ImDisk Virtual Disk Driver

I have decided to create RAM disk to relieve some pressure from SDD drive when compiling software to prevent wear out, but still something is writing a lot to the SSD during the compilation process with MSVC, despite the whole project being put in a RAM drive and compiled from there. The fact that it is done inside virtual machine is irrelevant, right?

I would like to make sure that I have correctly set up imDisk for this task. Could you take a look at the following screenshot and commands and tell that it indeed creates a partition in RAM and not anywhere else (like in the virtual memory of pagefile)?

Manually starting these drivers due to #8:

sc create imdisk type= kernel binpath= "C:\imdisk_test\imdisk.sys"
sc start imdisk
sc create awealloc type= kernel binpath= "C:\imdisk_test\awealloc.sys"
sc start awealloc

Then executing the cpl file:

• Windows 7 Pro x64

I wrote 2 scripts to mount/unmount:

mount.bat:

imdisk -a -f "d:\a\shared-store.vdi" -m "d:\a\shared-store"

unmount.bat:

imdisk -d -m "d:\a\shared-store"

The problem is that it can not mount nor unmount.

Some time ago it has mount/dismount correctly. But after that is something did happen and mounting now is broken.

The ProcessHacker shows that the System holds the vdi file: shared-store.vdi

The mount.bat reports access denied on mount. The unmount.bat reports that the file is not found. All scripts does execute under Administrator privileges.

If try to change the parameter from path d:\a\shared-store to drive e:, then the scripts reports almost the same.

If try to run MountImg.exe it shows nothing mounted.

The same vdi file mounts correctly by the VirtualBox Guest instance. If try to stop the VM instance with the vdi file attached and run scripts again, then the issue happens again.

I read this issue but did't found a solution: https://github.com/LTRData/ImDisk/issues/10

There is questions:

• How to correctly dismount buggy mounted vdi file?
• How mount vdi drive to a directory?

This is odd, but I triple-checked what I experienced, and here is the STR.

1. Install Firefox v95.0.2 on a Win7 box.
2. Install ImDisk v2.1.1.
3. Create an ImDisk volume using the command: imdisk.exe -a -m G: -s 1g -p "/fs:exFAT /v:ImDisk /q /y"
4. Create a Firefox profile on the ImDisk volume.
5. Install Bonjourr in the Firefox profile and configure it as the new tab page.
6. Open a new tab & notice that the Bonjourr page will never appear due to a storage error. (It works correctly on a non-ImDisk volume.)
7. Install Tabliss in the Firefox profile and configure it as the new tab page.
8. Open a new tab.
9. Notice that the Tabliss page does appear.
10. Now change settings in Tabliss.
11. Notice that none of the Tabliss settings will be saved due to a storage I/O error (IO error: Incorrect function. (os error 1) ExtensionStorageSync.jsm:66). This only happens on an ImDisk volume.

I also ran the same tests with the ImDisk volume formatted as NTFS. The same errors occurred.

• Windows x64 Pro SP1

Got the stack trace from the Process Hacker one of threads has having the cycles. It does change over the time, but the WriteFile been blocked on something. The whole system suffers from WriteFile blocking and none of application can be closed or terminated which had call to WriteFile recently. Even can not be terminated through the Process Hacker itself. The Process Hacker somehow survived that and I could take a single stack dump. Only the hard reset can reset the system, which means it is a critical bug.

Can't say how the notepad could save the stack trace into a file, may be the blocking is not stable.

