windows Archives - Tales of a Code Monkey

Writing to Event Log — the right way

roger — Sun, 27 Apr 2014 18:04:07 +0000

This one’s been on my mind for a long time. I know it’s very tempting to just use System.Diagnostics.EventLog.WriteEntry to write some string to the event log. But personally I never liked the fact that you write all that static text along with the variables like actual error messages etc. Why make your life harder analyzing events later on when there’s an easy way to fix that?

Instrumentation Manifests to the Rescue!

For a while now this has actually been quite easy, using instrumentation manifests. You can read more about it here: http://msdn.microsoft.com/en-us/library/windows/desktop/dd996930(v=vs.85).aspx. These manifests allow you to define events, templates for events, messages for events, even your own event channels (so you wouldn’t need to log into that crowded “Application” channel anymore) and a lot more. But let’s look at a little example.

Let’s start at the top. Lines 5-9 define some basic information about this instrumentation provider, like a name, a unique ID and a symbol (which will come in handy later). We can also define a friendly name for events logged this way (i.e. the event source). Let’s ignore the three xyzFileName attributes for now. On lines 11-22 we’re defining four events, some of them informational (like “the service started” or “the service stopped”), some are errors (e.g. configuration errors, or unhandled exceptions). If we wanted to define our own channel, we’d do so between lines 25 and 27. For now we’re just re-using (i.e. importing) the pre-defined “Application” channel.

Event Templates

Event templates are particularly handy if you want to write parameters with your events. Lines 29-31 define a template which has exactly one parameter, which happens to be a unicode string. We’ll use it to store exceptions. We can define more than one parameter and there’s a lot of types to use, but I’ll let you explore those on your own. This template, as you can see, is referred to by the two events with IDs 1002 and 1003.

Resources

The localization gods are with us to. Our event and template definitions so far were abstract, no actual UI strings were contained. We can define those per language, as you can see starting line 37. In the resources element and its sub-elements, we define the actual strings we want to show, including any parameters. Parameters are numbered (1-based) and are referred to with %1, %2, %3 and so on. As you can see on lines 51 and 53, we’re defining the strings for the two error events with one parameter each (“%1”), to contain the exception message. If you want line breaks, you’ll achieve those with “%n”.

Compile, with some Sugar added

So now we have a fancy manifest, but what can we do with it? Well, eventually we want to log events using the definitions from this manifest, so let’s get to it. The Windows SDK comes with two very handy tools, MC.exe (the message compiler) and RC.exe (the resource compiler). We’ll use the first to compile the manifest — and generate some c# code as a side effect — then use the second to compile the output of the first into a resource which can be linked into an executable. The commands are as follows.

mc.exe -css MyService.Events manifest.man -r obj\Debug
rc.exe obj\Debug\manifest.rc

MC.exe was nice enough to generate a file called manifest.cs for us. That file contains some code that you can use to log every event you defined in the manifest. This is why it was so handy to define the events (and templates): depending on how many parameters an event’s template has, the generated methods will ask you to provide just as many (typed) values for those parameters. Isn’t that great?! You’ll also find the compiled manifest.res file in obj\Debug. You can link that into its own executable (or your main assembly too, if you wanted), as follows:

csc.exe /out:MyService.Events.dll /target:library /win32res:obj\Debug\manifest.res

And you have a satellite assembly which holds the manifest you’ve built! CSC will log a warning about missing source files (because you didn’t add any .cs files to be compiled) but so far that doesn’t hurt anyone. We could probably also use link.exe but so far the C# compiler does a nice enough job.

Use that generated Code

Remember the code that was generated for us by MC.exe? Let’s go ahead and use it.

// ...
TheEventLog.EventWriteServiceStarted();
// ...
TheEventLog.EventWriteServiceConfigurationError(exception.Message); // ... or log the entire exception, including stack traces.
// ...

Wasn’t that very easy?

