透明代理的旁路由虚拟网关
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

575 lines
13 KiB

4 years ago
# TProxy
容器化的旁路由透明代理工具,使用Xray处理TProxy流量,实现独立MAC与IP地址的虚拟化代理网关,支持 `amd64`、`i386`、`arm64`、`armv7` 多种CPU架构,可代理全部TCP和UDP流量。
4 years ago
TProxy使用Docker容器化部署,在[Docker Hub](https://hub.docker.com/repository/docker/dnomd343/tproxy)或[Github Package](https://github.com/dnomd343/TProxy/pkgs/container/tproxy)可以查看已构建的镜像。
4 years ago
## 镜像获取
Docker镜像建议拉取 `latest` 版本,如果需要特定版本镜像,拉取时指定tag为版本号即可。
```
# latest版本
shell> docker pull dnomd343/tproxy
# 指定版本
shell> docker pull dnomd343/tproxy:v1.1
```
TProxy可以从多个镜像源拉取,其数据完全相同,国内用户建议首选阿里云镜像。
```
# Docker Hub
shell> docker pull docker.io/dnomd343/tproxy
# Github Package
shell> docker pull ghcr.io/dnomd343/tproxy
# 阿里云个人镜像
shell> docker pull registry.cn-shenzhen.aliyuncs.com/dnomd343/tproxy
```
## 开始部署
4 years ago
首先开启网卡混杂模式
```
shell> ip link set eth0 promisc on
```
创建 `macvlan` 网络
```
# 此处网段与网关信息需按实际网络指定
shell> docker network create -d macvlan \
--subnet=192.168.2.0/24 \
--gateway=192.168.2.1 \
-o parent=eth0 macvlan
```
4 years ago
选择一个目录存储数据,此处使用 `/etc/scutweb`
4 years ago
```
shell> mkdir /etc/scutweb
shell> cd /etc/scutweb
shell> vim custom.sh
```
`custom.sh` 将在容器启动时首先执行,可用于指定容器静态IP与主路由网关地址
4 years ago
```
# 指定网关为192.168.2.1,容器IP为192.168.2.2
4 years ago
ip addr flush dev eth0
ip addr add 192.168.2.2/24 brd 192.168.2.255 dev eth0
ip route add default via 192.168.2.1
```
由于TProxy只代理传输层TCP与UDP以上的流量,网络层及以下其他数据包将不被处理,其中最常见的是ICMP数据包,表现为ping流量不走代理。但可以在 `custom.sh` 中添加以下命令,回应所有发往外网的ICMP数据包,表现为ping成功且延迟为内网访问时间(ICMP数据包实际未到达,使用NAT方式假冒远程主机返回到达信息)
```
# DNAT目标指定为自身IP地址
iptables -t nat -N FAKE_PING
iptables -t nat -A FAKE_PING -j DNAT --to-destination 192.168.2.2
iptables -t nat -A PREROUTING -i eth0 -p icmp -j FAKE_PING
```
4 years ago
启动容器,此处映射时间与时区信息到容器中,可以与宿主机进行同步(容器内默认为UTC零时区),用于日志时间记录
4 years ago
```
# 容器名称和存储目录可自行指定
shell> docker run --restart always \
4 years ago
--name scutweb \
--network macvlan \
--privileged -d \
--volume /etc/scutweb/:/etc/xray/expose/ \
--volume /etc/timezone:/etc/timezone:ro \
--volume /etc/localtime:/etc/localtime:ro \
dnomd343/tproxy
# 此处为DockerHub镜像源,可按上文链接替换为其他源
4 years ago
```
使用以下命令查看容器运行状态
```
shell> docker ps -a
```
容器成功运行以后,将会在存储目录下生成多个文件和文件夹
+ `log`:文件夹,存储代理流量日志
4 years ago
+ `segment`:文件夹,存储不代理的网段信息
+ `outbounds.json`:指定流量出口信息
+ `routeing.json`:指定流量路由信息
`outbounds.json` 默认配置流量转发给网关,需要用户手动配置为上游接口,具体语法见[Xray文档](https://xtls.github.io/config/base/outbounds/)
```
{
"outbounds": [
{
"tag": "node",
"protocol": "freedom"
}
]
}
```
`routing.json` 默认配置将全部流量交由 `node` 接口,即 `outbounds.json` 中的 `freedom` 出口,具体语法见[Xray文档](https://xtls.github.io/config/base/routing/)
4 years ago
```
{
"routing": {
"domainStrategy": "IPIfNonMatch",
4 years ago
"rules": [
{
"type": "field",
"inboundTag": [
"proxy"
],
"outboundTag": "node"
},
4 years ago
{
"type": "field",
"network": "tcp,udp"
