252 lines
6.7 KiB
Plaintext
252 lines
6.7 KiB
Plaintext
---
|
||
title: 利用一台小鸡实现网络自由
|
||
date: '2022-10-06'
|
||
tags:
|
||
[
|
||
'SNI',
|
||
'TLS',
|
||
'Reverse Proxy',
|
||
'反向代理',
|
||
'正向代理',
|
||
‘内网穿透',
|
||
'Caddy',
|
||
'Xray',
|
||
'Vless',
|
||
]
|
||
draft: true
|
||
summary: SNI Proxy 进行 TLS 分流;Caddy 对网站和 Xray 进行反向代理;Xray 实现正向、反向代理(内网穿透)。
|
||
---
|
||
|
||
## 前言
|
||
|
||
由于条件所限,我在住所使用了一台 AMD 小主机运行了 PVE,上面跑了许多服务,他们有一个共同点——具有 Web UI。另一方面,我的宽带使用共享公网 IP,使用 Full Cone NAT,这就导致我在外难以方便地访问内网服务。为了解决这个令人不爽的问题,我准备使用 VPS 主机作为中间人,用于反射相关流量。
|
||
|
||
比较悲剧的是,在撰写本文时,VPS 的 443 端口被干掉……不过还好,VPS 即将到期,如果不能恢复,就换别的了。
|
||
|
||
## 网络拓扑图
|
||
|
||
简单画一下,大概网络拓扑长下面这样:
|
||
|
||
```mermaid
|
||
graph TB;
|
||
subgraph CLIENT
|
||
BROWSER(Browser)
|
||
XRAY-C(Xray)
|
||
GIT-C(Git)
|
||
end
|
||
|
||
subgraph VPS
|
||
SNI-PROXY(SNI Proxy)
|
||
CADDY-S(Caddy)
|
||
XRAY-S(Xray)
|
||
end
|
||
|
||
subgraph HOME
|
||
CADDY-H(Caddy)
|
||
XRAY-H(Xray)
|
||
GIT-S(Git)
|
||
end
|
||
|
||
BROWSER--https://www.example.com:443-->SNI-PROXY--https://www.example.com:443-->CADDY-S
|
||
BROWSER--https://home.example.com:443-->SNI-PROXY--https://home.example.com:443-->XRAY-H--https://home.example.com:443-->CADDY-H
|
||
XRAY-C--vless://vless.example.com:443-->SNI-PROXY--vless://vless.example.com:443-->XRAY-S
|
||
XRAY-C--vmess://vmess.example.com:443-->SNI-PROXY--vmess://vmess.example.com:443-->CADDY-S--vmess://vmess.example.com WITHOUT LTS-->XRAY-S
|
||
BROWSER--http://www.example.com:80-->CADDY-S
|
||
GIT-C--git://git.example.com:2222-->XRAY-S--git://git.example.com:2222-->XRAY-H--git://git.example.com:2222-->GIT-S
|
||
```
|
||
|
||
从上图中可以看到,VPS 公网 443 端口由 SNI Proxy 监听,80 端口由 Caddy 监听,2222 端口作为 家庭 Git 服务器的透传端口,由 Xray 直接监听并反射(relay)给家庭 Git 服务。
|
||
|
||
SNI Proxy 通过对 TLS 的 SNI 对流量进行区分,将访问部署在家庭中的网站的流量直接转发给 Xray,再由 Xray 根据规则转发给家庭中的 Caddy 服务,最后由家庭中的 Caddy 服务将 HTTPS 流量转为 HTTP 流量,并转发给目标服务。对于 VLESS 流量,将转发给 Xray 对应端口处理;对于 vmess 和其他网站流量,转发给 Caddy 的 443 端口处理。
|
||
|
||
Caddy 接收到 443 端口的流量,将根据域名等规则处理并转发给对应服务,如果是 vmess 流量,将扒掉 TLS 层并交由 Xray 处理。Caddy 还接收 80 端口的流量,这部分流量基本上都转发给对应 HTTP 服务进行处理。
|
||
|
||
## 部署
|
||
|
||
### Docker Compose
|
||
|
||
```yml
|
||
version: '3.9'
|
||
|
||
networks:
|
||
caddy:
|
||
xray:
|
||
|
||
volumes:
|
||
caddy-data:
|
||
caddy-config:
|
||
acme-sh-data:
|
||
|
||
services:
|
||
xray:
|
||
image: teddysun/xray
|
||
container_name: xray
|
||
restart: always
|
||
ports:
|
||
- '2222:2222'
|
||
networks:
|
||
- xray
|
||
volumes:
|
||
- ./xray/config.yml:/etc/xray/config.yml
|
||
- acme-sh-data:/certs
|
||
command: 'xray -c=/etc/xray/config.yml'
|
||
|
||
caddy:
|
||
image: caddy:2
|
||
container_name: caddy
|
||
restart: always
|
||
ports:
|
||
- '80:80'
|
||
networks:
|
||
- caddy
|
||
- xray
|
||
volumes:
|
||
- $PWD/caddy/Caddyfile:/etc/caddy/Caddyfile
|
||
- $PWD/site:/srv
|
||
- caddy-data:/data
|
||
- caddy-config:/config
|
||
- acme-sh-data:/certs
|
||
|
||
sniproxy:
|
||
image: tommylau/sniproxy
|
||
container_name: sniproxy