Install the Event Provider

There’s still something missing though: we’ll need to install our instrumentation/event provider with the system. It’s similar to creating the event source (which in fact will happen automatically when installing the manifest). This will typically happen in your application’s/service’s installer, using a command line as follows. But before that, remember the xyzFileName attributes we talked about? These need to be updated to point to the full path of the MyService.Events.dll assembly we generated. Otherwise the following command is going to fail.

wevtutil.exe im path\to\my\manifest.man

From now on, when your app or service starts and logs those events, they’ll show up in the event viewer. For the two events we defined with parameters, the values of the parameters are essentially the only thing that’s stored along with the ID of the event. Likewise, they’ll be the only thing that’s going to be exported with the event — so the files with the exported events you’re going to ask your customers to send you are going to be a lot smaller and won’t contain the static part of the events you already know anyway!

To uninstall the manifest, just run this command:

wevtutil.exe um path\to\my\manifest.man

Both commands need to run elevated (particularly important to remember when writing your installer).

Next Steps

As a next step, you’ll probably want to add the manual steps of compiling the manifest linking into the satellite assembly to the project file as automated targets. I’ll likely write another post about that in the future too.

Summary

As you can see, writing a manifest, compiling it and using the generated code to write to the event log is quite easy. So no more excuses to write each event as one big string (which is can be a lot harder to analyze when they come back from your customers because you first need to parse the strings).

The post Writing to Event Log — the right way appeared first on Tales of a Code Monkey.

Side-by-Side Tracing

roger — Tue, 28 Dec 2010 17:49:00 +0000

Whenever you run into problems with SxS, e.g. because an assembly cannot be loaded because of missing dependencies or similar, using the tracing tool which SxS offers can help you a lot. Hereâ€s how to use it.

Start Tracing

You start tracing by calling the following command:

sxstrace Trace â€“logfile:sxs.etl

This will keep the tracer running until you actually press enter. If you want to stop it through issuing another command, you should call

sxstrace Trace â€“logfile:sxs.etl -nostop

This will exit as soon as the tracing session was created.

Stop Tracing

In case you were using the nostop option, youâ€ll have to stop tracing through the following command.

sxstrace StopTrace

Transforming the Traces

As with most traces, youâ€ll need to transform the traces from the raw format in the ETL file to a human readable format. Youâ€ll do this with the following command.

sxstrace Parse â€“logfile:sxs.etl â€“outfile:sxs.log

In the out file you should then be able to easily figure out if your problem is related to SxS or not.

The post Side-by-Side Tracing appeared first on Tales of a Code Monkey.

A Basic Framework for Using Performance Counters with .Net Applications

roger — Sat, 03 Jul 2010 12:45:00 +0000

In my last post I promised to tell you more about the performance counters. So here we are: today I’m going to tell you about how to use the performance counter infrastructure offered by Windows in a .net application. On top of this I’m providing you with a basic framework that you can use to author performance counters through XML and then get the code to read/write the counters generated for you at build time. So let’s get started.

Some Basics

Before I go off to the code, here’s a little overview on performance counters in Windows. I’m sure if you’re reading this, you already know about Perfmon.exe, a nice little tool to look at various performance aspects of Windows machines. When plottig performance counters, you’re typically adding the counters you’re interested in. These counters are grouped in categories, for instance Processor or PhsyicalDisk. Each category can contain multiple counters, for instance Disk Read Bytes/sec or Disk Write Bytes/sec. And finally, each counter can have multiple instances, for instance in multi-processor machines, you’ll find one instance of the % Idle Time counter per processor.

What tools like Perfmon.exe do is to grab the values of the counters you selected every second (by default, but that can typically be changed) and record/plot the values. Your job here is to build performance counters that measure certain aspects of your application so these tools can help you in analyzing your performance. And that’s why you’re here, right?

Step 1: The XML Schema

In order to know what we talk about in the XML I announced, let me start with the XML schema. I use this mainly to make sure that what I have in the XML declaring the perf counters makes some sense and is legal input. And of course because Visual Studio tells you that something’s wrong when the schema is not adhered to.

