fix last commit

typo fix

ISSUE_TEMPLATE.md

@@ -0,0 +1,6 @@
+For English speaking user：
+https://github.com/wangyu-/UDPspeeder/wiki/Issue-Guide
+
+中文用户请看：
+https://github.com/wangyu-/UDPspeeder/wiki/发Issue前请看
+(否则Issue可能被忽略，或被直接关掉)

README.md

@@ -1,105 +1,150 @@
 # UDPspeeder
-![image0](images/Capture.PNG)
+Network Speed-Up Tool. Boost your Connection on a High Lantency High Packet-Loss Link by using Forward Error Correction.
-UDP加速器，降低UDP传输的丢包率。尤其适用于游戏和语音。
 
-这个是我自己稳定用了一个月的项目，用来玩服务器在美国的brawl stars和服务器在亚洲国外的mobile legend，效果不错。
+When used alone, UDPspeeder speeds-up only UDP connection. Nevertheless, if you used UDPspeeder + any UDP-based VPN together,
+you can speed-up any traffic(include TCP/UDP/ICMP), currently OpenVPN/L2TP/ShadowVPN are confirmed to be supported。
 
-#### 原理简介
+![](/images/en/udpspeeder.PNG)
-目前原理就是简单粗暴的多倍发包(以后会做Reed–Solomon code)。
 
-跟net-speeder比，优势在于client和server会把收到的多余包自动去掉，这个过程对上层透明，没有兼容性问题。而且发出的冗余数据包会做长度和内容的随机化，抓包是看不出发了冗余数据的，所以不用担心vps被封的问题。
+or
 
-每个冗余数据包都是间隔数毫秒（可配置）以后延迟发出的，可以避开中间路由器因为瞬时buffer长度过长而连续丢掉所有副本。
+![image_vpn](/images/en/udpspeeder+openvpn3.PNG)
 
-可以模拟延迟抖动,这样上层应用计算出来的RTT方差会更大，以等待后续冗余包的到达，不至于过早重传。
+Assume your local network to your server is lossy. Just establish a VPN connection to your server with UDPspeeder + any UDP-based VPN, access your server via this VPN connection, then your connection quality will be significantly improved. With well-tuned parameters , you can easily reduce IP or UDP/ICMP packet-loss-rate to less than 0.01% . Besides reducing packet-loss-rate, UDPspeeder can also significantly improve your TCP latency and TCP single-thread download speed.
 
-#### 适用场景
+[简体中文](/doc/README.zh-cn.md)(内容更丰富)
-绝大部分流量不高的情况。程序本身加速udp，但是配合openvpn可以加速任何流量。网络状况不好时，游戏卡得没法玩，或者网页卡得没法打开，使用起来效果最好。但是不适合大流量的场景，比如BT下载和在线看视频。 
 
-#### 其他功能
+###### Note
-输出UDP收发情况报告，便于分析网络。
+You can use udp2raw with UDPspeeder together to get better speed on some ISP with UDP QoS(UDP throttling).
 
-模拟丢包，模拟延迟，模拟jitter。便于通过实验找出应用卡顿的原因。
+udp2raw's repo：
 
-重复包过滤功能可以关掉，模拟网络本身有重复包的情况。用来测试应用对重复报的支持情况。
+https://github.com/wangyu-/udp2raw-tunnel
 
-目前有amd64,x86,ar71xx的binary
+You can also try tinyFecVPN, a lightweight high-performance VPN with build-in FEC support:
 
-# 简明操作说明
+tinyFecVPN's repo:
 
-### 环境要求
+https://github.com/wangyu-/tinyFecVPN
-Linux主机，可以使是openwrt路由器，也可以是树莓派。在windows和mac上可以开虚拟机（桥接模式测试可用）。
 
-### 安装
-下载编译好的二进制文件，解压到本地和服务器的任意目录。
 
-https://github.com/wangyu-/UDPspeeder/releases
+# Efficacy
+tested on a link with 100ms latency and 10% packet loss at both direction
 
-### 运行
+### Ping Packet Loss
-假设你有一个server，ip为44.55.66.77，有一个服务监听在udp 7777端口。 假设你需要加速本地到44.55.66.77:7777的流量
+![](/images/en/ping_compare_mode1.png)
 
-```
+### SCP Copy Speed
-在client端运行:
+![](/images/en/scp_compare2.PNG)
-./speeder_ar71xx -l0.0.0.0:3323 -r 44.55.66.77:8855 -c  -d2
 
-在server端运行:
+# Supported Platforms
-./speeder_amd64 -l0.0.0.0:8855 -r127.0.0.1:7777 -s -d2
+Linux host (including desktop Linux,Android phone/tablet, OpenWRT router, or Raspberry PI).
+
+For Windows and MacOS You can run UDPspeeder inside [this](https://github.com/wangyu-/udp2raw-tunnel/releases/download/20171108.0/lede-17.01.2-x86_virtual_machine_image.zip) 7.5mb virtual machine image.
+
+# How does it work
+
+UDPspeeder uses FEC(Forward Error Correction) to reduce packet loss rate, at the cost of addtional bandwidth. The algorithm for FEC is called Reed-Solomon.
+
+![image0](/images/en/fec.PNG)
+
+### Reed-Solomon
+
+`
+In coding theory, the Reed–Solomon code belongs to the class of non-binary cyclic error-correcting codes. The Reed–Solomon code is based on univariate polynomials over finite fields.
+`
+
+`
+It is able to detect and correct multiple symbol errors. By adding t check symbols to the data, a Reed–Solomon code can detect any combination of up to t erroneous symbols, or correct up to ⌊t/2⌋ symbols. As an erasure code, it can correct up to t known erasures, or it can detect and correct combinations of errors and erasures. Reed–Solomon codes are also suitable as multiple-burst bit-error correcting codes, since a sequence of b + 1 consecutive bit errors can affect at most two symbols of size b. The choice of t is up to the designer of the code, and may be selected within wide limits.
+`
+
+![](/images/en/rs.png)
+
+Check wikipedia for more info, https://en.wikipedia.org/wiki/Reed–Solomon_error_correction
+
+# Getting Started
+
+### Installing
+Download binary release from https://github.com/wangyu-/UDPspeeder/releases
+
+### Running (speed-up UDP only)
+Assume your server ip is 44.55.66.77, you have a service listening on udp port 7777.
+
+```bash
+# Run at server side:
+./speederv2 -s -l0.0.0.0:4096 -r 127.0.0.1:7777  -f20:10 -k "passwd"
+
+# Run at client side
+./speederv2 -c -l0.0.0.0:3333  -r44.55.66.77:4096 -f20:10 -k "passwd"
 ```
 
-现在client和server之间建立起了tunnel。想要连接44.55.66.77:7777，只需要连接 127.0.0.1:3333。来回的所有的udp流量会被加速。
+Now connecting to UDP port 3333 at the client side is equivalent to connecting to port 7777 at the server side, and the connection has been boosted by UDPspeeder.
 
-# 效果
+##### Note
-在模拟丢包的网络环境下，架设了vpn服务器(udp模式)。收发各有10%的丢包率
 
-#### 使用前
+`-f20:10` means sending 10 redundant packets for every 20 original packets.
-<img src="images/Capture4.PNG" width="500">
-#### 使用后
-3倍冗余数据。
 
-<img src="images/Capture6.PNG" width="500">
+`-k` enables simple XOR encryption
 
-# 进阶操作说明
+# Advanced Topic
-
+### Full Options
-### 命令选项
 ```
-UDPspeeder
+UDPspeeder V2
-version: Aug  9 2017 18:13:09
+git version: 6f55b8a2fc    build date: Nov 19 2017 06:11:23
 repository: https://github.com/wangyu-/UDPspeeder
 
 usage:
-    run as client : ./this_program -c -l local_listen_ip:local_port -r server_ip:server_port  [options]
+    run as client: ./this_program -c -l local_listen_ip:local_port -r server_ip:server_port  [options]
-    run as server : ./this_program -s -l server_listen_ip:server_port -r remote_ip:remote_port  [options]
+    run as server: ./this_program -s -l server_listen_ip:server_port -r remote_ip:remote_port  [options]
 
