XP/W2k3 x86 on Modern Hardware

[infuscomus]

@Mov AX, 0xDEAD

Windows 8.0 boot with patched XP outerspace acpi.sys - BSOD ATTEMPTED_EXECUTE_NOEXECUTE_MEMORY

Windows 8.0 boot with patched 7 acpi.sys - BSOD ACPI_BIOS_ERROR

Unfortunately it wouldn't show useful information like it usually does, just the title of the error, but nothing about where it occured.

[canonkong]

I think it is too hard to porting the win8.1 usb driver to work on win7/xp system. 

[infuscomus]

@canonkong

The windows 8.0 ACPI driver maybe? - vanilla versions gives A5 BSOD, but after patching it boots and the 149C controller works there properly at boot time..

It'd be one hell of a job to port it to XP though.

We'd first need to prove that it can work for the Vista RTM ACPI driver before we can even think of trying to port the Windows 8.0 version.

[canonkong]

@infuscomus

How to solve A5 BSOD? I found that many intel platform will A5 BSOD on win7/8/8.1, but AMD can replace the mod acpi.sys to solve it.

For AMD CPU inside XHCI contoller(Not only 149C), I think it is ASUS bios bug. Maybe cause by its power manage. On my testing, GIGABYTE, MSI, ASRock not have this problem, but ASUS, Lenovo mainboard will have this problem.

 

[infuscomus]

@canonkong

For A5 BSODs I asked @Dietmarfor help, he provided the patched windows 8.0 ACPI driver to me that help me get past the A5 BSOD there.

My system that has the 149C controller is a GIGABYTE TRX40 DESIGNARE and a 3960x - so it's not related to the ASUS bug and must be something else.

[canonkong]

@infuscomus

TRX40 have 148C and 149C, 149C is from chips, 148C form CPU, do the 148C have this problem?

If We want to porting the XHCI driver frome win8.1/10, we need @Mov AX, 0xDEAD help us to porting win8.1/10 usb driver. 

I can not solve those Timer funtions.

Edited December 3, 2021 by canonkong

[infuscomus]

@canonkong

I'm actually not sure which USB ports are connected to 148C, whenever I tested and checked through device manager which controller handled the device it was always 149C. I'll have to find which ones are connected to 148C.

[canonkong]

 @Mov AX, 0xDEAD@Dietmar

I already porting the usb driver from win8.1 and win10 1511, but they will code 10 error on win7.  What I need to do with the inf file?

 

Edited December 3, 2021 by canonkong

[Mov AX, 0xDEAD]

3 hours ago, infuscomus said:

The windows 8.0 ACPI driver maybe? - vanilla versions gives A5 BSOD, but after patching it boots and the 149C controller works there properly at boot time..

OK, original 8.0 usb3 driver works with 0x149c controller under Win8.0 environment (+ patched acpi), so problem is deeper and 8.1 driver in different OS may have same issue.

The first thing that comes to mind - DSDT, but on Linux open source usb3 driver (by Intel) uses acpi functions only in two cases - workaround for fewer Intel chipsets (Skylake and other), linux driver call special DSM methods to power up something in hardware, see xhci_pme_acpi_rtd3_enable()
second is powering ports(?) usb_acpi_set_power_state(), many subcalls.

[Mov AX, 0xDEAD]

33 minutes ago, canonkong said:

I already porting the usb driver from win8.1 and win10 1511, but they will code 10 error on win7.  What I need to do with the inf file?

Code 10 is common error -  unsupported hardware, unsupported OS, failed internal driver checks, etc...

Driver code starts, but windows receive some error from driver, you can't fix it with patching .inf

And don't forget to remove initial security cookie check, this check, will fail on pre-Win8 OSes

 

[infuscomus]

@Mov AX, 0xDEAD

With help from @skulltrail I've gotten the ntoskrnl source code to boot in vanilla SP3 XP.

However as soon as I try to use your ntoskrnl extender I get BSODs

First 0xDEADBEEF which I bypassed, but now I get BSOD 50 and 8E

any help to get your extender working with this?