On lines 21 to 23 you find some enumeration values. These map to the values defined in the PerformanceCounterType Enumeration. This is also where you can add support for more performance counter types when you need it. The other elements and attributes are used to describe the category for the performance counters and the performance counters themselves, including the help text that will show up in Perfmon.exe and also including the symbol that you can use in the code to access the counter.

Step 2: Declaring the Performance Counters

Now let’s use the schema we built above. To give you an idea of the context, I’m providing here some of the XML I use to build the perf counters for my SMTP server.

As you can see, I create seven perf counters, most of them ordinary counters that count the number of occurrences of a certain event (like when a client makes a connection to the server). I actually also use the NumberOfItems32 type to count the number of active connections â€” I basically increment the counter when a connection was established and decrement it when the connection is terminated. And I have a couple of counters which count the number of occurrences per second (the RateOfCountsPerSecond32 counter type). The good thing about these counters is that you don’t need to provide your own counter base (check out MSDN for more info on this, starting with the above mentioned PerformanceCounterType Enumeration).

Step 3: Building the XSLT to Generate Code

We already have our perf counters declared, so let’s generate some code from that XML. Using XSL stylesheets makes this very easy: We’ll use the XML as input and get c# code as output that we can simply include into our project. I won’t show the full XSLT here (it’s about 240 lines).



  

  
  
  

  
    
      
        The Parameter 'TargetNamespace' is undefined.
      
    
    
  

  
    
      private static PerformanceCounter _;

       static PerformanceCounter 
      {
        get
        {
          MakeSureCountersAreInitialized();
          return _;
        }
      }

  

  
    
      private static PerformanceCounter _;

       static PerformanceCounter 
      {
        get
        {
          MakeSureCountersAreInitialized();
          return _;
        }
      }

      private static PerformanceCounter _Base;

       static PerformanceCounter Base
      {
        get
        {
          MakeSureCountersAreInitialized();
          return _Base;
        }
      }

  

  
    
        _ = new PerformanceCounter(
          CategoryName,
          "",
          false);

  

  
    
        _ = new PerformanceCounter(
          CategoryName,
          "",
          false);

        _Base = new PerformanceCounter(
          CategoryName,
          " Base",
          false);

  

  
      
      {
        CounterCreationData counter = new CounterCreationData();

        counter.CounterName = "";
        counter.CounterHelp = "";
        counter.CounterType = PerformanceCounterType.;

        counters.Add(counter);
      }

  

  
      
      {
        CounterCreationData counter = new CounterCreationData();

        counter.CounterName = "";
        counter.CounterHelp = "";
        counter.CounterType = PerformanceCounterType.;

        CounterCreationData counterBase = new CounterCreationData();

        counterBase.CounterName = " Base";
        counterBase.CounterHelp = "";
        counterBase.CounterType = PerformanceCounterType.RawBase;

        counters.AddRange(new CounterCreationData[]{counter, counterBase});
      }

On lines 9 to 11 you’ll find the declaration for the parameters which you can use to change the namespace, the class name and the access modifiers (public vs internal really) for the generated classes.

Starting on lines 22 and 37 respectively, you find different templates for general counters and the RawFraction type counters. I use this to automatically generate the base counter so you won’t have to worry about it. You can do similar things for other counters which need a base counter. Then, starting on lines 63 and 72 respecctively, I initialize the counters, with specialization again for the RawFraction counter types. And starting on lines 86 and 100 respectively you’ll find the code to create and set the CounterCreationData which is used to register perf counters.

You can find the full code in the ZIP file linked to at the end of this post. With this we basically have all the tools we need to actually generate the code as part of the build. We just need to bring the pieces together.

Step 4: Bringing it Together aka Updating the Project

Well yes, I have silently assumed that you have a c# project (.csproj or any other project that is supported by MSBuild.exe) that you’re working on to update. All we need to do in the .csproj is to add the files we authored, then run the XSLT transformation and add the output .cs file to the list of source code files as well. And here’s how you can do that.



