Starting application from service running as SYSTE

I currently have a single application that needs to be started from a windows service that i am coding in .net 3.5. This application is currently running as the user who ran the service, in my case the SYSTEM user. If running as the SYSTEM user it does not show the application to the users desktop. Thoughts? advice?

//constructor
    private Process ETCHNotify = new Process();

//StartService()     
    ETCHNotify.StartInfo.FileName = baseDir + "\\EtchNotify.exe";
    ETCHNotify.StartInfo.UseShellExecute = false;

//BackgroundWorkerThread_DoWork()
    if (!systemData.GetUserName().Equals(""))
    {
        // start ETCHNotify
        try {
            ETCHNotify.Start();
        }
            catch (Exception ex)
        {
            systemData.Run("ERR: Notify can't start: " + ex.Message);
        }
    }

I only execute the try/catch if the function i have written GetUserName() (which determines the username of the user running explorer.exe) is not null

again to reiterate: desired functionality is that this starts ETCHNotify in a state that allows it to interact with the currently logged in user as determined by GetUserName()

回答1:

Collage of some post found around (this and this)

Note that as of Windows Vista, services are strictly forbidden from interacting directly with a user:

Important: Services cannot directly interact with a user as of Windows Vista. Therefore, the techniques mentioned in the section titled Using an Interactive Service should not be used in new code.

This "feature" is broken, and conventional wisdom dictates that you shouldn't have been relying on it anyway. Services are not meant to provide a UI or allow any type of direct user interaction. Microsoft has been cautioning that this feature be avoided since the early days of Windows NT because of the possible security risks.

There are some possible workarounds, however, if you absolutely must have this functionality. But I strongly urge you to consider its necessity carefully and explore alternative designs for your service.

Use WTSEnumerateSessions to find the right desktop, then CreateProcessAsUser to start the application on that desktop (you pass it the handle of the desktop as part of the STARTUPINFO structure) is correct.

However, I would strongly recommend against doing this. In some environments, such as Terminal Server hosts with many active users, determining which desktop is the 'active' one isn't easy, and may not even be possible.

A more conventional approach would be to put a shortcut to a small client app for your service in the global startup group. This app will then launch along with every user session, and can be used start other apps (if so desired) without any juggling of user credentials, sessions and/or desktops.

回答2:

I wasn't going to answer this since you already answered it, (and it's oh, what? going on 2.5 years OLD now!?) But there are ALWAYS those people who are searching for this same topic, and reading the answers...

In order to get my service to Interact with the Desktop, no matter WHAT desktop, nor, how MANY desktops, nor if the service was even running on the SAME COMPUTER as the desktop app!! None of that matters with what I got here... I won't bore you with the details, I'll just give you the meat and potatoes, and you and let me know if you want to see more...

Ok. First thing I did was create an Advertisement Service. This is a thread that the service runs, opens up a UDP socket to listen for broadcasts on the network. Then, using the same piece of code, I shared it with the client app, but it calls up Advertise.CLIENT, rather than Advertise.SERVER... The CLIENT opens the port I expect the service to be on, and broadcasts a message, "Hello... Is there anybody out there??", asking if they're there ANY servers listening, and if so, reply back to THIS IP address with your computer name, IP Address and port # where I can find the .NET remoting Services..." Then it waits a small amount of time-out time, gathers up the responses it gets, and if it's more than one, it presents the user with a dialog box and a list of services that responded... The Client then selects one, or, if only ONE responded, it will call Connect((TServerResponse) res); on that, to get connected up. At this point, the server is using Remoting Services with the WellKnownClientType, and WellKnownServerType to put itself out there...

I don't think you are too interested in my "Auto-Service locater", because a lot of people frown on UDP, even more so when your app start broadcasting on large networks. So, I'm assuming you'd be more interested in my RemotingHelper, that gets the client connected up to the server. It looks like this:

    public static Object GetObject(Type type) 
    {
        try {
            if(_wellKnownTypes == null) {
                InitTypeCache();
            }
            WellKnownClientTypeEntry entr = (WellKnownClientTypeEntry)_wellKnownTypes[type];

            if(entr == null) {
                throw new RemotingException("Type not found!");
            }
            return System.Activator.GetObject(entr.ObjectType, entr.ObjectUrl);
        } catch(System.Net.Sockets.SocketException sex) {
            DebugHelper.Debug.OutputDebugString("SocketException occured in RemotingHelper::GetObject().  Error: {0}.", sex.Message);
            Disconnect();
            if(Connect()) {
                return GetObject(type);
            }
        }
        return null;
    }

    private static void InitTypeCache() 
    {
        if(m_AdvertiseServer == null) {
            throw new RemotingException("AdvertisementServer cannot be null when connecting to a server.");
        }

        _wellKnownTypes = new Dictionary<Type, WellKnownClientTypeEntry>();

        Dictionary<string, object> channelProperties = new Dictionary<string, object>();
        channelProperties["port"] = 0;
        channelProperties["name"] = m_AdvertiseServer.ChannelName;

