdef ENABLE_IPV6
+ struct sockaddr_in6 ip6;
+#endif
+
+#ifdef ENABLE_IPV6
+ if(v6 == true)
+ fd = socket(AF_INET6, SOCK_STREAM, 0);
+ else
+#endif
+ fd = socket(AF_INET, SOCK_STREAM, 0);
+
+#ifndef ENABLE_IPV6
+ // check..
+ if(v6 == true){
+ ShowError("network_connect(%c, '%s', %u...): tried to create an ipv6 connection, IPV6 is not supported in this release.\n", (v6==true?'t':'f'), addr, port);
+ return -1;
+ }
+#endif
+
+ // check connection limits.
+ if(fd >= MAXCONN){
+ ShowError("network_connect(%c, '%s', %u...): cannot create new connection, exceeeds more than supported connections (%u)\n", (v6==true?'t':'f'), addr, port );
+ close(fd);
+ return -1;
+ }
+
+
+ // Originating IP/Port pair given ?
+ if(from_addr != NULL && *from_addr != 0){
+ //..
+ #ifdef SO_REUSEADDR
+ optval=1;
+ setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &optval, sizeof(optval));
+ #endif
+
+ #ifdef ENABLE_IPV6
+ if(v6 == true){
+ memset(&ip6, 0x00, sizeof(ip6));
+ ip6.sin6_family = AF_INET6;
+ ip6.sin6_port = htons(from_port);
+
+ if(inet_pton(AF_INET6, from_addr, &ip6.sin6_addr) != 1){
+ ShowError("network_connect(%c, '%s', %u...): cannot parse originating (from) IPV6 address (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+ close(fd);
+ return -1;
+ }
+
+ ret = bind(fd, (struct sockaddr*)&ip6, sizeof(ip6));
+ }else{
+ #endif
+ memset(&ip4, 0x00, sizeof(ip4));
+
+ ip4.sin_family = AF_INET;
+ ip4.sin_port = htons(from_port);
+ ip4.sin_addr.s_addr = inet_addr(from_addr);
+ ret = bind(fd, (struct sockaddr*)&ip4, sizeof(ip4));
+ #ifdef ENABLE_IPV6
+ }
+ #endif
+
+ }
+
+
+ // Set non block
+ if(_setnonblock(fd) == false){
+ ShowError("network_connect(%c, '%s', %u...): cannot set socket to nonblocking (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+ close(fd);
+ return -1;
+ }
+
+
+ // Create ip addr block to connect to ..
+#ifdef ENABLE_IPV6
+ if(v6 == true){
+ memset(&ip6, 0x00, sizeof(ip6));
+ ip6.sin6_family = AF_INET6;
+ ip6.sin6_port = htons(port);
+
+ if(inet_pton(AF_INET6, addr, &ip6.sin6_addr) != 1){
+ ShowError("network_connect(%c, '%s', %u...): cannot parse destination IPV6 address (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+ close(fd);
+ return -1;
+ }
+
+ }else{
+#endif
+ memset(&ip4, 0x00, sizeof(ip4));
+
+ ip4.sin_family = AF_INET;
+ ip4.sin_port = htons(port);
+ ip4.sin_addr.s_addr = inet_addr(addr);
+#ifdef ENABLE_IPV6
+ }
+#endif
+
+
+ // Assign Session..
+ s = &g_Session[fd];
+ s->type = NST_OUTGOING;
+ s->v6 = v6;
+ s->onConnect = onConnectionEstablishedHandler;
+ s->onDisconnect = onConnectionLooseHandler;
+ s->onRecv = NULL;
+ s->onSend = _network_connect_establishedHandler;
+#ifdef ENABLE_IPV6
+ if(v6 == true)
+ memcpy(&s->addr.v6, &ip6, sizeof(ip6));
+ else
+#endif
+ memcpy(&s->addr.v4, &ip4, sizeof(ip4));
+
+ // Register @ EVDP. as outgoing (see doc of the function)
+ if(evdp_addconnecting(fd, &s->evdp_data) == false){
+ ShowError("network_connect(%c, '%s', %u...): eventdispatcher subsystem returned an error.\n", (v6==true?'t':'f'), addr, port);
+
+ // cleanup session x.x..
+ s->type = NST_FREE;
+ s->onConnect = NULL;
+ s->onDisconnect = NULL;
+ s->onSend = NULL;
+
+ // close, return error code.