  
  
    
      Diag\PerfCounters.Generated.cs
      
        
      
      Designer
    
    
    
      true
      true
      PerfCounters.xml
    
  
  
    
      Designer

Add lines 4 to 32 to the end of the project file and make sure the paths you’re using point to the places where you actually stored the files. On line 8 you can see that I’m passing the target namespace parameter (called out in step 3) by concatenating the assemblies default root namespace and “.Diag”. This way you can actually put this stuff into a .targets file that you can include in all of the projects which take advantage of this framework.

Starting on line 25 I defined the actual target which runs the XslTransformation task that comes with .net 4.0. I also declared the inputs as being the XML file and the XSLT file â€” this way, while experimenting with the XSLT, the code gets regenerated also if the XML file hasn’t changed, but the XSLT has.

Using the Generated Classes

So now you have pretty much all the code you need to use these performance counters of yours. All that’s left is actually writing the code which

Installs the Performance Counters
Updates the Performance Counters
Uninstalls the Performance Counters

#1 and #3 would probably be used by your installer and #2 by your application at runtime. Incrementing and decrementing the counters is as easy as

PerfCounters.CymbeMail.ActiveConnections.Increment();
//...
PerfCounters.CymbeMail.ActiveConnections.Decrement();

Where PerfCounters is the name of the generated class which contains all the generated perf counters, CymbeMail is the name of the class for the category (the Symbol attribute, remember?) and ActiveConnections is of course the symbol name for the counter we’re modifying. Isn’t this simple?

As for setting things up (or removing them), the relevant piece of code gets generated for you, too. So all you need to do is actually call it, potentially from a small application that you run â€” as I mentioned before â€” from your installer.

// Setup the Performace Counters
PerfCounters.Setup();

// And for the Uninstaller:
// Remove the Performance Counters
PerfCounters.Remove();

Of course you’ll have to make sure that the version of the generated code that’s run by the uninstaller is the same version as the installer did run â€” else you may end up with stale categories / counters on the machine. But then again, if you’re installing through an MSI package, you get that for free … when done properly.

Summary

In just a few steps I have shown how you can build a framework to use performance counters in .Net applications. On top of it’s simplicity, it’s also quite easy to modify the framework and adapt it to your own needs or extend it to allow more performance counter types.

And finally, as promised, here’s the ZIP file which contains the relevant pieces. Please forgive me for not adding the .csproj file â€” that actually contains other relevant data that I didn’t want to share. Instead I added the generated c# code file for reference.

PerfCounters.zipÂ (4.08kb)

The post A Basic Framework for Using Performance Counters with .Net Applications appeared first on Tales of a Code Monkey.

You’ve got to restart that process …

roger — Sat, 06 Feb 2010 23:42:00 +0000

… in order to pick up the changes you’ve recently made to the environment variables.

Alright, to cut a long story short, I’m using the Microsoft Natural Ergonomic Keyboard at work. It has these (5) ‘My Favorits’ keys on top to which you can assign custom commands. I typically make heavy use of that to

Start a new console for the build environment
Start PowerShell
Start a lightweight editor with syntax hightlighting for many file types
Start PaintDotNet
Start a new (blank) console

I typically use #5 for checking out small things and starting remote user debugging sessions. From time to time, I tweak the environment a little bit by adding new environment variables or modifying existing ones (like e.g. adding new places to the %PATH% environment variable). To verify, I use the command line. Except that existing processes don’t automatically get the new variables. I was recently tricked by this, since I pressed #5 to get a new console and test the changes. Nothing there. Hmmm. When I start %COMSPEC% they are there. So why not with #5?

Well, the process which starts #5 was started when the changes to the environment variables where not yet made. That’s why also the processes started from this process don’t have the changes. So all I’ve got to do in this case is to kill itype.exe and then start it right away again. Now it works!

The post You’ve got to restart that process … appeared first on Tales of a Code Monkey.