        Dictionary<string, object> binFormatterProperties = new Dictionary<string, object>();
        binFormatterProperties["typeFilterLevel"] = "Full";

        if(Environment.UserInteractive) {
            BinaryServerFormatterSinkProvider binFormatterProvider = new BinaryServerFormatterSinkProvider(binFormatterProperties, null); 
            _serverChannel = new TcpServerChannel(channelProperties, binFormatterProvider);
            // LEF: Only if we are coming form OUTSIDE the SERVICE do we want to register the channel, since the SERVICE already has this
            //      channel registered in this AppDomain.
            ChannelServices.RegisterChannel(_serverChannel, false);
        }

        System.Diagnostics.Debug.Write(string.Format("Registering: {0}...\n", typeof(IPawnStatServiceStatus)));
        RegisterType(typeof(IPawnStatServiceStatus),m_AdvertiseServer.RunningStatusURL.ToString());
        System.Diagnostics.Debug.Write(string.Format("Registering: {0}...\n", typeof(IPawnStatService)));
        RegisterType(typeof(IPawnStatService),      m_AdvertiseServer.RunningServerURL.ToString());
        System.Diagnostics.Debug.Write(string.Format("Registering: {0}...\n", typeof(IServiceConfiguration)));
        RegisterType(typeof(IServiceConfiguration), m_AdvertiseServer.RunningConfigURL.ToString());
    }

    [SecurityPermission(SecurityAction.Demand, Flags=SecurityPermissionFlag.RemotingConfiguration, RemotingConfiguration=true)]
    public static void RegisterType(Type type, string serviceUrl)
    {
        WellKnownClientTypeEntry clientType = new WellKnownClientTypeEntry(type, serviceUrl);
        if(clientType != RemotingConfiguration.IsWellKnownClientType(type)) {
            RemotingConfiguration.RegisterWellKnownClientType(clientType);
        }
        _wellKnownTypes[type] = clientType;
    }

    public static bool Connect()
    {
        // Init the Advertisement Service, and Locate any listening services out there...
        m_AdvertiseServer.InitClient();
        if(m_AdvertiseServer.LocateServices(iTimeout)) {
            if(!Connected) {
                bConnected = true;
            }
        } else {
            bConnected = false;
        }
        return Connected;
    }

    public static void Disconnect()
    {
        if(_wellKnownTypes != null) {
            _wellKnownTypes.Clear();
        }
        _wellKnownTypes = null;
        if(_serverChannel != null) {
            if(Environment.UserInteractive) {
                // LEF: Don't unregister the channel, because we are running from the service, and we don't want to unregister the channel...
                ChannelServices.UnregisterChannel(_serverChannel);
                // LEF: If we are coming from the SERVICE, we do *NOT* want to unregister the channel, since it is already registered!
                _serverChannel = null;
            }
        }
        bConnected = false;
    }
}

So, THAT is meat of my remoting code, and allowed me to write a client that didn't have to be aware of where the services was installed, or how many services were running on the network. This allowed me to communicate with it over the network, or on the local machine. And it wasn't a problem to have two or more people running the app, however, yours might. Now, I have some complicated callback code in mine, where I register events to go across the remoting channel, so I have to have code that checks to see if the client is even still connected before I send the notification to the client that something happened. Plus, if you are running for more than one user, you might not want to use Singleton objects. It was fine for me, because the server OWNS the objects, and they are whatever the server SAYS they are. So, my STATS object, for example, is a Singleton. No reason to create an instance of it for EVERY connection, when everyone is going to see the same data, right?

I can provide more chunks of code if necessary. This is, of course, one TINY bit of the overall picture of what makes this work... Not to mention the subscription providers, and all that.

For the sake of completeness, I'm including the code chunk to keep your service connected for the life of the process.

public override object InitializeLifetimeService() 
{
    ILease lease = (ILease)base.InitializeLifetimeService();

    if(lease.CurrentState == LeaseState.Initial) {
        lease.InitialLeaseTime = TimeSpan.FromHours(24);
        lease.SponsorshipTimeout = TimeSpan.FromSeconds(30);
        lease.RenewOnCallTime = TimeSpan.FromHours(1);
    }
    return lease;
}

#region ISponsor Members

[SecurityPermissionAttribute(SecurityAction.LinkDemand, Flags=SecurityPermissionFlag.Infrastructure)]
public TimeSpan Renewal(ILease lease) 
{
    return TimeSpan.FromHours(12);
}
#endregion

If you include the ISponsor interface as part of your server object, you can implement the above code.

Hope SOME of this is useful.

回答3:

When you register your service, you can tell it to allow interactions with the desktop. You can read this oldie link http://www.codeproject.com/KB/install/cswindowsservicedesktop.aspx

Also, don't forget that you can have multiple users logged in at the same time.

Apparently on Windows Vista and newer interacting with the desktop has been made more difficult. Read this for a potential solution: http://www.codeproject.com/KB/cs/ServiceDesktopInteraction.aspx

回答4:

Ultimately in order to solve this i took the advice of @marco and the posts he mentioned. I have created the service to be entirely independent of the tray application that interacts with the user. I did however install the Tray application via registry 'start up' methods with the service. The Service installer will now install the application which interacts with the user as well... This was the safest and most complete method.

thanks for your help everyone.