+ close(fd);
+ return -1;
+ }
+
+
+#ifdef ENABLE_IPV6
+ if(v6 == true)
+ ret = connect(fd, (struct sockaddr*)&ip6, sizeof(ip6));
+ else
+#endif
+ ret = connect(fd, (struct sockaddr*)&ip4, sizeof(ip4));
+
+
+ //
+ if(ret != 0 && errno != EINPROGRESS){
+ ShowWarning("network_connect(%c, '%s', %u...): connection failed (errno: %u / %s)\n", (v6==true?'t':'f'), addr, port, errno, strerror(errno));
+
+ // Cleanup session ..
+ s->type = NST_FREE;
+ s->onConnect = NULL;
+ s->onDisconnect = NULL;
+ s->onSend = NULL;
+
+ // .. remove from evdp and close fd.
+ evdp_remove(fd, &s->evdp_data);
+ close(fd);
+ return -1;
+ }
+
+
+ // ! The Info Message :~D
+ ShowStatus("network_connect fd#%u (%s:%u) in progress.. \n", fd, addr, port);
+
+return fd;
+}//end: network_connect()
+
+
+static bool _onSend(int32 fd){
+ register SESSION *s = &g_Session[fd];
+ register netbuf buf, buf_next;
+ register uint32 szNeeded;
+ register int wLen;
+
+ if(s->type == NST_FREE)
+ return true; // Possible due to multipl non coalsced event notifications
+ // so onSend gets called after disconnect caused by an previous vent.
+ // we can ignore the call to onSend, then.
+
+ buf = s->write.buf;
+ while(1){
+ if(buf == NULL)
+ break;
+
+ buf_next = buf->next;
+
+
+ szNeeded = (buf->dataLen - s->write.dataPos); // using th session-local .dataPos member, due to shared write buffer support.
+
+ // try to write.
+ wLen = write(fd, &buf->buf[s->write.dataPos], szNeeded);
+ if(wLen == 0){
+ return false; // eof.
+ }else if(wLen == -1){
+ if(errno == EAGAIN || errno == EWOULDBLOCK)
+ return true; // dont disconnect / try again later.
+
+ // all other errors. .
+ return false;
+ }
+
+ // Wrote data.. =>
+ szNeeded -= wLen;
+ if(szNeeded > 0){
+ // still data left ..
+ //
+ s->write.dataPos += wLen; // fix offset.
+ return true;
+ }else{
+ // this buffer has been written successfully
+ // could be returned to pool.
+ netbuffer_put(buf);
+ s->write.n_outstanding--; // When threadsafe -> Interlocked here.
+ s->write.dataPos = 0;
+ }
+
+
+ buf = buf_next;
+ }
+
+ // okay,
+ // reaching this part means:
+ // while interrupted by break -
+ // which means all buffers are written, nothing left
+ //
+
+ s->write.buf_last = NULL;
+ s->write.buf = NULL;
+ s->write.n_outstanding = 0;
+ s->write.dataPos = 0;
+
+ // Remove from event dispatcher (write notification)
+ //
+ evdp_writable_remove(fd, &s->evdp_data);
+
+ return true;
+}//end: _onSend()
+
+
+static bool _onRORecv(int32 fd){
+ register SESSION *s = &g_Session[fd];
+ register uint32 szNeeded;
+ register char *p;
+ register int rLen;
+
+ if(s->type == NST_FREE)
+ return true; // Possible due to multiple non coalesced events by evdp.
+ // simply ignore this call returning positive result.
+
+ // Initialize p and szNeeded depending on change
+ //
+ switch(s->read.state){
+ case NRS_WAITOP:
+ szNeeded = s->read.head_left;
+ p = ((char*)&s->read.head[0]) + (2-szNeeded);
+ break;
+
+ case NRS_WAITLEN:
+ szNeeded = s->read.head_left;
+ p = ((char*)&s->read.head[1]) + (2-szNeeded);
+ break;
+
+ case NRS_WAITDATA:{
+ register netbuf buf = s->read.buf;
+
+ szNeeded = (buf->dataLen - buf->dataPos);
+ p = (char*)&buf->buf[ buf->dataPos ];
+ }
+ break;
+
+ default:
+ // .. the impossible gets possible ..