0, ntoskrnl.exe!KeInsertQueueApc+0x5cb
1, ntoskrnl.exe!ObfReferenceObject+0x99d
2, ntoskrnl.exe!KeWaitForSingleObject+0x1a3
3, ntoskrnl.exe!longjmp+0x1c470
4, Ntfs.sys+0x9bd48
5, fltmgr.sys+0x1102
6, fltmgr.sys!FltIsCallbackDataDirty+0x23ba
7, fltmgr.sys!FltDeletePushLock+0x3ee
8, ntoskrnl.exe!NtWriteFile+0x45f
9, ntoskrnl.exe!longjmp+0x5bf3
10, ntdll.dll!ZwWriteFile+0xa
11, KernelBase.dll!WriteFile+0x7b
12, kernel32.dll!WriteFile+0x36
13, DomainNeutralILStubClass.IL_STUB_PInvoke(Microsoft.Win32.SafeHandles.SafeFileHandle, Byte*, Int32, Int32 ByRef, IntPtr) + 0xc8 <-- mscorlib.ni.dll+0x63c9e8
14, System.IO.FileStream.WriteFileNative(Microsoft.Win32.SafeHandles.SafeFileHandle, Byte[], Int32, Int32, System.Threading.NativeOverlapped*, Int32 ByRef) + 0x83 <-- mscorlib.ni.dll+0x5ad683
15, System.IO.FileStream.WriteCore(Byte[], Int32, Int32) + 0x5d <-- mscorlib.ni.dll+0x5ad5dd
16, DiscUtils.Vmdk.HostedSparseExtentStream.Write(Byte[], Int32, Int32) + 0x12c <-- 0x7fe852b360c
17, LTR.IO.ImDisk.Devio.Server.Providers.DevioProviderFromStream.Write(Byte[], Int32, Int32, Int64) + 0x3e <-- 0x7fe852b34ae
18, LTR.IO.ImDisk.Devio.Server.Providers.DevioProviderManagedBase.Write(IntPtr, Int32, Int32, Int64) + 0x6b <-- 0x7fe852b343b
19, LTR.IO.ImDisk.Devio.Server.Services.DevioShmService.WriteData(System.Runtime.InteropServices.SafeBuffer) + 0x175 <-- 0x7fe852b2fa5
20, LTR.IO.ImDisk.Devio.Server.Services.DevioShmService.RunService() + 0x509 <-- 0x7fe852b17a9
21, LTR.IO.ImDisk.Devio.Server.Services.DevioServiceBase.ServiceThreadProcedure() + 0x24 <-- 0x7fe852b0e24
22, System.Threading.ExecutionContext.RunInternal(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object, Boolean) + 0x172 <-- mscorlib.ni.dll+0x58df12
23, System.Threading.ExecutionContext.Run(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object, Boolean) + 0x15 <-- mscorlib.ni.dll+0x58dd95
24, System.Threading.ExecutionContext.Run(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object) + 0x55 <-- mscorlib.ni.dll+0x58dd65
25, System.Threading.ThreadHelper.ThreadStart() + 0x55 <-- mscorlib.ni.dll+0x633ea5
26, clr.dll!LogHelp_LogAssert+0x1f13
27, clr.dll!LogHelp_LogAssert+0x1e28
28, clr.dll!LogHelp_LogAssert+0x2690
29, clr.dll!ClrCreateManagedInstance+0x20bf
30, clr.dll!LogHelp_LogAssert+0x3540
31, clr.dll!LogHelp_LogAssert+0x34b3
32, clr.dll!LogHelp_LogAssert+0x33f2
33, clr.dll!LogHelp_LogAssert+0x35cf
34, clr.dll!ClrCreateManagedInstance+0x1fa9
35, clr.dll!LogHelp_LogAssert+0x68c5
36, kernel32.dll!BaseThreadInitThunk+0xd
37, ntdll.dll!RtlUserThreadStart+0x1d

How to reproduce:

1. Create dynamic .vdi or .vmdk file of 1TB size.
2. Mount to drive B: and format for NTFS 4KB fast.
3. Start copying 300GB of files from an SSD into it.
4. After ~50GB it does hang and system from now can not be rebooted nor a process terminated.

I have reproduced it on both .vdi (fastes coping) and .vmdk (slower, the stack trace is taken from). The .vhd is too slow to reach the 50GB point (< 10MB/s on any files) and might be not affected because of slowness (I feel there is might be some kind of a race condition with high speed writes).

With the latest release version (v2.1.1), I can't run the installer on Windows 10 or 2019. I get the following error:

But if I unzip the sfx and run install.cmd, it works all fine. runwaitw.exe is where it is expected.