-common option,must be same on both sides:
+common options, must be same on both sides:
-    -k,--key              <string>        key for simple xor encryption,default:"secret key"
+    -k,--key              <string>        key for simple xor encryption. if not set, xor is disabled
 main options:
-    -d                    <number>        duplicated packet number, -d 0 means no duplicate. default value:0
+    -f,--fec              x:y             forward error correction, send y redundant packets for every x packets
-    -t                    <number>        duplicated packet delay time, unit: 0.1ms,default value:20(2ms)
+    --timeout             <number>        how long could a packet be held in queue before doing fec, unit: ms, default: 8ms
-    -j                    <number>        simulated jitter.randomly delay first packet for 0~jitter_value*0.1 ms,to
+    --report              <number>        turn on send/recv report, and set a period for reporting, unit: s
-                                          create simulated jitter.default value:0.do not use if you dont
-                                          know what it means
-    --report              <number>        turn on udp send/recv report,and set a time interval for reporting,unit:s
 advanced options:
-    -t                    tmin:tmax       simliar to -t above,but delay randomly between tmin and tmax
+    --mode                <number>        fec-mode,available values: 0,1; mode 0(default) costs less bandwidth,no mtu problem.
-    -j                    jmin:jmax       simliar to -j above,but create jitter randomly between jmin and jmax
+                                          mode 1 usually introduces less latency, but you have to care about mtu.
-    --random-drop         <number>        simulate packet loss ,unit 0.01%
+    --mtu                 <number>        mtu. for mode 0, the program will split packet to segment smaller than mtu value.
-    -m                    <number>        max pending packets,to prevent the program from eating up all your memory.
+                                          for mode 1, no packet will be split, the program just check if the mtu is exceed.
-other options:
+                                          default value: 1250. you typically shouldnt change this value.
-    --log-level           <number>        0:never    1:fatal   2:error   3:warn 
+    -q,--queue-len        <number>        fec queue len, only for mode 0, fec will be performed immediately after queue is full.
-                                          4:info (default)     5:debug   6:trace
+                                          default value: 200. 
-    --log-position                        enable file name,function name,line number in log
+    -j,--jitter           <number>        simulated jitter. randomly delay first packet for 0~<number> ms, default value: 0.
+                                          do not use if you dont know what it means.
+    -i,--interval         <number>        scatter each fec group to a interval of <number> ms, to protect burst packet loss.
+                                          default value: 0. do not use if you dont know what it means.
+    --random-drop         <number>        simulate packet loss, unit: 0.01%. default value: 0.
+    --disable-obscure     <number>        disable obscure, to save a bit bandwidth and cpu.
+developer options:
+    --fifo                <string>        use a fifo(named pipe) for sending commands to the running program, so that you
+                                          can change fec encode parameters dynamically, check readme.md in repository for
+                                          supported commands.
+    -j ,--jitter          jmin:jmax       similiar to -j above, but create jitter randomly between jmin and jmax
+    -i,--interval         imin:imax       similiar to -i above, but scatter randomly between imin and imax
+    --decode-buf          <number>        size of buffer of fec decoder,u nit: packet, default: 2000
+    --fix-latency         <number>        try to stabilize latency, only for mode 0
+    --delay-capacity      <number>        max number of delayed packets
+    --disable-fec         <number>        completely disable fec, turn the program into a normal udp tunnel
+    --sock-buf            <number>        buf size for socket, >=10 and <=10240, unit: kbyte, default: 1024
+log and help options:
+    --log-level           <number>        0: never    1: fatal   2: error   3: warn 
+                                          4: info (default)      5: debug   6: trace
+    --log-position                        enable file name, function name, line number in log
     --disable-color                       disable log color
-    --sock-buf            <number>        buf size for socket,>=10 and <=10240,unit:kbyte,default:512
     -h,--help                             print this help message
 
 ```
+#### `--fifo` option
+Use a fifo(named pipe) for sending commands to the running program. For example `--fifo fifo.file`, you can use following commands to change parameters dynamically:
+```
+echo fec 19:9 > fifo.file
+echo mtu 1100 > fifo.file
+echo timeout 5 > fifo.file
+echo queue-len 100 > fifo.file
+echo mode 0 > fifo.file
+```
+### Speed-Up any traffic with OpenVPN + UDPspeeder
 
-# 应用
+Check [UDPspeeder + openvpn config guide](/doc/udpspeeder_openvpn.md).
-
-#### UDPspeeder + openvpn加速任何流量
-如果你只是需要玩游戏，效果(可能/大概)会比kcp/finalspeed方案更好。可以优化tcp游戏的延迟（通过冗余发包，避免了上层的重传）。比如魔兽世界。
-![image0](images/Capture2.PNG)
-#### UDPspeeder + kcptun/finalspeed同时加速tcp和udp流量
-如果你需要用加速的tcp看视频和下载文件，这样效果比vpn方案更好。不论是速度，还是流量的耗费上。
-![image0](images/Capture3.PNG)
-

common.cpp

@@ -14,16 +14,45 @@ int about_to_exit=0;
 
 raw_mode_t raw_mode=mode_faketcp;
 unordered_map<int, const char*> raw_mode_tostring = {{mode_faketcp, "faketcp"}, {mode_udp, "udp"}, {mode_icmp, "icmp"}};
-int socket_buf_size=1024*1024;
-static int random_number_fd=-1;
-char iptables_rule[200]="";
-program_mode_t program_mode=unset_mode;//0 unset; 1client 2server
 
-u64_t get_current_time()
+
+//static int random_number_fd=-1;
+char iptables_rule[200]="";
+//int is_client = 0, is_server = 0;
+
+program_mode_t client_or_server=unset_mode;//0 unset; 1client 2server
+
+working_mode_t working_mode=tunnel_mode;
+
+int socket_buf_size=1024*1024;
+
+
+
+struct random_fd_t
+{
+	int random_number_fd;
+	random_fd_t()
+	{
+			random_number_fd=open("/dev/urandom",O_RDONLY);
+
+			if(random_number_fd==-1)
+			{
+				mylog(log_fatal,"error open /dev/urandom\n");
+				myexit(-1);
+			}
+			setnonblocking(random_number_fd);
+	}
+	int get_fd()
+	{
+		return random_number_fd;
+	}
+}random_fd;
+
+u64_t get_current_time()//ms
 {
 	timespec tmp_time;
 	clock_gettime(CLOCK_MONOTONIC, &tmp_time);
-	return tmp_time.tv_sec*1000+tmp_time.tv_nsec/(1000*1000l);
+	return ((u64_t)tmp_time.tv_sec)*1000llu+((u64_t)tmp_time.tv_nsec)/(1000*1000llu);
 }
 
 u64_t get_current_time_us()
@@ -49,6 +78,47 @@ u32_t get_u64_l(u64_t a)
 	return (a<<32u)>>32u;
 }
 
+void write_u16(char * p,u16_t w)
+{
+	*(unsigned char*)(p + 1) = (w & 0xff);
+	*(unsigned char*)(p + 0) = (w >> 8);
+}
+u16_t read_u16(char * p)
+{
+	u16_t res;
+	res = *(const unsigned char*)(p + 0);
+	res = *(const unsigned char*)(p + 1) + (res << 8);
+	return res;
+}
+
+void write_u32(char * p,u32_t l)
+{
+	*(unsigned char*)(p + 3) = (unsigned char)((l >>  0) & 0xff);
+	*(unsigned char*)(p + 2) = (unsigned char)((l >>  8) & 0xff);
+	*(unsigned char*)(p + 1) = (unsigned char)((l >> 16) & 0xff);
+	*(unsigned char*)(p + 0) = (unsigned char)((l >> 24) & 0xff);
+}
+u32_t read_u32(char * p)
+{
+	u32_t res;
+	res = *(const unsigned char*)(p + 0);
+	res = *(const unsigned char*)(p + 1) + (res << 8);
+	res = *(const unsigned char*)(p + 2) + (res << 8);
+	res = *(const unsigned char*)(p + 3) + (res << 8);
+	return res;
+}
+
+void write_u64(char * s,u64_t a)
+{
+	assert(0==1);
+}
+u64_t read_u64(char * s)
+{
+	assert(0==1);
+	return 0;
+}
+
+
 char * my_ntoa(u32_t ip)
 {
 	in_addr a;
@@ -94,22 +164,11 @@ int clear_iptables_rule()
 }
 
 
-void init_random_number_fd()
-{
 
-	random_number_fd=open("/dev/urandom",O_RDONLY);
-
-	if(random_number_fd==-1)
-	{
-		mylog(log_fatal,"error open /dev/urandom\n");
-		myexit(-1);
-	}
-	setnonblocking(random_number_fd);
-}
 u64_t get_true_random_number_64()
 {
 	u64_t ret;
-	int size=read(random_number_fd,&ret,sizeof(ret));
+	int size=read(random_fd.get_fd(),&ret,sizeof(ret));
 	if(size!=sizeof(ret))
 	{
 		mylog(log_fatal,"get random number failed %d\n",size);
@@ -122,7 +181,7 @@ u64_t get_true_random_number_64()
 u32_t get_true_random_number()
 {
 	u32_t ret;
-	int size=read(random_number_fd,&ret,sizeof(ret));
+	int size=read(random_fd.get_fd(),&ret,sizeof(ret));
 	if(size!=sizeof(ret))
 	{
 		mylog(log_fatal,"get random number failed %d\n",size);
@@ -201,28 +260,75 @@ unsigned short csum(const unsigned short *ptr,int nbytes) {
 
     return(answer);
 }
-int set_buf_size(int fd,int size)
+
-{
+
-	//int socket_buf_size=1024*1024;
+unsigned short tcp_csum(const pseudo_header & ph,const unsigned short *ptr,int nbytes) {//works both for big and little endian
-    if(setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &socket_buf_size, sizeof(socket_buf_size))<0)
+
-    //if(setsockopt(fd, SOL_SOCKET, SO_SNDBUFFORCE, &socket_buf_size, sizeof(socket_buf_size))<0)
+	register long sum;
+    unsigned short oddbyte;
+    register short answer;
+
+    sum=0;
+	unsigned short * tmp= (unsigned short *)&ph;
+	for(int i=0;i<6;i++)
 	{
-    	printf("set SO_SNDBUF fail\n");
+		sum+=*tmp++;
-    	exit(1);
+	}
+
+
+    while(nbytes>1) {
+        sum+=*ptr++;
+        nbytes-=2;
+    }
+    if(nbytes==1) {
+        oddbyte=0;
+        *((u_char*)&oddbyte)=*(u_char*)ptr;
+        sum+=oddbyte;
+    }
+
+    sum = (sum>>16)+(sum & 0xffff);
+    sum = sum + (sum>>16);
+    answer=(short)~sum;
+
+    return(answer);
+}
+
+int set_buf_size(int fd,int socket_buf_size,int force_socket_buf)
+{
+	if(force_socket_buf)
+	{
+		if(setsockopt(fd, SOL_SOCKET, SO_SNDBUFFORCE, &socket_buf_size, sizeof(socket_buf_size))<0)
+		{
+			mylog(log_fatal,"SO_SNDBUFFORCE fail  socket_buf_size=%d  errno=%s\n",socket_buf_size,strerror(errno));
+			myexit(1);
+		}
+		if(setsockopt(fd, SOL_SOCKET, SO_RCVBUFFORCE, &socket_buf_size, sizeof(socket_buf_size))<0)
+		{
+			mylog(log_fatal,"SO_RCVBUFFORCE fail  socket_buf_size=%d  errno=%s\n",socket_buf_size,strerror(errno));
+			myexit(1);
+		}
+	}
+	else
+	{
+		if(setsockopt(fd, SOL_SOCKET, SO_SNDBUF, &socket_buf_size, sizeof(socket_buf_size))<0)
+		{
+			mylog(log_fatal,"SO_SNDBUF fail  socket_buf_size=%d  errno=%s\n",socket_buf_size,strerror(errno));
+			myexit(1);
 		}
-    //if(setsockopt(fd, SOL_SOCKET, SO_RCVBUFFORCE, &socket_buf_size, sizeof(socket_buf_size))<0)
 		if(setsockopt(fd, SOL_SOCKET, SO_RCVBUF, &socket_buf_size, sizeof(socket_buf_size))<0)
 		{
-    	printf("set SO_RCVBUF fail\n");
+			mylog(log_fatal,"SO_RCVBUF fail  socket_buf_size=%d  errno=%s\n",socket_buf_size,strerror(errno));
-    	exit(1);
+			myexit(1);
+		}
 	}
 	return 0;
 }
+
 void myexit(int a)
 {
     if(enable_log_color)
    	 printf("%s\n",RESET);
-    clear_iptables_rule();
+   // clear_iptables_rule();
 	exit(a);
 }
 void  signal_handler(int sig)
@@ -322,10 +428,197 @@ bool larger_than_u16(uint16_t a,uint16_t b)
 
 void get_true_random_chars(char * s,int len)
 {
-	int size=read(random_number_fd,s,len);
+	int size=read(random_fd.get_fd(),s,len);
 	if(size!=len)
 	{
 		printf("get random number failed\n");
 		exit(-1);
 	}
 }
+
+int random_between(u32_t a,u32_t b)
+{
+	if(a>b)
+	{
+		mylog(log_fatal,"min >max?? %d %d\n",a ,b);
+		myexit(1);
+	}
+	if(a==b)return a;
+	else return a+get_true_random_number()%(b+1-a);
+}
+
+
+int set_timer_ms(int epollfd,int &timer_fd,u32_t timer_interval)
+{
+	int ret;
+	epoll_event ev;
+
+	itimerspec its;
+	memset(&its,0,sizeof(its));
+
+	if((timer_fd=timerfd_create(CLOCK_MONOTONIC,TFD_NONBLOCK)) < 0)
+	{
+		mylog(log_fatal,"timer_fd create error\n");
+		myexit(1);
+	}
+	its.it_interval.tv_sec=(timer_interval/1000);
+	its.it_interval.tv_nsec=(timer_interval%1000)*1000ll*1000ll;
+	its.it_value.tv_nsec=1; //imidiately
+	timerfd_settime(timer_fd,0,&its,0);
+
+
+	ev.events = EPOLLIN;
+	ev.data.fd = timer_fd;
+
+	ret=epoll_ctl(epollfd, EPOLL_CTL_ADD, timer_fd, &ev);
+	if (ret < 0) {
+		mylog(log_fatal,"epoll_ctl return %d\n", ret);
+		myexit(-1);
+	}
+	return 0;
+}
+/*
+int create_new_udp(int &new_udp_fd,int remote_address_uint32,int remote_port)
+{
+	struct sockaddr_in remote_addr_in;
+
+	socklen_t slen = sizeof(sockaddr_in);
+	memset(&remote_addr_in, 0, sizeof(remote_addr_in));
+	remote_addr_in.sin_family = AF_INET;
+	remote_addr_in.sin_port = htons(remote_port);
+	remote_addr_in.sin_addr.s_addr = remote_address_uint32;
+
+	new_udp_fd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
+	if (new_udp_fd < 0) {
+		mylog(log_warn, "create udp_fd error\n");
+		return -1;
+	}
+	setnonblocking(new_udp_fd);
+	set_buf_size(new_udp_fd);
+
+	mylog(log_debug, "created new udp_fd %d\n", new_udp_fd);
+	int ret = connect(new_udp_fd, (struct sockaddr *) &remote_addr_in, slen);
+	if (ret != 0) {
+		mylog(log_warn, "udp fd connect fail %d %s\n",ret,strerror(errno));
+		close(new_udp_fd);
+		return -1;
+	}
+	return 0;
+}*/
+void ip_port_t::from_u64(u64_t u64)
+{
+	ip=get_u64_h(u64);
+	port=get_u64_l(u64);
+}
+u64_t ip_port_t::to_u64()
+{
+	return pack_u64(ip,port);
+}
+char * ip_port_t::to_s()
+{
+	static char res[40];
+	sprintf(res,"%s:%d",my_ntoa(ip),port);
+	return res;
+}
+
+int round_up_div(int a,int b)
+{
+	return (a+b-1)/b;
+}
+
+int create_fifo(char * file)
+{
+	if(mkfifo (file, 0666)!=0)
+	{
+		if(errno==EEXIST)
+		{
+			mylog(log_warn,"warning fifo file %s exist\n",file);
+		}
+		else
+		{
+			mylog(log_fatal,"create fifo file %s failed\n",file);
+			myexit(-1);
+		}
+	}
+	int fifo_fd=open (file, O_RDWR);
+	if(fifo_fd<0)
+	{
+		mylog(log_fatal,"create fifo file %s failed\n",file);
+		myexit(-1);
+	}
+	struct stat st;
+	if (fstat(fifo_fd, &st)!=0)
+	{
+		mylog(log_fatal,"fstat failed for fifo file %s\n",file);
+		myexit(-1);
+	}
+
+	if(!S_ISFIFO(st.st_mode))
+	{
+		mylog(log_fatal,"%s is not a fifo\n",file);
+		myexit(-1);
+	}
+
+	setnonblocking(fifo_fd);
+	return fifo_fd;
+}
+
+
+int new_listen_socket(int &fd,u32_t ip,int port)
+{
+	fd =socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
+
+	int yes = 1;
+	//setsockopt(udp_fd, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(yes));
+
+	struct sockaddr_in local_me={0};
+
+	socklen_t slen = sizeof(sockaddr_in);
+	//memset(&local_me, 0, sizeof(local_me));
+	local_me.sin_family = AF_INET;
+	local_me.sin_port = htons(port);
+	local_me.sin_addr.s_addr = ip;
+
+	if (bind(fd, (struct sockaddr*) &local_me, slen) == -1) {
+		mylog(log_fatal,"socket bind error\n");
+		//perror("socket bind error");
+		myexit(1);
+	}
+	setnonblocking(fd);
+    set_buf_size(fd,socket_buf_size);
+
+    mylog(log_debug,"local_listen_fd=%d\n,",fd);
+
+	return 0;
+}
+int new_connected_socket(int &fd,u32_t ip,int port)
+{
+	char ip_port[40];
+	sprintf(ip_port,"%s:%d",my_ntoa(ip),port);
+
+	struct sockaddr_in remote_addr_in = { 0 };
+
+	socklen_t slen = sizeof(sockaddr_in);
+	//memset(&remote_addr_in, 0, sizeof(remote_addr_in));
+	remote_addr_in.sin_family = AF_INET;
+	remote_addr_in.sin_port = htons(port);
+	remote_addr_in.sin_addr.s_addr = ip;
+
+	fd = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
+	if (fd < 0) {
+		mylog(log_warn, "[%s]create udp_fd error\n", ip_port);
+		return -1;
+	}
+	setnonblocking(fd);
+	set_buf_size(fd, socket_buf_size);
+
+	mylog(log_debug, "[%s]created new udp_fd %d\n", ip_port, fd);
+	int ret = connect(fd, (struct sockaddr *) &remote_addr_in, slen);
+	if (ret != 0) {
+		mylog(log_warn, "[%s]fd connect fail\n",ip_port);
+		close(fd);
+		return -1;
+	}
+	return 0;
+}
+

common.h

@@ -7,7 +7,7 @@
 
 #ifndef COMMON_H_
 #define COMMON_H_
-#define __STDC_FORMAT_MACROS 1
+//#define __STDC_FORMAT_MACROS 1
 #include <inttypes.h>
 
 #include<stdio.h>
@@ -21,6 +21,7 @@
 #include <sys/wait.h>
 #include <sys/socket.h>    //for socket ofcourse
 #include <sys/types.h>
+#include <sys/stat.h>
 #include <stdlib.h> //for exit(0);
 #include <errno.h> //For errno - the error number
 #include <netinet/tcp.h>   //Provides declarations for tcp header
@@ -43,12 +44,11 @@
 #include <stdarg.h>
 #include <assert.h>
 #include <linux/if_packet.h>
-
+#include <linux/if_tun.h>
-
-
 
 #include<unordered_map>
 #include<unordered_set>
+#include<map>
 #include<list>
 using  namespace std;
 
@@ -59,16 +59,21 @@ typedef long long i64_t;
 typedef unsigned int u32_t;
 typedef int i32_t;
 
+typedef unsigned short u16_t;
+typedef short i16_t;
 
-const int max_data_len=1600;
+typedef u64_t my_time_t;
+
+const int max_data_len=2200;
 const int buf_len=max_data_len+200;
 
-const u32_t conv_clear_interval=200;
+
-const u32_t timer_interval=400;
+//const u32_t timer_interval=400;
-const int conv_clear_ratio=40;
+////const u32_t conv_timeout=180000;
-const int conv_clear_min=5;
+//const u32_t conv_timeout=40000;//for test
-const u32_t conv_timeout=20000;
+const u32_t conv_timeout=180000;
 const int max_conv_num=10000;
+const int max_conn_num=200;
 
 /*
 const u32_t max_handshake_conn_num=10000;
@@ -79,14 +84,14 @@ const u32_t max_ready_conn_num=1000;
 const u32_t client_handshake_timeout=5000;
 const u32_t client_retry_interval=1000;
 
-const u32_t server_handshake_timeout=10000;// this should be much longer than clients. client retry initially ,server retry passtively
+const u32_t server_handshake_timeout=10000;// this should be much longer than clients. client retry initially ,server retry passtively*/
 
-const int conv_clear_ratio=10;  //conv grabage collecter check 1/10 of all conv one time
+const int conv_clear_ratio=30;  //conv grabage collecter check 1/30 of all conv one time
-const int conn_clear_ratio=10;
+const int conn_clear_ratio=50;
-const int conv_clear_min=5;
+const int conv_clear_min=1;
 const int conn_clear_min=1;
 
-
+const u32_t conv_clear_interval=1000;
 const u32_t conn_clear_interval=1000;
 
 
@@ -97,24 +102,29 @@ const u32_t heartbeat_interval=1000;
 const u32_t timer_interval=400;//this should be smaller than heartbeat_interval and retry interval;
 
 //const uint32_t conv_timeout=120000; //120 second
-const u32_t conv_timeout=120000; //for test
+//const u32_t conv_timeout=120000; //for test
 
 const u32_t client_conn_timeout=10000;
 const u32_t client_conn_uplink_timeout=client_conn_timeout+2000;
 
 //const uint32_t server_conn_timeout=conv_timeout+60000;//this should be 60s+ longer than conv_timeout,so that conv_manager can destruct convs gradually,to avoid latency glicth
-const u32_t server_conn_timeout=conv_timeout+60000;//for test
+const u32_t server_conn_timeout=conv_timeout+20000;//for test
-*/
+
 
 extern int about_to_exit;
 
 enum raw_mode_t{mode_faketcp=0,mode_udp,mode_icmp,mode_end};
 extern raw_mode_t raw_mode;
 enum program_mode_t {unset_mode=0,client_mode,server_mode};
-extern program_mode_t program_mode;
+extern program_mode_t client_or_server;
 extern unordered_map<int, const char*> raw_mode_tostring ;
+
+enum working_mode_t {unset_working_mode=0,tunnel_mode,tun_dev_mode};
+extern working_mode_t working_mode;
+
 extern int socket_buf_size;
 
+
 typedef u32_t id_t;
 
 typedef u64_t iv_t;
@@ -123,6 +133,58 @@ typedef u64_t padding_t;
 
 typedef u64_t anti_replay_seq_t;
 
+typedef u64_t fd64_t;
+
+//enum dest_type{none=0,type_fd64_ip_port,type_fd64,type_fd64_ip_port_conv,type_fd64_conv/*,type_fd*/};
+enum dest_type{none=0,type_fd64_ip_port,type_fd64,type_fd,type_write_fd,type_fd_ip_port/*,type_fd*/};
+
+struct ip_port_t
+{
+	u32_t ip;
+	int port;
+	void from_u64(u64_t u64);
+	u64_t to_u64();
+	char * to_s();
+};
+
+struct fd64_ip_port_t
+{
+	fd64_t fd64;
+	ip_port_t ip_port;
+};
+struct fd_ip_port_t
+{
+	int fd;
+	ip_port_t ip_port;
+};
+union inner_t
+{
+	fd64_t fd64;
+	int fd;
+	fd64_ip_port_t fd64_ip_port;
+	fd_ip_port_t fd_ip_port;
+};
+struct dest_t
+{
+	dest_type type;
+	inner_t inner;
+	u32_t conv;
+	int cook=0;
+};
+
+struct fd_info_t
+{
+	ip_port_t ip_port;
+};
+
+struct pseudo_header {
+    u_int32_t source_address;
+    u_int32_t dest_address;
+    u_int8_t placeholder;
+    u_int8_t protocol;
+    u_int16_t tcp_length;
+};
+
 u64_t get_current_time();
 u64_t get_current_time_us();
 u64_t pack_u64(u32_t a,u32_t b);
@@ -131,6 +193,15 @@ u32_t get_u64_h(u64_t a);
 
 u32_t get_u64_l(u64_t a);
 
+void write_u16(char *,u16_t a);
+u16_t read_u16(char *);
+
+void write_u32(char *,u32_t a);
+u32_t read_u32(char *);
+
+void write_u64(char *,u64_t a);
+u64_t read_uu64(char *);
+
 char * my_ntoa(u32_t ip);
 
 void myexit(int a);
@@ -143,9 +214,10 @@ u64_t hton64(u64_t a);
 bool larger_than_u16(uint16_t a,uint16_t b);
 bool larger_than_u32(u32_t a,u32_t b);
 void setnonblocking(int sock);
-int set_buf_size(int fd,int size=socket_buf_size);
+int set_buf_size(int fd,int socket_buf_size,int force_socket_buf=0);
 
 unsigned short csum(const unsigned short *ptr,int nbytes);
+unsigned short tcp_csum(const pseudo_header & ph,const unsigned short *ptr,int nbytes);
 
 void  signal_handler(int sig);
 int numbers_to_char(id_t id1,id_t id2,id_t id3,char * &data,int &len);
@@ -157,5 +229,19 @@ int add_iptables_rule(char *);
 
 int clear_iptables_rule();
 void get_true_random_chars(char * s,int len);
+int random_between(u32_t a,u32_t b);
+
+int set_timer_ms(int epollfd,int &timer_fd,u32_t timer_interval);
+
+int round_up_div(int a,int b);
+
+int create_fifo(char * file);
+/*
+int create_new_udp(int &new_udp_fd,int remote_address_uint32,int remote_port);
+*/
+
+int new_listen_socket(int &fd,u32_t ip,int port);
+
+int new_connected_socket(int &fd,u32_t ip,int port);
 
 #endif /* COMMON_H_ */

connection.cpp

@@ -0,0 +1,293 @@
+/*
+ * connection.cpp
+ *
+ *  Created on: Sep 23, 2017
+ *      Author: root
+ */
+
+#include "connection.h"
+
+const int disable_conv_clear=0;//a udp connection in the multiplexer is called conversation in this program,conv for short.
+
+const int disable_conn_clear=0;//a raw connection is called conn.
+
+int report_interval=0;
+
+void server_clear_function(u64_t u64)//used in conv_manager in server mode.for server we have to use one udp fd for one conv(udp connection),
+//so we have to close the fd when conv expires
+{
+	fd64_t fd64=u64;
+	assert(fd_manager.exist(fd64));
+	fd_manager.fd64_close(fd64);
+}
+
+////////////////////////////////////////////////////////////////////
+
+conv_manager_t::conv_manager_t()
+{
+	clear_it=conv_last_active_time.begin();
+	long long last_clear_time=0;
+	reserve();
+}
+conv_manager_t::~conv_manager_t()
+{
+	clear();
+}
+int conv_manager_t::get_size()
+{
+	return conv_to_u64.size();
+}
+void conv_manager_t::reserve()
+{
+	u64_to_conv.reserve(10007);
+	conv_to_u64.reserve(10007);
+	conv_last_active_time.reserve(10007);
+}
+void conv_manager_t::clear()
+{
+	//if(disable_conv_clear) return ;//////what was the purpose of this code?
+
+	if(client_or_server==server_mode)
+	{
+		for(auto it=conv_to_u64.begin();it!=conv_to_u64.end();it++)
+		{
+			//int fd=int((it->second<<32u)>>32u);
+			server_clear_function(  it->second);
+		}
+	}
+	u64_to_conv.clear();
+	conv_to_u64.clear();
+	conv_last_active_time.clear();
+
+	clear_it=conv_last_active_time.begin();
+
+}
+u32_t conv_manager_t::get_new_conv()
+{
+	u32_t conv=get_true_random_number_nz();
+	while(conv_to_u64.find(conv)!=conv_to_u64.end())
+	{
+		conv=get_true_random_number_nz();
+	}
+	return conv;
+}
+int conv_manager_t::is_conv_used(u32_t conv)
+{
+	return conv_to_u64.find(conv)!=conv_to_u64.end();
+}
+int conv_manager_t::is_u64_used(u64_t u64)
+{
+	return u64_to_conv.find(u64)!=u64_to_conv.end();
+}
+u32_t conv_manager_t::find_conv_by_u64(u64_t u64)
+{
+	assert(is_u64_used(u64));
+	return u64_to_conv[u64];
+}
+u64_t conv_manager_t::find_u64_by_conv(u32_t conv)
+{
+	assert(is_conv_used(conv));
+	return conv_to_u64[conv];
+}
+int conv_manager_t::update_active_time(u32_t conv)
+{
+	assert(is_conv_used(conv));
+	return conv_last_active_time[conv]=get_current_time();
+}
+int conv_manager_t::insert_conv(u32_t conv,u64_t u64)//////todo add capacity ///done at upper level
+{
+	assert(!is_conv_used(conv));
+	int bucket_size_before=conv_last_active_time.bucket_count();
+	u64_to_conv[u64]=conv;
+	conv_to_u64[conv]=u64;
+	conv_last_active_time[conv]=get_current_time();
+	int bucket_size_after=conv_last_active_time.bucket_count();
+	if(bucket_size_after!=bucket_size_before)
+		clear_it=conv_last_active_time.begin();
+	return 0;
+}
+int conv_manager_t::erase_conv(u32_t conv)
+{
+	//if(disable_conv_clear) return 0;
+	assert(conv_last_active_time.find(conv)!=conv_last_active_time.end());
+	u64_t u64=conv_to_u64[conv];
+	if(client_or_server==server_mode)
+	{
+		server_clear_function(u64);
+	}
+	assert(conv_to_u64.find(conv)!=conv_to_u64.end());
+	conv_to_u64.erase(conv);
+	u64_to_conv.erase(u64);
+	conv_last_active_time.erase(conv);
+	return 0;
+}
+int conv_manager_t::clear_inactive(char * ip_port)
+{
+	if(get_current_time()-last_clear_time>conv_clear_interval)
