I am currently learning to use golang as a server side language. I'm learning how to handle forms, and so I wanted to see how I could prevent some malicious client from sending a very large (in the case of a form with multipart/form-data) file and causing the server to run out of memory. For now this is my code which I found in a question here on stackoverflow:

part, _ := ioutil.ReadAll(io.LimitReader(r.Body, 8388608))

r.Body = ioutil.NopCloser(io.MultiReader(bytes.NewReader(part), r.Body))

In my code r is equal to *http.Request. So, I think that code works well, but what happens is that when I send a file regardless of its size (according to my code, the maximum size is 8M) my code still receives the entire file, so I have doubts that my code actually works. So my question is. Does my code really work wrong? Is there a concept that I am missing and that is why I think my code is malfunctioning? How can I limit the size of an http request correctly?

Update

I tried to run the code that was shown in the answers, I mean, this code:

part, _ := ioutil.ReadAll(io.LimitReader(r.Body, 8388608))

r.Body = ioutil.NopCloser(bytes.NewReader(part))

But when I run that code, and when I send a file larger than 8M I get this message from my web browser:

The connection was reset

The connection to the server was reset while the page was loading.

How can I solve that? How can I read only 8M maximum but without getting that error?

CodePudding user response：

I would ask the question: "How is your service intended/expected to behave if it receives a request greater than the maximum size?"

Perhaps you could simply check the ContentLength of the request and immediately return a 400 Bad Request if it exceeds your maximum?

func MyHandler(rw http.ResponseWriter, rq *http.Request) {
   if rq.ContentLength > 8388608 {
      rw.WriteHeader(http.StatusBadRequest)
      rw.Write([]byte("request content limit exceeded"))
      return
   }

   // ... normal processing
}

This has the advantage of not reading anything and deciding not to proceed at the earliest possible opportunity (short of some throttling on the ingress itself), minimising cpu and memory load on your process.

It also simplifies your normal processing which then does not have to be concerned with catering for circumstances where a partial request might be involved, or aborting and possibly having to clean up processing if the request content limit is reached before all content has been processed..

CodePudding user response：

Your code reads:

r.Body = ioutil.NopCloser(io.MultiReader(bytes.NewReader(part), r.Body))

This means that you are assigned a new io.MultiReader to your body that:

1. reads at most 8388608 from a byte slice in memory
2. and then reads the rest of the body after those 8388608 bytes

To ensure that you only read 8388608 bytes at most, replace that line with:

r.Body = ioutil.NopCloser(bytes.NewReader(part))