+ ShowError("_onRORecv: fd #%u has unknown read.state (%d) - disconnecting\n", fd, s->read.state);
+ return false;
+ break;
+ }
+
+
+ //
+
+ rLen = read(fd, p, szNeeded);
+ if(rLen == 0){
+ // eof..
+ return false;
+ }else if(rLen == -1){
+
+ if(errno == EAGAIN || errno == EWOULDBLOCK){
+ // try again later .. (this case shouldnt happen, because we're event trigered.. but .. sometimes it happens :)
+ return true;
+ }
+
+ // an additional interesting case would be
+ // EINTR, this 'could' be handled .. but:
+ // posix says that its possible that data gets currupted during irq
+ // or data gor read and not reported.., so we'd have a data loss..
+ // (which shouldnt happen with stream based protocols such as tcp)
+ // its better to disonnect the client in that case.
+
+ return false;
+ }
+
+ //
+ // Got Data:
+ // next action also depends on current state ..
+ //
+ szNeeded -= rLen;
+ switch(s->read.state){
+ case NRS_WAITOP:
+
+ if(szNeeded > 0){
+ // still data missing ..
+ s->read.head_left = szNeeded;
+ return true; // wait for completion.
+ }else{
+ // complete ..
+ // next state depends on packet type.
+
+ s->read.head[1] = ((uint16*)s->netparser_data)[ s->read.head[0] ]; // store lenght of packet by opcode head[0] to head[1]
+
+ if(s->read.head[1] == ROPACKET_UNKNOWN){
+ // unknown packet - disconnect
+ ShowWarning("_onRORecv: fd #%u got unlnown packet 0x%04x - disconnecting.\n", fd, s->read.head[0]);
+ return false;
+ }
+ else if(s->read.head[1] == ROPACKET_DYNLEN){
+ // dynamic length
+ // next state: requrie len.
+ s->read.state = NRS_WAITLEN;
+ s->read.head_left = 2;
+ return true; //
+ }
+ else if(s->read.head[1] == 2){
+ // packet has no data (only opcode)
+ register netbuf buf = netbuffer_get(2); // :D whoohoo its giant!
+
+ NBUFW(buf, 0) = s->read.head[0]; // store opcode @ packet begin.
+ buf->dataPos = 2;
+ buf->dataLen = 2;
+ buf->next = NULL;
+
+ // Back to initial state -> Need opcode.
+ s->read.state = NRS_WAITOP;
+ s->read.head_left = 2;
+ s->read.buf = NULL;
+
+ // Call completion routine here.
+ s->onPacketComplete(fd, s->read.head[0], 2, buf);
+
+ return true; // done :)
+ }
+ else{
+ // paket needs .. data ..
+ register netbuf buf = netbuffer_get( s->read.head[1] );
+
+ NBUFW(buf, 0) = s->read.head[0]; // store opcode @ packet begin.
+ buf->dataPos = 2;
+ buf->dataLen = s->read.head[1];
+ buf->next = NULL;
+
+ // attach buffer.
+ s->read.buf = buf;
+
+ // set state:
+ s->read.state = NRS_WAITDATA;
+
+ return true;
+ }
+
+ }//endif: szNeeded > 0 (opcode read completed?)
+
+ break;
+
+
+ case NRS_WAITLEN:
+
+ if(szNeeded > 0){
+ // incomplete ..
+ s->read.head_left = szNeeded;
+ return true;
+ }else{
+
+ if(s->read.head[1] == 4){
+ // packet has no data (only opcode + length)
+ register netbuf buf = netbuffer_get( 4 );
+
+ NBUFL(buf, 0) = *((uint32*)&s->read.head[0]); // copy Opcode + length to netbuffer using MOVL
+ buf->dataPos = 4;
+ buf->dataLen = 4;
+ buf->next = NULL;
+
+ // set initial state (need opcode)
+ s->read.state = NRS_WAITOP;
+ s->read.head_left = 2;
+ s->read.buf = NULL;
+
+ // call completion routine.
+ s->onPacketComplete(fd, s->read.head[0], 4, buf);
+
+ return true;
+ }
+ else if(s->read.head[1] < 4){
+ // invalid header.
+ ShowWarning("_onRORecv: fd #%u invalid header - got packet 0x%04x, reported length < 4 - INVALID - disconnecting\n", fd, s->read.head[0]);
+ return false;
+ }
+ else{
+ // Data needed
+ // next state -> waitdata!