+	{
+		last_clear_time=get_current_time();
+		return clear_inactive0(ip_port);
+	}
+	return 0;
+}
+int conv_manager_t::clear_inactive0(char * ip_port)
+{
+	if(disable_conv_clear) return 0;
+
+	//map<uint32_t,uint64_t>::iterator it;
+	int cnt=0;
+	auto it=clear_it;
+	int size=conv_last_active_time.size();
+	int num_to_clean=size/conv_clear_ratio+conv_clear_min;   //clear 1/10 each time,to avoid latency glitch
+
+	num_to_clean=min(num_to_clean,size);
+
+	u64_t current_time=get_current_time();
+	for(;;)
+	{
+		if(cnt>=num_to_clean) break;
+		if(conv_last_active_time.begin()==conv_last_active_time.end()) break;
+
+		if(it==conv_last_active_time.end())
+		{
+			it=conv_last_active_time.begin();
+		}
+
+		if( current_time -it->second  >conv_timeout )
+		{
+			//mylog(log_info,"inactive conv %u cleared \n",it->first);
+			//auto old_it=it;
+			//it++;
+			u32_t conv= it->first;
+			it++;
+			erase_conv(conv);
+			if(ip_port==0)
+			{
+				mylog(log_info,"conv %x cleared\n",conv);
+			}
+			else
+			{
+				mylog(log_info,"[%s]conv %x cleared\n",ip_port,conv);
+			}
+		}
+		else
+		{
+			it++;
+		}
+		cnt++;
+	}
+	clear_it=it;
+	return 0;
+}
+
+
+////////////////////////////////////////////////////////////////////
+
+
+conn_manager_t::conn_manager_t()
+{
+	//ready_num=0;
+	mp.reserve(10007);
+	//fd64_mp.reserve(100007);
+	clear_it=mp.begin();
+	last_clear_time=0;
+}
+int conn_manager_t::exist(ip_port_t ip_port)
+{
+	u64_t u64=ip_port.to_u64();
+	if(mp.find(u64)!=mp.end())
+	{
+	 return 1;
+	}
+	return 0;
+}
+ conn_info_t *& conn_manager_t::find_p(ip_port_t ip_port) //todo capacity   ///done at upper level
+ //be aware,the adress may change after rehash
+ {
+	 assert(exist(ip_port));
+	 u64_t u64=ip_port.to_u64();
+	 return mp[u64];
+ }
+ conn_info_t & conn_manager_t::find(ip_port_t ip_port)  //be aware,the adress may change after rehash
+ {
+	 assert(exist(ip_port));
+	 u64_t u64=ip_port.to_u64();
+	 return *mp[u64];
+ }
+ int conn_manager_t::insert(ip_port_t ip_port)
+ {
+	assert(!exist(ip_port));
+	int bucket_size_before=mp.bucket_count();
+	mp[ip_port.to_u64()]=new conn_info_t;
+	int bucket_size_after=mp.bucket_count();
+	if(bucket_size_after!=bucket_size_before)
+		clear_it=mp.begin();
+	return 0;
+ }
+ int conn_manager_t::erase(unordered_map<u64_t,conn_info_t*>::iterator erase_it)
+ {
+	 ////////todo  close and erase timer_fd ,check fd64 empty   ///dont need
+		delete(erase_it->second);
+		mp.erase(erase_it->first);
+		return 0;
+ }
+int conn_manager_t::clear_inactive()
+{
+	if(get_current_time()-last_clear_time>conn_clear_interval)
+	{
+		last_clear_time=get_current_time();
+		return clear_inactive0();
+	}
+	return 0;
+}
+int conn_manager_t::clear_inactive0()
+{
+//mylog(log_info,"called\n");
+	 unordered_map<u64_t,conn_info_t*>::iterator it;
+	 unordered_map<u64_t,conn_info_t*>::iterator old_it;
+
+	if(disable_conn_clear) return 0;
+
+	//map<uint32_t,uint64_t>::iterator it;
+	int cnt=0;
+	it=clear_it;//TODO,write it back
+	int size=mp.size();
+	int num_to_clean=size/conn_clear_ratio+conn_clear_min;   //clear 1/10 each time,to avoid latency glitch
+
+	//mylog(log_trace,"mp.size() %d\n", size);
+
+	num_to_clean=min(num_to_clean,(int)mp.size());
+	u64_t current_time=get_current_time();
+
+	//mylog(log_info,"here size=%d\n",(int)mp.size());
+	for(;;)
+	{
+		if(cnt>=num_to_clean) break;
+		if(mp.begin()==mp.end()) break;
+		if(it==mp.end())
+		{
+			it=mp.begin();
+		}
+
+		if(it->second->conv_manager.get_size() >0)
+		{
+			//mylog(log_info,"[%s:%d]size %d \n",my_ntoa(get_u64_h(it->first)),get_u64_l(it->first),(int)it->second->conv_manager.get_size());
+			it++;
+		}
+		else if(current_time<it->second->last_active_time+server_conn_timeout)
+		{
+			it++;
+		}
+		else
+		{
+			mylog(log_info,"[%s:%d]inactive conn cleared \n",my_ntoa(get_u64_h(it->first)),get_u64_l(it->first));
+			old_it=it;
+			it++;
+			erase(old_it);
+		}
+		cnt++;
+	}
+	clear_it=it;
+	return 0;
+}
+

connection.h

@@ -0,0 +1,152 @@
+/*
+ * connection.h
+ *
+ *  Created on: Sep 23, 2017
+ *      Author: root
+ */
+
+#ifndef CONNECTION_H_
+#define CONNECTION_H_
+
+extern int disable_anti_replay;
+
+#include "connection.h"
+#include "common.h"
+#include "log.h"
+#include "delay_manager.h"
+#include "fd_manager.h"
+#include "fec_manager.h"
+
+extern int report_interval;
+
+struct conv_manager_t  // manage the udp connections
+{
+	//typedef hash_map map;
+	unordered_map<u64_t,u32_t> u64_to_conv;  //conv and u64 are both supposed to be uniq
+	unordered_map<u32_t,u64_t> conv_to_u64;
+	unordered_map<u32_t,u64_t> conv_last_active_time;
+
+	unordered_map<u32_t,u64_t>::iterator clear_it;
+
+	//void (*clear_function)(uint64_t u64) ;
+
+	long long last_clear_time;
+
+	conv_manager_t();
+	conv_manager_t(const conv_manager_t &b)
+	{
+		assert(0==1);
+	}
+	~conv_manager_t();
+	int get_size();
+	void reserve();
+	void clear();
+	u32_t get_new_conv();
+	int is_conv_used(u32_t conv);
+	int is_u64_used(u64_t u64);
+	u32_t find_conv_by_u64(u64_t u64);
+	u64_t find_u64_by_conv(u32_t conv);
+	int update_active_time(u32_t conv);
+	int insert_conv(u32_t conv,u64_t u64);
+	int erase_conv(u32_t conv);
+	int clear_inactive(char * ip_port=0);
+	int clear_inactive0(char * ip_port);
+};
+
+
+struct inner_stat_t
+{
+	u64_t input_packet_num;
+	u64_t input_packet_size;
+	u64_t output_packet_num;
+	u64_t output_packet_size;
+};
+struct stat_t
+{
+	u64_t last_report_time;
+	inner_stat_t normal_to_fec;
+	inner_stat_t fec_to_normal;
+	stat_t()
+	{
+		memset(this,0,sizeof(stat_t));
+	}
+	void report_as_client()
+	{
+		if(report_interval!=0 &&get_current_time()-last_report_time>u64_t(report_interval)*1000)
+		{
+			last_report_time=get_current_time();
+			inner_stat_t &a=normal_to_fec;
+			inner_stat_t &b=fec_to_normal;
+			mylog(log_info,"[report]client-->server:(original:%llu pkt;%llu byte) (fec:%llu pkt,%llu byte)  server-->client:(original:%llu pkt;%llu byte) (fec:%llu pkt;%llu byte)\n",
+					a.input_packet_num,a.input_packet_size,a.output_packet_num,a.output_packet_size,
+					b.output_packet_num,b.output_packet_size,b.input_packet_num,b.input_packet_size
+					);
+		}
+	}
+	void report_as_server(ip_port_t &ip_port)
+	{
+		if(report_interval!=0 &&get_current_time()-last_report_time>u64_t(report_interval)*1000)
+		{
+			last_report_time=get_current_time();
+			inner_stat_t &a=fec_to_normal;
+			inner_stat_t &b=normal_to_fec;
+			mylog(log_info,"[report][%s]client-->server:(original:%llu pkt;%llu byte) (fec:%llu pkt;%llu byte)  server-->client:(original:%llu pkt;%llu byte) (fec:%llu pkt;%llu byte)\n",
+					ip_port.to_s(),
+					a.output_packet_num,a.output_packet_size,a.input_packet_num,a.input_packet_size,
+					b.input_packet_num,b.input_packet_size,b.output_packet_num,b.output_packet_size
+					);
+		}
+	}
+};
+
+
+struct conn_info_t     //stores info for a raw connection.for client ,there is only one connection,for server there can be thousand of connection since server can
+//handle multiple clients
+{
+	conv_manager_t conv_manager;
+	fec_encode_manager_t fec_encode_manager;
+	fec_decode_manager_t fec_decode_manager;
+	my_timer_t timer;
+	//ip_port_t ip_port;
+	u64_t last_active_time;
+	stat_t stat;
+	conn_info_t()
+	{
+	}
+	void update_active_time()
+	{
+		last_active_time=get_current_time();
+	}
+	conn_info_t(const conn_info_t &b)
+	{
+		assert(0==1);
+	}
+};
+
+struct conn_manager_t  //manager for connections. for client,we dont need conn_manager since there is only one connection.for server we use one conn_manager for all connections
+{
+
+	unordered_map<u64_t,conn_info_t*> mp;//<ip,port> to conn_info_t;
+	unordered_map<u64_t,conn_info_t*>::iterator clear_it;
+	long long last_clear_time;
+
+	conn_manager_t();
+	conn_manager_t(const conn_info_t &b)
+	{
+		assert(0==1);
+	}
+	int exist(ip_port_t);
+	conn_info_t *& find_p(ip_port_t);  //be aware,the adress may change after rehash
+	conn_info_t & find(ip_port_t) ; //be aware,the adress may change after rehash
+	int insert(ip_port_t);
+
+	int erase(unordered_map<u64_t,conn_info_t*>::iterator erase_it);
+	int clear_inactive();
+	int clear_inactive0();
+
+};
+
+extern conn_manager_t conn_manager;
+
+
+#endif /* CONNECTION_H_ */

delay_manager.cpp

@@ -0,0 +1,123 @@
+/*
+ * delay_manager.cpp
+ *
+ *  Created on: Sep 15, 2017
+ *      Author: root
+ */
+#include "delay_manager.h"
+#include "log.h"
+#include "packet.h"
+
+int delay_data_t::handle()
+{
+	return my_send(dest,data,len)>=0;
+}
+
+
+delay_manager_t::delay_manager_t()
+{
+	capacity=0;
+
+	if ((timer_fd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK)) < 0)
+	{
+		mylog(log_fatal,"timer_fd create error");
+		myexit(1);
+	}
+
+	itimerspec zero_its;
+	memset(&zero_its, 0, sizeof(zero_its));
+
+	timerfd_settime(timer_fd, TFD_TIMER_ABSTIME, &zero_its, 0);
+
+}
+delay_manager_t::~delay_manager_t()
+{
+	//TODO ,we currently dont need to deconstruct it
+}
+
+int delay_manager_t::get_timer_fd()
+{
+	return timer_fd;
+}
+//int add(my_time_t delay,const dest_t &dest,const char *data,int len);
+int delay_manager_t::add(my_time_t delay,const dest_t &dest,char *data,int len)
+{
+	delay_data_t delay_data;
+	delay_data.dest=dest;
+	//delay_data.data=data;
+	delay_data.len=len;
+
+	if(capacity!=0&&int(delay_mp.size()) >=capacity)
+	{
+		mylog(log_warn,"max pending packet reached,ignored\n");
+		return -1;
+	}
+	if(delay==0)
+	{
+		static char buf[buf_len];
+		delay_data.data=buf;
+		memcpy(buf,data,len);
+		int ret=delay_data.handle();
+		if (ret != 0) {
+			mylog(log_trace, "handle() return %d\n", ret);
+		}
+		return 0;
+	}
+
+	delay_data_t tmp=delay_data;
+	tmp.data=(char *)malloc(delay_data.len+100);
+    if(!tmp.data)
+    {
+        mylog(log_warn, "malloc() returned null in delay_manager_t::add()");
+        return -1;
+    }
+	memcpy(tmp.data,data,delay_data.len);
+
+	my_time_t tmp_time=get_current_time_us();
+	tmp_time+=delay;
+
+	delay_mp.insert(make_pair(tmp_time,tmp));
+
+	return 0;
+}
+
+int delay_manager_t::check()
+{
+	if(!delay_mp.empty())
+	{
+		my_time_t current_time;
+
+		multimap<my_time_t,delay_data_t>::iterator it;
+		while(1)
+		{
+			int ret=0;
+			it=delay_mp.begin();
+			if(it==delay_mp.end()) break;
+
+			current_time=get_current_time_us();
+			if(it->first <= current_time)
+			{
+				ret=it->second.handle();
+				if (ret != 0) {
+					mylog(log_trace, "handle() return %d\n", ret);
+				}
+				free(it->second.data);
+				delay_mp.erase(it);
+			}
+			else
+			{
+				break;
+			}
+
+		}
+		if(!delay_mp.empty())
+		{
+			itimerspec its;
+			memset(&its.it_interval,0,sizeof(its.it_interval));
+			its.it_value.tv_sec=delay_mp.begin()->first/1000000llu;
+			its.it_value.tv_nsec=(delay_mp.begin()->first%1000000llu)*1000llu;
+			timerfd_settime(timer_fd,TFD_TIMER_ABSTIME,&its,0);
+		}
+	}
+	return 0;
+}

delay_manager.h

@@ -0,0 +1,131 @@
+/*
+ * delay_manager.h
+ *
+ *  Created on: Sep 15, 2017
+ *      Author: root
+ */
+
+#ifndef DELAY_MANAGER_H_
+#define DELAY_MANAGER_H_
+
+#include "common.h"
+#include "packet.h"
+#include "log.h"
+
+//enum delay_type_t {none=0,enum_sendto_u64,enum_send_fd,client_to_local,client_to_remote,server_to_local,server_to_remote};
+
+/*
+struct fd_ip_port_t
+{
+	int fd;
+	u32_t ip;
+	u32_t port;
+};
+union dest_t
+{
+	fd_ip_port_t fd_ip_port;
+	int fd;
+	u64_t u64;
+};
+*/
+
+struct my_timer_t
+{
+	int timer_fd;
+	fd64_t timer_fd64;
+	my_timer_t()
+	{
+		if ((timer_fd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK)) < 0)
+		{
+			mylog(log_fatal,"timer_fd create error");
+			myexit(1);
+		}
+		timer_fd64=fd_manager.create(timer_fd);
+	}
+	my_timer_t(const my_timer_t &b)
+	{
+		assert(0==1);
+	}
+	~my_timer_t()
+	{
+		fd_manager.fd64_close(timer_fd64);
+	}
+	int add_fd_to_epoll(int epoll_fd)
+	{
+		epoll_event ev;;
+		ev.events = EPOLLIN;
+		ev.data.u64 = timer_fd;
+		int ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, timer_fd, &ev);
+		if (ret!= 0) {
+			mylog(log_fatal,"add delay_manager.get_timer_fd() error\n");
+			myexit(-1);
+		}
+		return 0;
+	}
+	int add_fd64_to_epoll(int epoll_fd)
+	{
+		epoll_event ev;;
+		ev.events = EPOLLIN;
+		ev.data.u64 = timer_fd64;
+		int ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, timer_fd, &ev);
+		if (ret!= 0) {
+			mylog(log_fatal,"add delay_manager.get_timer_fd() error\n");
+			myexit(-1);
+		}
+		return 0;
+	}
+	int get_timer_fd()
+	{
+		return timer_fd;
+	}
+	fd64_t get_timer_fd64()
+	{
+		return timer_fd64;
+	}
+	int set_timer_repeat_us(my_time_t my_time)
+	{
+		itimerspec its;
+		memset(&its,0,sizeof(its));
+		its.it_interval.tv_sec=my_time/1000000llu;
+		its.it_interval.tv_nsec=my_time%1000000llu*1000llu;
+		its.it_value.tv_nsec=1; //imidiately
+		timerfd_settime(timer_fd,0,&its,0);
+		return 0;
+	}
+	int set_timer_abs_us(my_time_t my_time)
+	{
+		itimerspec its;
+		memset(&its,0,sizeof(its));
+		its.it_value.tv_sec=my_time/1000000llu;
+		its.it_value.tv_nsec=my_time%1000000llu*1000llu;
+		timerfd_settime(timer_fd,TFD_TIMER_ABSTIME,&its,0);
+		return 0;
+	}
+};
+struct delay_data_t
+{
+	dest_t dest;
+	//int left_time;//
+	char * data;
+	int len;
+	int handle();
+};
+
+struct delay_manager_t
+{
+	int timer_fd;
+	int capacity;
+	multimap<my_time_t,delay_data_t> delay_mp;  //unit us,1 us=0.001ms
+	delay_manager_t();
+	delay_manager_t(delay_manager_t &b)
+	{
+		assert(0==1);
+	}
+	int set_capacity(int a){capacity=a;return 0;}
+	~delay_manager_t();
+	int get_timer_fd();
+	int check();
+	int add(my_time_t delay,const dest_t &dest,char *data,int len);
+};
+
+#endif /* DELAY_MANAGER_H_ */

doc/README.zh-cn.md

@@ -0,0 +1,285 @@
+# UDPspeeder
+![image0](/images/cn/speedercn.PNG)
+
+双边网络加速工具，软件本身的功能是加速UDP；不过，配合vpn可以加速全流量(包括TCP/UDP/ICMP)。通过合理配置，可以加速游戏，降低游戏的丢包和延迟；也可以加速下载和看视频这种大流量的应用。用1.5倍的流量，就可以把10%的丢包率降低到万分之一以下。跟 kcptun/finalspeed/BBR 等现有方案比，主要优势是可以加速 UDP 和 ICMP，现有方案几乎都只能加速 TCP。
+
+我自己稳定用了几个月，用来加速美服的Brawl Stars和亚服的Mobile Legend，效果不错，加速前卡得几乎没法玩，加速后就没怎么卡过了。用来看视频也基本满速。
+
+最新的版本是v2版，在v1版的基础上增加了FEC功能，更省流量。如果你用的是v1版（路由器固件里自带的集成版很可能是v1版的），请看[v1版主页](/doc/README.zh-cn.v1.md)
+
+配合vpn加速全流量的原理图(已测试支持VPN的有OpenVPN、L2TP、$\*\*\*VPN):
+
+![image0](/images/Capture2.PNG)
+
+对于某些运营商，UDPspeeder跟udp2raw配合可以达到更好的速度，udp2raw负责把UDP伪装成TCP，来绕过运营商的UDP限速。
+
+udp2raw的repo:
+
+https://github.com/wangyu-/udp2raw-tunnel
+
+如果你嫌UDPspeeder+OpenVPN麻烦，你可以尝试tinyFecVPN，一个集成了UDPspeeder功能的VPN：
+
+tinyFecVPN的repo:
+
+https://github.com/wangyu-/tinyFecVPN
+
+
+#### 效果
+![image0](/images/cn/ping_compare_cn.PNG)
+
+![image0](/images/cn/scp_compare.PNG)
+#### 原理简介
+主要原理是通过冗余数据来对抗网络的丢包，发送冗余数据的方式支持FEC(Forward Error Correction)和多倍发包，其中FEC算法是Reed-Solomon。
+
+FEC方式的原理图:
+
+![image0](/images/en/fec.PNG)
+
+#### 其他功能
+对包的内容和长度做随机化（可以理解为混淆），从抓包看不出你发送了冗余数据，不用担心vps被封。
+
+在多个冗余包之间引入延迟（时间可配）来对抗突发性的丢包，避开中间路由器因为瞬时buffer长度过长而连续丢掉所有副本。
+
+模拟一定的延迟抖动（时间可配），这样上层应用计算出来的RTT方差会更大，以等待后续冗余包的到达，不至于发生在冗余包到达之前就触发重传的尴尬。
+
+输出UDP收发情况报告，可以看出丢包率。
+
+模拟丢包，模拟延迟，模拟jitter。便于通过实验找出应用卡顿的原因。
+
+client支持多个udp连接，server也支持多个client
+
+# 简明操作说明
+
+### 环境要求
+Linux主机，可以是桌面版，可以是android手机/平板，可以是openwrt路由器，也可以是树莓派。Release中提供了`amd64`、`x86`、`arm`、`mips_be`、`mips_le`的预编译binary.
+
+对于windows和mac用户，在虚拟机中可以稳定使用（speeder跑在Linux里，其他应用照常跑在window里，桥接模式测试可用）。可以使用[这个](https://github.com/wangyu-/udp2raw-tunnel/releases/download/20171108.0/lede-17.01.2-x86_virtual_machine_image.zip)虚拟机镜像，大小只有7.5mb，免去在虚拟机里装系统的麻烦；虚拟机自带ssh server，可以scp拷贝文件，可以ssh进去，可以复制粘贴，root密码123456。
+
+android版需要通过terminal运行。
+
+##### 注意
+在使用虚拟机时，建议手动指定桥接到哪个网卡，不要设置成自动。否则可能会桥接到错误的网卡。
+
+### 安装
+下载编译好的二进制文件，解压到本地和服务器的任意目录。
+
+https://github.com/wangyu-/UDPspeeder/releases
+
+### 运行
+假设你有一个server，ip为44.55.66.77，有一个服务监听在udp 7777端口。 假设你需要加速本地到44.55.66.77:7777的流量。
+```
+在server端运行:
+./speederv2 -s -l0.0.0.0:4096 -r127.0.0.1:7777  -f20:10 -k "passwd" --mode 0
+
+在client端运行:
+./speederv2 -c -l0.0.0.0:3333 -r44.55.66.77:4096 -f20:10 -k "passwd" --mode 0
+```
+
+现在client和server之间建立起了tunnel。想要连接44.55.66.77:7777，只需要连接 127.0.0.1:3333。来回的所有的udp流量会被加速。
+
+##### 备注:
+
+`-f20:10`表示对每20个原始数据发送10个冗余包。`-f20:10` 和`-f 20:10`都是可以的，空格可以省略，对于所有的单字节option都是如此。对于双字节option，例如后面会提到的`--mode 0`，空格不可以省略。
+
+`-k` 指定一个字符串，开启简单的异或加密
+
+推荐使用`--mode 0`选项，否则你可能需要考虑MTU问题。
+
+##### 注意
+
+这里推荐的参数是给日常/非游戏情况下使用的；玩游戏请用 [使用经验](https://github.com/wangyu-/UDPspeeder/wiki/使用经验) 里推荐的参数。
+
+# 进阶操作说明
+
+### 命令选项
+```
+UDPspeeder V2
+git version: 6f55b8a2fc    build date: Nov 19 2017 06:11:23
+repository: https://github.com/wangyu-/UDPspeeder
+
+usage:
+    run as client: ./this_program -c -l local_listen_ip:local_port -r server_ip:server_port  [options]
+    run as server: ./this_program -s -l server_listen_ip:server_port -r remote_ip:remote_port  [options]
+
+common options, must be same on both sides:
+    -k,--key              <string>        key for simple xor encryption. if not set, xor is disabled
+main options:
+    -f,--fec              x:y             forward error correction, send y redundant packets for every x packets
+    --timeout             <number>        how long could a packet be held in queue before doing fec, unit: ms, default: 8ms
+    --report              <number>        turn on send/recv report, and set a period for reporting, unit: s
+advanced options:
+    --mode                <number>        fec-mode,available values: 0,1; mode 0(default) costs less bandwidth,no mtu problem.
+                                          mode 1 usually introduces less latency, but you have to care about mtu.
+    --mtu                 <number>        mtu. for mode 0, the program will split packet to segment smaller than mtu value.
+                                          for mode 1, no packet will be split, the program just check if the mtu is exceed.
+                                          default value: 1250. you typically shouldnt change this value.
+    -q,--queue-len        <number>        fec queue len, only for mode 0, fec will be performed immediately after queue is full.
+                                          default value: 200. 
+    -j,--jitter           <number>        simulated jitter. randomly delay first packet for 0~<number> ms, default value: 0.
+                                          do not use if you dont know what it means.
+    -i,--interval         <number>        scatter each fec group to a interval of <number> ms, to protect burst packet loss.
+                                          default value: 0. do not use if you dont know what it means.
+    --random-drop         <number>        simulate packet loss, unit: 0.01%. default value: 0.
+    --disable-obscure     <number>        disable obscure, to save a bit bandwidth and cpu.
+developer options:
+    --fifo                <string>        use a fifo(named pipe) for sending commands to the running program, so that you
+                                          can change fec encode parameters dynamically, check readme.md in repository for
+                                          supported commands.
+    -j ,--jitter          jmin:jmax       similiar to -j above, but create jitter randomly between jmin and jmax
+    -i,--interval         imin:imax       similiar to -i above, but scatter randomly between imin and imax
+    --decode-buf          <number>        size of buffer of fec decoder,u nit: packet, default: 2000
+    --fix-latency         <number>        try to stabilize latency, only for mode 0
+    --delay-capacity      <number>        max number of delayed packets
+    --disable-fec         <number>        completely disable fec, turn the program into a normal udp tunnel
+    --sock-buf            <number>        buf size for socket, >=10 and <=10240, unit: kbyte, default: 1024
+log and help options:
+    --log-level           <number>        0: never    1: fatal   2: error   3: warn 
+                                          4: info (default)      5: debug   6: trace
+    --log-position                        enable file name, function name, line number in log
+    --disable-color                       disable log color
+    -h,--help                             print this help message
+
+
+
+```
+### 包发送选项，两端设置可以不同。 只影响本地包发送。
+##### `-f` 选项
+设置fec参数，影响数据的冗余度。
+##### `--timeout` 选项
+指定fec编码器在编码时候最多可以引入多大的延迟。越高fec越有效率，调低可以降低延迟，但是会牺牲效率。
+#####  `--mode` 选项 和 `--mtu`选项
+
+简单来说`--mode 0`更省流量，没有mtu问题；`--mode 1`可以稍微降低一点延迟，需要考虑mtu；另外还有个`--mode 0 -q1`模式，多倍发包专用，没有延迟，也没有mtu问题，适合游戏，但是最耗流量。
+
+具体见，https://github.com/wangyu-/UDPspeeder/wiki/mode和mtu选项
+
+对于新手，建议不要纠结这些参数的具体含义，就用我在`使用经验`里推荐的设置，不要乱改参数，尤其是不要改`--mtu`。
+
+##### `--report`  选项
+数据发送和接受报告。开启后可以根据此数据推测出包速和丢包率等特征。`--report 10`表示每10秒生成一次报告。
+
+##### `-i` 选项
+指定一个时间窗口，长度为n毫秒。同一个fec分组的数据在发送时候会被均匀分散到这n毫秒中。可以对抗突发性的丢包。默认值是0，因为这个功能需要用到时钟，在某些虚拟机里时钟不稳定，可能会导致个别包出现非常大的延迟，所以默认关掉了。这个功能很有用，默认参数效果不理想时可以尝试打开，比如用`-i 10`。这个选项的跟通信原理上常说的`交错fec` `交织fec`的原理是差不多的。
+
+##### `-j` 选项
+为原始数据的发送，增加一个延迟抖动值。这样上层应用计算出来的RTT方差会更大，以等待后续冗余包的到达，不至于发生在冗余包到达之前就触发重传的尴尬。配合-t选项使用。正常情况下跨国网络本身的延迟抖动就很大，可以不用设-j。这个功能也需要时钟，默认关掉了，不过一般情况应该不需要这个功能。
+
+-j选项不但可以模拟延迟抖动，也可以模拟延迟。
+
+##### `--random-drop` 选项
+随机丢包。模拟高丢包的网络环境时使用。 `--random-drop`和`-j`选项一起用，可以模拟高延迟（或者高延迟抖动）高丢包的网络，可用于测试FEC参数在各种网络环境下的表现。
+
+##### `-q` 选项
+仅对mode 0模式有用。设置fec编码器的最大队列长度。 比如`-q5`的意思是，在编码器积攒了5个数据包后，就立即发送。合理使用可以改善延迟。在下文的`使用经验`里有提到用`--mode 0 -q1` 来多倍发包。 
+
+`-q`和 `--timeout`的作用类似。`-q`决定fec编码器积攒了多少个数据包之后，立即发送。`--timeout`决定编码器收到第一个数据包以后，最多延迟多少毫秒后发送。
+
+默认值是200，也就是尽可能多得积攒数据。 
+
+建议不要自己调整这个参数，除非是用我在`使用经验`里推荐给你的形式。
+
+#### `--fifo` option
+用fifo(命名管道)向运行中的程序发送command。例如`--fifo fifo.file`，可用的command有：
+```
+echo fec 19:9 > fifo.file
+echo mtu 1100 > fifo.file
+echo timeout 5 > fifo.file
+echo queue-len 100 > fifo.file
+echo mode 0 > fifo.file
+```
+可以动态改变fec编码器参数。可以从程序的log里看到command是否发送成功。
+
+### 以下设置两端必须相同。 
+
+##### `-k`选项
+指定一个字符串，server/client间所有收发的包都会被异或，改变协议特征，防止UDPspeeder的协议被运营商针对。
+
+##### `--disable-obscure`
+UDPspeeder默认情况下会对每个发出的数据包随机填充和异或一些字节(4~32字节)，这样通过抓包难以发现你发了冗余数据，防止VPS被封。这个功能只是为了小心谨慎，即使你关掉这个功能，基本上也没问题，关掉可以省一些带宽和CPU。`--disable-obscure`可以关掉这个功能。
+
+# 使用经验
+
+https://github.com/wangyu-/UDPspeeder/wiki/使用经验
+
+# 应用
+
+#### UDPspeeder + OpenVPN加速任何流量，也适用于其他VPN
+![image0](/images/Capture2.PNG)
+
+可以和BBR/锐速叠加，不过BBR/锐速部署在VPS上只对本地和VPS间的流量有效，对从本地和第三方服务器间的流量无效。
+
+需要在服务端开启ipforward和NAT。在客户端改路由表（可以手动修改，也可以由OpenVPN的redirect-gateway选项自动加好）。
+
+Linux具体配置: [UDPspeeder + openvpn config guide](/doc/udpspeeder_openvpn.md).
+
+Windows具体配置: [win10系统UDPspeeder+OpenVPN的完整设置](https://github.com/wangyu-/UDPspeeder/wiki/win10系统UDPspeeder-OpenVPN的完整设置)
+
+如果UDPspeeder + OpenVPN对你来说显得太麻烦了，你可以尝试一下tinyFecVPN,一个集成了UDPspeeder功能的VPN:
+
+https://github.com/wangyu-/tinyFecVPN/
+
+#### UDPspeeder + kcptun/finalspeed + $*** 同时加速tcp和udp流量
+如果你需要用加速的tcp看视频和下载文件，这样效果可能比没有BBR的UDPspeeder+vpn方案更好。另外，如果你需要玩游戏，但是嫌配VPN麻烦，也可以用这种方案。
+![image0](/images/cn/speeder_kcptun.PNG)
+
+具体配置方法简介:
+
+假设$\*\*\*  server监听在在44.55.66.77的443端口(tcp和udp同时)。用kcptun把tcp 443映射到本地的tcp 1234；用UDPspeeder把udp 443的映射到本地的udp 1234。
+然后让$\*\*\* client 去连127.0.0.1:1234就可以了，tcp和udp都被加速了。完整命令：
+```
+run at server side:
+./kcp_server  -l ":4000" -t "127.0.0.1:443" -mode fast2
+./speederv2 -s -l0.0.0.0:4001 -r127.0.0.1:443  -f20:10 -k "passwd"
+
+run at client side:
+./kcp_client  -l ":1234" -r "44.55.66.77:4000" -mode fast2
+./speederv2 -c -l0.0.0.0:1234 -r44.55.66.77:4001 -f20:10 -k "passwd"
+```
+
+这就是全部的命令了。issue里有很多人困惑于怎么把tcp和udp流量"分开"，其实很简单就可以做到。
+
+如果只需要加速UDP，不需要加速TCP，可以把kcptun换成其他的任意端口转发方式，比如ncat/socat/ssh tunnel/iptables/[tinyPortMapper](https://github.com/wangyu-/tinyPortMapper/releases)。
+
+如果你没有kcptun只有BBR/锐速的话，也可以把kcptun换成ncat/socat/ssh tunnel/iptables/[tinyPortMapper](https://github.com/wangyu-/tinyPortMapper/releases)。这样，TCP流量由锐速/BBR加速，UDP由UDPspeeder加速。
+
+另外，即使你不想使用$\*\*\*的TCP功能，你也必须把$\*\*\*的TCP端口转发过来，否则无法使用UDP功能，这是socks5协议的工作方式决定的。($\*\*\*-redir方式不受此限制)
+
+#### UDPspeeder + openvpn + $*** 混合方案，也适用于其他VPN
+也是我正在用的方案。优点是可以随时在vpn和$\*\*\*方案间快速切换。
+实际部署起来比图中看起来的还要简单。不需要改路由表，不需要写iptables规则和开启NAT，需要做的只是用openvpn分配的ip访问$*** server。
+
+![image0](/images/cn/speeder_vpn_s.PNG)
+
+(也可以把图中的$*** server换成其他的socks5 server，这样就不需要$*** client了)
+
+可以和BBR/锐速叠加，BBR/锐速只要部署在VPS上就有效。
+
+也可以用[tinyFecVPN](https://github.com/wangyu-/tinyFecVPN/) + $\*\*\* ，配置起来更简单。
+
+# 应用实例
+
+#### win10系统UDPspeeder+OpenVPN的完整设置
+
+https://github.com/wangyu-/UDPspeeder/wiki/win10系统UDPspeeder-OpenVPN的完整设置
+
+
+#### UDPspeeder+OpenVPN运行在linux上，透明加速linux本机的网络
+
+https://github.com/wangyu-/tinyFecVPN/wiki/tinyFecVPN运行在linux上，透明加速linux本机的网络
+
+
+#### UDPspeeder+OpenVPN运行在虚拟机中，加速windows和局域网内其他主机的网络
+
+https://github.com/wangyu-/tinyFecVPN/wiki/tinyFecVPN运行在虚拟机中，加速windows和局域网内其他主机的网络
+
+
+
+#### 用树莓派做路由器，搭建透明代理，加速游戏主机的网络
+
+https://github.com/wangyu-/UDPspeeder/wiki/用树莓派做路由器，搭建透明代理，加速游戏主机的网络
+
+# 编译教程
+暂时先参考udp2raw的这篇教程，几乎一样的过程。
+
+https://github.com/wangyu-/udp2raw-tunnel/blob/master/doc/build_guide.zh-cn.md
+

doc/README.zh-cn.v1.md

@@ -0,0 +1,154 @@
+# UDPspeeder (v1)
+![image0](/images/Capture7.PNG)
+UDP双边加速工具，降低丢包率，配合vpn可以加速任何协议，尤其适用于加速游戏和网页打开速度；同时也是一个UDP连接的调试和统计工具。
+
+这个是我自己稳定用了一个月的项目，用来加速美服的Brawl Stars和亚服的Mobile Legend，效果不错。加速前卡得几乎没法玩，加速后就没怎么卡过了。
+
+注：目前最新版是v2版，这个是v1版的主页
+
+#### 效果
+![image0](/images/Capture8.PNG)
+#### 原理简介
+目前原理是多倍发包。以后会做各种优化，比如：对高频率的短包先合并再冗余；FEC（Forward Error Correction），在包速低的时候多倍发包，包速高时用FEC(这些功能在v2版里已经实现)。
+
+跟net-speeder比，优势在于client和server会把收到的多余包自动去掉，这个过程对上层透明，没有兼容性问题。而且发出的冗余数据包会做长度和内容的随机化，抓包是看不出发了冗余数据的，所以不用担心vps被封的问题。
