Exception from MVC API to MVC site gives error - c#

I have created an MVC (4) Web API which works fine. It has a simple login-method that throws an exception when the user cannot be found.
Besides the API I created a simple website that calls the API by HttpClient:
public T ExecutePost<T>(string apiUrl, IEnumerable<KeyValuePair<string, string>> postData) {
HttpContent content = null;
if (postData != null)
content = new FormUrlEncodedContent(postData);
var a = _client.PostAsync(apiUrl, content).ContinueWith(httpResponseMessage =>
JsonConvert.DeserializeObject<T>(httpResponseMessage.Result.Content.ReadAsStringAsync().Result)
);
return a.Result;
}
You can call this method as
ExecutePost<User>("url_to_controller_action_api", list_with_keys_and_values_to_post)
When this method calls the API with the postData-fiels username and password (both are correct and known by the system) an object called User will be returned... Works like a charm.
When I call this method with the wrong username and/or password, the API throws an exception (User not found), but the method ExecutePost throws an exception aswell and the web page shows a nice, yellow-isch, red-letter page with errors that the normal user does not understand. The reason is easy: The data sent back from the API is not the same as the data that can be put in the object User.
So, what I would like to do is deserialize the exception, from the API, in an object called Error and return that to the controller of the MVC website, so I can put the error "User not found" on the page with the correct design of the site.
What is the best way to do this? Try-catch in all actions? Doesn't feel right to me... So suggestions are more than welcome.
Most things I found were API-side stuff, but I want to fetch and handle the exception on the site-side.
Thanks in advance

On your Web API when you detect an invalid login, make sure that an HttpResponseException gets thrown with a status code of 401 (HttpStatusCode.Unauthorized).
//login failed:
var resp = new HttpResponseMessage(HttpStatusCode.Unauthorized)
{
Content = new StringContent("Invalid Username or password")
};
throw new HttpResponseException(resp);
In your calling code, you can then first check if httpResponseMessage.StatusCode==HttpStatusCode.OK before you attempt to deserialise the response into a User.
var a = _client.PostAsync(apiUrl, content).ContinueWith(httpResponseMessage => {
if (httpResponseMessage.Status!=HttpStatus.OK)
{
string ErrorMessage = httpResponseMessage.Result.Content.ReadAsStringAsync();
//and whatever you want to do with that error message here
}
else
{
try
{
var user = JsonConvert.DeserializeObject<T>(httpResponseMessage.Result.Content.ReadAsStringAsync().Result);
}
catch(Exception ex)
{
//honest to goodness unrecoverable failure on the webapi.
//all you can do here is fail gracefully to the caller.
}
}//endif (Status OK)
}//end anonymous function
);
If it's not a 200 then you can execute a second check to see if it's a 401 (and have some specific handling for invalid login) or a more general 500 (something went wrong on the server), etc, and get the actual error message of the exception ("Invalid Username or password") to send back to the client:
var errorMessage = httpResponseMessage.Content.ReadAsStringAsync().Result;
Finally, even if it is 200 then you should still try...catch the deserialisation operation so you can gracefully handle any issues with the returned data, if for whatever reason the data can't be turned into a User object.

Related

Response is not being written after Response.End() when request departs from asp:multiview or ajax

