[BUG] Calibreweb Nginx configuration doesn't work for IPv6

[Irishsmurf]

What happened?

Steps to reproduce:

1. Install nginx.
   • box install nginx
2. Install CalibreWeb
   • box install calibreweb
3. Attempt to load https://<website>/calibreweb from an IPv6 Address
4. 500 Internal Error
   • From Nginx Error logs:

   •  2023/09/08 16:45:16 [crit] 2382422#2382422: *490 bind(<IPv6 IP ADDRESS>) failed (97: Address family not supported by 
      protocol) while connecting to upstream, client: <IPv6 IP ADDRESS>, server: <SERVER HOSTNAME>, 
      request: "GET /calibreweb HTTP/2.0", upstream: "http://127.0.0.1:8083/calibreweb", host: "<SERVER HOSTNAME>", referrer: 
      "https://<SERVER HOSTNAME>/"
     

Expected results:

Page loads.

Results occur:

CalibreWeb is inaccessible if the client is using IPv6.

---

I fixed this by modifying the nginx configuration in calibreweb.conf to:

location /calibreweb {
        proxy_pass              http://127.0.0.1:8083;
        proxy_set_header        Host $proxy_host;
        proxy_set_header        X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header        X-Forwarded-Host $host;
        proxy_set_header        X-Forwarded-Proto $scheme;
        proxy_http_version      1.1;
        proxy_set_header        Upgrade $http_upgrade;
        proxy_set_header        Connection $http_connection;

        proxy_set_header        X-Scheme        $scheme;
        proxy_set_header        X-Script-Name   /calibreweb;  # IMPORTANT: path has NO trailing slash
}

Swizzin commit

9dc0c50

What OS are you using?

Debian 10 (Buster)

What architecture is your OS?

amd64

Relevant logs and output

2023/09/08 16:45:16 [crit] 2382422#2382422: *490 bind(<IPv6 IP ADDRESS>) failed (97: Address family not supported by 
       protocol) while connecting to upstream, client: <IPv6 IP ADDRESS>, server: <SERVER HOSTNAME>, 
       request: "GET /calibreweb HTTP/2.0", upstream: "http://127.0.0.1:8083/calibreweb", host: "<SERVER HOSTNAME>", referrer: 
       "https://<SERVER HOSTNAME>/"

[liaralabs]

        proxy_set_header        Upgrade $http_upgrade;
        proxy_set_header        Connection $http_connection;

Shouldn't be necessary -- these are for websocket upgrades, which I don't think calibre web makes use of.

It looks like the magic happens in translating

proxy_set_header        Host            $http_host;

to

proxy_set_header        Host $proxy_host;

And potentially adding the X-Forwarded-Host (though that mostly depends on whether calbre web even cares about this header)

[Irishsmurf]

Turns out it's not the Host header - http_host/proxy_host are essentially the same.

It's the proxy_bind bit:

1. Client hits the host via the IPv6 interface
2. nginx will then bind to $server_addr, which will be the server's IPv6 NIC Address.
3. Attempts to connect with calibreweb's upstream address - http://127.0.0.1:8083
4. This is an IPv4 address - the protocol mismatch between nginx's IPv6 bind() address, and the upstream server causes a critical internal error.
5. 500 Internal Error

---

I've sent a PR (#1072) which just removes this setting - and falls back to the default behavior (where it binds with the same protocol as upstream)