+
+每个冗余数据包都是间隔数毫秒（可配置）以后延迟发出的，可以避开中间路由器因为瞬时buffer长度过长而连续丢掉所有副本。
+
+模拟一定的延迟抖动，这样上层应用计算出来的RTT方差会更大，以等待后续冗余包的到达，不至于发生在冗余包到达之前就触发重传的尴尬。
+
+#### 适用场景
+绝大部分流量不高的情况。程序本身加速udp，但是配合openvpn可以加速任何流量。网络状况不好时，游戏卡得没法玩，或者网页卡得没法打开，使用起来效果最好。对于解决语音通话的断断续续效果也不错。不适合大流量的场景，比如BT下载和在线看视频。 无论从自己使用效果的角度，还是从国际出口带宽占用的角度，都建议不要在大流量环境使用。
+
+#### 其他功能
+输出UDP收发情况报告，可以看出丢包率。
+
+模拟丢包，模拟延迟，模拟jitter。便于通过实验找出应用卡顿的原因。
+
+重复包过滤功能可以关掉，模拟网络本身有重复包的情况。用来测试应用对重复报的支持情况。
+
+client支持多个udp连接，server也支持多个client
+
+目前有amd64,x86,ar71xx,树莓派armv7和android的binary
+
+如果你需要绕过UDP屏蔽/QoS，或者需要连接复用/连接保持功能，或者是加密。解决方案在另一个repo(可以跟UDPspeeder一起使用)：
+
+https://github.com/wangyu-/udp2raw-tunnel
+
+# 简明操作说明
+
+### 环境要求
+Linux主机，可以是桌面版，可以是android手机/平板，可以是openwrt路由器，也可以是树莓派。在windows和mac上配合虚拟机可以稳定使用（speeder跑在Linux里，其他应用照常跑在window里，桥接模式测试可用）。
+
+android版需要通过terminal运行。
+
+### 安装
+下载编译好的二进制文件，解压到本地和服务器的任意目录。
+
+https://github.com/wangyu-/UDPspeeder/releases
+
+### 运行
+假设你有一个server，ip为44.55.66.77，有一个服务监听在udp 7777端口。 假设你需要加速本地到44.55.66.77:7777的流量。
+```
+在client端运行:
+./speeder_ar71xx -l0.0.0.0:3333 -r 44.55.66.77:8855 -c  -d2 -k "passwd"
+
+在server端运行:
+./speeder_amd64 -l0.0.0.0:8855 -r127.0.0.1:7777 -s -d2 -k "passwd"
+```
+
+现在client和server之间建立起了tunnel。想要连接44.55.66.77:7777，只需要连接 127.0.0.1:3333。来回的所有的udp流量会被加速。
+
+###### 注:
+
+-d2 表示除了本来的包以外，额外再发2个冗余包。可调。
+
+-k 指定一个字符串，server/client间所有收发的包都会被异或，改变协议特征，防止UDPspeeder的协议被运营商针对。
+
+# 进阶操作说明
+
+### 命令选项
+```
+UDPspeeder
+git version:b4bd385e88    build date:Sep 11 2017 10:29:25
+repository: https://github.com/wangyu-/UDPspeeder
+
+usage:
+    run as client : ./this_program -c -l local_listen_ip:local_port -r server_ip:server_port  [options]
+    run as server : ./this_program -s -l server_listen_ip:server_port -r remote_ip:remote_port  [options]
+
+common option,must be same on both sides:
+    -k,--key              <string>        key for simple xor encryption,default:"secret key"
+main options:
+    -d                    <number>        duplicated packet number, -d 0 means no duplicate. default value:0
+    -t                    <number>        duplicated packet delay time, unit: 0.1ms,default value:20(2ms)
+    -j                    <number>        simulated jitter.randomly delay first packet for 0~jitter_value*0.1 ms,to
+                                          create simulated jitter.default value:0.do not use if you dont
+                                          know what it means
+    --report              <number>        turn on udp send/recv report,and set a time interval for reporting,unit:s
+advanced options:
+    -t                    tmin:tmax       simliar to -t above,but delay randomly between tmin and tmax
+    -j                    jmin:jmax       simliar to -j above,but create jitter randomly between jmin and jmax
+    --random-drop         <number>        simulate packet loss ,unit:0.01%
+    --disable-filter                      disable duplicate packet filter.
+    -m                    <number>        max pending packets,to prevent the program from eating up all your memory,
+                                          default value:0(disabled).
+other options:
+    --log-level           <number>        0:never    1:fatal   2:error   3:warn 
+                                          4:info (default)     5:debug   6:trace
+    --log-position                        enable file name,function name,line number in log
+    --disable-color                       disable log color
+    --sock-buf            <number>        buf size for socket,>=10 and <=10240,unit:kbyte,default:1024
+    -h,--help                             print this help message
+
+```
+### 包发送选项，两端设置可以不同。 只影响本地包发送。
+##### -d 选项
+设置冗余包数量。
+##### -t 选项
+为冗余包的发送，增加一个延迟.对中间路由buffer做优化，应对瞬时Buffer过长导致的连续丢包.对于多个冗余包，依次在前一个包的基础上增加这个延迟。
+##### -j 选项
+为原始数据的发送，增加一个延迟抖动值。这样上层应用计算出来的RTT方差会更大，以等待后续冗余包的到达，不至于发生在冗余包到达之前就触发重传的尴尬。配合-t选项使用。正常情况下跨国网络本身的延迟抖动就很大。可以不用设-j
+
+##### --report  选项
+数据发送和接受报告。开启后可以根据此数据推测出包速和丢包率等特征。
+
+##### 加强版 -t 选项
+跟普通-t类似，允许设置最大值最小值，用随机延迟发送冗余包。
+
+##### 加强版 -j 选项
+允许给jitter选项设置最大值最小值。在这个区间随机化jitter。如果最大值最小值一样就是模拟延迟。可以模拟高延迟、高jitter的网络环境。
+
+##### --random-drop 选项
+随机丢包。模拟恶劣的网络环境时使用。
+
+### 包接收选项，两端设置可以不同。只影响本地包接受
+##### --disable-filter    
+关闭重复包过滤器。这样配合-d 选项可以模拟有重复包的网络环境。
+
+# 应用
+
+#### UDPspeeder + openvpn加速任何流量
+如果你只是需要玩游戏，效果预期会kcp/finalspeed方案更好。可以优化tcp游戏的延迟（通过冗余发包，避免了上层的重传）。比如魔兽世界用的是tcp连接。
+![image0](/images/Capture2.PNG)
+
+跟openvpn via kcptun方式的对比：
+
+kcptun在udp层有RS code，也是一种冗余传输，通过openvpn把流量转成tcp，再通过kcptun加速是有一定效果的。但是tcp只支持按序到达。按序到达的意思是，如果你发了1 2 3 4 5 6，6个包，如果第一个包丢了，那么必须等第一个包重传成功以后 2 3 4 5 6 才能到达；只要有一个包不到，后续数据包就要一直等待。用tcp承载udp流量会破坏udp的实时性。会造成游戏卡顿更严重。
+
+udp协议本身是ip协议加上了端口之后的直接封装，udp继承了ip协议的实时/乱序到达特性，更适合中转vpn。
+
+#### UDPspeeder + kcptun/finalspeed + $*** 同时加速tcp和udp流量
+如果你需要用加速的tcp看视频和下载文件，这样效果比vpn方案更好。不论是速度，还是流量的耗费上。
+![image0](/images/cn/speeder_kcptun.PNG)
+
+#### UDPspeeder + openvpn + $*** 混合方案
+也是我正在用的方案。优点是可以随时在vpn和$*** 方案间快速切换。
+实际部署起来比图中看起来的还要简单。不需要改路由表，需要做的只是用openvpn的ip访问$*** server。
+
+![image0](/images/cn/speeder_vpn_s.PNG)
+(也可以把图中的$*** server换成其他的socks5 server，这样连$*** client也不需要了)
+# 编译教程
+暂时先参考udp2raw的这篇教程，几乎一样的过程。
+
+https://github.com/wangyu-/udp2raw-tunnel/blob/master/doc/build_guide.zh-cn.md

doc/udpspeeder_openvpn.md

@@ -0,0 +1,107 @@
+
+# UDPspeeder + openvpn config guide
+![image_vpn](/images/en/udpspeeder+openvpn3.PNG)
+
+# UDPspeeder command
+
+#### run at server side
+```
+./speederv2 -s -l0.0.0.0:8855 -r 127.0.0.1:7777 -f20:10
+```
+
+#### run at client side
+assume server ip is 45.66.77.88
+```
+./speederv2 -c -l0.0.0.0:3333 -r 45.66.77.88:8855 -f20:10
+```
+
+# openvpn config
+
+#### client side config
+```
+client
+dev tun100
+proto udp
+
+remote 127.0.0.1 3333
+resolv-retry infinite 
+nobind 
+persist-key 
+persist-tun  
+
+ca /root/add-on/openvpn/ca.crt
+cert /root/add-on/openvpn/client.crt
+key /root/add-on/openvpn/client.key
+
+keepalive 3 20
+verb 3
+mute 20
+
+comp-lzo no
+
+fragment 1200       ##### very important    you can turn it up a bit. but, the lower the safer
+mssfix 1200         ##### very important
+
+sndbuf 2000000      ##### important
+rcvbuf 2000000      ##### important
+txqueuelen 4000     ##### suggested
+```
+
+
+#### server side config
+```
+local 0.0.0.0
+port 7777 
+proto udp
+dev tun 
+
+ca /etc/openvpn/easy-rsa/2.0/keys/ca.crt
+cert /etc/openvpn/easy-rsa/2.0/keys/server.crt
+key /etc/openvpn/easy-rsa/2.0/keys/server.key
+dh /etc/openvpn/easy-rsa/2.0/keys/dh1024.pem
+
+server 10.222.2.0 255.255.255.0 
+ifconfig 10.222.2.1 10.222.2.6
+
+client-to-client
+duplicate-cn 
+keepalive 10 60 
+
+max-clients 50
+
+persist-key
+persist-tun
+
+status /etc/openvpn/openvpn-status.log
+
+verb 3
+mute 20  
+
+comp-lzo no
+
+fragment 1200       ##### very important    you can turn it up a bit. but, the lower the safer
+mssfix 1200         ##### very important
+
+sndbuf 2000000      ##### important
+rcvbuf 2000000      ##### important
+txqueuelen 4000     ##### suggested
+```
+
+##### Note:
+If you use the `redirect-gateway` option of OpenVPN, you may need to add a route exception for your remote server ip at client side.Otherwise OpenVPN may hijack UDPspeeder 's traffic.
+
+For example, depend on your network environment, the command may looks like:
+```
+ip route add 44.55.66.77 via 44.55.66.1
+```
+or
+
+```
+ip route add 44.55.66.77 dev XXX
+```
+(run at client side)
+
+##### Other Info
+You can also use tinyFecVPN,a lightweight VPN with build-in FEC support:
+
+https://github.com/wangyu-/tinyFecVPN

fd_manager.cpp

@@ -0,0 +1,63 @@
+/*
+ * fd_manager.cpp
+ *
+ *  Created on: Sep 25, 2017
+ *      Author: root
+ */
+
+
+#include "fd_manager.h"
+int fd_manager_t::fd_exist(int fd)
+{
+	return fd_to_fd64_mp.find(fd)!=fd_to_fd64_mp.end();
+}
+int fd_manager_t::exist(fd64_t fd64)
+{
+	return fd64_to_fd_mp.find(fd64)!=fd64_to_fd_mp.end();
+}
+int fd_manager_t::to_fd(fd64_t fd64)
+{
+	assert(exist(fd64));
+	return fd64_to_fd_mp[fd64];
+}
+void fd_manager_t::fd64_close(fd64_t fd64)
+{
+	assert(exist(fd64));
+	int fd=fd64_to_fd_mp[fd64];
+	fd64_to_fd_mp.erase(fd64);
+	fd_to_fd64_mp.erase(fd);
+	if(exist_info(fd64))
+	{
+		fd_info_mp.erase(fd64);
+	}
+	close(fd);
+}
+void fd_manager_t::reserve(int n)
+{
+	fd_to_fd64_mp.reserve(n);
+	fd64_to_fd_mp.reserve(n);
+	fd_info_mp.reserve(n);
+}
+u64_t fd_manager_t::create(int fd)
+{
+	assert(!fd_exist(fd));
+	fd64_t fd64=counter++;
+	fd_to_fd64_mp[fd]=fd64;
+	fd64_to_fd_mp[fd64]=fd;
+	return fd64;
+}
+fd_manager_t::fd_manager_t()
+{
+	counter=u32_t(-1);
+	counter+=100;
+	reserve(10007);
+}
+fd_info_t & fd_manager_t::get_info(fd64_t fd64)
+{
+	assert(exist(fd64));
+	return fd_info_mp[fd64];
+}
+int fd_manager_t::exist_info(fd64_t fd64)
+{
+	return fd_info_mp.find(fd64)!=fd_info_mp.end();
+}

fd_manager.h

@@ -0,0 +1,38 @@
+/*
+ * fd_manager.h
+ *
+ *  Created on: Sep 25, 2017
+ *      Author: root
+ */
+
+#ifndef FD_MANAGER_H_
+#define FD_MANAGER_H_
+
+#include "common.h"
+#include "packet.h"
+
+
+
+struct fd_manager_t   //conver fd to a uniq 64bit number,avoid fd value conflict caused by close and re-create
+//this class is not strictly necessary,it just makes epoll fd handling easier
+{
+	fd_info_t & get_info(fd64_t fd64);
+	int exist_info(fd64_t);
+	int exist(fd64_t fd64);
+	int to_fd(fd64_t);
+	void fd64_close(fd64_t fd64);
+	void reserve(int n);
+	u64_t create(int fd);
+	fd_manager_t();
+private:
+	u64_t counter;
+	unordered_map<int,fd64_t> fd_to_fd64_mp;
+	unordered_map<fd64_t,int> fd64_to_fd_mp;
+	unordered_map<fd64_t,fd_info_t> fd_info_mp;
+	int fd_exist(int fd);
+	//void remove_fd(int fd);
+	//fd64_t fd_to_fd64(int fd);
+};
+
+extern fd_manager_t fd_manager;
+#endif /* FD_MANAGER_H_ */

fec_manager.cpp

@@ -0,0 +1,846 @@
+/*
+ * fec_manager.cpp
+ *
+ *  Created on: Sep 27, 2017
+ *      Author: root
+ */
+
+#include "fec_manager.h"
+#include "log.h"
+#include "common.h"
+#include "lib/rs.h"
+#include "fd_manager.h"
+
+int g_fec_data_num=20;
+int g_fec_redundant_num=10;
+int g_fec_mtu=1250;
+int g_fec_queue_len=200;
+int g_fec_timeout=8*1000; //8ms
+int g_fec_mode=0;
+
+int dynamic_update_fec=1;
+
+const int encode_fast_send=1;
+const int decode_fast_send=1;
+
+int short_packet_optimize=1;
+int header_overhead=40;
+
+u32_t fec_buff_num=2000;// how many packet can fec_decode_manager hold. shouldnt be very large,or it will cost huge memory
+
+blob_encode_t::blob_encode_t()
+{
+	clear();
+}
+int blob_encode_t::clear()
+{
+	counter=0;
+	current_len=(int)sizeof(u32_t);
+	return 0;
+}
+
+int blob_encode_t::get_num()
+{
+	return counter;
+}
+int blob_encode_t::get_shard_len(int n)
+{
+	return round_up_div(current_len,n);
+}
+
+int blob_encode_t::get_shard_len(int n,int next_packet_len)
+{
+	return round_up_div(current_len+(int)sizeof(u16_t)+next_packet_len,n);
+}
+
+int blob_encode_t::input(char *s,int len)
+{
+	assert(current_len+len+sizeof(u16_t) +100<sizeof(input_buf));
+	assert(len<=65535&&len>=0);
+	counter++;
+	assert(counter<=max_blob_packet_num);
+	write_u16(input_buf+current_len,len);
+	current_len+=sizeof(u16_t);
+	memcpy(input_buf+current_len,s,len);
+	current_len+=len;
+	return 0;
+}
+
+int blob_encode_t::output(int n,char ** &s_arr,int & len)
+{
+	len=round_up_div(current_len,n);
+	write_u32(input_buf,counter);
+	for(int i=0;i<n;i++)
+	{
+		output_buf[i]=input_buf+len*i;
+	}
+	s_arr=output_buf;
+	return 0;
+}
+blob_decode_t::blob_decode_t()
+{
+	clear();
+}
+int blob_decode_t::clear()
+{
+	current_len=0;
+	last_len=-1;
+	counter=0;
+	return 0;
+}
+int blob_decode_t::input(char *s,int len)
+{
+	if(last_len!=-1)
+	{
+		assert(last_len==len);
+	}
+	counter++;
+	assert(counter<=max_fec_packet_num);
+	last_len=len;
+	assert(current_len+len+100<(int)sizeof(input_buf));//avoid overflow
+	memcpy(input_buf+current_len,s,len);
+	current_len+=len;
+	return 0;
+}
+int blob_decode_t::output(int &n,char ** &s_arr,int *&len_arr)
+{
+
+	int parser_pos=0;
+
+	if(parser_pos+(int)sizeof(u32_t)>current_len) {mylog(log_info,"failed 0\n");return -1;}
+
+	n=(int)read_u32(input_buf+parser_pos);
+	if(n>max_blob_packet_num) {mylog(log_info,"failed 1\n");return -1;}
+	s_arr=output_buf;
+	len_arr=output_len;
+
+	parser_pos+=sizeof(u32_t);
+	for(int i=0;i<n;i++)
+	{
+		if(parser_pos+(int)sizeof(u16_t)>current_len) {mylog(log_info,"failed2 \n");return -1;}
+		len_arr[i]=(int)read_u16(input_buf+parser_pos);
+		parser_pos+=(int)sizeof(u16_t);
+		if(parser_pos+len_arr[i]>current_len) {mylog(log_info,"failed 3 %d  %d %d\n",parser_pos,len_arr[i],current_len);return -1;}
+		s_arr[i]=input_buf+parser_pos;
+		parser_pos+=len_arr[i];
+	}
+	return 0;
+}
+
+
+fec_encode_manager_t::~fec_encode_manager_t()
+{
+	fd_manager.fd64_close(timer_fd64);
+}
+u64_t fec_encode_manager_t::get_timer_fd64()
+{
+	return timer_fd64;
+}
+fec_encode_manager_t::fec_encode_manager_t()
+{
+	//int timer_fd;
+	if ((timer_fd = timerfd_create(CLOCK_MONOTONIC, TFD_NONBLOCK)) < 0)
+	{
+		mylog(log_fatal,"timer_fd create error");
+		myexit(1);
+	}
+	timer_fd64=fd_manager.create(timer_fd);
+
+	reset_fec_parameter(g_fec_data_num,g_fec_redundant_num,g_fec_mtu,g_fec_queue_len,g_fec_timeout,g_fec_mode);
+
+
+}
+int fec_encode_manager_t::reset_fec_parameter(int data_num,int redundant_num,int mtu,int queue_len,int timeout,int mode)
+{
+	fec_data_num=data_num;
+	fec_redundant_num=redundant_num;
+	fec_mtu=mtu;
+	fec_queue_len=queue_len;
+	fec_timeout=timeout;
+	fec_mode=mode;
+
+	assert(data_num+redundant_num<max_fec_packet_num);
+
+	clear();
+
+	return 0;
+}
+int fec_encode_manager_t::append(char *s,int len/*,int &is_first_packet*/)
+{
+	if(counter==0)
+	{
+		itimerspec its;
+		memset(&its.it_interval,0,sizeof(its.it_interval));
+		first_packet_time=get_current_time_us();
+		my_time_t tmp_time=fec_timeout+first_packet_time;
+		its.it_value.tv_sec=tmp_time/1000000llu;
+		its.it_value.tv_nsec=(tmp_time%1000000llu)*1000llu;
+		timerfd_settime(timer_fd,TFD_TIMER_ABSTIME,&its,0);
+	}
+	if(fec_mode==0)//for type 0 use blob
+	{
+		assert(blob_encode.input(s,len)==0);
+	}
+	else if(fec_mode==1)//for tpe 1 use  input_buf and counter
+	{
+		mylog(log_trace,"counter=%d\n",counter);
+		assert(len<=65535&&len>=0);
+		//assert(len<=fec_mtu);//relax this limitation
+		char * p=input_buf[counter]+sizeof(u32_t)+4*sizeof(char);//copy directly to final position,avoid unnecessary copy.
+		//remember to change this,if protocol is modified
+
+		write_u16(p,(u16_t)((u32_t)len));  //TODO  omit this u16 for data packet while sending
+		p+=sizeof(u16_t);
+		memcpy(p,s,len);
+		input_len[counter]=len+sizeof(u16_t);
+	}
+	else
+	{
+		assert(0==1);
+	}
+	counter++;
+	return 0;
+}
+int fec_encode_manager_t::input(char *s,int len/*,int &is_first_packet*/)
+{
+	if(counter==0&&dynamic_update_fec)
+	{
+		fec_data_num=g_fec_data_num;
+		fec_redundant_num=g_fec_redundant_num;
+		fec_mtu=g_fec_mtu;
+		fec_queue_len=g_fec_queue_len;
+		fec_timeout=g_fec_timeout;
+		fec_mode=g_fec_mode;
+	}
+
+	int about_to_fec=0;
+	int delayed_append=0;
+	//int counter_back=counter;
+	assert(fec_mode==0||fec_mode==1);
+
+	if(fec_mode==0&& s!=0 &&counter==0)
+	{
+		int out_len=blob_encode.get_shard_len(fec_data_num,len);
+		if(out_len>fec_mtu)
+		{
+			mylog(log_warn,"message too long ori_len=%d out_len=%d fec_mtu=%d,ignored\n",len,out_len,fec_mtu);
+			return -1;
+		}
+	}
+	if(fec_mode==1&&s!=0&&len>fec_mtu)
+	{
+		mylog(log_warn,"mode==1,message len=%d,len>fec_mtu,fec_mtu=%d,packet may not be delivered\n",len,fec_mtu);
+		//return -1;
+	}
+	if(s==0&&counter==0)
+	{
+		mylog(log_warn,"unexpected s==0&&counter==0\n");
+		return -1;
+	}
+	if(s==0) about_to_fec=1;//now
+
+	if(fec_mode==0&& blob_encode.get_shard_len(fec_data_num,len)>fec_mtu) {about_to_fec=1; delayed_append=1;}//fec then add packet
+
+	if(fec_mode==0) assert(counter<fec_queue_len);//counter will never equal fec_pending_num,if that happens fec should already been done.
+	if(fec_mode==1) assert(counter<fec_data_num);
+
+
+	if(s!=0&&!delayed_append)
+	{
+		append(s,len);
+	}
+
+	if(fec_mode==0&& counter==fec_queue_len) about_to_fec=1;
+
+	if(fec_mode==1&& counter==fec_data_num) about_to_fec=1;
+
+
+    if(about_to_fec)
+	{
+    	char ** blob_output=0;
+    	int fec_len=-1;
+    	mylog(log_trace,"counter=%d\n",counter);
+
+    	if(counter==0)
+    	{
+    		mylog(log_warn,"unexpected counter==0 here\n");
+    		return -1;
+    	}
+
+    	int actual_data_num;
+    	int actual_redundant_num;
+
+    	if(fec_mode==0)
+    	{
+
+    		actual_data_num=fec_data_num;
+    		actual_redundant_num=fec_redundant_num;
+
+    		if(short_packet_optimize)
+    		{
+    			u32_t best_len=(blob_encode.get_shard_len(fec_data_num,0)+header_overhead)*(fec_data_num+fec_redundant_num);
+    			int best_data_num=fec_data_num;
+    			for(int i=1;i<actual_data_num;i++)
+    			{
+    				u32_t shard_len=blob_encode.get_shard_len(i,0);
+    				if(shard_len>(u32_t)fec_mtu) continue;
+
+    				u32_t new_len=(shard_len+header_overhead)*(i+fec_redundant_num);
+    				if(new_len<best_len)
+    				{
+    					best_len=new_len;
+    					best_data_num=i;
+    				}
+    			}
+    			actual_data_num=best_data_num;
+    			actual_redundant_num=fec_redundant_num;
+    			mylog(log_trace,"actual_data_num=%d actual_redundant_num=%d\n",best_data_num,fec_redundant_num);
+    		}
+
+        	assert(blob_encode.output(actual_data_num,blob_output,fec_len)==0);
+    	}
+    	else
+    	{
+    		actual_data_num=counter;
+    		actual_redundant_num=fec_redundant_num;
+
+    		for(int i=0;i<counter;i++)
+    		{
+    			assert(input_len[i]>=0);
+    			if(input_len[i]>fec_len) fec_len=input_len[i];
+    		}
+
+    	}
+    	mylog(log_trace,"%d %d %d\n",actual_data_num,actual_redundant_num,fec_len);
+
+    	char *tmp_output_buf[max_fec_packet_num+5]={0};
+    	for(int i=0;i<actual_data_num+actual_redundant_num;i++)
+    	{
+    		int tmp_idx=0;
+
+        	write_u32(input_buf[i] + tmp_idx, seq);
+			tmp_idx += sizeof(u32_t);
+			input_buf[i][tmp_idx++] = (unsigned char) fec_mode;
+			if (fec_mode == 1 && i < actual_data_num)
+			{
+				input_buf[i][tmp_idx++] = (unsigned char) 0;
+				input_buf[i][tmp_idx++] = (unsigned char) 0;
+			} else
+			{
+				input_buf[i][tmp_idx++] = (unsigned char) actual_data_num;
+				input_buf[i][tmp_idx++] = (unsigned char) actual_redundant_num;
+			}
+			input_buf[i][tmp_idx++] = (unsigned char) i;
+
+    		tmp_output_buf[i]=input_buf[i]+tmp_idx; //////caution ,trick here.
+
+    		if(fec_mode==0)
+    		{
+        		output_len[i]=tmp_idx+fec_len;
+        		if(i<actual_data_num)
+        		{
+        			memcpy(input_buf[i]+tmp_idx,blob_output[i],fec_len);
+        		}
+    		}
+    		else
+    		{
+    			if(i<actual_data_num)
+    			{
+    				output_len[i]=tmp_idx+input_len[i];
+    				memset(tmp_output_buf[i]+input_len[i],0,fec_len-input_len[i]);
+    			}
+    			else
+    				output_len[i]=tmp_idx+fec_len;
+
+    		}
+    		output_buf[i]=input_buf[i];//output_buf points to same block of memory with different offset
+
+    	}
+
+    	if(0)
+    	{
+			printf("seq=%u,fec_len=%d,%d %d,before fec\n",seq,fec_len,actual_data_num,actual_redundant_num);
+
+			for(int i=0;i<actual_data_num;i++)
+			{
+				printf("{");
+				for(int j=0;j<8+fec_len;j++)
+				{
+					log_bare(log_warn,"0x%02x,",(u32_t)(unsigned char)input_buf[i][j]);
+				}
+				printf("},\n");
+				//log_bare(log_warn,"")
+			}
+    	}
+    	//output_len=blob_len+sizeof(u32_t)+4*sizeof(char);/////remember to change this 4,if modified the protocol
+		rs_encode2(actual_data_num,actual_data_num+actual_redundant_num,tmp_output_buf,fec_len);
+
+		if(0)
+		{
+			printf("seq=%u,fec_len=%d,%d %d,after fec\n",seq,fec_len,actual_data_num,actual_redundant_num);
+			for(int i=0;i<actual_data_num+actual_redundant_num;i++)
+			{
+				printf("{");
+				for(int j=0;j<8+fec_len;j++)
+				{
+					log_bare(log_warn,"0x%02x,",(u32_t)(unsigned char)output_buf[i][j]);
+				}
+				printf("},\n");
+				//log_bare(log_warn,"")
+			}
+		}
+
+		//mylog(log_trace,"!!! s= %d\n");
+		assert(ready_for_output==0);
+    	ready_for_output=1;
+    	first_packet_time_for_output=first_packet_time;
+    	first_packet_time=0;
+    	seq++;
+    	counter=0;
+    	output_n=actual_data_num+actual_redundant_num;
+    	blob_encode.clear();
+
+		itimerspec its;
+		memset(&its,0,sizeof(its));
+		timerfd_settime(timer_fd,TFD_TIMER_ABSTIME,&its,0);
+
+    	if(encode_fast_send&&fec_mode==1)
+    	{
+			int packet_to_send[max_fec_packet_num+5]={0};
+			int packet_to_send_counter=0;
+
+			//assert(counter!=0);
+			if(s!=0)
+				packet_to_send[packet_to_send_counter++]=actual_data_num-1;
+			for(int i=actual_data_num;i<actual_data_num+actual_redundant_num;i++)
+			{
+
+					packet_to_send[packet_to_send_counter++]=i;
+			}
+			output_n=packet_to_send_counter;//re write
+			for(int i=0;i<packet_to_send_counter;i++)
+			{
+				output_buf[i]=output_buf[packet_to_send[i]];
+				output_len[i]=output_len[packet_to_send[i]];
+			}
+
+    	}
+	}
+    else
+    {
+    	if(encode_fast_send&&s!=0&&fec_mode==1)
+    	{
+    		assert(counter>=1);
+    		assert(counter<=255);
+    		int input_buf_idx=counter-1;
+    		assert(ready_for_output==0);
+    		ready_for_output=1;
+    		first_packet_time_for_output=0;
+    		output_n=1;
+
+
+    		int tmp_idx=0;
+    		write_u32(input_buf[input_buf_idx]+tmp_idx,seq);
+    		tmp_idx+=sizeof(u32_t);
+
+    		input_buf[input_buf_idx][tmp_idx++]=(unsigned char)fec_mode;
+    		input_buf[input_buf_idx][tmp_idx++]=(unsigned char)0;
+    		input_buf[input_buf_idx][tmp_idx++]=(unsigned char)0;
+    		input_buf[input_buf_idx][tmp_idx++]=(unsigned char)((u32_t)input_buf_idx);
+
+    		output_len[0]=input_len[input_buf_idx]+tmp_idx;
+    		output_buf[0]=input_buf[input_buf_idx];
+
+    		if(0)
+    		{
+				printf("seq=%u,buf_idx=%d\n",seq,input_buf_idx);
+				for(int j=0;j<output_len[0];j++)
+				{
+					log_bare(log_warn,"0x%02x,",(u32_t)(unsigned char)output_buf[0][j]);
+				}
+				printf("\n");
+    		}
+    	}
+    }
+
+	if(s!=0&&delayed_append)
+	{
+		assert(fec_mode!=1);
+		append(s,len);
+	}
+
+	return 0;
+}
+
+int fec_encode_manager_t::output(int &n,char ** &s_arr,int *&len)
+{
+	if(!ready_for_output)
+	{
+		n=-1;
+		len=0;
+		s_arr=0;
+	}
+	else
+	{
+		n=output_n;
+		len=output_len;
+		s_arr=output_buf;
+		ready_for_output=0;
+	}
+	return 0;
+}
+/*
+int fec_decode_manager_t::re_init()
+{
+	clear();
+	return 0;
+}*/
+
+int fec_decode_manager_t::input(char *s,int len)
+{
+	assert(s!=0);
+	assert(len+100<buf_len);//guarenteed by upper level
+
+	int tmp_idx=0;
+	int tmp_header_len=sizeof(u32_t)+sizeof(char)*4;
+	if(len<tmp_header_len)
+	{
+		mylog(log_warn,"len =%d\n",len);
+		return -1;
+	}
+	u32_t seq=read_u32(s+tmp_idx);
+	tmp_idx+=sizeof(u32_t);
+	int type=(unsigned char)s[tmp_idx++];
+	int data_num=(unsigned char)s[tmp_idx++];
+	int redundant_num=(unsigned char)s[tmp_idx++];
+	int inner_index=(unsigned char)s[tmp_idx++];
+	len=len-tmp_idx;
+
+	//mylog(log_trace,"input\n");
+
+	if(len<0)
+	{
+		mylog(log_warn,"len<0\n");
+		return -1;
+	}
+
+	if(type==1)
+	{
+		if(len<(int)sizeof(u16_t))
+		{
+			mylog(log_warn,"type==1&&len<2\n");
+			return -1;
+		}
+		if(data_num==0&&(int)( read_u16(s+tmp_idx)+sizeof(u16_t))!=len)
+		{
+			mylog(log_warn,"inner_index<data_num&&read_u16(s+tmp_idx)+sizeof(u16_t)!=len    %d %d\n",(int)( read_u16(s+tmp_idx)+sizeof(u16_t)),len);
+			return -1;
+		}
+	}
+
+	if(type==0&&data_num==0)
+	{
+		mylog(log_warn,"unexpected type==0&&data_num==0\n");
+		return -1;
+	}
+	if(data_num+redundant_num>=max_fec_packet_num)
+	{
+		mylog(log_warn,"data_num+redundant_num>=max_fec_packet_num\n");
+		return -1;
+	}
+	if(!anti_replay.is_vaild(seq))
+	{
+		mylog(log_trace,"!anti_replay.is_vaild(seq) ,seq =%u\n",seq);
+		return 0;
+	}
+
+	if(mp[seq].group_mp.find(inner_index)!=mp[seq].group_mp.end() )
+	{
+		mylog(log_debug,"dup fec index\n");//duplicate can happen on  a normal network, so its just log_debug
+		return -1;
+	}
+
+
+	if(mp[seq].type==-1)
+		mp[seq].type=type;
+	else
+	{
+		if(mp[seq].type!=type)
+		{
+			mylog(log_warn,"type mismatch\n");
+			return -1;
+		}
+	}
+
+	if(data_num!=0)
+	{
+		//mp[seq].data_counter++;
+
+		if(mp[seq].data_num==-1)
+		{
+			mp[seq].data_num=data_num;
+			mp[seq].redundant_num=redundant_num;
+			mp[seq].len=len;
+		}
+		else
+		{
+			if(mp[seq].data_num!=data_num||mp[seq].redundant_num!=redundant_num||mp[seq].len!=len)
+			{
+				mylog(log_warn,"unexpected mp[seq].data_num!=data_num||mp[seq].redundant_num!=redundant_num||mp[seq].len!=len\n");
+				return -1;
+			}
+		}
+	}
+
+	//mylog(log_info,"mp.size()=%d index=%d\n",mp.size(),index);
+
+	if(fec_data[index].used!=0)
+	{
+		u32_t tmp_seq=fec_data[index].seq;
+		anti_replay.set_invaild(tmp_seq);
+
+		if(mp.find(tmp_seq)!=mp.end())
+		{
+			mp.erase(tmp_seq);
+		}
+		if(tmp_seq==seq)
+		{
+			mylog(log_warn,"unexpected tmp_seq==seq ,seq=%d\n",seq);
+			return -1;
+		}
+	}
+
+	fec_data[index].used=1;
+	fec_data[index].seq=seq;
+	fec_data[index].type=type;
+	fec_data[index].data_num=data_num;
+	fec_data[index].redundant_num=redundant_num;
+	fec_data[index].idx=inner_index;
+	fec_data[index].len=len;
+	assert(0<=index&&index<(int)fec_buff_num);
+	assert(len+100<buf_len);