4 years ago
"outboundTag": "node"
}
]
}
}
```
`dns.json` 指定路由匹配时的DNS服务器,具体语法见[Xray文档](https://xtls.github.io/config/base/dns/)
```
{
"dns": {
"servers": [
"223.5.5.5",
"119.29.29.29"
]
}
}
```
`segment` 文件夹下默认有 `ipv4``ipv6` 两个文件,其中存储不代理的网段信息,建议绕过内网地址、本地回环地址、链路本地地址、组播地址等网段
4 years ago
```
# IPv4与IPv6均默认绕过组播地址
shell> cat /etc/scutweb/segment/ipv4
127.0.0.0/8
169.254.0.0/16
224.0.0.0/3
shell> cat /etc/scutweb/segment/ipv6
4 years ago
::1/128
FC00::/7
FE80::/10
FF00::/8
```
4 years ago
配置完成后重启容器生效
4 years ago
```
shell> docker restart scutweb
```
此时宿主机无法与macvlan网络直接连接,需要手动配置桥接,这里以Debian系Linux发行版为例
```
shell> vim /etc/network/interfaces
```
补充如下配置
```
# 具体网络信息需要按实际情况指定
auto eth0
iface eth0 inet manual
auto macvlan
iface macvlan inet static
address 192.168.2.34
netmask 255.255.255.0
gateway 192.168.2.2
dns-nameservers 192.168.2.3
pre-up ip link add macvlan link eth0 type macvlan mode bridge
post-down ip link del macvlan link eth0 type macvlan mode bridge
```
重启宿主机网络生效(重启宿主机亦可)
```
shell> /etc/init.d/networking restart
[ ok ] Restarting networking (via systemctl): networking.service.
```
配置完成后,TProxy容器的IP地址可视为旁路由IP,需要使用TProxy代理的设备修改其网关为该IP地址,若想让内网全部设备均可使用,则需修改路由器DHCP设置,将网关指向容器IP(仅对动态IP地址设备生效,配置过静态IP的设备仍需手动修改)
## 实例演示
### 示例1-全局科学上网
> 代理全部流量并进行分流,国内流量直连,国外流量走科学上网节点
初始化命令中指定容器IP地址
```
# custom.sh
ip addr flush dev eth0
ip addr add 192.168.2.4/24 brd 192.168.2.255 dev eth0
ip route add default via 192.168.2.2
```
绕过内网IP地址,添加 `192.168.2.0/24` 网段
```
# segment/ipv4
127.0.0.0/8
169.254.0.0/16
192.168.2.0/24
224.0.0.0/3
```
此处将DNS指向本地的无污染[ClearDNS](https://github.com/dnomd343/ClearDNS)服务,如果未部署该服务,修改为 `8.8.8.8` 即可(Xray会将其重新路由至国外节点,不存在污染问题)
```
# dns.json
{
"dns": {
"servers": [
"192.168.2.3"
]
}
}
```
可以配置多个服务节点负载均衡,提高科学上网速度,这里设置了三个VLESS+XTLS节点。
```
# outbounds.json
{
"outbounds": [
{
"tag": "proxy01",
"protocol": "vless",
"settings": {
"vnext": [
{
"address": "···",
"port": 443,
"users": [
{
"id": "···",
"encryption": "none",
"flow": "xtls-rprx-direct"
}
]
}
]
},
"streamSettings": {
"network": "tcp",
"security": "xtls",
"xtlsSettings": {
"allowInsecure": false,
"serverName": "···"
}
}
},
{
"tag": "proxy02",
"protocol": "vless",
"settings": {
"vnext": [
{
"address": "···",
"port": 443,
"users": [
{
"id": "···",
"encryption": "none",
"flow": "xtls-rprx-direct"
}
]
}
]
},
"streamSettings": {
"network": "tcp",
"security": "xtls",
"xtlsSettings": {
"allowInsecure": false,
"serverName": "···"
}
}
},
{
"tag": "proxy03",
"protocol": "vless",
"settings": {
"vnext": [
{
"address": "···",
"port": 443,
"users": [
{
"id": "···",
"encryption": "none",
"flow": "xtls-rprx-direct"
}
]
}
]
},
"streamSettings": {
"network": "tcp",
"security": "xtls",
"xtlsSettings": {
"allowInsecure": false,
"serverName": "···"
}
}
},
{
"tag": "direct",
"protocol": "freedom",
"settings": {}
},
{
"tag": "block",
"protocol": "blackhole",
"settings": {}
}
]
}
4 years ago
