Serving a websocket in Go - websocket

I'm starting to play around with websockets + go and well I think I'm misunderstanding something quite basic with websockets in Go.
I'd like to simply listen for a websocket connection and process accordingly. However all examples I see in Go using websocket is serving the web page that then connects to the websocket, is this a requirement?
The following is a basic echo server I have setup:
package main
import (
"fmt"
"code.google.com/p/go.net/websocket"
"net/http"
)
func webHandler(ws *websocket.Conn) {
var s string
fmt.Fscan(ws, &s)
fmt.Println("Received: ", s)
}
func main() {
fmt.Println("Starting websock server: ")
http.Handle("/echo", websocket.Handler(webHandler))
err := http.ListenAndServe(":8080", nil)
if err != nil {
panic("ListenAndServe: " + err.Error())
}
}
This is the javascript used to connect:
ws = new WebSocket("ws://localhost:8080/echo");
ws.onmessage = function(e) {
console.log("websock: " + e.data);
};
However this results in:
WebSocket connection to 'ws://localhost:8080/echo' failed: Unexpected response code: 403

When working with websockets from Javascript, you will seldom have to read the frames directly. To be honest, I am not even sure how to do that.
Fortunately, the websocket package already has a type, Codec that does this for you. My suggestion is to use the predefined websocket.Message codec to Recieve and Send messages instead.
Message is a codec to send/receive text/binary data in a frame on WebSocket connection. To send/receive text frame, use string type. To send/receive binary frame, use []byte type.
Using websocket.Message, your webHandler would look something like this:
func webHandler(ws *websocket.Conn) {
var in []byte
if err := websocket.Message.Receive(ws, &in); err != nil {
return
}
fmt.Printf("Received: %s\n", string(in))
websocket.Message.Send(ws, in)
}
And, no, it is not a requirement that Go serves the webpage. The 403 error you received does not have to do with Go or the websocket package.

I had a similar issue and the 403 error problem is related to how Go treats the http Origin header:
Handler is a simple interface to a WebSocket browser client. It checks if Origin header is valid URL by default. You might want to verify websocket.Conn.Config().Origin in the func. If you use Server instead of Handler, you could call websocket.Origin and check the origin in your Handshake func. So, if you want to accept non-browser client, which doesn't send Origin header, you could use Server . that doesn't check origin in its Handshake.
In order to disable the Origin check, you must use something like:
http.HandleFunc("/echo",
func (w http.ResponseWriter, req *http.Request) {
s := websocket.Server{Handler: websocket.Handler(webHandler)}
s.ServeHTTP(w, req)
});
At least that solved the issue for me (server to server WebSocket communication) and I think it could solve the problem as well if the origin header does not match.