+	memcpy(fec_data[index].buf,s+tmp_idx,len);
+	mp[seq].group_mp[inner_index]=index;
+	//index++ at end of function
+
+	map<int,int> &inner_mp=mp[seq].group_mp;
+
+
+	int about_to_fec=0;
+	if(type==0)
+	{
+		//assert((int)inner_mp.size()<=data_num);
+		if((int)inner_mp.size()>data_num)
+		{
+			mylog(log_warn,"inner_mp.size()>data_num\n");
+			anti_replay.set_invaild(seq);
+			goto end;
+		}
+		if((int)inner_mp.size()==data_num)
+			about_to_fec=1;
+	}
+	else
+	{
+		if(mp[seq].data_num!=-1)
+		{
+			if((int)inner_mp.size()>mp[seq].data_num+1)
+			{
+				mylog(log_warn,"inner_mp.size()>data_num+1\n");
+				anti_replay.set_invaild(seq);
+				goto end;
+			}
+			if((int)inner_mp.size()>=mp[seq].data_num)
+			{
+				about_to_fec=1;
+			}
+		}
+	}
+
+
+	if(about_to_fec)
+	{
+		int group_data_num=mp[seq].data_num;
+		int group_redundant_num=mp[seq].redundant_num;
+
+		//mylog(log_error,"fec here!\n");
+		if(type==0)
+		{
+			char *fec_tmp_arr[max_fec_packet_num+5]={0};
+			for(auto it=inner_mp.begin();it!=inner_mp.end();it++)
+			{
+				fec_tmp_arr[it->first]=fec_data[it->second].buf;
+			}
+			assert(rs_decode2(group_data_num,group_data_num+group_redundant_num,fec_tmp_arr,len)==0); //the input data has been modified in-place
+			//this line should always succeed
+
+			blob_decode.clear();
+			for(int i=0;i<group_data_num;i++)
+			{
+				blob_decode.input(fec_tmp_arr[i],len);
+			}
+			if(blob_decode.output(output_n,output_s_arr,output_len_arr)!=0)
+			{
+				mylog(log_warn,"blob_decode failed\n");
+				//ready_for_output=0;
+				anti_replay.set_invaild(seq);
+				goto end;
+			}
+			assert(ready_for_output==0);
+			ready_for_output=1;
+			anti_replay.set_invaild(seq);
+		}
+		else//type==1
+		{
+
+
+			int max_len=-1;
+			int fec_result_ok=1;
+			int data_check_ok=1;
+			int debug_num=inner_mp.size();
+
+			int missed_packet[max_fec_packet_num+5];
+			int missed_packet_counter=0;
+
+			//outupt_s_arr_buf[max_fec_packet_num+5]={0};
+
+			//memset(output_s_arr_buf,0,sizeof(output_s_arr_buf));//in efficient
+
+			for(int i=0;i<group_data_num+group_redundant_num;i++)
+			{
+				output_s_arr_buf[i]=0;
+			}
+			for(auto it=inner_mp.begin();it!=inner_mp.end();it++)
+			{
+				output_s_arr_buf[it->first]=fec_data[it->second].buf;
+				if(fec_data[it->second].len<(int)sizeof(u16_t))
+				{
+					mylog(log_warn,"fec_data[it->second].len<(int)sizeof(u16_t)");
+					data_check_ok=0;
+				}
+
+				if(fec_data[it->second].len > max_len)
+					max_len=fec_data[it->second].len;
+			}
+			if(max_len!=mp[seq].len)
+			{
+				data_check_ok=0;
+				mylog(log_warn,"max_len!=mp[seq].len");
+			}
+			if(data_check_ok==0)
+			{
+				//ready_for_output=0;
+				mylog(log_warn,"data_check_ok==0\n");
+				anti_replay.set_invaild(seq);
+				goto end;
+			}
+			for(auto it=inner_mp.begin();it!=inner_mp.end();it++)
+			{
+				int tmp_idx=it->second;
+				assert(max_len>=fec_data[tmp_idx].len);//guarenteed by data_check_ok
+				memset(fec_data[tmp_idx].buf+fec_data[tmp_idx].len,0,max_len-fec_data[tmp_idx].len);
+			}
+
+			for(int i=0;i<group_data_num;i++)
+			{
+				if(output_s_arr_buf[i]==0 ||i==inner_index) //only missed packet +current packet
+				{
+					missed_packet[missed_packet_counter++]=i;
+				}
+			}
+			mylog(log_trace,"fec done,%d %d,missed_packet_counter=%d\n",group_data_num,group_redundant_num,missed_packet_counter);
+
+			assert(rs_decode2(group_data_num,group_data_num+group_redundant_num,output_s_arr_buf,max_len)==0);//this should always succeed
+
+			for(int i=0;i<group_data_num;i++)
+			{
+				output_len_arr_buf[i]=read_u16(output_s_arr_buf[i]);
+				output_s_arr_buf[i]+=sizeof(u16_t);
+				if(output_len_arr_buf[i]>max_data_len)
+				{
+					mylog(log_warn,"invaild len %d,seq= %u,data_num= %d r_num= %d,i= %d\n",output_len_arr_buf[i],seq,group_data_num,group_redundant_num,i);
+					fec_result_ok=0;
+					for(int i=0;i<missed_packet_counter;i++)
+					{
+						log_bare(log_warn,"%d ",missed_packet[i]);
+					}
+					log_bare(log_warn,"\n");
+					//break;
+				}
+			}
+			if(fec_result_ok)
+			{
+
+				output_n=group_data_num;
+
+				if(decode_fast_send)
+				{
+					output_n=missed_packet_counter;
+					for(int i=0;i<missed_packet_counter;i++)
+					{
+						output_s_arr_buf[i]=output_s_arr_buf[missed_packet[i]];
+						output_len_arr_buf[i]=output_len_arr_buf[missed_packet[i]];
+					}
+				}
+
+
+				output_s_arr=output_s_arr_buf;
+				output_len_arr=output_len_arr_buf;
+				assert(ready_for_output==0);
+				ready_for_output=1;
+			}
+			else
+			{
+				//fec_not_ok:
+				ready_for_output=0;
+			}
+			anti_replay.set_invaild(seq);
+		}// end of type==1
+	}
+	else //not about_to_fec
+	{
+
+		if(decode_fast_send)
+		{
+			if(type==1&&data_num==0)
+			{
+				assert(ready_for_output==0);
+				output_n=1;
+				int check_len=read_u16(fec_data[index].buf);
+				output_s_arr_buf[0]=fec_data[index].buf+sizeof(u16_t);
+				output_len_arr_buf[0]=fec_data[index].len-sizeof(u16_t);
+
+				if(output_len_arr_buf[0]!=check_len)
+				{
+					mylog(log_warn,"len mismatch %d %d\n",output_len_arr_buf[0],check_len);
+				}
+				output_s_arr=output_s_arr_buf;
+				output_len_arr=output_len_arr_buf;
+
+				ready_for_output=1;
+			}
+		}
+	}
+
+	end:
+	index++;
+	if(index==int(fec_buff_num)) index=0;
+
+	return 0;
+}
+int fec_decode_manager_t::output(int &n,char ** &s_arr,int* &len_arr)
+{
+	if(!ready_for_output)
+	{
+		n=-1;
+		s_arr=0;
+		len_arr=0;
+	}
+	else
+	{
+		ready_for_output=0;
+		n=output_n;
+		s_arr=output_s_arr;
+		len_arr=output_len_arr;
+	}
+	return 0;
+}

fec_manager.h

@@ -0,0 +1,254 @@
+/*
+ * fec_manager.h
+ *
+ *  Created on: Sep 27, 2017
+ *      Author: root
+ */
+
+#ifndef FEC_MANAGER_H_
+#define FEC_MANAGER_H_
+
+#include "common.h"
+#include "log.h"
+#include "lib/rs.h"
+
+const int max_blob_packet_num=30000;//how many packet can be contain in a blob_t ,can be set very large
+const u32_t anti_replay_buff_size=30000;//can be set very large
+
+const int max_fec_packet_num=255;// this is the limitation of the rs lib
+extern u32_t fec_buff_num;
+
+
+/*begin for first time init or dynamic update*/
+extern int g_fec_data_num;
+extern int g_fec_redundant_num;
+extern int g_fec_mtu;
+extern int g_fec_queue_len;
+extern int g_fec_timeout; //8ms
+extern int g_fec_mode;
+extern int dynamic_update_fec;
+/*end for first time init or dynamic update*/
+
+
+struct anti_replay_t
+{
+
+	u64_t replay_buffer[anti_replay_buff_size];
+	unordered_set<u32_t> st;
+	int index;
+	anti_replay_t()
+	{
+		clear();
+	}
+	int clear()
+	{
+		memset(replay_buffer,-1,sizeof(replay_buffer));
+		st.clear();
+		st.rehash(anti_replay_buff_size*3);
+		index=0;
+		return 0;
+	}
+	void set_invaild(u32_t seq)
+	{
+
+		if(st.find(seq)!=st.end() )
+		{
+			mylog(log_trace,"seq %u exist\n",seq);
+			return;
+			//return 0;
+		}
+		if(replay_buffer[index]!=u64_t(i64_t(-1)))
+		{
+			assert(st.find(replay_buffer[index])!=st.end());
+			st.erase(replay_buffer[index]);
+		}
+		replay_buffer[index]=seq;
+		st.insert(seq);
+		index++;
+		if(index==int(anti_replay_buff_size)) index=0;
+		//return 1; //for complier check
+	}
+	int is_vaild(u32_t seq)
+	{
+		return st.find(seq)==st.end();
+	}
+};
+
+struct blob_encode_t
+{
+	char input_buf[(max_fec_packet_num+5)*buf_len];
+	int current_len;
+	int counter;
+
+	char *output_buf[max_fec_packet_num+100];
+
+	blob_encode_t();
+
+    int clear();
+
+    int get_num();
+    int get_shard_len(int n);
+    int get_shard_len(int n,int next_packet_len);
+
+	int input(char *s,int len);  //len=use len=0 for second and following packet
+	int output(int n,char ** &s_arr,int & len);
+};
+
+struct blob_decode_t
+{
+	char input_buf[(max_fec_packet_num+5)*buf_len];
+	int current_len;
+	int last_len;
+	int counter;
+
+	char *output_buf[max_blob_packet_num+100];
+	int output_len[max_blob_packet_num+100];
+
+	blob_decode_t();
+	int clear();
+	int input(char *input,int len);
+	int output(int &n,char ** &output,int *&len_arr);
+};
+
+class fec_encode_manager_t
+{
+private:
+	u32_t seq;
+
+	int fec_mode;
+	int fec_data_num,fec_redundant_num;
+	int fec_mtu;
+	int fec_queue_len;
+	int fec_timeout;
+
+	my_time_t first_packet_time;
+	my_time_t first_packet_time_for_output;
+
+
+	blob_encode_t blob_encode;
+	char input_buf[max_fec_packet_num+5][buf_len];
+	int input_len[max_fec_packet_num+100];
+
+	char *output_buf[max_fec_packet_num+100];
+	int output_len[max_fec_packet_num+100];
+
+	int counter;
+	int timer_fd;
+	u64_t timer_fd64;
+
+	int ready_for_output;
+	u32_t output_n;
+
+
+	int append(char *s,int len);
+
+public:
+	fec_encode_manager_t();
+	~fec_encode_manager_t();
+
+	int clear()
+	{
+		counter=0;
+		blob_encode.clear();
+		ready_for_output=0;
+
+		itimerspec zero_its;
+		memset(&zero_its, 0, sizeof(zero_its));
+
+		timerfd_settime(timer_fd, TFD_TIMER_ABSTIME, &zero_its, 0);
+
+		seq=(u32_t)get_true_random_number(); //TODO temp solution for a bug.
+
+		return 0;
+	}
+
+	my_time_t get_first_packet_time()
+	{
+		return first_packet_time_for_output;
+	}
+
+	int get_pending_time()
+	{
+		return fec_timeout;
+	}
+
+	int get_type()
+	{
+		return fec_mode;
+	}
+	u64_t get_timer_fd64();
+	int reset_fec_parameter(int data_num,int redundant_num,int mtu,int pending_num,int pending_time,int type);
+	int input(char *s,int len/*,int &is_first_packet*/);
+	int output(int &n,char ** &s_arr,int *&len);
+};
+struct fec_data_t
+{
+	int used;
+	u32_t seq;
+	int type;
+	int data_num;
+	int redundant_num;
+	int idx;
+	char buf[buf_len];
+	int len;
+};
+struct fec_group_t
+{
+	int type=-1;
+	int data_num=-1;
+	int redundant_num=-1;
+	int len=-1;
+	//int data_counter=0;
+	map<int,int>  group_mp;
+};
+class fec_decode_manager_t
+{
+	anti_replay_t anti_replay;
+	fec_data_t *fec_data;
+	unordered_map<u32_t, fec_group_t> mp;
+	blob_decode_t blob_decode;
+
+	int index;
+
+	int output_n;
+	char ** output_s_arr;
+	int * output_len_arr;
+	int ready_for_output;
+
+	char *output_s_arr_buf[max_fec_packet_num+100];//only for type=1,for type=0 the buf inside blot_t is used
+	int output_len_arr_buf[max_fec_packet_num+100];//same
+
+public:
+	fec_decode_manager_t()
+	{
+		fec_data=new fec_data_t[fec_buff_num+5];
+		clear();
+	}
+	fec_decode_manager_t(const fec_decode_manager_t &b)
+	{
+		assert(0==1);//not allowed to copy
+	}
+	~fec_decode_manager_t()
+	{
+		delete fec_data;
+	}
+	int clear()
+	{
+		anti_replay.clear();
+		mp.clear();
+		mp.rehash(fec_buff_num*3);
+
+		for(int i=0;i<(int)fec_buff_num;i++)
+			fec_data[i].used=0;
+		ready_for_output=0;
+		index=0;
+
+		return 0;
+	}
+
+	//int re_init();
+	int input(char *s,int len);
+	int output(int &n,char ** &s_arr,int* &len_arr);
+};
+
+#endif /* FEC_MANAGER_H_ */

images/cn/11

@@ -0,0 +1 @@
+

images/en/11

@@ -0,0 +1 @@
+

lib/fec.c

@@ -0,0 +1,917 @@
+/*
+ * fec.c -- forward error correction based on Vandermonde matrices
+ * 980624
+ * (C) 1997-98 Luigi Rizzo (luigi@iet.unipi.it)
+ *
+ * Portions derived from code by Phil Karn (karn@ka9q.ampr.org),
+ * Robert Morelos-Zaragoza (robert@spectra.eng.hawaii.edu) and Hari
+ * Thirumoorthy (harit@spectra.eng.hawaii.edu), Aug 1995
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above
+ *    copyright notice, this list of conditions and the following
+ *    disclaimer in the documentation and/or other materials
+ *    provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+ * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
+ * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY,
+ * OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
+ * OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
+ * TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
+ * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
+ * OF SUCH DAMAGE.
+ */
+
+/*
+ * The following parameter defines how many bits are used for
+ * field elements. The code supports any value from 2 to 16
+ * but fastest operation is achieved with 8 bit elements
+ * This is the only parameter you may want to change.
+ */
+#ifndef GF_BITS
+#define GF_BITS  8	/* code over GF(2**GF_BITS) - change to suit */
+#endif
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+typedef unsigned long u_long;
+/*
+ * compatibility stuff
+ */
+#ifdef MSDOS	/* but also for others, e.g. sun... */
+#define NEED_BCOPY
+#define bcmp(a,b,n) memcmp(a,b,n)
+#endif
+
+#ifdef NEED_BCOPY
+#define bcopy(s, d, siz)        memcpy((d), (s), (siz))
+#define bzero(d, siz)   memset((d), '\0', (siz))
+#endif
+
+/*
+ * stuff used for testing purposes only
+ */
+
+#ifdef	TEST
+#define DEB(x)
+#define DDB(x) x
+#define	DEBUG	0	/* minimal debugging */
+#ifdef	MSDOS
+#include <time.h>
+struct timeval {
+    unsigned long ticks;
+};
+#define gettimeofday(x, dummy) { (x)->ticks = clock() ; }
+#define DIFF_T(a,b) (1+ 1000000*(a.ticks - b.ticks) / CLOCKS_PER_SEC )
+typedef unsigned long u_long ;
+typedef unsigned short u_short ;
+#else /* typically, unix systems */
+#include <sys/time.h>
+#define DIFF_T(a,b) \
+	(1+ 1000000*(a.tv_sec - b.tv_sec) + (a.tv_usec - b.tv_usec) )
+#endif
+
+#define TICK(t) \
+	{struct timeval x ; \
+	gettimeofday(&x, NULL) ; \
+	t = x.tv_usec + 1000000* (x.tv_sec & 0xff ) ; \
+	}
+#define TOCK(t) \
+	{ u_long t1 ; TICK(t1) ; \
+	  if (t1 < t) t = 256000000 + t1 - t ; \
+	  else t = t1 - t ; \
+	  if (t == 0) t = 1 ;}
+	
+u_long ticks[10];	/* vars for timekeeping */
+#else
+#define DEB(x)
+#define DDB(x)
+#define TICK(x)
+#define TOCK(x)
+#endif /* TEST */
+
+/*
+ * You should not need to change anything beyond this point.
+ * The first part of the file implements linear algebra in GF.
+ *
+ * gf is the type used to store an element of the Galois Field.
+ * Must constain at least GF_BITS bits.
+ *
+ * Note: unsigned char will work up to GF(256) but int seems to run
+ * faster on the Pentium. We use int whenever have to deal with an
+ * index, since they are generally faster.
+ */
+#if (GF_BITS < 2  && GF_BITS >16)
+#error "GF_BITS must be 2 .. 16"
+#endif
+#if (GF_BITS <= 8)
+typedef unsigned char gf;
+#else
+typedef unsigned short gf;
+#endif
+
+#define	GF_SIZE ((1 << GF_BITS) - 1)	/* powers of \alpha */
+
+/*
+ * Primitive polynomials - see Lin & Costello, Appendix A,
+ * and  Lee & Messerschmitt, p. 453.
+ */
+static const char *allPp[] = {    /* GF_BITS	polynomial		*/
+    NULL,		    /*  0	no code			*/
+    NULL,		    /*  1	no code			*/
+    "111",		    /*  2	1+x+x^2			*/
+    "1101",		    /*  3	1+x+x^3			*/
+    "11001",		    /*  4	1+x+x^4			*/
+    "101001",		    /*  5	1+x^2+x^5		*/
+    "1100001",		    /*  6	1+x+x^6			*/
+    "10010001",		    /*  7	1 + x^3 + x^7		*/
+    "101110001",	    /*  8	1+x^2+x^3+x^4+x^8	*/
+    "1000100001",	    /*  9	1+x^4+x^9		*/
+    "10010000001",	    /* 10	1+x^3+x^10		*/
+    "101000000001",	    /* 11	1+x^2+x^11		*/
+    "1100101000001",	    /* 12	1+x+x^4+x^6+x^12	*/
+    "11011000000001",	    /* 13	1+x+x^3+x^4+x^13	*/
+    "110000100010001",	    /* 14	1+x+x^6+x^10+x^14	*/
+    "1100000000000001",	    /* 15	1+x+x^15		*/
+    "11010000000010001"	    /* 16	1+x+x^3+x^12+x^16	*/
+};
+
+
+
+/*
+ * To speed up computations, we have tables for logarithm, exponent
+ * and inverse of a number. If GF_BITS <= 8, we use a table for
+ * multiplication as well (it takes 64K, no big deal even on a PDA,
+ * especially because it can be pre-initialized an put into a ROM!),
+ * otherwhise we use a table of logarithms.
+ * In any case the macro gf_mul(x,y) takes care of multiplications.
+ */
+
+static gf gf_exp[2*GF_SIZE];	/* index->poly form conversion table	*/
+static int gf_log[GF_SIZE + 1];	/* Poly->index form conversion table	*/
+static gf inverse[GF_SIZE+1];	/* inverse of field elem.		*/
+				/* inv[\alpha**i]=\alpha**(GF_SIZE-i-1)	*/
+
+/*
+ * modnn(x) computes x % GF_SIZE, where GF_SIZE is 2**GF_BITS - 1,
+ * without a slow divide.
+ */
+static inline gf
+modnn(int x)
+{
+    while (x >= GF_SIZE) {
+	x -= GF_SIZE;
+	x = (x >> GF_BITS) + (x & GF_SIZE);
+    }
+    return x;
+}
+
+#define SWAP(a,b,t) {t tmp; tmp=a; a=b; b=tmp;}
+
+/*
+ * gf_mul(x,y) multiplies two numbers. If GF_BITS<=8, it is much
+ * faster to use a multiplication table.
+ *
+ * USE_GF_MULC, GF_MULC0(c) and GF_ADDMULC(x) can be used when multiplying
+ * many numbers by the same constant. In this case the first
+ * call sets the constant, and others perform the multiplications.
+ * A value related to the multiplication is held in a local variable
+ * declared with USE_GF_MULC . See usage in addmul1().
+ */
+#if (GF_BITS <= 8)
+static gf gf_mul_table[GF_SIZE + 1][GF_SIZE + 1];
+
+#define gf_mul(x,y) gf_mul_table[x][y]
+
+#define USE_GF_MULC register gf * __gf_mulc_
+#define GF_MULC0(c) __gf_mulc_ = gf_mul_table[c]
+#define GF_ADDMULC(dst, x) dst ^= __gf_mulc_[x]
+
+static void
+init_mul_table()
+{
+    int i, j;
+    for (i=0; i< GF_SIZE+1; i++)
+	for (j=0; j< GF_SIZE+1; j++)
+	    gf_mul_table[i][j] = gf_exp[modnn(gf_log[i] + gf_log[j]) ] ;
+
+    for (j=0; j< GF_SIZE+1; j++)
+	    gf_mul_table[0][j] = gf_mul_table[j][0] = 0;
+}
+#else	/* GF_BITS > 8 */
+static inline gf
+gf_mul(x,y)
+{
+    if ( (x) == 0 || (y)==0 ) return 0;
+     
+    return gf_exp[gf_log[x] + gf_log[y] ] ;
+}
+#define init_mul_table()
+
+#define USE_GF_MULC register gf * __gf_mulc_
+#define GF_MULC0(c) __gf_mulc_ = &gf_exp[ gf_log[c] ]
+#define GF_ADDMULC(dst, x) { if (x) dst ^= __gf_mulc_[ gf_log[x] ] ; }
+#endif
+
+/*
+ * Generate GF(2**m) from the irreducible polynomial p(X) in p[0]..p[m]
+ * Lookup tables:
+ *     index->polynomial form		gf_exp[] contains j= \alpha^i;
+ *     polynomial form -> index form	gf_log[ j = \alpha^i ] = i
+ * \alpha=x is the primitive element of GF(2^m)
+ *
+ * For efficiency, gf_exp[] has size 2*GF_SIZE, so that a simple
+ * multiplication of two numbers can be resolved without calling modnn
+ */
+
+/*
+ * i use malloc so many times, it is easier to put checks all in
+ * one place.
+ */
+static void *
+my_malloc(int sz, const char *err_string)
+{
+    void *p = malloc( sz );
+    if (p == NULL) {
+	fprintf(stderr, "-- malloc failure allocating %s\n", err_string);
+	exit(1) ;
+    }
+    return p ;
+}
+
+#define NEW_GF_MATRIX(rows, cols) \
+    (gf *)my_malloc(rows * cols * sizeof(gf), " ## __LINE__ ## " )
+
+/*
+ * initialize the data structures used for computations in GF.
+ */
+static void
+generate_gf(void)
+{
+    int i;
+    gf mask;
+    const char *Pp =  allPp[GF_BITS] ;
+
+    mask = 1;	/* x ** 0 = 1 */
+    gf_exp[GF_BITS] = 0; /* will be updated at the end of the 1st loop */
+    /*
+     * first, generate the (polynomial representation of) powers of \alpha,
+     * which are stored in gf_exp[i] = \alpha ** i .
+     * At the same time build gf_log[gf_exp[i]] = i .
+     * The first GF_BITS powers are simply bits shifted to the left.
+     */
+    for (i = 0; i < GF_BITS; i++, mask <<= 1 ) {
+	gf_exp[i] = mask;
+	gf_log[gf_exp[i]] = i;
+	/*
+	 * If Pp[i] == 1 then \alpha ** i occurs in poly-repr
+	 * gf_exp[GF_BITS] = \alpha ** GF_BITS
+	 */
+	if ( Pp[i] == '1' )
+	    gf_exp[GF_BITS] ^= mask;
+    }
+    /*
+     * now gf_exp[GF_BITS] = \alpha ** GF_BITS is complete, so can als
+     * compute its inverse.
+     */
+    gf_log[gf_exp[GF_BITS]] = GF_BITS;
+    /*
+     * Poly-repr of \alpha ** (i+1) is given by poly-repr of
+     * \alpha ** i shifted left one-bit and accounting for any
+     * \alpha ** GF_BITS term that may occur when poly-repr of
+     * \alpha ** i is shifted.
+     */
+    mask = 1 << (GF_BITS - 1 ) ;
+    for (i = GF_BITS + 1; i < GF_SIZE; i++) {
+	if (gf_exp[i - 1] >= mask)
+	    gf_exp[i] = gf_exp[GF_BITS] ^ ((gf_exp[i - 1] ^ mask) << 1);
+	else
+	    gf_exp[i] = gf_exp[i - 1] << 1;
+	gf_log[gf_exp[i]] = i;
+    }
+    /*
+     * log(0) is not defined, so use a special value
+     */
+    gf_log[0] =	GF_SIZE ;
+    /* set the extended gf_exp values for fast multiply */
+    for (i = 0 ; i < GF_SIZE ; i++)
+	gf_exp[i + GF_SIZE] = gf_exp[i] ;
+
+    /*
+     * again special cases. 0 has no inverse. This used to
+     * be initialized to GF_SIZE, but it should make no difference
+     * since noone is supposed to read from here.
+     */
+    inverse[0] = 0 ;
+    inverse[1] = 1;
+    for (i=2; i<=GF_SIZE; i++)
+	inverse[i] = gf_exp[GF_SIZE-gf_log[i]];
+}
+
+/*
+ * Various linear algebra operations that i use often.
+ */
+
+/*
+ * addmul() computes dst[] = dst[] + c * src[]
+ * This is used often, so better optimize it! Currently the loop is
+ * unrolled 16 times, a good value for 486 and pentium-class machines.
+ * The case c=0 is also optimized, whereas c=1 is not. These
+ * calls are unfrequent in my typical apps so I did not bother.
+ * 
+ * Note that gcc on
+ */
+#define addmul(dst, src, c, sz) \
+    if (c != 0) addmul1(dst, src, c, sz)
+
+#define UNROLL 16 /* 1, 4, 8, 16 */
+static void
+addmul1(gf *dst1, gf *src1, gf c, int sz)
+{
+    USE_GF_MULC ;
+    register gf *dst = dst1, *src = src1 ;
+    gf *lim = &dst[sz - UNROLL + 1] ;
+
+    GF_MULC0(c) ;
+
+#if (UNROLL > 1) /* unrolling by 8/16 is quite effective on the pentium */
+    for (; dst < lim ; dst += UNROLL, src += UNROLL ) {
+	GF_ADDMULC( dst[0] , src[0] );
+	GF_ADDMULC( dst[1] , src[1] );
+	GF_ADDMULC( dst[2] , src[2] );
+	GF_ADDMULC( dst[3] , src[3] );
+#if (UNROLL > 4)
+	GF_ADDMULC( dst[4] , src[4] );
+	GF_ADDMULC( dst[5] , src[5] );
+	GF_ADDMULC( dst[6] , src[6] );
+	GF_ADDMULC( dst[7] , src[7] );
+#endif
+#if (UNROLL > 8)
+	GF_ADDMULC( dst[8] , src[8] );
+	GF_ADDMULC( dst[9] , src[9] );
+	GF_ADDMULC( dst[10] , src[10] );
+	GF_ADDMULC( dst[11] , src[11] );
+	GF_ADDMULC( dst[12] , src[12] );
+	GF_ADDMULC( dst[13] , src[13] );
+	GF_ADDMULC( dst[14] , src[14] );
+	GF_ADDMULC( dst[15] , src[15] );
+#endif
+    }
+#endif
+    lim += UNROLL - 1 ;
+    for (; dst < lim; dst++, src++ )		/* final components */
+	GF_ADDMULC( *dst , *src );
+}
+
+/*
+ * computes C = AB where A is n*k, B is k*m, C is n*m
+ */
+static void
+matmul(gf *a, gf *b, gf *c, int n, int k, int m)
+{
+    int row, col, i ;
+
+    for (row = 0; row < n ; row++) {
+	for (col = 0; col < m ; col++) {
+	    gf *pa = &a[ row * k ];
+	    gf *pb = &b[ col ];
+	    gf acc = 0 ;
+	    for (i = 0; i < k ; i++, pa++, pb += m )
+		acc ^= gf_mul( *pa, *pb ) ;
+	    c[ row * m + col ] = acc ;
+	}
+    }
+}
+
+#ifdef DEBUG
+/*
+ * returns 1 if the square matrix is identiy
+ * (only for test)
+ */
+static int
+is_identity(gf *m, int k)
+{
+    int row, col ;
+    for (row=0; row<k; row++)
+	for (col=0; col<k; col++)
+	    if ( (row==col && *m != 1) ||
+		 (row!=col && *m != 0) )
+		 return 0 ;
+	    else
+		m++ ;
+    return 1 ;
+}
+#endif /* debug */
+
+/*
+ * invert_mat() takes a matrix and produces its inverse
+ * k is the size of the matrix.
+ * (Gauss-Jordan, adapted from Numerical Recipes in C)
+ * Return non-zero if singular.
+ */
+DEB( int pivloops=0; int pivswaps=0 ; /* diagnostic */)
+static int
+invert_mat(gf *src, int k)
+{
+    gf c, *p ;
+    int irow, icol, row, col, i, ix ;
+
+    int error = 1 ;
+    int *indxc = (int*)my_malloc(k*sizeof(int), "indxc");
+    int *indxr = (int*)my_malloc(k*sizeof(int), "indxr");
+    int *ipiv = (int*)my_malloc(k*sizeof(int), "ipiv");
+    gf *id_row = NEW_GF_MATRIX(1, k);
+    gf *temp_row = NEW_GF_MATRIX(1, k);
+
+    bzero(id_row, k*sizeof(gf));
+    DEB( pivloops=0; pivswaps=0 ; /* diagnostic */ )
+    /*
+     * ipiv marks elements already used as pivots.
+     */
+    for (i = 0; i < k ; i++)
+	ipiv[i] = 0 ;
+
+    for (col = 0; col < k ; col++) {
+	gf *pivot_row ;
+	/*
+	 * Zeroing column 'col', look for a non-zero element.
+	 * First try on the diagonal, if it fails, look elsewhere.
+	 */
+	irow = icol = -1 ;
+	if (ipiv[col] != 1 && src[col*k + col] != 0) {
+	    irow = col ;
+	    icol = col ;
+	    goto found_piv ;
+	}
+	for (row = 0 ; row < k ; row++) {
+	    if (ipiv[row] != 1) {
+		for (ix = 0 ; ix < k ; ix++) {
+		    DEB( pivloops++ ; )
+		    if (ipiv[ix] == 0) {
+			if (src[row*k + ix] != 0) {
+			    irow = row ;
+			    icol = ix ;
+			    goto found_piv ;
+			}
+		    } else if (ipiv[ix] > 1) {
+			fprintf(stderr, "singular matrix\n");
+			goto fail ; 
+		    }
+		}
+	    }
+	}
+	if (icol == -1) {
+	    fprintf(stderr, "XXX pivot not found!\n");
+	    goto fail ;
+	}
+found_piv:
+	++(ipiv[icol]) ;
+	/*
+	 * swap rows irow and icol, so afterwards the diagonal
+	 * element will be correct. Rarely done, not worth
+	 * optimizing.
+	 */
+	if (irow != icol) {
+	    for (ix = 0 ; ix < k ; ix++ ) {
+		SWAP( src[irow*k + ix], src[icol*k + ix], gf) ;
+	    }
+	}
+	indxr[col] = irow ;
+	indxc[col] = icol ;
+	pivot_row = &src[icol*k] ;
+	c = pivot_row[icol] ;
+	if (c == 0) {
+	    fprintf(stderr, "singular matrix 2\n");
+	    goto fail ;
+	}
+	if (c != 1 ) { /* otherwhise this is a NOP */
+	    /*
+	     * this is done often , but optimizing is not so
+	     * fruitful, at least in the obvious ways (unrolling)
+	     */
+	    DEB( pivswaps++ ; )
+	    c = inverse[ c ] ;
+	    pivot_row[icol] = 1 ;
+	    for (ix = 0 ; ix < k ; ix++ )
+		pivot_row[ix] = gf_mul(c, pivot_row[ix] );
+	}
+	/*
+	 * from all rows, remove multiples of the selected row
+	 * to zero the relevant entry (in fact, the entry is not zero
+	 * because we know it must be zero).