Any idea?

Hi, I'm sorry about the undescriptive bug report title, but I couldn't think of anything better.

A while ago I filed a report at ImDisk Toolkit's tracker, through which I'm using the ImDisk driver.

Basically, I have setup a RAM disk that's always recreated at startup with user temporary directories pointing to it (i.e., R:\Temp), 8 GB, using AWE and being dynamic in terms of memory. It works great most of the times, but some applications seem to have a hard time working with it.

These are some of the error messages that are shown:

The first one belongs to the current graphic driver installer from Intel, it states that the application is in an unauthorized location, whatever that means. Also from Intel, their oneAPI compiler package installer complains X MB of space are needed, even when it was virtually empty. In both cases I had to move things out of the RAM disk for them to work.

The behavior is similar in a way to what I encountered years ago when Process Hacker was updating and %TEMP% pointed to inside the RAM disk, I reported it over at their forums, and their developer kind of explained what the problem was, so I could report it properly, but I completely forgot about it. The reply is here, in case it's still relevant (it seems to have to do with IOCTLs): https://wj32.org/processhacker/forums/viewtopic.php?t=3419#p10644

I don't know if you guys are aware of the issue(s), but I just stumbled upon this project here and seeing there seems to be active development I thought of referencing it here too. The third picture belongs to another case that exhibits the same behavior, I had to temporarily switch the TEMP environmental variables to a real disk to be able to continue with that installation.

Could you please look into it? Don't hesitate to let me know if you need more data samples, or you want logs of some sort, I would gladly do trial and error for you with debug builds of the driver. It needn't be signed, I could self sign them and make sure Windows trusts that signature to do the testing, or enable installation of unsigned drivers for the tests.

Thank you for your time by the way.

When I access the mounted volume and then try to unmount it, sometimes it says access denied and I'm unable to do so, even though I'm not using the volume anymore. The only option in such a case is to restart my computer. What could be the problem?

Hello!

When installation comes to the point of installing the inf file a message box with text "Installation failed." pops up. The installer, the script in zip file, and choosing the "Install" option when clicking on the inf file all fail with this error. All attempts were performed with administrator privileges. Files were downloaded from https://www.ltr-data.se/

Manually loading the drivers via the sc create and sc start and executing the corresponding cpl file works though, I have managed to get them all to load that way and create partition in RAM.

Not sure if that matters, but the system is a guest OS in a virtual machine which is disconnected from the Internet.

Hello,

I am exploring ImDisk for use in a potential project and think that it is really an amazing piece of work from the testing that I have done so far.

In addition to using it as a normal RAMDISK, I also have some ideas on how it might be used as part of a specialized IPC communication channel across independent applications written in different languages as well as for some other exciting work that I am trying to do as well.

On thing that I was wondering is that I know ImDisk can save images since I was playing with the ImDisk TK (toolkit) but was wondering if there might be some type of compression settings so that upon saving the image then it would be compressed and upon loading then it would be decompressed.

The other question is that I would like to investigate if there was any type of encryption functionality already built into ImDisk as that is a feature that I would like to use, but if it is not then maybe I could investigate adding it in some way but wanted to check first.

Truly awesome work you have done along with the original developer and it still holds great potential for a lot of good use.

Thanks and have a great day.

Hi Olof,

I downloaded https://ltr-data.se/files/imdiskinst.exe from https://ltr-data.se/opencode.html/ (that's your site, correct?) the other day, and the site indicated it was v2.1.1, built on 2021-11-05 (IIRC). The download was 613KB.

I then returned to https://ltr-data.se/opencode.html/ a few hours ago, and it indicated the version number was still v2.1.1, but it now indicated it was built on 2021-11-07 (as opposed to 2021-11-05). The download size increased substantially to 641KB.

Also, 4 security vendors currently flag this file as malicious on virustotal.com, whereas the release from a couple days ago only currently has 1 malware flag.

Do you know what's going on?

Why is the download from https://ltr-data.se/opencode.html/ changing without the version number changing?

Also, why do none of the file timestamps match the build date?

Thank you for your help.