Windows Authentication through Forms

roger — Sun, 11 Oct 2009 11:36:00 +0000

Let’s assume you have a web site which is exposed to the internet (i.e. to the public) but the site itself works with windows accounts internally. If you have problems visualizing this scenario, think about a web mail interface for a mail server system which offers mail services for user accounts in active directory. Users inside your corporate network can use integrated Windows authentication to access this site, so they don’t really have a problem. But when they want to access the interface from outside the corporate network, or from a machine/device which doesn’t understand integrated Windows authentication, you’ll realize that it doesn’t just work like that.

So you’ll want to do authentication based on the credentials the user enters in a simple form on a web page. The challenge now is that the rest of the site (i.e. apart from the one form which does the initial authentication) will likely still work with the HttpContext.Current.User property to determine which user is currently authenticated and provide actions and data based on that identity, because you don’t want to re-implement the entire logic. The good news is, you can do that! The bad news is, you don’t just get it for free. But that’s why you came here, and that’s why I’ll try to help you with this.

Let’s first arrange a few things on the web site facing the public — you don’t necessarily need to change this on your internal site. First, the login page will need to be accessible for anyone, i.e. anonymous access must be turned on on the site. Second, from ASP.net’s point of view, all the pages which require the user to be authenticated should be in/under the same directory. If you don’t want to do this, you’ll simply have to add a element for all pages requiring authentication in your web.config or do the opposite and add all pages which don’t require authentication in the same way. With the directories however, the basic structure can be as simple as

WebSiteRoot/
+--Default.aspx        Could redirect to /ActualSite/Default.aspx
+--ActualSite/
   +--Default.aspx     The actual home page with all its logic
+--Auth/
   +--Login.aspx       The login page

So all the logic you had in the site’s root directly would be under ActualSite or whatever name you chose.

Updating Web.config

Now for the above scenario with different directories, the web.config in the site’s root could look as follows. Please note that for the pages under ActualSite we are simply disabling anonymous access, while for everything else, we allow it. Also, in the authentication element we’re setting the mode to None because we’re not going to use any predefined authentication mechanism as is.

But of course that’s not all yet. You’ll see that if now you wanted to look at the actual site, you’ll get a 401 because you’re not authenticated. So let’s take a look at that.

Validating credentials

Next, let’s create the login page which the user will use to enter his credentials. Fortunately, ASP.net offers the Login control which does almost everything we need. So add that one to your login page plus add an event handler for the OnAuthenticate event of that control. Alternatively, you can also derive your own control from the Login control and override the OnAuthenticate method. This event handler is where we’ll do our custom logic to check that the credentials really map to an existing Windows user. Below’s the code which does that.

protected override void OnAuthenticate(AuthenticateEventArgs args)
{
    string[] parts = UserName.Split('\\');
    string password = Password;
    string username;
    string domain = null;

    args.Authenticated = false;

    if (parts.Length == 1)
    {
        username = parts[0];
    }
    else if (parts.Length == 2)
    {
        domain = parts[0];
        username = parts[1];
    }
    else
    {
        return;
    }

    if (WebAuthenticationModule.AuthenticateUser(username, domain, password))
    {
        string userData = String.Format("{0}\n{1}\n{2}",
            username, domain, password);

        FormsAuthenticationTicket ticket = new FormsAuthenticationTicket(
            2                           /* version */,
            username,
            DateTime.Now                /* issueDate */,
            DateTime.Now.AddMinutes(30) /* expiration */,
            true                        /* isPersistent */,
            userData,
            FormsAuthentication.FormsCookiePath);

        HttpCookie ticketCookie = new HttpCookie(
            FormsAuthentication.FormsCookieName,
            FormsAuthentication.Encrypt(ticket));

        Context.Response.Cookies.Add(ticketCookie);

        Context.Response.Redirect(
            FormsAuthentication.GetRedirectUrl(username, false), false);
    }
}

The AuthenticateUser method from WebAuthenticationModule is a wrapper on the LogonUser function from the Win32 API which will return true if the user could be logged on. So if the credentials are valid, we’re going to pass them in to the FormsAuthenticationTicket in the UserData property so later on, we’ll be able to use them again. At least, we don’t want the consumers of the site have to enter credentials for every request their making, right? Also, we’re encrypting the entire ticket because we’re going to send it over the wire. The Encrypt method from the FormsAuthentication class does this. However, you’ll have to make sure that the protection attribute of the forms element in the web.config is set to All which actually is the default (but it can be inherited, so watch out!).