+ register netbuf buf = netbuffer_get( s->read.head[1] );
+
+ NBUFL(buf, 0) = *((uint32*)&s->read.head[0]); // copy Opcode + length to netbuffer using MOVL
+ buf->dataPos = 4;
+ buf->dataLen = s->read.head[1];
+ buf->next = NULL;
+
+ // attach to session:
+ s->read.buf = buf;
+ s->read.state = NRS_WAITDATA;
+
+ return true;
+ }
+
+ }//endif: szNeeded > 0 (length read complete?)
+
+ break;
+
+
+ case NRS_WAITDATA:
+
+ if(szNeeded == 0){
+ // Packet finished!
+ // compltion.
+ register netbuf buf = s->read.buf;
+
+ // set initial state.
+ s->read.state = NRS_WAITOP;
+ s->read.head_left = 2;
+ s->read.buf = NULL;
+
+ // Call completion routine.
+ s->onPacketComplete(fd, NBUFW(buf, 0), buf->dataLen, buf);
+
+ return true;
+ }else{
+ // still data needed
+ s->read.buf->dataPos += rLen;
+
+ return true;
+ }
+ break;
+
+
+ //
+ default:
+ ShowError("_onRORecv: fd #%u has unknown read.state (%d) [2] - disconnecting\n", fd, s->read.state);
+ return false;
+ break;
+ }
+
+
+ return false;
+}//end: _onRORecv()
+
+
+void network_send(int32 fd, netbuf buf){
+ register SESSION *s = &g_Session[fd];
+
+#ifdef PARANOID_CHECKS
+ if(fd >= MAXCONN){
+ ShowError("network_send: tried to attach buffer to connection idientifer #%u which is out of bounds.\n", fd);
+ _network_free_netbuf_async(buf);
+ return;
+ }
+#endif
+
+
+ if(s->type == NST_FREE)
+ return;
+
+ // Check Max Outstanding buffers limit.
+ if( (s->write.max_outstanding > 0) &&
+ (s->write.n_outstanding >= s->write.max_outstanding) ){
+
+ ShowWarning("network_send: fd #%u max Outstanding buffers exceeded. - disconnecting.\n", fd);
+ network_disconnect(fd);
+ //
+ _network_free_netbuf_async(buf);
+ return;
+ }
+
+
+ // Attach to the end:
+ buf->next = NULL;
+ if(s->write.buf_last != NULL){
+ s->write.buf_last->next = buf;
+ s->write.buf_last = buf;
+
+ }else{
+ // currently no buffer attached.
+ s->write.buf = s->write.buf_last = buf;
+
+ // register @ evdp for writable notification.
+ evdp_writable_add(fd, &s->evdp_data); //
+ }
+
+
+ //
+ s->write.n_outstanding++;
+
+}//end: network_send()
+
+
+void network_parser_set_ro(int32 fd,
+ int16 *packetlentable,
+ void (*onPacketCompleteProc)(int32 fd, uint16 op, uint16 len, netbuf buf)
+ ){
+ register SESSION *s = &g_Session[fd];
+ register netbuf b, nb; // used for potential free attached buffers.
+
+ if(s->type == NST_FREE)
+ return;
+
+ s->onPacketComplete = onPacketCompleteProc;
+
+ s->onRecv = _onRORecv; // ..
+ s->onSend = _onSend; // Using the normal generic netbuf based send function.
+
+ s->netparser_data = packetlentable;
+
+ // Initial State -> Need Packet OPCode.
+ s->read.state = NRS_WAITOP;
+ s->read.head_left = 2;
+
+
+ // Detach (if..) all buffers.
+ if(s->read.buf != NULL){
+ _network_free_netbuf_async(s->read.buf); //
+ s->read.buf = NULL;
+ }
+
+ if(s->write.buf != NULL){
+ b = s->write.buf;
+ while(1){
+ nb = b->next;
+
+ _network_free_netbuf_async(b);
+
+ b = nb;
+ }
+
+ s->write.buf = NULL;
+ s->write.buf_last = NULL;
+ s->write.n_outstanding = 0;
+ }
+
+ // not changing any limits on outstanding ..
+ //
+
+}//end: network_parser_set_ro()
+