Crash on NT4.0 using VMware NVMe Controller

[Win-DVD]

I tried to load the driver on NT4 with the VMware NVMe and it seems to crash with a vague error in event viewer (unless I installed this wrong somehow?)
"Driver detect an internal error in its data structures for . "

Is this user error or a driver fault?

VM allocated 4 CPU cores and 3.5GB ram

[ages2001]

@Win-DVD did you try with standard Hal instead of MPS Hal?

[Win-DVD]

@Win-DVD did you try with standard Hal instead of MPS Hal?

I haven't yet, is there a way to switch HAL in an existing installation or do i need to clean install?

[the-r3dacted]

I haven't yet, is there a way to switch HAL in an existing installation or do i need to clean install?

From my understanding this is possible, in device manager, under Computer, it should say like MPS Multiprocessor PC or similar. Double click that, Driver tab, Update Driver, then click next on the Wizard and tick Display a list of the known drivers for this device, and next.

Now you should be able to pick another HAL, like MPS Uniprocessor PC or Standard PC or something else. (Probably take a snapshot here in case anything breaks.) Then pick the HAL you want, next, let it install, and reboot.

[Win-DVD]

I haven't yet, is there a way to switch HAL in an existing installation or do i need to clean install?

From my understanding this is possible, in device manager, under Computer, it should say like MPS Multiprocessor PC or similar. Double click that, Driver tab, Update Driver, then click next on the Wizard and tick Display a list of the known drivers for this device, and next.

I don't believe NT4 has Device Manager

[the-r3dacted]

I think my brain saw nvme2k, forgot the issue was about NT4, and provided instructions on how to do it for 2000.

Here's instructions about that direct from Microsoft
https://web.archive.org/web/19990221160925/http://support.microsoft.com/support/kb/articles/q156/3/58.asp

[Win-DVD]

I decided to just do a clean installation and interestingly got this

I wonder why it's not picking it up

[ages2001]

@Win-DVD use Universal NT Installer (UNTI). It supports Hal switching without reinstalling the OS. Install vanilla edition.

UNTI will support installing NT3.5x and NT4 directly to NVMe disk in near future. For now, install it to an IDE or AHCI disk. Then, install NVMe driver manually.

[red-ray]

I tried to load the driver on NT4 with the VMware NVMe and it seems to crash with a vague error in event viewer

I have never managed to get NVMe2K running on any of my NT4 systems and get exactly the same entry in the event log using the 1.2.0.0 release. Looking at the NVMe2K service to me it seems like it started and existed, I guess from DriverEntry(), with an NT error that got translated to WIN32 error 31.

    set DeviceValues     = { +
            {NumberOfRequests, 0, $(!REG_VT_DWORD), 32}, +
            {MaximumSGList, 0, $(!REG_VT_DWORD), 512}  +

I also noted that these two values in the NT4 .INF file are probably incorrect, they got fixed in the W2K + WDM NVMe2K.inf files and I guess this one got overlooked.

I am running on real hardware and using the MPS 1.4 HAL which is the only sensible option for this system. The same system can also be booted into XP-x32 + 2003-x32 + W10-x64 and the 2.2.0.0 driver runs on all three of those systems.

As I see it NVMe2K is broken for NT4 and needs fixing.

[ages2001]

I tried to load the driver on NT4 with the VMware NVMe and it seems to crash with a vague error in event viewer

I have never managed to get NVMe2K running on any of my NT4 systems and get exactly the same entry in the event log using the 1.2.0.0 release. Looking at the NVMe2K service to me it seems like it started and existed, I guess from DriverEntry(), with an NT error that got translated to WIN32 error 31.

``` set DeviceValues = { + {NumberOfRequests, 0, $(!REG_VT_DWORD), 32}, + {MaximumSGList, 0, $(!REG_VT_DWORD), 512} + ```

I also noted that these two values in the NT4 .INF file are probably incorrect, they got fixed in the W2K + WDM NVMe2K.inf files and I guess this one got overlooked.

I am running on real hardware and using the MPS 1.4 HAL which is the only sensible option for this system. The same system can also be booted into XP-x32 + 2003-x32 + W10-x64 and the 2.2.0.0 driver runs on all three of those systems.

As I see it NVMe2K is broken for NT4 and needs fixing.

Maybe some updates broke the compatibility. I have a driver which works with A620M motherboard and Samsung 980 Pro NVMe SSD in NT3.50 and NT4. I can share driver if you want.

[techomancer]

I tested the driver on my NT4 checked build in VirtualBox before release.
I dont think HAL matters here.
These registry entries are correct, but they are not strictly needed anyway

[red-ray]

Maybe some updates broke the compatibility. I have a driver which works with A620M motherboard in NT3.50 and NT4.

Thank you for the offer, but as I am pondering building a PowerPC NVMe2K driver for NT4 SP2 the only sensible options is to get the latest source working.

Which HAL are you using?

[techomancer]

The changes from 1.0.0.3 to 1.2.0.0 are quite minimal and mostly affecting NVME SMART to support SIV.
The init sequence is very much the same. It did change quite a bit between 1.0.0.2 and 1.0.0.3 because it has been moved earlier when then device is discovered so MDTS derived maximum data transfer can be reported to scsiport.

[ages2001]

Maybe some updates broke the compatibility. I have a driver which works with A620M motherboard in NT3.50 and NT4.

Thank you for the offer, but as I am pondering building a PowerPC NVMe2K driver for NT4 SP2 the only sensible options is to get the latest source working.

Which HAL are you using?

I used the standard HAL for NT3.50 and NT4. That mobo doesn't support MPS, so the only choice is standard HAL.

[red-ray]

I used the standard HAL for NT3.50 and NT4. That mobo doesn't support MPS, so the only choice is standard HAL.

Thank you, all my x32 PCI/PCIe NT4 systems have MPS 1.4 HAL. I wonder, does my SIV utility report NVMe SMART on NT4 with the 1.2.0.0 NVMe2K driver? The latest SIV32N V5.85 Beta-03 (https://workupload.com/file/BYkhW2BkhuM) and later have code to do this, but I have not been able to test it. It would be handy if from a command window you did SIV32N -DBGSDM -EXIT=20 > SIV_DBGOUT.log | more and posted the .LOG so I can see how it get's on. With luck the Menu->Devices->SMART Drives panel will report it and you should see similar to:

I dont think HAL matters here.

Which HAL were you using? So far in this thread no success has been reported for the MPS 1.4 HAL, does anyone have it working with the MPS 1.4 HAL?

The changes from 1.0.0.3 to 1.2.0.0 are quite minimal and mostly affecting NVME SMART to support SIV.

The changes are to support way more programs than SIV, I also tested them with https://www.hdsentinel.com/ and suspect most NVMe SMART reporting programs will now work.

The init sequence is very much the same. It did change quite a bit between 1.0.0.2 and 1.0.0.3 because it has been moved earlier when then device is discovered so MDTS derived maximum data transfer can be reported to scsiport.

NVMe2K has never worked for me on NT4, as I recall 1.0.0.2 was the first version I tried. All the x32 systems I have NT4 installed on which have PCI or PCIe are MP systems and use the MPS 1.4 HAL

[techomancer]

I think HAL thing may be a red herring and there is something different that causes these problems.
The driver is well insulated from the HAL it runs on by scsiport layer so it does not call any HAL functions directly.
I will try creating VM with multiple CPUs to see if it makes any difference.