Websocket data will be truncated if HTTP header is too long

[ccpz]

Since libwebsockets V1.7.0, it adds a max_http_header_data in info struct:

https://github.com/warmcat/libwebsockets/blob/master/changelog#L370

• max_http_header_data: 0 for default (1024) or set the maximum amount of known
  http header payload that lws can deal with. Payload in unknown http
  headers is dropped silently. If for some reason you need to send huge
  cookies or other HTTP-level headers, you can now increase this at context-
  creation time.

Therefore if mosquitto's websocket interface receives a HTTP header over 1024 bytes, all additional bytes will be dropped. Maybe it is better to make this value configurable at mosquitto config file?

[ralight]

Is this a problem you are seeing in practice? If so, could you give me an example? A single header >1024 bytes is pretty big.

[ccpz]

We have a web site based on Laravel framework 5.2, when user login, it will open a websockets channel using paho-javascript library from browser to Nginx reverse proxy, it forwards to mosquitto websockets server (in order to prevent CSRF problem):

location /websockets/ {
        proxy_pass http:https://127.0.0.1:9001/;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";
        proxy_connect_timeout 7d;
        proxy_send_timeout 7d;
        proxy_read_timeout 7d;
    }

Therefore it will also forwards the full HTTP headers to mosquitto, And here is the PCAP file:
websockets.zip

[ralight]

I've had it confirmed that max_http_header_data is for the entirety of the headers, I'll see about making this an option.

[Hades32]

This is a very practical issue. We share the same domain for all services (including the "/mqtt" websocket endpoint). Therefor all requests will get all the cookies. Cookies with several KB are really nothing special. So, I think the default should be at least 4-8KB.

Currently I work around it with a HAProxy config:
reqidel ^Cookie:.*

[memoryhole]

It doesn't looks like this made it into a release. What would it take to get it there?

[ralight]

@memoryhole I'm pleased to say it is in version 1.6 (so I'm closing this now).