I wrote a HttpModule to intercept, evaluate and authorize requests, checking if logged user has appropriate access to the url being requested, in a pretty old legacy system written in ASP.NET 2.0(Web pages, not Web app), whose customer does not want to port to a newer framework. Restrictions have been loaded and cached at login time.
Everything works fine, except when some page contains an <asp:MultiView> component or when there is a button that launch an ajax method. When one of these situations occur, and user doesn't have rights to access that url, an alert box pops up with an "Unknown error" message, that came from a ThreadAbortException thrown by Response.End() method.
The question is: Why does my "Unauthorized" message is being overwritten by "Unknown Error" from the exception, only on these two situations?
Is there a way of doing an Url Authorization system, using database and caching and without cluttering Web.config with roles like those older ASP.NET samples?
// My module init method.
public void Init(HttpApplication context)
{
context.PreRequestHandlerExecute += new EventHandler(context_PreRequestHandlerExecute);
// PreRequestHandlerExecute is the first stage at ASP.NET Pipeline
// where we could get a fulfilled Session variable
}
private void context_PreRequestHandlerExecute(object sender, EventArgs e)
{
HttpApplication application = (HttpApplication)sender;
HttpContext context = application.Context;
// additional request filtering/validation/etc.
LoggedUser user = (LoggedUser)application.Session["user"];
string path = context.Request.Path;
// more checks and rules...
if (!checkUserAuthorization(path, user))
{
context.Response.Write("<script>alert('Unauthorized. Contact your manager.');</script>");
context.Response.Write("<script>window.history.back();</script>");
context.Response.StatusCode = 403;
context.Response.End();
}
}
EDIT: What I've already tried (with no goal):
Response.OutputStream.Close();
Response.Flush();
HttpApplication.CompleteRequest();
it's by design. you must ignore it and add a catch for that exception.
try {
context.Response.End();
}
catch{}
Foreword
After a lot of research, finally I got it. Considering ASP.NET 2.0, concerning AJAX operations, the project I'm working uses a Microsoft component called "Atlas", which in turn got renamed to ASP.NET AJAX. At the time this system was written, the developers used the beta ASP.NET AJAX (codename "Atlas") to address all ajax and partial rendering needs.
I needed to dig deeper in source code (thanks to Reflector), to understand and inspect from where does that "Unknown Error" comes.
Inside the Microsoft.Web.Atlas, there is a file named Microsoft.Web.Resources.ScriptLibrary.*.Atlas.js (where * could be Debug or Release) which is rendered at runtime through a WebResource.axd "proxy".
This javascript file have a bug, because it expects to ASP.NET request always return an HTTP 200 (OK) response code, which in my code it's not happening (I'm returning a 403 Forbidden code at my module).
Code
From Microsoft.Web.Resources.ScriptLibrary.*.Atlas.js taken from WebResource.axd:
this._onFormSubmitCompleted = function(sender, eventArgs) {
var isErrorMode = true;
var errorNode;
var delta;
if (sender.get_statusCode() == 200) {
delta = sender.get_xml();
if (delta) {
errorNode = delta.selectSingleNode("/delta/pageError");
if (!errorNode) {
isErrorMode = false;
}
}
}
if (isErrorMode) {
if (errorNode) {
pageErrorMessage = errorNode.attributes.getNamedItem('message').nodeValue;
}
else {
pageErrorMessage = 'Unknown error';
}
this._enterErrorMode(pageErrorMessage);
return;
}
// Code continues.
}
From this code, we can see that since response code is not an 200 OK, that errorNode variable won't be set, and this if (errorNode) statement will always be false.
In this case, I was left with two options: Always return HTTP 200 and modify all pages that have an <atlas:ScriptManager> with and add an ErrorTemplate tag on each, or supersede that script with one that consider non-HTTP 200 responses, loading it below </form> tag at the Master page.
There is a lot of tutorials on how to do a proper error handling when using ScriptManager and UpdatePanels (an official one here), by subscribing to the AsyncPostBackError event), but this beta version (Atlas) simply don't have this event.