|
||
restart: always
|
||
networks:
|
||
- caddy
|
||
- xray
|
||
ports:
|
||
- '443:443'
|
||
volumes:
|
||
- $PWD/sniproxy:/etc/sniproxy
|
||
- /var/log/sniproxy:/var/log/sniproxy
|
||
acme.sh:
|
||
image: neilpang/acme.sh:dev
|
||
container_name: acme.sh
|
||
restart: always
|
||
volumes:
|
||
- acme-sh-data:/acme.sh
|
||
env_file: acme.env
|
||
command: 'daemon'
|
||
```
|
||
|
||
### SNI Proxy 配置
|
||
|
||
文件位于 `sniproxy/sniproxy.conf
|
||
|
||
```conf
|
||
user nobody
|
||
group nogroup
|
||
|
||
# PID file
|
||
pidfile /var/run/sniproxy.pid
|
||
|
||
error_log {
|
||
# Log to the daemon syslog facility
|
||
# syslog deamon
|
||
|
||
# Alternatively we could log to file
|
||
filename /var/log/sniproxy/sniproxy.log
|
||
|
||
# Control the verbosity of the log
|
||
priority notice
|
||
}
|
||
|
||
listen 443 {
|
||
proto tls
|
||
table https_hosts
|
||
|
||
access_log {
|
||
filename /var/log/sniproxy/https_access.log
|
||
priority notice
|
||
}
|
||
}
|
||
|
||
table https_hosts {
|
||
vmess.example.com caddy:443 # vmess
|
||
vless.example.com xray:443 # VLESS
|
||
.*\.example\.com xray:8443 # HTTPS tunnel
|
||
www.example.com caddy:443 # VPS vhost
|
||
}
|
||
```
|
||
|
||
### Caddy 配置
|
||
|
||
文件位于 `caddy/Caddyfile`。
|
||
|
||
```caddyfile
|
||
www.example.com {
|
||
encode zstd gzip
|
||
reverse_proxy some-host:80
|
||
}
|
||
|
||
vmess.example.com {
|
||
reverse_proxy h2c://xray:80 // vmess
|
||
}
|
||
```
|
||
|
||
### Xray 配置
|
||
|
||
文件位于 `xray/config.yml`
|
||
|
||
```yml
|
||
---
|
||
inbounds:
|
||
- tag: vless-xtls.in
|
||
listen: 0.0.0.0
|
||
port: 443
|
||
protocol: vless
|
||
settings:
|
||
clients:
|
||
- id: <your-uuid>
|
||
flow: xtls-rprx-direct
|
||
decryption: none
|
||
streamSettings:
|
||
network: tcp
|
||
security: xtls
|
||
xtlsSettings:
|
||
serverName: vless.example.com
|
||
alpn:
|
||
- h2
|
||
- http/1.1
|
||
certificates:
|
||
- certificateFile: /certs/*.example.com/fullchain.cer
|
||
keyFile: /certs/*.example.com/*.example.com.key
|
||
falklbacks:
|
||
- name: 'vmess.ivanli.cc'
|
||
dist: 80
|
||
xver: 1
|
||
- dest: 'caddy:80'
|
||
xver: 1
|
||
- listen: 0.0.0.0
|
||
port: 80
|
||
protocol: vmess
|
||
settings:
|
||
clients:
|
||
- id: <your-uuid>
|
||
decryption: none
|
||
streamSettings:
|
||
network: h2
|
||
security: none
|
||
httpSettings:
|
||
path: /
|
||
host:
|
||
- vmess.example.com
|
||
# ... 其他配置 ...
|
||
```
|
||
|
||
关于反向代理及泛域名证书相关配置,参考[使用 Xray、acme.sh、Docker Compose 搭建内网穿透服务](https://ivanli.cc/blog/build-an-frp-using-xray-acme.sh-docker-compose)。
|
||
|
||
### 启动服务
|
||
|
||
启动服务前,确保你的配置都完成了。之后在 `docker-compose.yml` 同级目录下执行下面的命令:
|
||
|
||
```bash
|
||
docker compose up -d
|
||
```
|
||
|
||
如果有防火墙,注意放行端口。
|
||
|
||
## 方案浅析
|
||
|
||
上述方案主要解决了一个问题,443 端口的复用。使得所有**“HTTPS 流量”**都流向 443 端口,显得非常治愈。因为使用了境外服务器,为了避免被误杀,实现了站点伪装,这样在一定程度上避免被主动探测发现问题。可是在撰写本文前,翻车了,原因可能是客户端指纹特征导致的。因为除了穿透服务的客户端是完全可控的,能通过配置[开启 uTLS 来规避](https://t.ivanli.cc/UFImlX),但是在网关上的 Clash 客户端却没提供相关配置选项。大意了。
|