What you see is that we’re heavily using the functionality offered by the FormsAuthentication class and related classes to handle the tickets, encryption, settings, etc. This is not mandatory but it helps a lot. Plus it’s better anyway than coming with your own ticketing and encryption mechanisms; unless you have a degree in maths and/or cryptography, chances are that that’s not so secure as you think it is.

Authenticating users

Then, we need to authenticate the user for all the requests he makes after providing the credentials. Thus, we need to add some custom logic to the AuthenticateRequest event of the HttpApplication. There’s multiple ways to do that:

Add a file called global.asax to your site’s root
Create and register a new HttpModule by implementing the System.Web.IHttpModule interface and adding an entry in the httpModules section in your root’s web.config

Personally, I like the approach with the custom module more, but adding this stuff to the global.asax can be done a little bit faster. In either case, make sure you can handle the AuthenticateRequest event. My code proposal for the handler is given below. I’m intentionally omitting most error handling code here.

private static void OnAuthenticateRequest(object sender, EventArgs args)
{
    HttpApplication application = sender as HttpApplication;

    HttpContext context = application.Context;
    HttpRequest request = context.Request;
    HttpResponse response = context.Response;

    if (!request.IsAuthenticated &&
        !context.SkipAuthorization)
    {
        if (request.CurrentExecutionFilePath.Equals(FormsAuthentication.LoginUrl,
                                                    StringComparison.OrdinalIgnoreCase) ||
            request.CurrentExecutionFilePath.EndsWith(".axd"))
        {
            context.SkipAuthorization = true;
        }
        else
        {
            HttpCookie cookie = request.Cookies.Get(FormsAuthentication.FormsCookieName);
            FormsAuthenticationTicket ticket = FormsAuthentication.Decrypt(cookie.Value);

            if (!ticket.Expired)
            {
                IntPtr hToken = LogonUserFromTicket(ticket);

                WindowsIdentity identity = new WindowsIdentity(
                    hToken, "Win/Forms", WindowsAccountType.Normal, true);

                context.User = new WindowsPrincipal(identity);

                if (FormsAuthentication.SlidingExpiration)
                {
                    ticket = FormsAuthentication.RenewTicketIfOld(ticket);
                    cookie.Value = FormsAuthentication.Encrypt(ticket);
                    response.Cookies.Set(cookie);
                }

                return;
            }

            FormsAuthentication.RedirectToLoginPage();
        }
    }
}

So let’s go through the important things here. First we check (lines 9/10) if the request is already authenticated or authorization is to be skipped completely. If either of those is true, we’re out of the picture already. Else it’s our job to do the authentication. Lines 12-17 are used to prevent authentication on the login page as well as the *.axd handlers, which are typically used to return resources like scripts for ASP.net components. Lines 20-42 do the actual authentication: we first retrieve the cookie from the previous credential validation and decrypt it to get the ticket. If the ticket has not expired, we log the user on (the call to LogonUserFromTicket is again only a wrapper for the LogonUser function from the Win32 API; it uses the data from the UserData property of the ticket) to get the logon token which we’ll pass in to the WindowsIdentity constructor to get the WindowsIdentity object all the code in ActualSite will use to determine which user is making the request. Then of course we need to update the context with the new identity. If sliding expiration is turned on in the web.config, we renew the ticket. And if the ticket has expired, of course we don’t authenticate the user but instead we redirect him to the login page.