Check if route is valid in RequestStartup in NancyFX

I have a self-hosted REST API in my wpf application, built using NancyFX. The application is used for updating firmware and running diagnostics on different consumer products that are connected to the computer with a USB cable.
A product must be connected to the computer in order to use the API. So I thought it would be smart to do this check in the overriden RequestStartup() method in WindsorNancyBootstrapper which means that the check can be done in one location, instead of in every module. It worked as expected. No modules will handle the request if a product isn't connected.
But this led to an unwanted side effect in the following scenario:
A product is not connected to the computer
The path is invalid
This will always return a 404 with a message saying that the device is not connected, instead of a "bad url" message. I could move the check to each module, but i'd hate doing so. What I want:
If the url is invalid, no matter if there's a connected device, always return a 404 "Bad url" response without involving any modules
if the url is valid, but there is no connected device, return a 400 "no connected device" without involving any modules
And I would like to do this in one place. I have looked around for a solution but I haven't found anything. I'm thinking that maybe my approach is a dead end. After all, I'm using the BeforeRequest pipeline which could mean that there's no way of validating the URL yet?
My method (simplified) looks like this at the moment:
protected override void RequestStartup(IWindsorContainer container, IPipelines pipelines, NancyContext context)
{
pipelines.BeforeRequest.AddItemToEndOfPipeline(ctx =>
{
// TODO: Here I would like to check if the url is valid in order to be able to return a 404 "bad url" response
if (!_hasConnectedDevice)
{
// ResponseBase is my base class for all my JSON responses
var response = new ResponseBase(ctx.Request.Url, Messages.DeviceNotConnected);
return new JsonResponse(response, new DefaultJsonSerializer())
{
StatusCode = HttpStatusCode.NotFound
};
}
if (!_deviceIsReady)
{
var response = new ResponseBase(ctx.Request.Url, Messages.DeviceNotReady);
return new JsonResponse(response, new DefaultJsonSerializer())
{
StatusCode = HttpStatusCode.BadRequest
};
}
return null;
});
// Catch all unhandled exceptions here.
pipelines.OnError += (ctx, ex) =>
{
var response = new ResponseBase(ctx.Request.Url, ex.Message);
return new JsonResponse(response, new DefaultJsonSerializer())
{
StatusCode = HttpStatusCode.InternalServerError
};
};
}
The best way I can think of to do this would be to do something similar to what we do with security where you create an extension method on INancyModule that adds something to the pipline that checks if the device is plugged in and ready and returns the relevant status code if it's not, otherwise it just returns null, then in each module you want to only work with the USB devices you add:
this.RequiresUsbThings(); // Or whatever you want to call it :)

Server cannot set status after HTTP headers have been sent - ELMAH

I have been receiving this error through ELMAH. Even though the program completes it's intended actions, I still receive this error through ELMAH and I'd like to know why and to fix it. I've gone through other threads and tried to use those suggestions but nothing I've read so far doesn't seem to work.
It's intended action is to create an excel document and then redirect the user to the page they were just on.
ActionResult:
public ActionResult ExportClaimNumberReport(int ClientID, string ClaimNo) {
ClaimNumberViewModel model = ClaimNumberReport(ClientID, ClaimNo);
CreateExcelFile.CreateExcelDocument(
model.ReportData.ToList(),
model.ReportDescription + (".xlsx"),
HttpContext.ApplicationInstance.Response);
ViewBag.client = client;
Response.Buffer = true;
Response.Redirect(Request.UrlReferrer.ToString());
if (!Response.IsRequestBeingRedirected) {
Response.Redirect("/Error/ErrorHandler");
}
return RedirectToAction("ErrorHandler", "Error");
}
If you need anymore info, just let me know
You will get the error because you are doing 2 redirects.
Once here
Response.Redirect(Request.UrlReferrer.ToString());
And then again here:
return RedirectToAction("ErrorHandler", "Error");
So the first redirect will write a redirect header to the response stream, then the second would try to do it again, but obviously you can't send http headers to the browser twice so it throws an exception. However the user won't notice because by the time the exception is thrown the browser has already been told to redirect elsewhere.
What you want to do is just call the Redirect method as the return statement from your controller action.
So replace all this:
Response.Redirect(Request.UrlReferrer.ToString());
if (!Response.IsRequestBeingRedirected) // this would always be false anyway
{
Response.Redirect("/Error/ErrorHandler");
}
return RedirectToAction("ErrorHandler", "Error");
With this:
return Redirect(Request.UrlReferrer.ToString())
Although why you are redirecting the browser back to the referring page is unclear.