Related

Go GRPC Refresh token for a bidirectional stream

TLDR: I am looking for a way to update headers on an open stream for each call to stream.Send(msg) without closing the stream and opening a new one.
Summary
I have a GRPC client and server built to handle bidirectional streams. To authenticate with the server the client must send a JWT in the request headers, set as "authorization". The token is valid for 30 minutes. After the token has expired, the server will terminate the connection.
I am looking for a way to refresh my authorization token from the client, and keep the stream open. The client should run in a loop executing a new request every 30 minutes with the updated token, and the updated payload. I have not seen a way to update a header from the client side for an already opened stream.
Let's look at some code to get an idea of what the client side looks like. The code below has a function to create a new instance of the client, and another function to establish the connection to the GRPC server.
func NewWatchClient(config *Config, logger *logrus.Logger) (*WatchClient, error) {
cc, err := newConnection(config, logger)
if err != nil {
return nil, err
}
service := proto.NewWatchServiceClient(cc)
return &WatchClient{
config: config,
conn: cc,
logger: entry,
service: service,
}, nil
}
func newConnection(config *Config, logger *logrus.Logger) (*grpc.ClientConn, error) {
address := fmt.Sprintf("%s:%d", config.Host, config.Port)
// rpcCredential implements credentials.PerRPCCredentials
rpcCredential := newTokenAuth(config.Auth, config.TenantID)
return grpc.Dial(
address,
grpc.WithPerRPCCredentials(rpcCredential),
)
}
Looking at the newConnection function above I can see that there is a call to another function, newTokenAuth, to create an auth token. This func returns a struct that implements the PerRPCCredentials interface.
There are two ways to set the authorization for a request.
Use grpc.WithPerRPCCredentials to add the authorization at the time of creating the connection to the server.
Use grpc.PerRPCCredentials to add the authorization to each stream opened on the connection to the server.
In this case, I am using grpc.WithPerRPCCredentials to attach the token at the time of creating the connection to the server.
Now, let's take a look at the definition of PerRPCCredentials.
type PerRPCCredentials interface {
// GetRequestMetadata gets the current request metadata, refreshing
// tokens if required. This should be called by the transport layer on
// each request, and the data should be populated in headers or other
// context. If a status code is returned, it will be used as the status
// for the RPC. uri is the URI of the entry point for the request.
// When supported by the underlying implementation, ctx can be used for
// timeout and cancellation. Additionally, RequestInfo data will be
// available via ctx to this call.
// TODO(zhaoq): Define the set of the qualified keys instead of leaving
// it as an arbitrary string.
GetRequestMetadata(ctx context.Context, uri ...string) (map[string]string, error)
// RequireTransportSecurity indicates whether the credentials requires
// transport security.
RequireTransportSecurity() bool
}
The interface requires that you define two methods. The documentation of GetRequestMetadata says
GetRequestMetadata gets the current request metadata, refreshing tokens if required
So, it looks like my implementation of PerRPCCredentials should be able to handle a token refresh for my stream or connection. Let's take a look at my implementation of PerRPCCredentials.
// tokenAuth implements the PerRPCCredentials interface
type tokenAuth struct {
tenantID string
tokenRequester auth.PlatformTokenGetter
token string
}
// RequireTransportSecurity leave as false for now
func (tokenAuth) RequireTransportSecurity() bool {
return false
}
// GetRequestMetadata sets the http header prior to transport
func (t tokenAuth) GetRequestMetadata(_ context.Context, _ ...string) (map[string]string, error) {
token, err := t.tokenRequester.GetToken()
if err != nil {
return nil, err
}
t.token = token
go func() {
time.Sleep(25 * time.Minute)
token, _ := t.tokenRequester.GetToken()
t.token = token
}()
return map[string]string{
"tenant-id": t.tenantID,
"authorization": "Bearer " + t.token,
}, nil
}
As you can see, the call to GetRequestMetadata will establish a go routine that will attempt to refresh a token every 25 minutes. Adding a go routine right here is probably not the right way to do it. It was an attempt to get the auth header to refresh, which doesn't work.
Let's take a look at the stream.
func (w WatchClient) CreateWatch() error {
topic := &proto.Request{SelfLink: w.config.TopicSelfLink}
stream, err := w.service.CreateWatch(context.Background())
if err != nil {
return err
}
for {
err = stream.Send(topic)
if err != nil {
return err
}
time.Sleep(25 * time.Minute)
}
}
The client sends a message on the stream every 25 minutes. All I'm looking to get here is that when stream.Send is called, the updated token is also sent.
This function, GetRequestMetadata only gets called once, regardless if I am setting the auth through grpc.WithPerRPCCredentials or grpc.PerRPCCredsCallOption so there appears to be no way to update the authorization header.
If you have any idea what I have missed in my attempt to utilize the PerRPCCredentials for token refresh then please let me know.
Thank you.
Headers are sent at the beginning of an RPC, and cannot be updated during the RPC. If you need to send data during the life of a stream, it needs to be part of the request message in your proto definition.

307 redirect with Authorization header

In looking at the Go docs for http it looks like the Authorization header is removed when a response is a 307. Obviously it makes sense for almost every case but is there a way not to remove the Authorization header?
You can modify your http.Client to add the header again after it has been removed using CheckRedirect:
CheckRedirect func(req *Request, via []*Request) error
Since req is the upcoming request, it can be modified before it is sent. After making the changes, return nil to indicate that the request should still be sent.
Since this is a change to the http client instead of the request, you should check that this redirect is only used for the one URL where you need it (in case you use that client to do other requests).
You client definition could look like this:
http.Client{
CheckRedirect: func(req *http.Request, via []*http.Request) error {
// you can check old responses for a status code
if len(via) != 0 && via[0].Response.StatusCode == http.StatusTemporaryRedirect {
req.Header.Add("Authorization", "some-value")
}
return nil
},
}

Echo CORS w/ Proxy middlewares causes problems w/ Access-Allow-Origins response header