```
在路由中配置分流与负载均衡,其中还开启了广告拦截功能
```
# routing.json
{
"routing": {
"domainStrategy": "IPOnDemand",
"rules": [
{
"type": "field",
"inboundTag": [
"proxy"
],
"balancerTag": "balancer"
},
{
"type": "field",
"domain": [
"geosite:category-ads-all"
],
"outboundTag": "block"
},
{
"type": "field",
"domain": [
"geosite:cn"
],
"outboundTag": "direct"
},
{
"type": "field",
"ip": [
"geoip:private",
"geoip:cn"
],
"outboundTag": "direct"
},
{
"type": "field",
"network": "tcp,udp",
"balancerTag": "balancer"
}
],
"balancers": [
{
"tag": "balancer",
"selector": [
"proxy"
]
}
]
}
}
```
配置完成后重启容器即可使用
### 示例2-校园网绕过认证
> 部分校园网存在TCP/53或UDP/53端口无认证漏洞,可将全部流量代理并转发到个人服务器上,实现免认证、无限速的上网
初始化命令中指定容器IP地址,同时模拟外网ICMP数据包响应(否则所有ping均为超时)
```
# custom.sh
ip addr flush dev eth0
ip addr add 192.168.2.2/24 brd 192.168.2.255 dev eth0
ip route add default via 192.168.2.1
iptables -t nat -N SCUT_PING
iptables -t nat -A SCUT_PING -j DNAT --to-destination 192.168.2.2
iptables -t nat -A PREROUTING -i eth0 -p icmp -j SCUT_PING
```
绕过内网IP地址,添加 `192.168.2.0/24` 网段
```
# segment/ipv4
127.0.0.0/8
169.254.0.0/16
192.168.2.0/24
224.0.0.0/3
```
此时所有流量将被代理,不存在域名分流需求,因此无需设置DNS服务器
```
# dns.json
{
"dns": {}
}
```
这里配置了三个节点,平时使用一台即可,其他两台作为备用容灾
```
# outbounds.json
{
"outbounds": [
{
"tag": "nodeA",
"protocol": "vless",
"settings": {
"vnext": [
{
"address": "···",
"port": 53,
"users": [
{
"id": "···",
"encryption": "none",
"flow": "xtls-rprx-direct"
}
]
}
]
},
"streamSettings": {
"network": "tcp",
"security": "xtls",
"xtlsSettings": {
"allowInsecure": false,
"serverName": "···"
}
}
},
{
"tag": "nodeB",
"protocol": "vmess",
"settings": {
"vnext": [
{
"address": "···",
"port": 53,
"users": [
{
"id": "···",
"alterId": 0,
"security": "auto"
}
]
}
]
},
"streamSettings": {
"network": "tcp",
"security": "tls",
"tlsSettings": {
"allowInsecure": false,
"serverName": "···"
}
}
},
{
"tag": "nodeC",
"protocol": "shadowsocks",
"settings": {
"servers": [
{
"address": "···",
"method": "aes-256-gcm",
"ota": false,
"password": "···",
"port": 53
}
]
},
"streamSettings": {
"network": "tcp"
}
}
]
}
```
路由核心接管全部流量并转发给 `nodeA` 节点
```
# routing.json
{
"routing": {
"domainStrategy": "AsIs",
"rules": [
{
"type": "field",
"network": "tcp,udp",
"outboundTag": "nodeA"
}
]
}
}
```
配置完成后重启容器即可使用
## 开发相关
### 预设接口
+ `Socks5代理`: 1080端口,支持UDP,无授权,标志为 `socks`
+ `HTTP代理`: 1081端口,无授权,标志为 `http`
+ `全局Socks5代理`: 10808端口,支持UDP,无授权,标志为 `proxy`
### 容器构建
**本地构建**
4 years ago
```
4 years ago
# 克隆仓库
shell> git clone https://github.com/dnomd343/TProxy.git
shell> cd TProxy
# 构建镜像
shell> docker build -t tproxy .
4 years ago
```
**交叉构建**
4 years ago
```
4 years ago
# 构建并推送至Docker Hub
shell> docker buildx build -t dnomd343/tproxy --platform="linux/amd64,linux/arm64,linux/386,linux/arm/v7" https://github.com/dnomd343/TProxy.git#master --push
4 years ago
```
## 许可证
4 years ago
MIT ©2021 [@dnomd343](https://github.com/dnomd343)