+	 * (Here, if we know that the pivot_row is the identity,
+	 * we can optimize the addmul).
+	 */
+	id_row[icol] = 1;
+	if (bcmp(pivot_row, id_row, k*sizeof(gf)) != 0) {
+	    for (p = src, ix = 0 ; ix < k ; ix++, p += k ) {
+		if (ix != icol) {
+		    c = p[icol] ;
+		    p[icol] = 0 ;
+		    addmul(p, pivot_row, c, k );
+		}
+	    }
+	}
+	id_row[icol] = 0;
+    } /* done all columns */
+    for (col = k-1 ; col >= 0 ; col-- ) {
+	if (indxr[col] <0 || indxr[col] >= k)
+	    fprintf(stderr, "AARGH, indxr[col] %d\n", indxr[col]);
+	else if (indxc[col] <0 || indxc[col] >= k)
+	    fprintf(stderr, "AARGH, indxc[col] %d\n", indxc[col]);
+	else
+	if (indxr[col] != indxc[col] ) {
+	    for (row = 0 ; row < k ; row++ ) {
+		SWAP( src[row*k + indxr[col]], src[row*k + indxc[col]], gf) ;
+	    }
+	}
+    }
+    error = 0 ;
+fail:
+    free(indxc);
+    free(indxr);
+    free(ipiv);
+    free(id_row);
+    free(temp_row);
+    return error ;
+}
+
+/*
+ * fast code for inverting a vandermonde matrix.
+ * XXX NOTE: It assumes that the matrix
+ * is not singular and _IS_ a vandermonde matrix. Only uses
+ * the second column of the matrix, containing the p_i's.
+ *
+ * Algorithm borrowed from "Numerical recipes in C" -- sec.2.8, but
+ * largely revised for my purposes.
+ * p = coefficients of the matrix (p_i)
+ * q = values of the polynomial (known)
+ */
+
+int
+invert_vdm(gf *src, int k)
+{
+    int i, j, row, col ;
+    gf *b, *c, *p;
+    gf t, xx ;
+
+    if (k == 1) 	/* degenerate case, matrix must be p^0 = 1 */
+	return 0 ;
+    /*
+     * c holds the coefficient of P(x) = Prod (x - p_i), i=0..k-1
+     * b holds the coefficient for the matrix inversion
+     */
+    c = NEW_GF_MATRIX(1, k);
+    b = NEW_GF_MATRIX(1, k);
+
+    p = NEW_GF_MATRIX(1, k);
+   
+    for ( j=1, i = 0 ; i < k ; i++, j+=k ) {
+	c[i] = 0 ;
+	p[i] = src[j] ;    /* p[i] */
+    }
+    /*
+     * construct coeffs. recursively. We know c[k] = 1 (implicit)
+     * and start P_0 = x - p_0, then at each stage multiply by
+     * x - p_i generating P_i = x P_{i-1} - p_i P_{i-1}
+     * After k steps we are done.
+     */
+    c[k-1] = p[0] ;	/* really -p(0), but x = -x in GF(2^m) */
+    for (i = 1 ; i < k ; i++ ) {
+	gf p_i = p[i] ; /* see above comment */
+	for (j = k-1  - ( i - 1 ) ; j < k-1 ; j++ )
+	    c[j] ^= gf_mul( p_i, c[j+1] ) ;
+	c[k-1] ^= p_i ;
+    }
+
+    for (row = 0 ; row < k ; row++ ) {
+	/*
+	 * synthetic division etc.
+	 */
+	xx = p[row] ;
+	t = 1 ;
+	b[k-1] = 1 ; /* this is in fact c[k] */
+	for (i = k-2 ; i >= 0 ; i-- ) {
+	    b[i] = c[i+1] ^ gf_mul(xx, b[i+1]) ;
+	    t = gf_mul(xx, t) ^ b[i] ;
+	}
+	for (col = 0 ; col < k ; col++ )
+	    src[col*k + row] = gf_mul(inverse[t], b[col] );
+    }
+    free(c) ;
+    free(b) ;
+    free(p) ;
+    return 0 ;
+}
+
+static int fec_initialized = 0 ;
+static void
+init_fec()
+{
+    TICK(ticks[0]);
+    generate_gf();
+    TOCK(ticks[0]);
+    DDB(fprintf(stderr, "generate_gf took %ldus\n", ticks[0]);)
+    TICK(ticks[0]);
+    init_mul_table();
+    TOCK(ticks[0]);
+    DDB(fprintf(stderr, "init_mul_table took %ldus\n", ticks[0]);)
+    fec_initialized = 1 ;
+}
+
+/*
+ * This section contains the proper FEC encoding/decoding routines.
+ * The encoding matrix is computed starting with a Vandermonde matrix,
+ * and then transforming it into a systematic matrix.
+ */
+
+#define FEC_MAGIC	0xFECC0DEC
+
+struct fec_parms {
+    u_long magic ;
+    int k, n ;		/* parameters of the code */
+    gf *enc_matrix ;
+} ;
+
+void
+fec_free(void *p0)
+{
+	struct fec_parms *p= (struct fec_parms *) p0;
+    if (p==NULL ||
+       p->magic != ( ( (FEC_MAGIC ^ p->k) ^ p->n) ^ (int)((long)p->enc_matrix)) ) {
+	fprintf(stderr, "bad parameters to fec_free\n");
+	return ;
+    }
+    free(p->enc_matrix);
+    free(p);
+}
+
+/*
+ * create a new encoder, returning a descriptor. This contains k,n and
+ * the encoding matrix.
+ */
+struct fec_parms *
+fec_new(int k, int n)
+{
+    int row, col ;
+    gf *p, *tmp_m ;
+
+    struct fec_parms *retval ;
+
+    if (fec_initialized == 0)
+	init_fec();
+
+    if (k > GF_SIZE + 1 || n > GF_SIZE + 1 || k > n ) {
+	fprintf(stderr, "Invalid parameters k %d n %d GF_SIZE %d\n",
+		k, n, GF_SIZE );
+	return NULL ;
+    }
+    retval = (struct fec_parms *)my_malloc(sizeof(struct fec_parms), "new_code");
+    retval->k = k ;
+    retval->n = n ;
+    retval->enc_matrix = NEW_GF_MATRIX(n, k);
+    retval->magic = ( ( FEC_MAGIC ^ k) ^ n) ^ (int)((long)retval->enc_matrix) ;
+    tmp_m = NEW_GF_MATRIX(n, k);
+    /*
+     * fill the matrix with powers of field elements, starting from 0.
+     * The first row is special, cannot be computed with exp. table.
+     */
+    tmp_m[0] = 1 ;
+    for (col = 1; col < k ; col++)
+	tmp_m[col] = 0 ;
+    for (p = tmp_m + k, row = 0; row < n-1 ; row++, p += k) {
+	for ( col = 0 ; col < k ; col ++ )
+	    p[col] = gf_exp[modnn(row*col)];
+    }
+
+    /*
+     * quick code to build systematic matrix: invert the top
+     * k*k vandermonde matrix, multiply right the bottom n-k rows
+     * by the inverse, and construct the identity matrix at the top.
+     */
+    TICK(ticks[3]);
+    invert_vdm(tmp_m, k); /* much faster than invert_mat */
+    matmul(tmp_m + k*k, tmp_m, retval->enc_matrix + k*k, n - k, k, k);
+    /*
+     * the upper matrix is I so do not bother with a slow multiply
+     */
+    bzero(retval->enc_matrix, k*k*sizeof(gf) );
+    for (p = retval->enc_matrix, col = 0 ; col < k ; col++, p += k+1 )
+	*p = 1 ;
+    free(tmp_m);
+    TOCK(ticks[3]);
+
+    DDB(fprintf(stderr, "--- %ld us to build encoding matrix\n",
+	    ticks[3]);)
+    DEB(pr_matrix(retval->enc_matrix, n, k, "encoding_matrix");)
+    return retval ;
+}
+
+/*
+ * fec_encode accepts as input pointers to n data packets of size sz,
+ * and produces as output a packet pointed to by fec, computed
+ * with index "index".
+ */
+void
+fec_encode(void *code0, void *src0[], void *fec0, int index, int sz)
+//fec_encode(struct fec_parms *code0, gf *src[], gf *fec, int index, int sz)
+{
+	struct fec_parms * code= (struct fec_parms *)code0;
+	gf **src=(gf**) src0;
+	gf* fec=(gf*)fec0;
+    int i, k = code->k ;
+    gf *p ;
+
+    if (GF_BITS > 8)
+	sz /= 2 ;
+
+    if (index < k)
+         bcopy(src[index], fec, sz*sizeof(gf) ) ;
+    else if (index < code->n) {
+	p = &(code->enc_matrix[index*k] );
+	bzero(fec, sz*sizeof(gf));
+	for (i = 0; i < k ; i++)
+	    addmul(fec, src[i], p[i], sz ) ;
+    } else
+	fprintf(stderr, "Invalid index %d (max %d)\n",
+	    index, code->n - 1 );
+}
+
+/*
+ * shuffle move src packets in their position
+ */
+static int
+shuffle(gf *pkt[], int index[], int k)
+{
+    int i;
+
+    for ( i = 0 ; i < k ; ) {
+	if (index[i] >= k || index[i] == i)
+	    i++ ;
+	else {
+	    /*
+	     * put pkt in the right position (first check for conflicts).
+	     */
+	    int c = index[i] ;
+
+	    if (index[c] == c) {
+		DEB(fprintf(stderr, "\nshuffle, error at %d\n", i);)
+		return 1 ;
+	    }
+	    SWAP(index[i], index[c], int) ;
+	    SWAP(pkt[i], pkt[c], gf *) ;
+	}
+    }
+    DEB( /* just test that it works... */
+    for ( i = 0 ; i < k ; i++ ) {
+	if (index[i] < k && index[i] != i) {
+	    fprintf(stderr, "shuffle: after\n");
+	    for (i=0; i<k ; i++) fprintf(stderr, "%3d ", index[i]);
+	    fprintf(stderr, "\n");
+	    return 1 ;
+	}
+    }
+    )
+    return 0 ;
+}
+
+/*
+ * build_decode_matrix constructs the encoding matrix given the
+ * indexes. The matrix must be already allocated as
+ * a vector of k*k elements, in row-major order
+ */
+static gf *
+build_decode_matrix(struct fec_parms *code, gf *pkt[], int index[])
+{
+    int i , k = code->k ;
+    gf *p, *matrix = NEW_GF_MATRIX(k, k);
+
+    TICK(ticks[9]);
+    for (i = 0, p = matrix ; i < k ; i++, p += k ) {
+#if 1 /* this is simply an optimization, not very useful indeed */
+	if (index[i] < k) {
+	    bzero(p, k*sizeof(gf) );
+	    p[i] = 1 ;
+	} else
+#endif
+	if (index[i] < code->n )
+	    bcopy( &(code->enc_matrix[index[i]*k]), p, k*sizeof(gf) ); 
+	else {
+	    fprintf(stderr, "decode: invalid index %d (max %d)\n",
+		index[i], code->n - 1 );
+	    free(matrix) ;
+	    return NULL ;
+	}
+    }
+    TICK(ticks[9]);
+    if (invert_mat(matrix, k)) {
+	free(matrix);
+	matrix = NULL ;
+    }
+    TOCK(ticks[9]);
+    return matrix ;
+}
+
+/*
+ * fec_decode receives as input a vector of packets, the indexes of
+ * packets, and produces the correct vector as output.
+ *
+ * Input:
+ *	code: pointer to code descriptor
+ *	pkt:  pointers to received packets. They are modified
+ *	      to store the output packets (in place)
+ *	index: pointer to packet indexes (modified)
+ *	sz:    size of each packet
+ */
+int
+fec_decode(void *code0, void *pkt0[], int index[], int sz)
+//fec_decode(struct fec_parms *code, gf *pkt[], int index[], int sz)
+{
+	struct fec_parms * code=(struct fec_parms*)code0;
+	gf **pkt=(gf**)pkt0;
+    gf *m_dec ; 
+    gf **new_pkt ;
+    int row, col , k = code->k ;
+
+    if (GF_BITS > 8)
+	sz /= 2 ;
+
+    if (shuffle(pkt, index, k))	/* error if true */
+	return 1 ;
+    m_dec = build_decode_matrix(code, pkt, index);
+
+    if (m_dec == NULL)
+	return 1 ; /* error */
+    /*
+     * do the actual decoding
+     */
+    new_pkt = (gf** )my_malloc (k * sizeof (gf * ), "new pkt pointers" );
+    for (row = 0 ; row < k ; row++ ) {
+	if (index[row] >= k) {
+	    new_pkt[row] = (gf*)my_malloc (sz * sizeof (gf), "new pkt buffer" );
+	    bzero(new_pkt[row], sz * sizeof(gf) ) ;
+	    for (col = 0 ; col < k ; col++ )
+		addmul(new_pkt[row], pkt[col], m_dec[row*k + col], sz) ;
+	}
+    }
+    /*
+     * move pkts to their final destination
+     */
+    for (row = 0 ; row < k ; row++ ) {
+	if (index[row] >= k) {
+	    bcopy(new_pkt[row], pkt[row], sz*sizeof(gf));
+	    free(new_pkt[row]);
+	}
+    }
+    free(new_pkt);
+    free(m_dec);
+
+    return 0;
+}
+int get_n(void *code0)
+{
+	struct fec_parms * code= (struct fec_parms *)code0;
+	return code->n;
+}
+int get_k(void *code0)
+{
+	struct fec_parms * code= (struct fec_parms *)code0;
+	return code->k;
+}
+/*********** end of FEC code -- beginning of test code ************/
+
+#if (TEST || DEBUG)
+void
+test_gf()
+{
+    int i ;
+    /*
+     * test gf tables. Sufficiently tested...
+     */
+    for (i=0; i<= GF_SIZE; i++) {
+        if (gf_exp[gf_log[i]] != i)
+	    fprintf(stderr, "bad exp/log i %d log %d exp(log) %d\n",
+		i, gf_log[i], gf_exp[gf_log[i]]);
+
+        if (i != 0 && gf_mul(i, inverse[i]) != 1)
+	    fprintf(stderr, "bad mul/inv i %d inv %d i*inv(i) %d\n",
+		i, inverse[i], gf_mul(i, inverse[i]) );
+	if (gf_mul(0,i) != 0)
+	    fprintf(stderr, "bad mul table 0,%d\n",i);
+	if (gf_mul(i,0) != 0)
+	    fprintf(stderr, "bad mul table %d,0\n",i);
+    }
+}
+#endif /* TEST */

lib/fec.h

@@ -0,0 +1,56 @@
+/*
+ * fec.c -- forward error correction based on Vandermonde matrices
+ * 980614
+ * (C) 1997-98 Luigi Rizzo (luigi@iet.unipi.it)
+ *
+ * Portions derived from code by Phil Karn (karn@ka9q.ampr.org),
+ * Robert Morelos-Zaragoza (robert@spectra.eng.hawaii.edu) and Hari
+ * Thirumoorthy (harit@spectra.eng.hawaii.edu), Aug 1995
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above
+ *    copyright notice, this list of conditions and the following
+ *    disclaimer in the documentation and/or other materials
+ *    provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+ * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
+ * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY,
+ * OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
+ * OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR
+ * TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
+ * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY
+ * OF SUCH DAMAGE.
+ */
+
+/*
+ * The following parameter defines how many bits are used for
+ * field elements. The code supports any value from 2 to 16
+ * but fastest operation is achieved with 8 bit elements
+ * This is the only parameter you may want to change.
+ */
+#ifndef GF_BITS
+#define GF_BITS  8	/* code over GF(2**GF_BITS) - change to suit */
+#endif
+
+#define	GF_SIZE ((1 << GF_BITS) - 1)	/* powers of \alpha */
+void fec_free(void *p) ;
+void * fec_new(int k, int n) ;//n>=k
+
+void init_fec() ;  //if you never called this,it will be automatically called in fec_new()
+void fec_encode(void *code, void *src[], void *dst, int index, int sz) ;
+int fec_decode(void *code, void *pkt[], int index[], int sz) ;
+
+int get_k(void *code);
+int get_n(void *code);
+
+/* end of file */

lib/rs.c

@@ -0,0 +1,76 @@
+/*
+ * rs.c
+ *
+ *  Created on: Sep 14, 2017
+ *      Author: root
+ */
+#include "rs.h"
+#include "stdlib.h"
+#include "string.h"
+
+void rs_encode(void *code,char *data[],int size)
+{
+	int k=get_k(code);
+	int n=get_n(code);
+	for(int i=k;i<n;i++)
+	{
+		fec_encode(code, (void **)data, data[i],i, size);
+	}
+
+	return ;
+}
+
+int rs_decode(void *code,char *data[],int size)
+{
+	int k=get_k(code);
+	int n=get_n(code);
+	int index[n];
+	int count=0;
+	for(int i=0;i<n;i++)
+	{
+		if(data[i]!=0)
+		{
+			index[count++]=i;
+		}
+	}
+	if(count<k)
+		return -1;
+	for(int i=0;i<n;i++)
+	{
+		if(i<count)
+			data[i]=data[index[i]];
+		else
+			data[i]=0;
+	}
+	return fec_decode(code,(void**)data,index,size);
+}
+
+static void * (*table)[256]=0;
+void* get_code(int k,int n)
+{
+	if (table==0)
+	{
+		table=(void* (*)[256]) malloc(sizeof(void*)*256*256);
+		if(!table)
+		{
+		    return table;
+		}
+		memset(table,0,sizeof(void*)*256*256);
+	}
+	if(table[k][n]==0)
+	{
+		table[k][n]=fec_new(k,n);
+	}
+	return table[k][n];
+}
+void rs_encode2(int k,int n,char *data[],int size)
+{
+	void* code=get_code(k,n);
+	rs_encode(code,data,size);
+}
+
+int rs_decode2(int k,int n,char *data[],int size)
+{
+	void* code=get_code(k,n);
+	return rs_decode(code,data,size);
+}

lib/rs.h

@@ -0,0 +1,51 @@
+/*
+ * rs.h
+ *
+ *  Created on: Sep 14, 2017
+ *      Author: root
+ */
+
+#ifndef LIB_RS_H_
+#define LIB_RS_H_
+
+#include "fec.h"
+
+// input:
+// code, generated by fec_new() function from fec.h
+// data[0....k-1 ], points to original data
+// size, data length
+//
+// output:
+// data[k....n-1], points to generated redundant data
+//
+// info:
+// the function will always succeed,except malloc fail.if malloc fail,it will call exit()
+void rs_encode(void *code,char *data[],int size); 
+
+
+// input:
+// data[0.....n-1] points to original data and redundate data,in right order
+// if data[i] is missing ,set poniter data[i] to 0 (point it to null)
+//
+// outout:
+// data[0.....k-1] will point to the recovered original data.
+//
+// info:
+// return zero on success
+// if the number of no-zero pointers is less than k,the function will fail and return non-zero
+//
+// advanced info:
+// 1. rs_decode wont malloc memory for those zero pointers in data[0.....k-1]. instead it will re-use the memory of other non-zero pointers (and let data[0.....k-1] point to those memory).
+// 2. if the input data[0.....n-1] contains x non-zero pointers,after called rs_decode,there will still be exactly x non-zero poninters in data[0.....n-1],just the order may change.
+int rs_decode(void *code,char *data[],int size);
+
+
+void rs_encode2(int k,int n,char *data[],int size);
+
+int rs_decode2(int k,int n,char *data[],int size);
+
+
+
+
+
+#endif /* LIB_RS_H_ */

makefile

@@ -1,18 +1,69 @@
-cross_cc=mips-openwrt-linux-g++
+cc_cross=/home/wangyu/Desktop/arm-2014.05/bin/arm-none-linux-gnueabi-g++
-FLAGS=-Wall -Wextra -Wno-unused-variable -Wno-unused-parameter -ggdb
+cc_local=g++
-FLAGS2= -O3
+#cc_mips34kc=/toolchains/OpenWrt-SDK-ar71xx-for-linux-x86_64-gcc-4.8-linaro_uClibc-0.9.33.2/staging_dir/toolchain-mips_34kc_gcc-4.8-linaro_uClibc-0.9.33.2/bin/mips-openwrt-linux-g++
-all:
+cc_mips24kc_be=/toolchains/lede-sdk-17.01.2-ar71xx-generic_gcc-5.4.0_musl-1.1.16.Linux-x86_64/staging_dir/toolchain-mips_24kc_gcc-5.4.0_musl-1.1.16/bin/mips-openwrt-linux-musl-g++
-	g++ main.cpp common.cpp log.cpp -I. -o speeder -static -lrt -std=c++11  ${FLAGS} ${FLAGS2}
+cc_mips24kc_le=/toolchains/lede-sdk-17.01.2-ramips-mt7621_gcc-5.4.0_musl-1.1.16.Linux-x86_64/staging_dir/toolchain-mipsel_24kc_gcc-5.4.0_musl-1.1.16/bin/mipsel-openwrt-linux-musl-g++
-release:
+#cc_arm= /toolchains/gcc-linaro-4.9.4-2017.01-x86_64_arm-linux-gnueabi/bin/arm-linux-gnueabi-g++ -march=armv6 -marm 
-	g++ main.cpp common.cpp log.cpp -I. -o speeder_amd64 -static -lrt -std=c++11  ${FLAGS} ${FLAGS2} 
+cc_arm= /toolchains/arm-2014.05/bin/arm-none-linux-gnueabi-g++
-	g++ -m32 main.cpp common.cpp log.cpp -I. -o speeder_x86 -static -lrt -std=c++11  ${FLAGS} ${FLAGS2}
+#cc_bcm2708=/home/wangyu/raspberry/tools/arm-bcm2708/gcc-linaro-arm-linux-gnueabihf-raspbian/bin/arm-linux-gnueabihf-g++ 
-	${cross_cc} main.cpp common.cpp log.cpp -I. -o speeder_ar71xx -lrt -std=c++11  ${FLAGS} ${FLAGS2}
+FLAGS= -std=c++11   -Wall -Wextra -Wno-unused-variable -Wno-unused-parameter -Wno-missing-field-initializers -ggdb
-	tar -zcvf udp_speeder_binaries.tar.gz speeder_amd64  speeder_x86  speeder_ar71xx
-cross:
-	${cross_cc} main.cpp common.cpp log.cpp -I. -o speeder_cross -lrt -std=c++11  ${FLAGS} ${FLAGS2}
 
-debug:
+SOURCES=main.cpp log.cpp common.cpp lib/fec.c lib/rs.c packet.cpp delay_manager.cpp fd_manager.cpp connection.cpp fec_manager.cpp misc.cpp tunnel.cpp
-	g++ main.cpp common.cpp log.cpp -I. -o speeder -static -lrt -std=c++11  ${FLAGS} -Wformat-nonliteral -D MY_DEBUG
+
+NAME=speederv2
+TARGETS=amd64 arm mips24kc_be x86  mips24kc_le
+
+TAR=${NAME}_binaries.tar.gz `echo ${TARGETS}|sed -r 's/([^ ]+)/speederv2_\1/g'` version.txt
+
+all:git_version
+	rm -f ${NAME}
+	${cc_local}   -o ${NAME}          -I. ${SOURCES} ${FLAGS} -lrt -ggdb -static -O3
+debug: git_version
+	rm -f ${NAME}
+	${cc_local}   -o ${NAME}          -I. ${SOURCES} ${FLAGS} -lrt -Wformat-nonliteral -D MY_DEBUG 
+debug2: git_version
+	rm -f ${NAME}
+	${cc_local}   -o ${NAME}          -I. ${SOURCES} ${FLAGS} -lrt -Wformat-nonliteral -ggdb
+
+mips24kc_be: git_version
+	${cc_mips24kc_be}  -o ${NAME}_$@   -I. ${SOURCES} ${FLAGS} -lrt -lgcc_eh -static -O3
+
+mips24kc_be_debug: git_version
+	${cc_mips24kc_be}  -o ${NAME}_$@   -I. ${SOURCES} ${FLAGS} -lrt -lgcc_eh -static -ggdb
+
+mips24kc_le: git_version
+	${cc_mips24kc_le}  -o ${NAME}_$@   -I. ${SOURCES} ${FLAGS} -lrt -lgcc_eh -static -O3
+
+amd64:git_version
+	${cc_local}   -o ${NAME}_$@    -I. ${SOURCES} ${FLAGS} -lrt -static -O3
+amd64_debug:git_version
+	${cc_local}   -o ${NAME}_$@    -I. ${SOURCES} ${FLAGS} -lrt -static -ggdb
+x86:git_version
+	${cc_local}   -o ${NAME}_$@      -I. ${SOURCES} ${FLAGS} -lrt -static -O3 -m32
+arm:git_version
+	${cc_arm}   -o ${NAME}_$@      -I. ${SOURCES} ${FLAGS} -lrt -static -O3
+
+arm_debug:git_version
+	${cc_arm}   -o ${NAME}_$@      -I. ${SOURCES} ${FLAGS} -lrt -static -ggdb
+
+cross:git_version
+	${cc_cross}   -o ${NAME}_cross    -I. ${SOURCES} ${FLAGS} -lrt -O3
+
+cross2:git_version
+	${cc_cross}   -o ${NAME}_cross    -I. ${SOURCES} ${FLAGS} -lrt -static -lgcc_eh -O3   
+
+cross3:git_version
+	${cc_cross}   -o ${NAME}_cross    -I. ${SOURCES} ${FLAGS} -lrt -static -O3
+
+release: ${TARGETS} 
+	cp git_version.h version.txt
+	tar -zcvf ${TAR}
+
+clean:	
+	rm -f ${TAR}
+	rm -f speeder speeder_cross
+	rm -f git_version.h
+
+git_version:
+	    echo "const char * const gitversion = \"$(shell git rev-parse HEAD)\";" > git_version.h
 	
-#g++ forward.cpp aes.c -o forward -static
-#	${ccarm} forward.cpp aes.c  -o forwardarm   -static -lgcc_eh

misc.cpp

@@ -0,0 +1,974 @@
+/*
+ * misc.cpp
+ *
+ *  Created on: Oct 26, 2017
+ *      Author: root
+ */
+
+
+#include "misc.h"
+
+
+char fifo_file[1000]="";
+
+int mtu_warn=1350;
+
+int disable_mtu_warn=1;
+int disable_fec=0;
+int disable_checksum=0;
+
+int debug_force_flush_fec=0;
+
+
+int jitter_min=0*1000;
+int jitter_max=0*1000;
+
+int output_interval_min=0*1000;
+int output_interval_max=0*1000;
+
+int fix_latency=0;
+
+u32_t local_ip_uint32,remote_ip_uint32=0;
+char local_ip[100], remote_ip[100];
+int local_port = -1, remote_port = -1;
+
+conn_manager_t conn_manager;
+delay_manager_t delay_manager;
+fd_manager_t fd_manager;
+
+int time_mono_test=0;
+
+int delay_capacity=0;
+
+
+char sub_net[100]="10.22.22.0";
+u32_t sub_net_uint32=0;
+
+char tun_dev[100]="";
+
+
+int keep_reconnect=0;
+
+int tun_mtu=1500;
+
+int mssfix=1;
+
+
+int from_normal_to_fec(conn_info_t & conn_info,char *data,int len,int & out_n,char **&out_arr,int *&out_len,my_time_t *&out_delay)
+{
+
+	static my_time_t out_delay_buf[max_fec_packet_num+100]={0};
+	//static int out_len_buf[max_fec_packet_num+100]={0};
+	//static int counter=0;
+	out_delay=out_delay_buf;
+	//out_len=out_len_buf;
+	inner_stat_t &inner_stat=conn_info.stat.normal_to_fec;
+	if(disable_fec)
+	{
+		if(data==0)
+		{
+			out_n=0;
+			return 0;
+		}
+		//assert(data!=0);
+		inner_stat.input_packet_num++;
+		inner_stat.input_packet_size+=len;
+		inner_stat.output_packet_num++;
+		inner_stat.output_packet_size+=len;
+
+		out_n=1;
+		static char *data_static;
+		data_static=data;
+		static int len_static;
+		len_static=len;
+		out_arr=&data_static;
+		out_len=&len_static;
+		out_delay[0]=0;
+
+	}
+	else
+	{
+		if(data!=0)
+		{
+			inner_stat.input_packet_num++;
+			inner_stat.input_packet_size+=len;
+		}
+		//counter++;
+
+		conn_info.fec_encode_manager.input(data,len);
+
+		//if(counter%5==0)
+			//conn_info.fec_encode_manager.input(0,0);
+
+		//int n;
+		//char **s_arr;
+		//int s_len;
+
+
+		conn_info.fec_encode_manager.output(out_n,out_arr,out_len);
+
+		if(out_n>0)
+		{
+			my_time_t common_latency=0;
+			my_time_t first_packet_time=conn_info.fec_encode_manager.get_first_packet_time();
+
+			if(fix_latency==1&&conn_info.fec_encode_manager.get_type()==0)
+			{
+				my_time_t current_time=get_current_time_us();
+				my_time_t tmp;
+				assert(first_packet_time!=0);
+				//mylog(log_info,"current_time=%llu first_packlet_time=%llu   fec_pending_time=%llu\n",current_time,first_packet_time,(my_time_t)fec_pending_time);
+				if((my_time_t)conn_info.fec_encode_manager.get_pending_time() >=(current_time - first_packet_time))
+				{
+					tmp=(my_time_t)conn_info.fec_encode_manager.get_pending_time()-(current_time - first_packet_time);
+					//mylog(log_info,"tmp=%llu\n",tmp);
+				}
+				else
+				{
+					tmp=0;
+					//mylog(log_info,"0\n");
+				}
+				common_latency+=tmp;
+			}
+
+			common_latency+=random_between(jitter_min,jitter_max);
+
+			out_delay_buf[0]=common_latency;
+
+			for(int i=1;i<out_n;i++)
+			{
+				out_delay_buf[i]=out_delay_buf[i-1]+ (my_time_t)( random_between(output_interval_min,output_interval_max)/(out_n-1)  );
+			}
+		}
+
+	}
+
+	mylog(log_trace,"from_normal_to_fec input_len=%d,output_n=%d\n",len,out_n);
+
+	if(out_n>0)
+	{
+		log_bare(log_trace,"seq= %u ",read_u32(out_arr[0]));
+	}
+	for(int i=0;i<out_n;i++)
+	{
+		inner_stat.output_packet_num++;
+		inner_stat.output_packet_size+=out_len[i];
+
+		log_bare(log_trace,"%d ",out_len[i]);
+
+	}
+
+	log_bare(log_trace,"\n");
+	//for(int i=0;i<n;i++)
+	//{
+		//delay_send(0,dest,s_arr[i],s_len);
+	//}
+	//delay_send(0,dest,data,len);
+	//delay_send(1000*1000,dest,data,len);
+	return 0;
+}
+int from_fec_to_normal(conn_info_t & conn_info,char *data,int len,int & out_n,char **&out_arr,int *&out_len,my_time_t *&out_delay)
+{
+	static my_time_t out_delay_buf[max_blob_packet_num+100]={0};
+	out_delay=out_delay_buf;
+	inner_stat_t &inner_stat=conn_info.stat.fec_to_normal;
+	if(disable_fec)
+	{
+		assert(data!=0);
+		inner_stat.input_packet_num++;
+		inner_stat.input_packet_size+=len;
+		inner_stat.output_packet_num++;
+		inner_stat.output_packet_size+=len;
+
+		if(data==0)
+		{
+			out_n=0;
+			return 0;
+		}
+		out_n=1;
+		static char *data_static;
+		data_static=data;
+		static int len_static;
+		len_static=len;
+		out_arr=&data_static;
+		out_len=&len_static;
+		out_delay[0]=0;
+	}
+	else
+	{
+
+		if(data!=0)
+		{
+			inner_stat.input_packet_num++;
+			inner_stat.input_packet_size+=len;
+		}
+
+		conn_info.fec_decode_manager.input(data,len);
+
+		//int n;char ** s_arr;int* len_arr;
+		conn_info.fec_decode_manager.output(out_n,out_arr,out_len);
+		for(int i=0;i<out_n;i++)
+		{
+			out_delay_buf[i]=0;
+
+			inner_stat.output_packet_num++;
+			inner_stat.output_packet_size+=out_len[i];
+		}
+
+
+	}
+
+	mylog(log_trace,"from_fec_to_normal input_len=%d,output_n=%d,input_seq=%u\n",len,out_n,read_u32(data));
+
+
+//	printf("<n:%d>",n);
+	/*
+	for(int i=0;i<n;i++)
+	{
+		delay_send(0,dest,s_arr[i],len_arr[i]);
+		//s_arr[i][len_arr[i]]=0;
+		//printf("<%s>\n",s_arr[i]);
+	}*/
+	//my_send(dest,data,len);
+	return 0;
+}
+
+
+int delay_send(my_time_t delay,const dest_t &dest,char *data,int len)
+{
+	//int rand=random()%100;
+	//mylog(log_info,"rand = %d\n",rand);
+
+	if (dest.cook&&random_drop != 0) {
+		if (get_true_random_number() % 10000 < (u32_t) random_drop) {
+			return 0;
+		}
+	}
+	return delay_manager.add(delay,dest,data,len);;
+}
+
+int print_parameter()
+{
+	mylog(log_info,"jitter_min=%d jitter_max=%d output_interval_min=%d output_interval_max=%d fec_timeout=%d fec_data_num=%d fec_redundant_num=%d fec_mtu=%d fec_queue_len=%d fec_mode=%d\n",