Cleanup

Finally, we still have a tiny problem here. We have used the LogonUser function, but according to its documentation, we should call the CloseHandle function from the Win32 API. For the method AuthenticateUser I have already done that; there we don’t need the token/handle anymore when we have verified the credentials. But what about the authentication we’re doing in AuthenticateRequest? We’re setting the newly created WindowsIdentity (which is here used to represent the user token) to the current HttpContext because we’ll need it there to ultimately handle the request. But once the request handler is done, we don’t need it anymore. Luckily, there’s also an event for this purpose. It’s the last event that there is and it’s called EndRequest. So let’s add the following code in the handler for that event.

private static void OnEndRequest(object sender, EventArgs args)
{
    HttpApplication application = sender as HttpApplication;

    HttpContext context = application.Context;

    if (null != context.User)
    {
        WindowsIdentity identity = context.User.Identity as WindowsIdentity;

        if (null != identity)
        {
            NativeAuth.CloseHandle(identity.Token);
        }
    }
}

Basically all it does, if the request was given a WindowsIdentity, it’ll call the CloseHandle function from the Win32 API on that identity’s token handle. That should do the trick and we shouldn’t leak handles anymore.

Summary

I have shown here how you can make use of the forms authentication mechanisms which come with ASP.net to do Windows authentication behind the scenes. This can be very useful in cases when not all users have access to systems which know how to do Windows authentication.

Please note that I do not claim that this solution is going to work for every challenge you may be facing. The solution shown here is neither claimed to be complete nor suitable for every web application.

The post Windows Authentication through Forms appeared first on Tales of a Code Monkey.

Windows 7 to RTM

roger — Thu, 23 Jul 2009 07:24:00 +0000

Microsoft has announced yesterday the release of Windows 7 (and Windows Server 2008 R2) to manufacturing, Windows 7 will be available on October 22^nd. And just to show you how much fun we have when we RTM a product, check out the Windows 7 RTM Sign-off video on YouTube …

The post Windows 7 to RTM appeared first on Tales of a Code Monkey.

Debugging Service Startup

roger — Thu, 09 Jul 2009 07:06:00 +0000

If you are like me working on server development you may run into the situation that a service fails early during startup, i.e. within the first couple of seconds. You’ll soon realize that manually attaching a debugger doesn’t work well, if at all. Even if you’ll be running

windbg -pn MyService.exe

you may not actually be fast enough. Or there are multiple instances of that image running (e.g. SvcHost.exe) and the above command becomes kind of useless. Now if you have been reading the ‘Debugging Tools for Windows’ documentation (or have debugged services before) you’ll already know what I am about to tell you here.

As I mentioned above, the first problem you’re fighting with is that the failure happens very early. Another problem you may have is that your service is running e.g. as ‘NT AUTHORITY\NETWORK SERVICE’ and thus may not be able to interact with your desktop. But Windows’ right here to help you out with ‘Image File Execution Options’. This basically allows you to execute a command when a particular image is being executed.

You start by creating a key for the image file in the registry:

HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Image File Execution Options\MyService.exe

There you create a new string value of the name ‘Debugger’ and set its value to the path of the debugger you want to invoke, e.g. something like ‘C:\Debuggers\cdb.exe’. To mitigate the problem of the non-interactive service, you’ll probably want to use the debugger remotely, so you’ll create a debugging server: ‘C:\Debuggers\cdb.exe -server npipe:pipe=MyDebuggingSession’. What you have now is a debugger which is attached as soon as the service starts and which is accessible remotely. Thus, you can either on the server itself or another machine which has access run for instance

windbg -remote npipe:server=my-machine,pipe=MyDebuggingSession

and there you go: you are now debugging the service. Use the various command line options for the debuggers to

ignore the initial break point (-g)
run commands right after the debugger is attached (-c)
create a script which sets some useful breakpoints and run it when the debugger is attached (e.g. -c “$

You may actually have to adjust the service control timeout. To do so, add a DWORD called ‘ServicesPipeTimeout’ to the registry key

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control

and set its value to the number of milliseconds you want the service to wait before timing out.

You’ll find pretty much all of this information also in the help file for ‘Debugging Tools for Windows’ under ‘Debugging a Service Application’. Enjoy!

The post Debugging Service Startup appeared first on Tales of a Code Monkey.