AccessViolationException thrown for MSI

[broken]

I'm getting the error:

Exception thrown: 'System.AccessViolationException' in RGB.NET.Devices.Msi.dll
An unhandled exception of type 'System.AccessViolationException' occurred in RGB.NET.Devices.Msi.dll
Attempted to read or write protected memory. This is often an indication that other memory is corrupt.

This occurs when trying to load the device. My code:

static void Main(string[] argv)
        {
            RGBSurface surface = RGBSurface.Instance;
            surface.Exception += args => Debug.Print(args.Exception.Message);
            surface.LoadDevices(MsiDeviceProvider.Instance, throwExceptions: true);
...
}

The exception is thrown in the MsiDeviceProvider at line 98:

if ((errorCode = _MsiSDK.GetDeviceInfo(out string[] deviceTypes, out int[] ledCounts)) != 0)

What's odd is that it should be successfully initialized by this point, and GetDevicesInfo throws an exception rather than returning an error code.

I've tried x86 & x64 builds, and running VS as administrator and not as admin.

I installed the RGB.NET.Resources.* using the "NuGet Package Manager" using RBG.NET as the package source. While I'm new to Visual Studio, I believe this was done properly as I see the dlls in the correct file locations.

My MSI hardware:

• nvidia geforce gtx 1060
• x470 gaming pro carbon

From other threads I realize you have no MSI hardware to test with, but still wanted to start an issue in hopes others have found a solution, or maybe a place for future me to write how I fixed it. :)

As a side note, the Corsair dll appears to load correctly, but it cannot find my RGB Vengeance RAM. I'm simply not having any luck..

[akmadian]

I've had similar issues with the ASUS device provider that I haven't been able to fix (even try catch doesn't work), my guess is that the manufacturer's SDK is broken somewhere. Not much we can do unless MSI/ ASUS fix their SDKs...

[DarthAffe]

Hey, that's somehow interesting. the fact that you're getting an exception is at least different to what we got until now. Just to verify that it's really something in the SDK (which is not too unlikely) could you run this test-application? http://uploads.wyrez.org/MsiSDKTest.zip While running it's expected to run, return two error-codes (please report back which one) and a led-count if it's able to get a device (but we never got that to work until now).

[broken]

My apologies, but I'm just now seeing the end of a house move and surgery recovery, so while I haven't gotten my pc back up and running yet, I should be able to run your test after a couple weeks.

[deckertron9000]

@DarthAffe I just started playing around with this project today. I've noticed that I'm only able to get a non-zero pointer for the sdk dll if I run in x86. I downloaded your sdk test and ran that. Here's the output:

Starting ... Initialized! (ErrorCode: 0)

Unhandled Exception: System.AccessViolationException: Attempted to read or write protected memory. This is often an indication that other memory is corrupt. at System.StubHelpers.MngdNativeArrayMarshaler.ConvertContentsToManaged(IntPtr pMarshalState, Object& pManagedHome, IntPtr pNativeHome) at MsiSDKTest.Program.MLAPI_GetDeviceInfo(String[]& pDevType, Int32[]& pLedCount) at MsiSDKTest.Program.Main(String[] args)

[deckertron9000]

Okay, more info from playing around with the marshaling.

[UnmanagedFunctionPointer(CallingConvention.Cdecl, CharSet = CharSet.Ansi)]
private delegate int GetDeviceInfoPointer([Out, MarshalAs(UnmanagedType.SafeArray, SafeArraySubType = VarEnum.VT_BSTR)] out string[] pDevType, 
[Out, MarshalAs(UnmanagedType.SafeArray, SafeArraySubType = VarEnum.VT_BSTR)] out int[] pLedCount);

Marshaling as safe arrays is required to get this to work as you would expect.

[DarthAffe]

Yeah, i just changed that too and uploaded a new test-version^^ does it work like this? I thought about that

[DllImport("MysticLight_SDK.dll", CallingConvention = CallingConvention.Cdecl)]
public static extern int MLAPI_GetDeviceInfo([MarshalAs(UnmanagedType.SafeArray, SafeArraySubType = VarEnum.VT_BSTR)] out string[] pDevType,
                                             [MarshalAs(UnmanagedType.SafeArray, SafeArraySubType = VarEnum.VT_INT)] out int[] pLedCount);

[deckertron9000]

I can't get it to work with the SafeArraySubType = VarEnum.VT_INT on the pLedCount. I've only been able to get that working with VT_BSTR. But then you have a mismatched array (pLedCount thinks it's an array of int but it's actually an array of string). To work around that I modified GetDeviceInfo to parse the int:

internal static int GetDeviceInfo(out string[] pDevType, out int[] pLedCount)
{    
    int result = _getDeviceInfoPointer(out pDevType, out string[] pLedCount1);
    pLedCount = new int[pLedCount1.Length];

    for (int i = 0; i < pLedCount1.Length; i++)
    pLedCount[i] = int.Parse(pLedCount1[i]);

    return result;
}

[DarthAffe]

Hmm ok that's weird - but as long as it works ... Are you getting a reasonable result if you call it?

[deckertron9000]

Yeah, the only compatible MSI device in my current system is the MPG Z390 Gaming Pro Carbon. The pDevType on that is "MSI_MB" and the ledCount is 9. Calling GetLedInfo with each of the various indexes 0-8 gives the various RGB locations on the motherboard currently as well as motherboard headers that can be connected to various RGB products.

[DarthAffe]

Ok, sounds good. With that knowledge it should be possible to get this further. Are you working on it? If not I'll continue implementing it based on the documentation.

[deckertron9000]

I can submit a PR with my updates to _MsiSDK class if that would help.

[DarthAffe]

Hey, finally got some time to continue on what you've done. Since the documentation doesn't describe the parameters at all I'd like to make sure I got it correct before integrating it in RGB.NET. Could you compile/run http://uploads.wyrez.org/MsiSDKTest.zip and see if it correctly sets LED-colors and copy me the output? That should be all of the SDK-methods we need (at least I hope so^^)

[deckertron9000]

I'll give it a shot hopefully today sometime and report back!

[deckertron9000]

Starting ... Initialized! (ErrorCode: 0) Got devices: (ErrorCode: 0) MSI_MB - LedCount: 9 Set ColorsDevice: MSI_MBLED 0: IO Cover - Off, Rainbow, Meteor, Stack, Breathing, Flashing, Double Flashing, Steady, Lightning, CPU Temperature, Random, POP, RAP, JAZZ, Play, Movie, Planetary, Double Meteor, Energy, Blink, Clock, Color Pulse, Color Shift, Color Wave, Marquee, Rainbow Wave, Visor, Rainbow Flashing, Rainbow Double Flashing, Default Get ColorsDevice: MSI_MBLED 0: [255, 0, 0]

Note that I had to build the application and then run as administrator from command line. Debugging straight from VS 2019 seems to not run the program as admin. Also, pretty sure x64 is not supported based on the readme here.

[deckertron9000]

Another thing worth noting is that you'll need to set the led style to one that supports direct color setting otherwise you'll get an "unsupported" error. "Steady" is the only one I've tried so far that works but there may be others.

Also in the inner loop you need to check j < ledCounts[i] to correctly loop through all addressable leds.