How to handle PrincipalPermission Security exceptions

I have a simple method that is secured
[PrincipalPermission(SecurityAction.Demand, Role = "Administrator")]
protected void lnkClearCache_Click(object sender, EventArgs e)
{
...
}
If this is clicked without the role, it generates a System.Security.SecurityException: Request for principal permission failed. as expected.
I use ELMAH to handle logging for my errors, and I have a custom ELMAH event in my global.asax to transfer to the error pages in ways that preserve status codes which works correctly.
private void ErrorLog_Logged(object sender, ErrorLoggedEventArgs args)
{
var customErrorsSection = GetCustomErrorsSection();
var error = args.Entry;
string statusCode = error.Error.StatusCode.ToString();
if (statusCode == "0" && error is security exception)
statusCode = "403";
var errorSection = customErrorsSection.Errors[statusCode];
string redirectUrl = errorSection == null ?
customErrorsSection.DefaultRedirect : errorSection.Redirect;
RespondWithServerError(error.Id, redirectUrl, statusCode);
}
This works all well and fine and redirects to my error page which works properly, however instead of displaying the content as expected. I immediately get a second request for the error page but this time using the value of customErrorsSection.DefaultRedirect that does not come from my code in any way that I can see.
As far as I can tell it's almost as if when .NET raises an exception for PrincipalPermission and then lets the entire request complete, then after the request is complete it throws away the application response and instead responds with the default custom error.
When I'm debugging I do break on 2 separate exceptions for PrincipalPermission, whether this is a just a rethrow by .NET I'm not sure but my .NET code never sees the 2nd throw, nor does ELMAH. I always end up with a single response, single error logged, but that the url that finally renders to the browser is the default url and not 403 url that I specifically server.transferred to. If I browse to a /location that is secure I properly get the 403 error page.
I don´t excactly know where the problem is. But i´´m using something similar and for me this solution (minimized) works great.
Sub Application_Error(ByVal sender As Object, ByVal e As EventArgs)
Dim er = HttpContext.Current.Error
If er.GetType.Equals(GetType(System.Security.SecurityException)) Then
HttpContext.Current.Response.Redirect(FormsAuthentication.LoginUrl & "?ReturnUrl=" & HttpContext.Current.Request.Path)
End If
End Sub
That´s out of global.asax
But on some places I don´´t redirect, and just using try and catch the securityexception to display the user, that he is not allowed to perform such action.

Throwing a XML Web Service Exception for invalid username / password

I've created a web service and am using a Soap header for authentication, as described here:
http://aspalliance.com/805
I've adapted it, so that in every method, it calls a seperate "authenticate" method, which searches username and password in the db, and returns true or false.
My question is, within this method, if it returns false (ie, the user isn't validated) how should i throw an exception, that filters back to the consumer application?
First of all, you'd do better to add a Login operation that takes your username/password header as input, authenticates the user, then returns an authorization token of some kind in a return SOAP Header. This header should then be supplied as in input header in all subsequent operations.
Second, you should throw a SOAPException. This will translate more or less directly into a SOAP Fault. A SOAP Fault is the appropriate way to indicate an error with a web service operation for the same reason that Exceptions are better than return status in a normal method - you don't have to check the return status at the point of the call.
Finally, were you aware that Microsoft has declared ASMX web services to be "legacy" code, and that they are no longer fixing bugs in it? It's time to move to WCF.
i have used soap exceptions for login fails:
[WebMethod]
[SoapHeader("authentication")]
public User Authenticate()
{
try
{
authentication.Roles = new string[] { UserRoles.Users };
ConfigureAuthentication();
Service<ISecurity>.Interface.Authenticate();
Guid userId = Service<ISecurity>.Interface.GetUserId(authentication.UserName);
return Service<IObjectRetriever>.Interface.Retrieve<User>(userId);
}
catch (Exception ex)
{
WriteException(ex);
throw new SoapException(ex.Message, new XmlQualifiedName(SoapException.ServerFaultCode.Name), ex);
}
}

Categories

Resources