I'm using LabStack's Golang Echo Framework to build out a service.
One of the routes, needs to proxy requests and responses to and from a backend service.
But I also need CORS to work on this service as well.
So I'm using middleware.CORSWithConfig along w/ a middleware.ProxyWithConfig in my request/response stack.
I'm seeing some oddness w/ the Access-Control-Allow-Origins header where the value for that header on the response from the proxied service to my Echo server *, but once it passes through the proxy, it changes to *, * by the time it gets back into the client.
Upon which I start seeing the following browser errors related to CORS violations:
VM1627:362 Access to XMLHttpRequest at 'http://localhost:6273/' from origin 'http://localhost:8002' has been blocked by CORS policy: The 'Access-Control-Allow-Origin' header contains multiple values '*, *', but only one is allowed.
Has anyone come across this? Anyone have any idea why this might be happening and maybe a way around it?
Here's some example code:
package main
func singleTargetBalancer(url *url.URL) middleware.ProxyBalancer {
targetURL := []*middleware.ProxyTarget{
{
URL: url,
},
}
return middleware.NewRoundRobinBalancer(targetURL)
}
func Noop(ctx echo.Context) (err error) {
ctx.String(
http.StatusNotImplemented,
"No op handler should never be reached!",
)
return err
}
func main() {
e := echo.New()
e.HideBanner = true
e.Use(
middleware.CORSWithConfig(middlewares.CustomCorsConfig),
middlewares.ThriftMetrics(),
)
// Have to use a Noop handler since we're not trying to set up a full-on proxy for the backend service. We only want this one route to be proxied.
e.POST(
"/",
handlers.Noop,
middleware.ProxyWithConfig(middleware.ProxyConfig{
Balancer: singleTargetBalancer("[backend service URL]"),
})
)
}
I ultimately solved this by writing a custom Echo middleware to hook into the response before Echo's proxy middleware could send the headers back to the client.
func setResponseACAOHeaderFromRequest (req http.Request, resp echo.Response) {
resp.Header().Set(echo.HeaderAccessControlAllowOrigin,
req.Header.Get(echo.HeaderOrigin))
}
func ACAOHeaderOverwriteMiddleware(next echo.HandlerFunc) echo.HandlerFunc {
return func(ctx echo.Context) error {
ctx.Response().Before(func() {
setResponseACAOHeaderFromRequest(*ctx.Request(), *ctx.Response())
})
return next(ctx)
}
}
Then just put this middleware in e.Use() right before your proxy middleware:
e.POST(
"/",
handlers.Noop,
ACAOHeaderOverwriteMiddleware,
middleware.ProxyWithConfig(middleware.ProxyConfig{
Balancer: singleTargetBalancer("[backend service URL]"),
})
)
Docs for Echo's Request::Before() hook: https://echo.labstack.com/guide/response#before-response

how to proxy GET request in golang with URL manipulation

I want to build a golang service which will listen for GET request, do some URL manipulation and then proxy a new request (to the manipulated URL) back to the browser:
(from browser -> server) GET http://www.example.com/7fbsjfhfh93hdkwhfbf398fhkef93..
(server manipulates URL - decrypts "7fbsjfhfh93hdkwhfbf398fhkef93.." -> "my-super-resource")
(server -> URL resource) GET http://www.somewhereelse.com/my-super-resource
(server -> browser) Response from http://www.somewhereelse.com/my-super-resource passed on to browser (using cors)
The whole chain will need to be synchronous which is ok. Is there a decent proxy library which allows for this sort of thing?
You can do something like this in less than 10 lines of code with the Sling package:
type Foo struct {
Bar string `json:"bar"`
}
func Handler(w http.ResponseWriter, r *http.Request) {
// Get the URL and decrypt it
url := getUrl(r)
decryptedUrl := decryptUrl(url)
// Use the decrypted URL to make a request to another service
var data *Foo
req, err := sling.New().Get(decryptedUrl).Receive(data)
if err != nil {
// Handle error...
}
// Respond to the original request using the data from the other service
respond(w, http.StatusOK, data)
}

Golang websocket nil memory error

Below is the code of my websocket server.
http.Handle("/gatewayconnector", websocket.Handler(socketHandler))
Method socketHandler has the code below:
func socketHandler(ws *websocket.Conn) {
LoadClient(ws)
var msg []byte
for {
if err := websocket.Message.Receive(ws, &msg); err != nil {
log.Error("Error in socketHandler: ", err)
break
}
validateMessage(msg)
}
}
The socket handler call back method get called when handshake happened from client and it also initiate the websocket object. So at very first step of this method i stored this websocket object into an array, so that i can retrieve it at the send method call (used to send message to client).
The issue is sometimes I'm getting nil memory reference in ws i.e. the websocket while trying to send some message to client.
Below is the send method code:
func Send(msg interface{}) error {
ws := webSocketClients[0]
if (ws == nil) {
log.Error("Websocket connection is nil, gateway should initiate the connection.")
}
return websocket.JSON.Send(ws, msg)
}
This issue arises when websocket connection remanis idle for a longer persiod of time and I call directly the send method.
Right now i have put one hack that client will ping my websocket server at regular interval to avoid this.
How to avoid this kind of problem?
Pinging websocket connection on interval is normal thing to do, please check how it's done in Gorilla Websocket Chat Example. I know you're using net/websocket, but the idea should be similar. The problem is that you can't rely that your long-time TCP connection will remain active, especially in public networks. Some firewalls silently RST connections without traffic flow after some period of time.

Resources