+			jitter_min/1000,jitter_max/1000,output_interval_min/1000,output_interval_max/1000,g_fec_timeout/1000,
+			g_fec_data_num,g_fec_redundant_num,g_fec_mtu,g_fec_queue_len,g_fec_mode);
+	return 0;
+}
+int handle_command(char *s)
+{
+	int len=strlen(s);
+	while(len>=1&&s[len-1]=='\n')
+		s[len-1]=0;
+	mylog(log_info,"got data from fifo,len=%d,s=[%s]\n",len,s);
+	int a=-1,b=-1;
+	if(strncmp(s,"fec",strlen("fec"))==0)
+	{
+		mylog(log_info,"got command [fec]\n");
+		sscanf(s,"fec %d:%d",&a,&b);
+		if(a<1||b<0||a+b>254)
+		{
+			mylog(log_warn,"invaild value\n");
+			return -1;
+		}
+		g_fec_data_num=a;
+		g_fec_redundant_num=b;
+	}
+	else if(strncmp(s,"mtu",strlen("mtu"))==0)
+	{
+		mylog(log_info,"got command [mtu]\n");
+		sscanf(s,"mtu %d",&a);
+		if(a<100||a>2000)
+		{
+			mylog(log_warn,"invaild value\n");
+			return -1;
+		}
+		g_fec_mtu=a;
+	}
+	else if(strncmp(s,"queue-len",strlen("queue-len"))==0)
+	{
+		mylog(log_info,"got command [queue-len]\n");
+		sscanf(s,"queue-len %d",&a);
+		if(a<1||a>10000)
+		{
+			mylog(log_warn,"invaild value\n");
+			return -1;
+		}
+		g_fec_queue_len=a;
+	}
+	else if(strncmp(s,"mode",strlen("mode"))==0)
+	{
+		mylog(log_info,"got command [mode]\n");
+		sscanf(s,"mode %d",&a);
+		if(a!=0&&a!=1)
+		{
+			mylog(log_warn,"invaild value\n");
+			return -1;
+		}
+		g_fec_mode=a;
+	}
+	else if(strncmp(s,"timeout",strlen("timeout"))==0)
+	{
+		mylog(log_info,"got command [timeout]\n");
+		sscanf(s,"timeout %d",&a);
+		if(a<0||a>1000)
+		{
+			mylog(log_warn,"invaild value\n");
+			return -1;
+		}
+		g_fec_timeout=a*1000;
+	}
+	else
+	{
+		mylog(log_info,"unknown command\n");
+	}
+	print_parameter();
+
+	return 0;
+}
+
+int unit_test()
+{
+	int i,j,k;
+	void *code=fec_new(3,6);
+	char arr[6][100]=
+	{
+		"aaa","bbb","ccc"
+		,"ddd","eee","fff"
+	};
+	char *data[6];
+	for(i=0;i<6;i++)
+	{
+		data[i]=arr[i];
+	}
+	rs_encode2(3,6,data,3);
+	//printf("%d %d",(int)(unsigned char)arr[5][0],(int)('a'^'b'^'c'^'d'^'e'));
+
+	for(i=0;i<6;i++)
+	{
+		printf("<%s>",data[i]);
+	}
+
+	data[0]=0;
+	//data[1]=0;
+	//data[5]=0;
+
+	int ret=rs_decode2(3,6,data,3);
+	printf("ret:%d\n",ret);
+
+	for(i=0;i<6;i++)
+	{
+		printf("<%s>",data[i]);
+	}
+	fec_free(code);
+
+
+	char arr2[6][100]=
+	{
+		"aaa11111","","ccc333333333"
+		,"ddd444","eee5555","ff6666"
+	};
+	blob_encode_t blob_encode;
+	for(int i=0;i<6;i++)
+		blob_encode.input(arr2[i],strlen(arr2[i]));
+
+	char **output;
+	int shard_len;
+	blob_encode.output(7,output,shard_len);
+
+
+	printf("<shard_len:%d>",shard_len);
+	blob_decode_t blob_decode;
+	for(int i=0;i<7;i++)
+	{
+		blob_decode.input(output[i],shard_len);
+	}
+
+	char **decode_output;
+	int * len_arr;
+	int num;
+
+
+	ret=blob_decode.output(num,decode_output,len_arr);
+
+	printf("<num:%d,ret:%d>\n",num,ret);
+	for(int i=0;i<num;i++)
+	{
+		char buf[1000]={0};
+		memcpy(buf,decode_output[i],len_arr[i]);
+		printf("<%d:%s>",len_arr[i],buf);
+	}
+	printf("\n");
+	static fec_encode_manager_t fec_encode_manager;
+	static fec_decode_manager_t fec_decode_manager;
+
+	{
+
+		string a = "11111";
+		string b = "22";
+		string c = "33333333";
+
+		fec_encode_manager.input((char *) a.c_str(), a.length());
+		fec_encode_manager.input((char *) b.c_str(), b.length());
+		fec_encode_manager.input((char *) c.c_str(), c.length());
+		fec_encode_manager.input(0, 0);
+
+		int n;
+		char **s_arr;
+		int *len;
+
+
+		fec_encode_manager.output(n,s_arr,len);
+		printf("<n:%d,len:%d>",n,len[0]);
+
+		for(int i=0;i<n;i++)
+		{
+			fec_decode_manager.input(s_arr[i],len[i]);
+		}
+
+		{
+			int n;char ** s_arr;int* len_arr;
+			fec_decode_manager.output(n,s_arr,len_arr);
+			printf("<n:%d>",n);
+			for(int i=0;i<n;i++)
+			{
+				s_arr[i][len_arr[i]]=0;
+				printf("<%s>\n",s_arr[i]);
+			}
+		}
+
+
+	}
+
+	{
+		string a = "aaaaaaa";
+		string b = "bbbbbbbbbbbbb";
+		string c = "ccc";
+
+
+		fec_encode_manager.input((char *) a.c_str(), a.length());
+		fec_encode_manager.input((char *) b.c_str(), b.length());
+		fec_encode_manager.input((char *) c.c_str(), c.length());
+		fec_encode_manager.input(0, 0);
+
+		int n;
+		char **s_arr;
+		int * len;
+
+
+		fec_encode_manager.output(n,s_arr,len);
+		printf("<n:%d,len:%d>",n,len[0]);
+
+		for(int i=0;i<n;i++)
+		{
+			if(i==1||i==3||i==5||i==0)
+			fec_decode_manager.input(s_arr[i],len[i]);
+		}
+
+		{
+			int n;char ** s_arr;int* len_arr;
+			fec_decode_manager.output(n,s_arr,len_arr);
+			printf("<n:%d>",n);
+			for(int i=0;i<n;i++)
+			{
+				s_arr[i][len_arr[i]]=0;
+				printf("<%s>\n",s_arr[i]);
+			}
+		}
+	}
+
+	for(int i=0;i<10;i++)
+	{
+		string a = "aaaaaaaaaaaaaaaaaaaaaaa";
+		string b = "bbbbbbbbbbbbb";
+		string c = "cccccccccccccccccc";
+
+
+
+		printf("======\n");
+		int n;
+		char **s_arr;
+		int * len;
+		fec_decode_manager.output(n,s_arr,len);
+
+		fec_encode_manager.reset_fec_parameter(3,2,g_fec_mtu,g_fec_queue_len,g_fec_timeout,1);
+
+		fec_encode_manager.input((char *) a.c_str(), a.length());
+		fec_encode_manager.output(n,s_arr,len);
+		assert(n==1);
+
+		fec_decode_manager.input(s_arr[0],len[0]);
+
+		fec_decode_manager.output(n,s_arr,len);
+		assert(n==1);
+		printf("%s\n",s_arr[0]);
+
+		fec_encode_manager.input((char *) b.c_str(), b.length());
+		fec_encode_manager.output(n,s_arr,len);
+		assert(n==1);
+		//fec_decode_manager.input(s_arr[0],len[0]);
+
+
+		fec_encode_manager.input((char *) c.c_str(), c.length());
+		fec_encode_manager.output(n,s_arr,len);
+
+		assert(n==3);
+
+		fec_decode_manager.input(s_arr[0],len[0]);
+		//printf("n=%d\n",n);
+
+
+		{
+			int n;
+			char **s_arr;
+			int * len;
+			fec_decode_manager.output(n,s_arr,len);
+			assert(n==1);
+			printf("%s\n",s_arr[0]);
+		}
+
+		fec_decode_manager.input(s_arr[1],len[1]);
+
+		{
+			int n;
+			char **s_arr;
+			int * len;
+			fec_decode_manager.output(n,s_arr,len);
+			assert(n==1);
+			printf("n=%d\n",n);
+			s_arr[0][len[0]]=0;
+			printf("%s\n",s_arr[0]);
+		}
+
+	}
+
+	return 0;
+}
+
+
+void process_arg(int argc, char *argv[])
+{
+	int is_client=0,is_server=0;
+	int i, j, k;
+	int opt;
+    static struct option long_options[] =
+      {
+		{"log-level", required_argument,    0, 1},
+		{"log-position", no_argument,    0, 1},
+		{"disable-color", no_argument,    0, 1},
+		{"disable-filter", no_argument,    0, 1},
+		{"disable-fec", no_argument,    0, 1},
+		{"disable-obscure", no_argument,    0, 1},
+		{"disable-xor", no_argument,    0, 1},
+		{"disable-checksum", no_argument,    0, 1},
+		{"fix-latency", no_argument,    0, 1},
+		{"sock-buf", required_argument,    0, 1},
+		{"random-drop", required_argument,    0, 1},
+		{"report", required_argument,    0, 1},
+		{"delay-capacity", required_argument,    0, 1},
+		{"mtu", required_argument,    0, 1},
+		{"mode", required_argument,   0,1},
+		{"timeout", required_argument,   0,1},
+		{"decode-buf", required_argument,   0,1},
+		{"queue-len", required_argument,   0,'q'},
+		{"fec", required_argument,   0,'f'},
+		{"jitter", required_argument,   0,'j'},
+		{"fifo", required_argument,    0, 1},
+		{"sub-net", required_argument,    0, 1},
+		{"tun-dev", required_argument,    0, 1},
+		{"tun-mtu", required_argument,    0, 1},
+		{"disable-mssfix", no_argument,    0, 1},
+		{"keep-reconnect", no_argument,    0, 1},
+		{NULL, 0, 0, 0}
+      };
+    int option_index = 0;
+
+	for (i = 0; i < argc; i++)
+	{
+		if(strcmp(argv[i],"--unit-test")==0)
+		{
+			unit_test();
+			myexit(0);
+		}
+	}
+
+	for (i = 0; i < argc; i++)
+	{
+		if(strcmp(argv[i],"--log-level")==0)
+		{
+			if(i<argc -1)
+			{
+				sscanf(argv[i+1],"%d",&log_level);
+				if(0<=log_level&&log_level<log_end)
+				{
+				}
+				else
+				{
+					log_bare(log_fatal,"invalid log_level\n");
+					myexit(-1);
+				}
+			}
+		}
+		if(strcmp(argv[i],"--disable-color")==0)
+		{
+			enable_log_color=0;
+		}
+	}
+
+    mylog(log_info,"argc=%d ", argc);
+
+	for (i = 0; i < argc; i++) {
+		log_bare(log_info, "%s ", argv[i]);
+	}
+	log_bare(log_info, "\n");
+
+
+	int no_l = 1, no_r = 1;
+	while ((opt = getopt_long(argc, argv, "l:r:hcsk:j:f:p:n:i:q:",long_options,&option_index)) != -1)
+	{
+		//string opt_key;
+		//opt_key+=opt;
+		switch (opt)
+		{
+		case 'k':
+			sscanf(optarg,"%s\n",key_string);
+			mylog(log_debug,"key=%s\n",key_string);
+			if(strlen(key_string)==0)
+			{
+				mylog(log_fatal,"key len=0??\n");
+				myexit(-1);
+			}
+			break;
+		case 'j':
+			if (strchr(optarg, ':') == 0)
+			{
+				int jitter;
+				sscanf(optarg,"%d\n",&jitter);
+				if(jitter<0 ||jitter>1000*10)
+				{
+					mylog(log_fatal,"jitter must be between 0 and 10,000(10 second)\n");
+					myexit(-1);
+				}
+				jitter_min=0;
+				jitter_max=jitter;
+
+
+			}
+			else
+			{
+				sscanf(optarg,"%d:%d\n",&jitter_min,&jitter_max);
+				if(jitter_min<0 ||jitter_max<0||jitter_min>jitter_max)
+				{
+					mylog(log_fatal," must satisfy  0<=jmin<=jmax\n");
+					myexit(-1);
+				}
+			}
+			jitter_min*=1000;
+			jitter_max*=1000;
+			break;
+		case 'i':
+			if (strchr(optarg, ':') == 0)
+			{
+				int output_interval=-1;
+				sscanf(optarg,"%d\n",&output_interval);
+				if(output_interval<0||output_interval>1000*10)
+				{
+					mylog(log_fatal,"output_interval must be between 0 and 10,000(10 second)\n");
+					myexit(-1);
+				}
+				output_interval_min=output_interval_max=output_interval;
+			}
+			else
+			{
+				sscanf(optarg,"%d:%d\n",&output_interval_min,&output_interval_max);
+				if(output_interval_min<0 ||output_interval_max<0||output_interval_min>output_interval_max)
+				{
+					mylog(log_fatal," must satisfy  0<=output_interval_min<=output_interval_max\n");
+					myexit(-1);
+				}
+			}
+			output_interval_min*=1000;
+			output_interval_max*=1000;
+			break;
+		case 'f':
+			if (strchr(optarg, ':') == 0)
+			{
+				mylog(log_fatal,"invalid format for f");
+				myexit(-1);
+			}
+			else
+			{
+				sscanf(optarg,"%d:%d\n",&g_fec_data_num,&g_fec_redundant_num);
+				if(g_fec_data_num<1 ||g_fec_redundant_num<0||g_fec_data_num+g_fec_redundant_num>254)
+				{
+					mylog(log_fatal,"fec_data_num<1 ||fec_redundant_num<0||fec_data_num+fec_redundant_num>254\n");
+					myexit(-1);
+				}
+			}
+			break;
+		case 'q':
+			sscanf(optarg,"%d",&g_fec_queue_len);
+			if(g_fec_queue_len<1||g_fec_queue_len>10000)
+			{
+
+					mylog(log_fatal,"fec_pending_num should be between 1 and 10000\n");
+					myexit(-1);
+			}
+			break;
+		case 'c':
+			is_client = 1;
+			break;
+		case 's':
+			is_server = 1;
+			break;
+		case 'l':
+			no_l = 0;
+			if (strchr(optarg, ':') != 0)
+			{
+				sscanf(optarg, "%[^:]:%d", local_ip, &local_port);
+			}
+			else
+			{
+				mylog(log_fatal," -r ip:port\n");
+				myexit(1);
+				strcpy(local_ip, "127.0.0.1");
+				sscanf(optarg, "%d", &local_port);
+			}
+			break;
+		case 'r':
+			no_r = 0;
+			if (strchr(optarg, ':') != 0)
+			{
+				//printf("in :\n");
+				//printf("%s\n",optarg);
+				sscanf(optarg, "%[^:]:%d", remote_ip, &remote_port);
+				//printf("%d\n",remote_port);
+			}
+			else
+			{
+				mylog(log_fatal," -r ip:port\n");
+				myexit(1);
+				strcpy(remote_ip, "127.0.0.1");
+				sscanf(optarg, "%d", &remote_port);
+			}
+			break;
+		case 'h':
+			break;
+		case 1:
+			if(strcmp(long_options[option_index].name,"log-level")==0)
+			{
+			}
+			else if(strcmp(long_options[option_index].name,"disable-filter")==0)
+			{
+				disable_replay_filter=1;
+				//enable_log_color=0;
+			}
+			else if(strcmp(long_options[option_index].name,"disable-color")==0)
+			{
+				//enable_log_color=0;
+			}
+			else if(strcmp(long_options[option_index].name,"disable-fec")==0)
+			{
+				disable_fec=1;
+			}
+			else if(strcmp(long_options[option_index].name,"disable-obscure")==0)
+			{
+				mylog(log_info,"obscure disabled\n");
+				disable_obscure=1;
+			}
+			else if(strcmp(long_options[option_index].name,"disable-xor")==0)
+			{
+				mylog(log_info,"xor disabled\n");
+				disable_xor=1;
+			}
+			else if(strcmp(long_options[option_index].name,"disable-checksum")==0)
+			{
+				disable_checksum=1;
+				mylog(log_warn,"checksum disabled\n");
+			}
+			else if(strcmp(long_options[option_index].name,"fix-latency")==0)
+			{
+				mylog(log_info,"fix-latency enabled\n");
+				fix_latency=1;
+			}
+
+			else if(strcmp(long_options[option_index].name,"log-position")==0)
+			{
+				enable_log_position=1;
+			}
+			else if(strcmp(long_options[option_index].name,"random-drop")==0)
+			{
+				sscanf(optarg,"%d",&random_drop);
+				if(random_drop<0||random_drop>10000)
+				{
+					mylog(log_fatal,"random_drop must be between 0 10000 \n");
+					myexit(-1);
+				}
+				mylog(log_info,"random_drop=%d\n",random_drop);
+			}
+			else if(strcmp(long_options[option_index].name,"delay-capacity")==0)
+			{
+				sscanf(optarg,"%d",&delay_capacity);
+
+				if(delay_capacity<0)
+				{
+					mylog(log_fatal,"delay_capacity must be >=0 \n");
+					myexit(-1);
+				}
+			}
+			else if(strcmp(long_options[option_index].name,"report")==0)
+			{
+				sscanf(optarg,"%d",&report_interval);
+
+				if(report_interval<=0)
+				{
+					mylog(log_fatal,"report_interval must be >0 \n");
+					myexit(-1);
+				}
+			}
+			else if(strcmp(long_options[option_index].name,"sock-buf")==0)
+			{
+				int tmp=-1;
+				sscanf(optarg,"%d",&tmp);
+				if(10<=tmp&&tmp<=10*1024)
+				{
+					socket_buf_size=tmp*1024;
+				}
+				else
+				{
+					mylog(log_fatal,"sock-buf value must be between 1 and 10240 (kbyte) \n");
+					myexit(-1);
+				}
+			}
+			else if(strcmp(long_options[option_index].name,"decode-buf")==0)
+			{
+				sscanf(optarg,"%d",&fec_buff_num);
+				if(fec_buff_num<300 || fec_buff_num>20000)
+				{
+					mylog(log_fatal,"decode-buf value must be between 300 and 20000 (kbyte) \n");
+					myexit(-1);
+				}
+				mylog(log_info,"decode-buf=%d\n",fec_buff_num);
+			}
+			else if(strcmp(long_options[option_index].name,"mode")==0)
+			{
+				sscanf(optarg,"%d",&g_fec_mode);
+				if(g_fec_mode!=0&&g_fec_mode!=1)
+				{
+					mylog(log_fatal,"mode should be 0 or 1\n");
+					myexit(-1);
+				}
+			}
+			else if(strcmp(long_options[option_index].name,"mtu")==0)
+			{
+				sscanf(optarg,"%d",&g_fec_mtu);
+				if(g_fec_mtu<100||g_fec_mtu>2000)
+				{
+					mylog(log_fatal,"fec_mtu should be between 100 and 2000\n");
+					myexit(-1);
+				}
+			}
+			else if(strcmp(long_options[option_index].name,"timeout")==0)
+			{
+				sscanf(optarg,"%d",&g_fec_timeout);
+				if(g_fec_timeout<0||g_fec_timeout>1000)
+				{
+
+						mylog(log_fatal,"fec_pending_time should be between 0 and 1000(1s)\n");
+						myexit(-1);
+				}
+				g_fec_timeout*=1000;
+			}
+			else if(strcmp(long_options[option_index].name,"fifo")==0)
+			{
+				sscanf(optarg,"%s",fifo_file);
+
+				mylog(log_info,"fifo_file =%s \n",fifo_file);
+			}
+			else if(strcmp(long_options[option_index].name,"keep-reconnect")==0)
+			{
+				keep_reconnect=1;
+				mylog(log_info,"keep_reconnect enabled\n");
+			}
+			else if(strcmp(long_options[option_index].name,"sub-net")==0)
+			{
+				sscanf(optarg,"%s",sub_net);
+				mylog(log_info,"sub_net %s\n",sub_net);
+
+			}
+			else if(strcmp(long_options[option_index].name,"tun-dev")==0)
+			{
+				sscanf(optarg,"%s",tun_dev);
+				mylog(log_info,"tun_dev=%s\n",tun_dev);
+
+				mylog(log_info,"running at tun-dev mode\n");
+				working_mode=tun_dev_mode;
+			}
+			else if(strcmp(long_options[option_index].name,"tun-mtu")==0)
+			{
+				sscanf(optarg,"%d",&tun_mtu);
+				mylog(log_warn,"changed tun_mtu,tun_mtu=%d\n",tun_mtu);
+			}
+			else if(strcmp(long_options[option_index].name,"disable-mssfix")==0)
+			{
+				mssfix=0;
+				mylog(log_warn,"mssfix disabled\n");
+			}
+			else
+			{
+				mylog(log_fatal,"unknown option\n");
+				myexit(-1);
+			}
+			break;
+		default:
+			mylog(log_fatal,"unknown option <%x>", opt);
+			myexit(-1);
+		}
+	}
+
+	if (is_client == 0 && is_server == 0)
+	{
+		mylog(log_fatal,"-s -c hasnt been set\n");
+		myexit(-1);
+	}
+	if (is_client == 1 && is_server == 1)
+	{
+		mylog(log_fatal,"-s -c cant be both set\n");
+		myexit(-1);
+	}
+	if(is_client==1)
+	{
+		client_or_server=client_mode;
+	}
+	else
+	{
+		client_or_server=server_mode;
+	}
+
+
+	if(working_mode==tunnel_mode)
+	{
+		if (no_l)
+			mylog(log_fatal,"error: -l not found\n");
+		if (no_r)
+			mylog(log_fatal,"error: -r not found\n");
+		if (no_l || no_r)
+			myexit(-1);
+	}
+	else if(working_mode==tun_dev_mode)
+	{
+		if(client_or_server==client_mode&&no_r)
+		{
+			mylog(log_fatal,"error: -r not found\n");
+			myexit(-1);
+		}
+		else if(client_or_server==server_mode&&no_l)
+		{
+			mylog(log_fatal,"error: -l not found\n");
+			myexit(-1);
+		}
+	}
+
+	print_parameter();
+
+}
+

misc.h

@@ -0,0 +1,74 @@
+/*
+ * misc.h
+ *
+ *  Created on: Oct 26, 2017
+ *      Author: root
+ */
+
+#ifndef MISC_H_
+#define MISC_H_
+
+#include "common.h"
+#include "connection.h"
+#include "fd_manager.h"
+#include "delay_manager.h"
+#include "fec_manager.h"
+
+
+
+extern char fifo_file[1000];
+
+extern int mtu_warn;
+
+extern int disable_mtu_warn;
+extern int disable_fec;
+extern int disable_checksum;
+
+extern int debug_force_flush_fec;
+
+
+extern int jitter_min;
+extern int jitter_max;
+
+extern int output_interval_min;
+extern int output_interval_max;
+
+extern int fix_latency;
+
+extern u32_t local_ip_uint32,remote_ip_uint32;
+extern char local_ip[100], remote_ip[100];
+extern int local_port, remote_port;
+
+extern conn_manager_t conn_manager;
+extern delay_manager_t delay_manager;
+extern fd_manager_t fd_manager;
+
+extern int time_mono_test;
+
+extern int delay_capacity;
+
+extern int keep_reconnect;
+
+extern int tun_mtu;
+
+extern int mssfix;
+
+
+int from_normal_to_fec(conn_info_t & conn_info,char *data,int len,int & out_n,char **&out_arr,int *&out_len,my_time_t *&out_delay);
+int from_fec_to_normal(conn_info_t & conn_info,char *data,int len,int & out_n,char **&out_arr,int *&out_len,my_time_t *&out_delay);
+
+int delay_send(my_time_t delay,const dest_t &dest,char *data,int len);
+int print_parameter();
+int handle_command(char *s);
+
+int unit_test();
+
+//void print_help();
+
+void process_arg(int argc, char *argv[]);
+
+extern char sub_net[100];
+extern u32_t sub_net_uint32;
+extern char tun_dev[100];
+
+#endif /* MISC_H_ */

packet.cpp

@@ -0,0 +1,402 @@
+/*
+ * packet.cpp
+ *
+ *  Created on: Sep 15, 2017
+ *      Author: root
+ */
+
+
+#include "common.h"
+#include "log.h"
+#include "packet.h"
+#include "misc.h"
+
+int iv_min=4;
+int iv_max=32;//< 256;
+u64_t packet_send_count=0;
+u64_t dup_packet_send_count=0;
+u64_t packet_recv_count=0;
+u64_t dup_packet_recv_count=0;
+
+typedef u64_t anti_replay_seq_t;
+int disable_replay_filter=0;
+
+int disable_obscure=0;
+int disable_xor=0;
+
+int random_drop=0;
+
+char key_string[1000]= "";
+
+//int local_listen_fd=-1;
+
+
+void encrypt_0(char * input,int &len,char *key)
+{
+	int i,j;
+	if(key[0]==0) return;
+	for(i=0,j=0;i<len;i++,j++)
+	{
+		if(key[j]==0)j=0;
+		input[i]^=key[j];
+	}
+}
+
+void decrypt_0(char * input,int &len,char *key)
+{
+	int i,j;
+	if(key[0]==0) return;
+	for(i=0,j=0;i<len;i++,j++)
+	{
+		if(key[j]==0)j=0;
+		input[i]^=key[j];
+	}
+}
+int do_obscure_old(const char * input, int in_len,char *output,int &out_len)
+{
+	//memcpy(output,input,in_len);
+//	out_len=in_len;
+	//return 0;
+
+	int i, j, k;
+	if (in_len > 65535||in_len<0)
+		return -1;
+	int iv_len=iv_min+rand()%(iv_max-iv_min);
+	get_true_random_chars(output,iv_len);
+	memcpy(output+iv_len,input,in_len);
+
+	output[iv_len+in_len]=(uint8_t)iv_len;
+
+	output[iv_len+in_len]^=output[0];
+	output[iv_len+in_len]^=key_string[0];
+
+	for(i=0,j=0,k=1;i<in_len;i++,j++,k++)
+	{
+		if(j==iv_len) j=0;
+		if(key_string[k]==0)k=0;
+		output[iv_len+i]^=output[j];
+		output[iv_len+i]^=key_string[k];
+	}
+
+
+	out_len=iv_len+in_len+1;
+	return 0;
+}
+
+int do_obscure(char * data,int &len)
+{
+	assert(len>=0);
+	assert(len<buf_len);
+
+	int iv_len=random_between(iv_min,iv_max);
+	get_true_random_chars(data+len,iv_len);
+	data[iv_len+len]=(uint8_t)iv_len;
+	for(int i=0,j=0;i<len;i++,j++)
+	{
+		if(j==iv_len)j=0;
+		data[i]^=data[len+j];
+	}
+
+	len=len+iv_len+1;
+	return 0;
+}
+
+int de_obscure(char * data,int &len)
+{
+	if(len<1) return -1;
+	int iv_len=int ((uint8_t) data[len-1]);
+
+	if(len<1+iv_len) return -1;
+
+	len=len-1-iv_len;
+	for(int i=0,j=0;i<len;i++,j++)
+	{
+		if(j==iv_len)j=0;
+		data[i]^=data[len+j];
+	}
+
+	return 0;
+}
+int de_obscure_old(const char * input, int in_len,char *output,int &out_len)
+{
+	//memcpy(output,input,in_len);
+	//out_len=in_len;
+	//return 0;
+
+	int i, j, k;
+	if (in_len > 65535||in_len<0)
+	{
+		mylog(log_debug,"in_len > 65535||in_len<0 ,  %d",in_len);
+		return -1;
+	}
+	int iv_len= int ((uint8_t)(input[in_len-1]^input[0]^key_string[0]) );
+	out_len=in_len-1-iv_len;
+	if(out_len<0)
+	{
+		mylog(log_debug,"%d %d\n",in_len,out_len);
+		return -1;
+	}
+	for(i=0,j=0,k=1;i<in_len;i++,j++,k++)
+	{
+		if(j==iv_len) j=0;
+		if(key_string[k]==0)k=0;
+		output[i]=input[iv_len+i]^input[j]^key_string[k];
+
+	}
+	dup_packet_recv_count++;
+	return 0;
+}
+
+
+int sendto_fd_ip_port (int fd,u32_t ip,int port,char * buf, int len,int flags)
+{
+
+	sockaddr_in tmp_sockaddr;
+
+	memset(&tmp_sockaddr,0,sizeof(tmp_sockaddr));
+	tmp_sockaddr.sin_family = AF_INET;
+	tmp_sockaddr.sin_addr.s_addr = ip;
+	tmp_sockaddr.sin_port = htons(uint16_t(port));
+
+	return sendto(fd, buf,
+			len , 0,
+			(struct sockaddr *) &tmp_sockaddr,
+			sizeof(tmp_sockaddr));
+}
+/*
+int sendto_ip_port (u32_t ip,int port,char * buf, int len,int flags)
+{
+	return sendto_fd_ip_port(local_listen_fd,ip,port,buf,len,flags);
+}*/
+
+int send_fd (int fd,char * buf, int len,int flags)
+{
+	return send(fd,buf,len,flags);
+}
+
+int my_send(const dest_t &dest,char *data,int len)
+{
+	if(dest.cook)
+	{
+		do_cook(data,len);
+	}
+	switch(dest.type)
+	{
+		case type_fd_ip_port:
+		{
+			return sendto_fd_ip_port(dest.inner.fd,dest.inner.fd_ip_port.ip_port.ip,dest.inner.fd_ip_port.ip_port.port,data,len,0);
+			break;
+		}
+		case type_fd64_ip_port:
+		{
+			if(!fd_manager.exist(dest.inner.fd64)) return -1;
+			int fd=fd_manager.to_fd(dest.inner.fd64);
+
+			return sendto_fd_ip_port(fd,dest.inner.fd64_ip_port.ip_port.ip,dest.inner.fd64_ip_port.ip_port.port,data,len,0);
+			break;
+		}
+		case type_fd:
+		{
+			return send_fd(dest.inner.fd,data,len,0);
+			break;
+		}
+		case type_write_fd:
+		{
+			return write(dest.inner.fd,data,len);
+			break;
+		}
+		case type_fd64:
+		{
+
+			if(!fd_manager.exist(dest.inner.fd64)) return -1;
+			int fd=fd_manager.to_fd(dest.inner.fd64);
+
+			return send_fd(fd,data,len,0);
+			break;
+		}
+		/*
+		case type_fd64_ip_port_conv:
+		{
+			if(!fd_manager.exist(dest.inner.fd64)) return -1;
+			int fd=fd_manager.to_fd(dest.inner.fd64);
+
+			char *new_data;
+			int new_len;
+
+			put_conv(dest.conv,data,len,new_data,new_len);
+			return sendto_fd_ip_port(fd,dest.inner.fd64_ip_port.ip_port.ip,dest.inner.fd64_ip_port.ip_port.port,new_data,new_len,0);
+			break;
+		}*/
+
+		/*
+		case type_fd64_conv:
+		{
+			char *new_data;
+			int new_len;
+			put_conv(dest.conv,data,len,new_data,new_len);
+
+			if(!fd_manager.exist(dest.inner.fd64)) return -1;
+			int fd=fd_manager.to_fd(dest.inner.fd64);
+			return send_fd(fd,new_data,new_len,0);
+		}*/
+		/*
+		case type_fd:
+		{
+			send_fd(dest.inner.fd,data,len,0);
+			break;
+		}*/
+		default:
+			assert(0==1);
+	}
+	return 0;
+}
+
+/*
+ *  this function comes from  http://www.hackersdelight.org/hdcodetxt/crc.c.txt
+ */
+unsigned int crc32h(unsigned char *message,int len) {
+	assert(len>=0);
+   int i, crc;
+   unsigned int byte, c;
+   const unsigned int g0 = 0xEDB88320,    g1 = g0>>1,
+      g2 = g0>>2, g3 = g0>>3, g4 = g0>>4, g5 = g0>>5,
+      g6 = (g0>>6)^g0, g7 = ((g0>>6)^g0)>>1;
+
+   i = 0;
+   crc = 0xFFFFFFFF;
+   while (i!=len) {    // Get next byte.
+	   byte = message[i];
+      crc = crc ^ byte;
+      c = ((crc<<31>>31) & g7) ^ ((crc<<30>>31) & g6) ^
+          ((crc<<29>>31) & g5) ^ ((crc<<28>>31) & g4) ^
+          ((crc<<27>>31) & g3) ^ ((crc<<26>>31) & g2) ^
+          ((crc<<25>>31) & g1) ^ ((crc<<24>>31) & g0);
+      crc = ((unsigned)crc >> 8) ^ c;
+      i = i + 1;
+   }
+   return ~crc;
+}
+
+int put_conv0(u32_t conv,const char * input,int len_in,char *&output,int &len_out)
+{
+	assert(len_in>=0);
+	static char buf[buf_len];
+	output=buf;
+	u32_t n_conv=htonl(conv);
+	memcpy(output,&n_conv,sizeof(n_conv));
+	memcpy(output+sizeof(n_conv),input,len_in);
+	u32_t crc32=crc32h((unsigned char *)output,len_in+sizeof(crc32));
+	u32_t crc32_n=htonl(crc32);
+	len_out=len_in+(int)(sizeof(n_conv))+(int)sizeof(crc32_n);
+	memcpy(output+len_in+(int)(sizeof(n_conv)),&crc32_n,sizeof(crc32_n));
+	return 0;
+}
+int get_conv0(u32_t &conv,const char *input,int len_in,char *&output,int &len_out )
+{
+	assert(len_in>=0);
+	u32_t n_conv;
+	memcpy(&n_conv,input,sizeof(n_conv));
+	conv=ntohl(n_conv);
+	output=(char *)input+sizeof(n_conv);
+	u32_t crc32_n;
+	len_out=len_in-(int)sizeof(n_conv)-(int)sizeof(crc32_n);
+	if(len_out<0)
+	{
+		mylog(log_debug,"len_out<0\n");
+		return -1;
+	}
+	memcpy(&crc32_n,input+len_in-(int)sizeof(crc32_n),sizeof(crc32_n));
+	u32_t crc32=ntohl(crc32_n);
+	if(crc32!=crc32h((unsigned char *)input,len_in-(int)sizeof(crc32_n)))
+	{
+		mylog(log_debug,"crc32 check failed\n");
+		return -1;
+	}
+	return 0;
+}
+int put_crc32(char * s,int &len)
+{
+	if(disable_checksum)return 0;
+	assert(len>=0);
+	//if(len<0) return -1;
+	u32_t crc32=crc32h((unsigned char *)s,len);
+	write_u32(s+len,crc32);
+	len+=sizeof(u32_t);
+
+
+	return 0;
+}
+
+int do_cook(char * data,int &len)
+{
+	put_crc32(data,len);
+	if(!disable_obscure)do_obscure(data,len);
+	if(!disable_xor)encrypt_0(data,len,key_string);
+	return 0;
+}
+
+int de_cook(char * s,int &len)
+{
+	if(!disable_xor)decrypt_0(s,len,key_string);
+	if(!disable_obscure)
+	{
+		int ret=de_obscure(s,len);
+		if(ret!=0)
+		{
+			mylog(log_debug,"de_obscure fail\n");
+			return ret;
+		}
+	}
+	int ret=rm_crc32(s,len);
+	if(ret!=0)
+	{
+		mylog(log_debug,"rm_crc32 fail\n");
+		return ret;
+	}
+	return 0;
+}
+int rm_crc32(char * s,int &len)
+{
+	if(disable_checksum)return 0;
+	assert(len>=0);
+	len-=sizeof(u32_t);
+	if(len<0) return -1;
+	u32_t crc32_in=read_u32(s+len);
+	u32_t crc32=crc32h((unsigned char *)s,len);
+	if(crc32!=crc32_in) return -1;
+	return 0;
+}
+/*
+int do_obs()
+{
+
+}
+int de_obs()*/
+int put_conv(u32_t conv,const char * input,int len_in,char *&output,int &len_out)
+{
+	static char buf[buf_len];
+	output=buf;
+	u32_t n_conv=htonl(conv);
+	memcpy(output,&n_conv,sizeof(n_conv));
+	memcpy(output+sizeof(n_conv),input,len_in);
+	len_out=len_in+(int)(sizeof(n_conv));
+
+	return 0;
+}
+int get_conv(u32_t &conv,const char *input,int len_in,char *&output,int &len_out )
+{
+	u32_t n_conv;
+	memcpy(&n_conv,input,sizeof(n_conv));
+	conv=ntohl(n_conv);
+	output=(char *)input+sizeof(n_conv);
+	len_out=len_in-(int)sizeof(n_conv);
+	if(len_out<0)
+	{
+		mylog(log_debug,"len_out<0\n");
+		return -1;
+	}
+	return 0;
+}
+
+
+

packet.h

@@ -0,0 +1,47 @@
+/*
+ * packet.h
+ *
+ *  Created on: Sep 15, 2017
+ *      Author: root
+ */
+
+#ifndef PACKET_H_
+#define PACKET_H_
+
+#include "common.h"
+#include "fd_manager.h"
+
+extern int iv_min;
+extern int iv_max;//< 256;
+
+extern u64_t packet_send_count;
+extern u64_t dup_packet_send_count;
+extern u64_t packet_recv_count;
+extern u64_t dup_packet_recv_count;
+extern char key_string[1000];
+extern int disable_replay_filter;
+extern int random_drop;
+extern int disable_obscure;
+extern int disable_xor;
+
+
+int my_send(const dest_t &dest,char *data,int len);
+
+void encrypt_0(char * input,int &len,char *key);
+void decrypt_0(char * input,int &len,char *key);
+int add_seq(char * data,int &data_len );
+int remove_seq(char * data,int &data_len);
+int do_obscure(const char * input, int in_len,char *output,int &out_len);
+int de_obscure(const char * input, int in_len,char *output,int &out_len);
+
+//int sendto_fd_u64 (int fd,u64_t u64,char * buf, int len,int flags);
+int sendto_ip_port (u32_t ip,int port,char * buf, int len,int flags);
+int send_fd (int fd,char * buf, int len,int flags);
+
+int put_conv(u32_t conv,const char * input,int len_in,char *&output,int &len_out);
+int get_conv(u32_t &conv,const char *input,int len_in,char *&output,int &len_out );
+int put_crc32(char * s,int &len);
+int rm_crc32(char * s,int &len);
+int do_cook(char * data,int &len);
+int de_cook(char * s,int &len);
+#endif /* PACKET_H_ */

tunnel.cpp

@@ -0,0 +1,705 @@
+/*
+ * tunnel.cpp
+ *
+ *  Created on: Oct 26, 2017
+ *      Author: root
+ */
+
+#include "tunnel.h"
+
+
+int tunnel_client_event_loop()
+{
+	//char buf[buf_len];
+	int i, j, k;int ret;
+	int yes = 1;
+	int epoll_fd;
+	int remote_fd;
+	fd64_t remote_fd64;
+
+    conn_info_t *conn_info_p=new conn_info_t;
+    conn_info_t &conn_info=*conn_info_p;  //huge size of conn_info,do not allocate on stack
+    //conn_info.conv_manager.reserve();
+	//conn_info.fec_encode_manager.re_init(fec_data_num,fec_redundant_num,fec_mtu,fec_pending_num,fec_pending_time,fec_type);
+
+
+	int local_listen_fd;
+	//fd64_t local_listen_fd64;
+    new_listen_socket(local_listen_fd,local_ip_uint32,local_port);
+    //local_listen_fd64=fd_manager.create(local_listen_fd);
+
+	epoll_fd = epoll_create1(0);
+	assert(epoll_fd>0);
+
+	const int max_events = 4096;
+	struct epoll_event ev, events[max_events];
+	if (epoll_fd < 0) {
+		mylog(log_fatal,"epoll return %d\n", epoll_fd);
+		myexit(-1);
+	}
+
+	ev.events = EPOLLIN;
+	ev.data.u64 = local_listen_fd;
+	ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, local_listen_fd, &ev);
+	if (ret!=0) {
+		mylog(log_fatal,"add  udp_listen_fd error\n");
+		myexit(-1);
+	}
+
+	assert(new_connected_socket(remote_fd,remote_ip_uint32,remote_port)==0);
+	remote_fd64=fd_manager.create(remote_fd);
+
+	mylog(log_debug,"remote_fd64=%llu\n",remote_fd64);
+
+	ev.events = EPOLLIN;
+	ev.data.u64 = remote_fd64;
+
+	ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, remote_fd, &ev);
+	if (ret!= 0) {
+		mylog(log_fatal,"add raw_fd error\n");
+		myexit(-1);
+	}
+
+	ev.events = EPOLLIN;
+	ev.data.u64 = delay_manager.get_timer_fd();
+
+	mylog(log_debug,"delay_manager.get_timer_fd()=%d\n",delay_manager.get_timer_fd());
+	ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, delay_manager.get_timer_fd(), &ev);
+	if (ret!= 0) {
+		mylog(log_fatal,"add delay_manager.get_timer_fd() error\n");
+		myexit(-1);
+	}
+
+	u64_t tmp_fd64=conn_info.fec_encode_manager.get_timer_fd64();
+	ev.events = EPOLLIN;
+	ev.data.u64 = tmp_fd64;
+
+	mylog(log_debug,"conn_info.fec_encode_manager.get_timer_fd64()=%llu\n",conn_info.fec_encode_manager.get_timer_fd64());
+	ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, fd_manager.to_fd(tmp_fd64), &ev);
+	if (ret!= 0) {
+		mylog(log_fatal,"add fec_encode_manager.get_timer_fd64() error\n");
+		myexit(-1);
+	}
+
+	//my_timer_t timer;
+	conn_info.timer.add_fd_to_epoll(epoll_fd);
+	conn_info.timer.set_timer_repeat_us(timer_interval*1000);
+
+	mylog(log_debug,"conn_info.timer.get_timer_fd()=%d\n",conn_info.timer.get_timer_fd());
+
+
+
+	int fifo_fd=-1;
+
+	if(fifo_file[0]!=0)
+	{
+		fifo_fd=create_fifo(fifo_file);
+		ev.events = EPOLLIN;
+		ev.data.u64 = fifo_fd;
+
+		ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, fifo_fd, &ev);
+		if (ret!= 0) {
+			mylog(log_fatal,"add fifo_fd to epoll error %s\n",strerror(errno));
+			myexit(-1);
+		}
+		mylog(log_info,"fifo_file=%s\n",fifo_file);
+	}
+
+	while(1)////////////////////////
+	{
+		if(about_to_exit) myexit(0);
+
+		int nfds = epoll_wait(epoll_fd, events, max_events, 180 * 1000);
+		if (nfds < 0) {  //allow zero
+			if(errno==EINTR  )
+			{
+				mylog(log_info,"epoll interrupted by signal continue\n");
+				//myexit(0);
+			}
+			else
+			{
+				mylog(log_fatal,"epoll_wait return %d,%s\n", nfds,strerror(errno));
+				myexit(-1);
+			}
+		}
+		int idx;
+		for (idx = 0; idx < nfds; ++idx) {
+			if(events[idx].data.u64==(u64_t)conn_info.timer.get_timer_fd())
+			{
+				uint64_t value;
+				read(conn_info.timer.get_timer_fd(), &value, 8);
+				conn_info.conv_manager.clear_inactive();
+				mylog(log_trace,"events[idx].data.u64==(u64_t)conn_info.timer.get_timer_fd()\n");
+
+				conn_info.stat.report_as_client();
+
+				if(debug_force_flush_fec)
+				{
+				int  out_n;char **out_arr;int *out_len;my_time_t *out_delay;
+				dest_t dest;
+				dest.type=type_fd64;
+				dest.inner.fd64=remote_fd64;
+				dest.cook=1;
+				from_normal_to_fec(conn_info,0,0,out_n,out_arr,out_len,out_delay);
+				for(int i=0;i<out_n;i++)
+				{
+					delay_send(out_delay[i],dest,out_arr[i],out_len[i]);
+				}
+				}
+			}
+			else if (events[idx].data.u64 == (u64_t)fifo_fd)
+			{
+				char buf[buf_len];
+				int len=read (fifo_fd, buf, sizeof (buf));
+				if(len<0)
+				{
+					mylog(log_warn,"fifo read failed len=%d,errno=%s\n",len,strerror(errno));
+					continue;
+				}
+				buf[len]=0;
+				handle_command(buf);
+			}
+			else if (events[idx].data.u64 == (u64_t)local_listen_fd||events[idx].data.u64 == conn_info.fec_encode_manager.get_timer_fd64())
+			{
+				char data[buf_len];
+				int data_len;
+				ip_port_t ip_port;
+				u32_t conv;
+				int  out_n;char **out_arr;int *out_len;my_time_t *out_delay;
+				dest_t dest;
+				dest.type=type_fd64;
+				dest.inner.fd64=remote_fd64;
+				dest.cook=1;
+
+				if(events[idx].data.u64 == conn_info.fec_encode_manager.get_timer_fd64())
+				{
+					fd64_t fd64=events[idx].data.u64;
+					mylog(log_trace,"events[idx].data.u64 == conn_info.fec_encode_manager.get_timer_fd64()\n");
+
+					//mylog(log_info,"timer!!!\n");
+					uint64_t value;
+					if(!fd_manager.exist(fd64))   //fd64 has been closed
+					{
+						mylog(log_trace,"!fd_manager.exist(fd64)");
+						continue;
+					}
+					if((ret=read(fd_manager.to_fd(fd64), &value, 8))!=8)
+					{
+						mylog(log_trace,"(ret=read(fd_manager.to_fd(fd64), &value, 8))!=8,ret=%d\n",ret);
+						continue;
+					}
+					if(value==0)
+					{
+						mylog(log_debug,"value==0\n");
+						continue;
+					}
+					assert(value==1);
+					from_normal_to_fec(conn_info,0,0,out_n,out_arr,out_len,out_delay);
+					//from_normal_to_fec(conn_info,0,0,out_n,out_arr,out_len,out_delay);
+				}
+				else//events[idx].data.u64 == (u64_t)local_listen_fd
+				{
+					mylog(log_trace,"events[idx].data.u64 == (u64_t)local_listen_fd\n");
+					struct sockaddr_in udp_new_addr_in={0};
+					socklen_t udp_new_addr_len = sizeof(sockaddr_in);
+					if ((data_len = recvfrom(local_listen_fd, data, max_data_len, 0,
+							(struct sockaddr *) &udp_new_addr_in, &udp_new_addr_len)) == -1) {
+						mylog(log_error,"recv_from error,this shouldnt happen,err=%s,but we can try to continue\n",strerror(errno));
+						continue;
+						//mylog(log_error,"recv_from error,this shouldnt happen at client\n");
+						//myexit(1);
+					};
+
+					if(!disable_mtu_warn&&data_len>=mtu_warn)
+					{
+						mylog(log_warn,"huge packet,data len=%d (>=%d).strongly suggested to set a smaller mtu at upper level,to get rid of this warn\n ",data_len,mtu_warn);
+					}
+					mylog(log_trace,"Received packet from %s:%d,len: %d\n", inet_ntoa(udp_new_addr_in.sin_addr),
+							ntohs(udp_new_addr_in.sin_port),data_len);
+
+					ip_port.ip=udp_new_addr_in.sin_addr.s_addr;
+					ip_port.port=ntohs(udp_new_addr_in.sin_port);
+
+					u64_t u64=ip_port.to_u64();
+
+					if(!conn_info.conv_manager.is_u64_used(u64))
+					{
+						if(conn_info.conv_manager.get_size() >=max_conv_num)
+						{
+							mylog(log_warn,"ignored new udp connect bc max_conv_num exceed\n");
+							continue;
+						}
+						conv=conn_info.conv_manager.get_new_conv();
+						conn_info.conv_manager.insert_conv(conv,u64);
+						mylog(log_info,"new packet from %s:%d,conv_id=%x\n",inet_ntoa(udp_new_addr_in.sin_addr),ntohs(udp_new_addr_in.sin_port),conv);
+					}
+					else
+					{
+						conv=conn_info.conv_manager.find_conv_by_u64(u64);
+						mylog(log_trace,"conv=%d\n",conv);
+					}
+					conn_info.conv_manager.update_active_time(conv);
+					char * new_data;
+					int new_len;
+					put_conv(conv,data,data_len,new_data,new_len);
+
+
+					mylog(log_trace,"data_len=%d new_len=%d\n",data_len,new_len);
+					//dest.conv=conv;
+					from_normal_to_fec(conn_info,new_data,new_len,out_n,out_arr,out_len,out_delay);
+
+				}
+				mylog(log_trace,"out_n=%d\n",out_n);
+				for(int i=0;i<out_n;i++)
+				{
+					delay_send(out_delay[i],dest,out_arr[i],out_len[i]);
+				}
+				//my_send(dest,data,data_len);
+			}
+		    else if (events[idx].data.u64 == (u64_t)delay_manager.get_timer_fd()) {
+				uint64_t value;
+				read(delay_manager.get_timer_fd(), &value, 8);
+				mylog(log_trace,"events[idx].data.u64 == (u64_t)delay_manager.get_timer_fd()\n");
+				//printf("<timerfd_triggered, %d>",delay_mp.size());
+				//fflush(stdout);
+			}
+			else if(events[idx].data.u64>u32_t(-1) )
+			{
+				char data[buf_len];
+				if(!fd_manager.exist(events[idx].data.u64))   //fd64 has been closed
+				{
+					mylog(log_trace,"!fd_manager.exist(events[idx].data.u64)");
+					continue;
+				}
+				assert(events[idx].data.u64==remote_fd64);
+				int fd=fd_manager.to_fd(remote_fd64);
+				int data_len =recv(fd,data,max_data_len,0);
+				mylog(log_trace, "received data from udp fd %d, len=%d\n", remote_fd,data_len);
+				if(data_len<0)
+				{
+					if(errno==ECONNREFUSED)
+					{
+						//conn_manager.clear_list.push_back(udp_fd);
+						mylog(log_debug, "recv failed %d ,udp_fd%d,errno:%s\n", data_len,remote_fd,strerror(errno));
+					}
+
+					mylog(log_warn, "recv failed %d ,udp_fd%d,errno:%s\n", data_len,remote_fd,strerror(errno));
+					continue;
+				}
+				if(!disable_mtu_warn&&data_len>mtu_warn)
+				{
+					mylog(log_warn,"huge packet,data len=%d (>%d).strongly suggested to set a smaller mtu at upper level,to get rid of this warn\n ",data_len,mtu_warn);
+				}
+
+				if(de_cook(data,data_len)!=0)
+				{
+					mylog(log_debug,"de_cook error");
+					continue;
+				}
+
+				int  out_n;char **out_arr;int *out_len;my_time_t *out_delay;
+				from_fec_to_normal(conn_info,data,data_len,out_n,out_arr,out_len,out_delay);
+
+				mylog(log_trace,"out_n=%d\n",out_n);
+
+				for(int i=0;i<out_n;i++)
+				{
+					u32_t conv;
+					char *new_data;
+					int new_len;
+					if(get_conv(conv,out_arr[i],out_len[i],new_data,new_len)!=0)
+					{
+						mylog(log_debug,"get_conv(conv,out_arr[i],out_len[i],new_data,new_len)!=0");
+						continue;
+					}
+					if(!conn_info.conv_manager.is_conv_used(conv))
+					{
+						mylog(log_trace,"!conn_info.conv_manager.is_conv_used(conv)");
+						continue;
+					}
+
+					conn_info.conv_manager.update_active_time(conv);
+
+					u64_t u64=conn_info.conv_manager.find_u64_by_conv(conv);
+					dest_t dest;
+					dest.inner.fd_ip_port.fd=local_listen_fd;
+					dest.inner.fd_ip_port.ip_port.from_u64(u64);
+					dest.type=type_fd_ip_port;
+					//dest.conv=conv;
+
+					delay_send(out_delay[i],dest,new_data,new_len);
+				}
+				//mylog(log_trace,"[%s] send packet\n",dest.inner.ip_port.to_s());
+			}
+			else
+			{
+				mylog(log_fatal,"unknown fd,this should never happen\n");
+				myexit(-1);
+			}
+		}
+		delay_manager.check();
+	}
+	return 0;
+}
+
+int tunnel_server_event_loop()
+{
+	//char buf[buf_len];
+	int i, j, k;int ret;
+	int yes = 1;
+	int epoll_fd;
+	int remote_fd;
+
+//    conn_info_t conn_info;
+	int local_listen_fd;
+//	fd64_t local_listen_fd64;
+    new_listen_socket(local_listen_fd,local_ip_uint32,local_port);
+   // local_listen_fd64=fd_manager.create(local_listen_fd);
+
+	epoll_fd = epoll_create1(0);
+	assert(epoll_fd>0);
+
+	const int max_events = 4096;
+	struct epoll_event ev, events[max_events];
+	if (epoll_fd < 0) {
+		mylog(log_fatal,"epoll return %d\n", epoll_fd);
+		myexit(-1);
+	}
+
+	ev.events = EPOLLIN;
+	ev.data.u64 = local_listen_fd;
+	ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, local_listen_fd, &ev);
+	if (ret!=0) {
+		mylog(log_fatal,"add  udp_listen_fd error\n");
+		myexit(-1);
+	}
+
+	ev.events = EPOLLIN;
+	ev.data.u64 = delay_manager.get_timer_fd();
+	ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, delay_manager.get_timer_fd(), &ev);
+	if (ret!= 0) {
+		mylog(log_fatal,"add delay_manager.get_timer_fd() error\n");
+		myexit(-1);
+	}
+
+	mylog(log_debug," delay_manager.get_timer_fd() =%d\n", delay_manager.get_timer_fd());
+
+	mylog(log_info,"now listening at %s:%d\n",my_ntoa(local_ip_uint32),local_port);
+
+	my_timer_t timer;
+	timer.add_fd_to_epoll(epoll_fd);
+	timer.set_timer_repeat_us(timer_interval*1000);
+
+	mylog(log_debug," timer.get_timer_fd() =%d\n",timer.get_timer_fd());
+
+	int fifo_fd=-1;
+
+	if(fifo_file[0]!=0)
+	{
+		fifo_fd=create_fifo(fifo_file);
+		ev.events = EPOLLIN;
+		ev.data.u64 = fifo_fd;
+
+		ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, fifo_fd, &ev);
+		if (ret!= 0) {
+			mylog(log_fatal,"add fifo_fd to epoll error %s\n",strerror(errno));
+			myexit(-1);
+		}
+		mylog(log_info,"fifo_file=%s\n",fifo_file);
+	}
+
+	while(1)////////////////////////
+	{
+
+		if(about_to_exit) myexit(0);
+
+		int nfds = epoll_wait(epoll_fd, events, max_events, 180 * 1000);
+		if (nfds < 0) {  //allow zero
+			if(errno==EINTR  )
+			{
+				mylog(log_info,"epoll interrupted by signal,continue\n");
+				//myexit(0);
+			}
+			else
+			{
+				mylog(log_fatal,"epoll_wait return %d,%s\n", nfds,strerror(errno));
+				myexit(-1);
+			}
+		}
+		int idx;
+		for (idx = 0; idx < nfds; ++idx)
+		{
+			/*
+			if ((events[idx].data.u64 ) == (u64_t)timer_fd)
+			{
+				conn_manager.clear_inactive();
+				u64_t dummy;
+				read(timer_fd, &dummy, 8);
+				//current_time_rough=get_current_time();
+			}
+			else */
+			if(events[idx].data.u64==(u64_t)timer.get_timer_fd())
+			{
+				uint64_t value;
+				read(timer.get_timer_fd(), &value, 8);
+				conn_manager.clear_inactive();
+				mylog(log_trace,"events[idx].data.u64==(u64_t)timer.get_timer_fd()\n");
+				//conn_info.conv_manager.clear_inactive();
+			}
+
+			else if (events[idx].data.u64 == (u64_t)fifo_fd)
+			{
+				char buf[buf_len];
+				int len=read (fifo_fd, buf, sizeof (buf));
+				if(len<0)
+				{
+					mylog(log_warn,"fifo read failed len=%d,errno=%s\n",len,strerror(errno));
+					continue;
+				}
+				buf[len]=0;
+				handle_command(buf);
+			}
+
+			else if (events[idx].data.u64 == (u64_t)local_listen_fd)
+			{
+
+				mylog(log_trace,"events[idx].data.u64 == (u64_t)local_listen_fd\n");
+				//int recv_len;
+				char data[buf_len];
+				int data_len;
+				struct sockaddr_in udp_new_addr_in={0};
+				socklen_t udp_new_addr_len = sizeof(sockaddr_in);
+				if ((data_len = recvfrom(local_listen_fd, data, max_data_len, 0,
+						(struct sockaddr *) &udp_new_addr_in, &udp_new_addr_len)) == -1) {
+					mylog(log_error,"recv_from error,this shouldnt happen,err=%s,but we can try to continue\n",strerror(errno));
+					continue;
+					//myexit(1);
+				};
+				mylog(log_trace,"Received packet from %s:%d,len: %d\n", inet_ntoa(udp_new_addr_in.sin_addr),
+						ntohs(udp_new_addr_in.sin_port),data_len);
+
+				if(!disable_mtu_warn&&data_len>=mtu_warn)///////////////////////delete this for type 0 in furture
+				{
+					mylog(log_warn,"huge packet,data len=%d (>=%d).strongly suggested to set a smaller mtu at upper level,to get rid of this warn\n ",data_len,mtu_warn);
+				}
+
+
+				if(de_cook(data,data_len)!=0)
+				{
+					mylog(log_debug,"de_cook error");
+					continue;
+				}
+
+
+				ip_port_t ip_port;
+				ip_port.ip=udp_new_addr_in.sin_addr.s_addr;
+				ip_port.port=ntohs(udp_new_addr_in.sin_port);
+				mylog(log_trace,"ip_port= %s\n",ip_port.to_s());
+				if(!conn_manager.exist(ip_port))
+				{
+					if(conn_manager.mp.size() >=max_conn_num)
+					{
+						mylog(log_warn,"new connection %s ignored bc max_conn_num exceed\n",ip_port.to_s());
+						continue;
+					}
+
+					conn_manager.insert(ip_port);
+					conn_info_t &conn_info=conn_manager.find(ip_port);
+					//conn_info.fec_encode_manager.re_init(fec_data_num,fec_redundant_num,fec_mtu,fec_pending_num,fec_pending_time,fec_type);
+					//conn_info.conv_manager.reserve();  //already reserved in constructor
+
+					u64_t fec_fd64=conn_info.fec_encode_manager.get_timer_fd64();
+					mylog(log_debug,"fec_fd64=%llu\n",fec_fd64);
+					ev.events = EPOLLIN;
+					ev.data.u64 = fec_fd64;
+					ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, fd_manager.to_fd(fec_fd64), &ev);
+
+					fd_manager.get_info(fec_fd64).ip_port=ip_port;
+
+					conn_info.timer.add_fd64_to_epoll(epoll_fd);
+					conn_info.timer.set_timer_repeat_us(timer_interval*1000);
+
+					mylog(log_debug,"conn_info.timer.get_timer_fd64()=%llu\n",conn_info.timer.get_timer_fd64());
+
+					u64_t timer_fd64=conn_info.timer.get_timer_fd64();
+					fd_manager.get_info(timer_fd64).ip_port=ip_port;
+
+					mylog(log_info,"new connection from %s\n",ip_port.to_s());
+
+				}
+				conn_info_t &conn_info=conn_manager.find(ip_port);
+
+				conn_info.update_active_time();
+				int  out_n;char **out_arr;int *out_len;my_time_t *out_delay;
+				from_fec_to_normal(conn_info,data,data_len,out_n,out_arr,out_len,out_delay);
+
+				mylog(log_trace,"out_n= %d\n",out_n);
+				for(int i=0;i<out_n;i++)
+				{
+					u32_t conv;
+					char *new_data;
+					int new_len;
+					if(get_conv(conv,out_arr[i],out_len[i],new_data,new_len)!=0)
+					{
+						mylog(log_debug,"get_conv failed");
+						continue;
+					}
+
+					/*
+					id_t tmp_conv_id;
+					memcpy(&tmp_conv_id,&data_[0],sizeof(tmp_conv_id));
+					tmp_conv_id=ntohl(tmp_conv_id);*/
+
+					if (!conn_info.conv_manager.is_conv_used(conv))
+					{
+						if(conn_info.conv_manager.get_size() >=max_conv_num)
+						{
+							mylog(log_warn,"ignored new udp connect bc max_conv_num exceed\n");
+							continue;
+						}
+
+						int new_udp_fd;
+						ret=new_connected_socket(new_udp_fd,remote_ip_uint32,remote_port);
+
+						if (ret != 0) {
+							mylog(log_warn, "[%s]new_connected_socket failed\n",ip_port.to_s());
+							continue;
+						}
+
+						fd64_t fd64 = fd_manager.create(new_udp_fd);
+						ev.events = EPOLLIN;
+						ev.data.u64 = fd64;
+						ret = epoll_ctl(epoll_fd, EPOLL_CTL_ADD, new_udp_fd, &ev);
+
+						conn_info.conv_manager.insert_conv(conv, fd64);
+						fd_manager.get_info(fd64).ip_port=ip_port;
+
+
+						mylog(log_info,"[%s]new conv %x,fd %d created,fd64=%llu\n",ip_port.to_s(),conv,new_udp_fd,fd64);
+						//assert(!conn_manager.exist_fd64(fd64));
+
+						//conn_manager.insert_fd64(fd64,ip_port);
+					}
+					conn_info.conv_manager.update_active_time(conv);
+					fd64_t fd64= conn_info.conv_manager.find_u64_by_conv(conv);
+					//int fd=fd_manager.fd64_to_fd(fd64);
+					dest_t dest;
+					dest.type=type_fd64;
+					dest.inner.fd64=fd64;
+					//dest.conv=conv;
+					delay_send(out_delay[i],dest,new_data,new_len);
+				}
+			}
+		    else if (events[idx].data.u64 == (u64_t)delay_manager.get_timer_fd()) {
+				uint64_t value;
+				read(delay_manager.get_timer_fd(), &value, 8);
+				mylog(log_trace,"events[idx].data.u64 == (u64_t)delay_manager.get_timer_fd()\n");
+			}
+			else if (events[idx].data.u64 >u32_t(-1))
+			{
+				char data[buf_len];
+				int data_len;
+				u32_t conv;
+				fd64_t fd64=events[idx].data.u64;
+				mylog(log_trace,"events[idx].data.u64 >u32_t(-1),%llu\n",(u64_t)events[idx].data.u64);
+				if(!fd_manager.exist(fd64))   //fd64 has been closed
+				{
+					mylog(log_trace,"!fd_manager.exist(fd64)\n");
+					continue;
+				}
+
+				assert(fd_manager.exist_info(fd64));
+				ip_port_t ip_port=fd_manager.get_info(fd64).ip_port;
+				assert(conn_manager.exist(ip_port));
+
+				conn_info_t &conn_info=conn_manager.find(ip_port);
+				//conn_info.update_active_time(); //cant put it here
+
+				int  out_n=-2;char **out_arr;int *out_len;my_time_t *out_delay;
+
+				dest_t dest;
+				dest.inner.fd_ip_port.fd=local_listen_fd;
+				dest.inner.fd_ip_port.ip_port=ip_port;
+				dest.type=type_fd_ip_port;
+				dest.cook=1;
+
+				if(fd64==conn_info.fec_encode_manager.get_timer_fd64())
+				{
+					//mylog(log_infol,"timer!!!\n");
+					uint64_t value;
+					if((ret=read(fd_manager.to_fd(fd64), &value, 8))!=8)
+					{
+						mylog(log_trace,"fd_manager.to_fd(fd64), &value, 8)!=8 ,%d\n",ret);
+						continue;
+					}
+					if(value==0)
+					{
+						mylog(log_trace,"value==0\n");
+						continue;
+					}
+					assert(value==1);
+					from_normal_to_fec(conn_info,0,0,out_n,out_arr,out_len,out_delay);
+				}
+				else if(fd64==conn_info.timer.get_timer_fd64())
+				{
+					uint64_t value;
+					read(conn_info.timer.get_timer_fd(), &value, 8);
+					conn_info.conv_manager.clear_inactive();
+					if(debug_force_flush_fec)
+					{
+					from_normal_to_fec(conn_info,0,0,out_n,out_arr,out_len,out_delay);
+					}
+
+					conn_info.stat.report_as_server(ip_port);
+					continue;
+				}
+				else
+				{
+					assert(conn_info.conv_manager.is_u64_used(fd64));
+
+					conv=conn_info.conv_manager.find_conv_by_u64(fd64);
+					conn_info.conv_manager.update_active_time(conv);
+					conn_info.update_active_time();
+
+					int fd=fd_manager.to_fd(fd64);
+					data_len=recv(fd,data,max_data_len,0);
+
+					mylog(log_trace,"received a packet from udp_fd,len:%d,conv=%d\n",data_len,conv);
+
+					if(data_len<0)
+					{
+						mylog(log_debug,"udp fd,recv_len<0 continue,%s\n",strerror(errno));
+
+						continue;
+					}
+
+					if(!disable_mtu_warn&&data_len>=mtu_warn)
+					{
+						mylog(log_warn,"huge packet,data len=%d (>=%d).strongly suggested to set a smaller mtu at upper level,to get rid of this warn\n ",data_len,mtu_warn);
+					}
+
+					char * new_data;
+					int new_len;
+					put_conv(conv,data,data_len,new_data,new_len);
+
+					from_normal_to_fec(conn_info,new_data,new_len,out_n,out_arr,out_len,out_delay);
+				}
+
+				mylog(log_trace,"out_n=%d\n",out_n);
+				for(int i=0;i<out_n;i++)
+				{
+					delay_send(out_delay[i],dest,out_arr[i],out_len[i]);
+				}
+				//mylog(log_trace,"[%s] send packet\n",ip_port.to_s());
+
+			}
+			else
+			{
+				mylog(log_fatal,"unknown fd,this should never happen\n");
+				myexit(-1);
+			}
+		}
+		delay_manager.check();
+	}
+	return 0;
+}

tunnel.h

@@ -0,0 +1,17 @@
+/*
+ * tunnel.h
+ *
+ *  Created on: Oct 26, 2017
+ *      Author: root
+ */
+
+#ifndef TUNNEL_H_
+#define TUNNEL_H_
+
+
+#include "misc.h"
+
+int tunnel_client_event_loop();
+int tunnel_server_event_loop();
+
+#endif /* TUNNEL_H_ */