[Mov AX, 0xDEAD]

3 minutes ago, infuscomus said:

@Mov AX, 0xDEAD

First 0xDEADBEEF which I bypassed, but now I get BSOD 50 and 8E

any help to get your extender working with this?

I can explain only 0xDEADBEEF bsod, because this is extender's internal assertion checks

0xDEADBEEF, 0x3 - bug in PsSetCreateProcessNotifyRoutineEx_k8, overflowed list or some similar

0xDEADBEEF, 0x5 - not found signature of existing but not exported functions in ntoskrnl.exe,  just comment KeBugCheckEx()

50 or 8E - can't help with this, when you mix parts of different OS - you get mess

[infuscomus]

1 hour ago, Mov AX, 0xDEAD said:

The first thing that comes to mind - DSDT, but on Linux open source usb3 driver (by Intel) uses acpi functions only in two cases - workaround for fewer Intel chipsets (Skylake and other), linux driver call special DSM methods to power up something in hardware, see xhci_pme_acpi_rtd3_enable() second is powering ports(?) usb_acpi_set_power_state(), many subcalls.

@Mov AX, 0xDEAD

So do you think the 149C controller is failing because of a missing function in our patched acpi.sys that the windows 8.0 acpi.sys has?

[infuscomus]

20 minutes ago, Mov AX, 0xDEAD said:

50 or 8E - can't help with this, when you mix parts of different OS - you get mess

@Mov AX, 0xDEAD

I have an OEM SP3 XP ISO where I swapped out the kernel with my compiled version if you want to try it out? This version installs in a VM at least.

[Mov AX, 0xDEAD]

21 minutes ago, infuscomus said:

@Mov AX, 0xDEAD

So do you think the 149C controller is failing because of a missing function in our patched acpi.sys that the windows 8.0 acpi.sys has?

I dont know, MS usb3 driver is very complex due "usb stack", maybe it require specific acpi functionality. Or maybe bios uses specific registers of amd usb controller to reprogram it to custom mode so DSDT is only way to control this "custom" mode.

[Mov AX, 0xDEAD]

14 minutes ago, infuscomus said:

I have an OEM SP3 XP ISO where I swapped out the kernel with my compiled version if you want to try it out? This version installs in a VM at least.

We already have compiled sp3 kernel from MS, so No. If you have custom kernel, you need all related files compiled from same source - hal, acpi, win32k(?), pci(?) ...

[infuscomus]

12 hours ago, Mov AX, 0xDEAD said:

We already have compiled sp3 kernel from MS, so No. If you have custom kernel, you need all related files compiled from same source - hal, acpi, win32k(?), pci(?) ...

@Mov AX, 0xDEAD

I haven't needed to compile anything else other than the kernel and it works when I swap ntoskrnl/ntkrnlmp/ntkrnlpa/ntkrpamp out on a SP3 ISO at least, but ntoskrnl extender it seems refuses to work with it.

[Dietmar]

@infuscomus

Can you send to me your modded hal and ntoskrnl build from Sources

Dietmar

Edited December 4, 2021 by Dietmar

[infuscomus]

@Dietmar

here you go.

https://ufile.io/3gzbkrf1

 

P.S I included HALs too but have not yet tested them.

[Dietmar]

@infuscomus

I try your hal and ntoskrnl files from Sources on the Asrock z370 Fatalty Gaming k6 board.

With Ntos8 drivers together I get DEADBEEF Bsod.

When I replace ntoskrnl and hal on a XP SP3 with iastor 1006,

I get very late in boot process Bsod 0xC000021A, also in Safe Mode.

The same Bsod happens from an XP SP3 Install CD on 3. reboot(last), when I changed hal and ntoskrnl

about yours from Sources  on Setup CD (also in SP3.cab I replace them all).

I think, that this Bsod goes to WinLogon or CSRSS.exe .

So it is not enough for only to replace hal and ntoskrnl from Sources.

About BSOD with Extender it may be, that some functions for storport are missed in XP SP1 ntoskrnl.

I talked with @daniel_k about this and he makes an storport.sys for me, which works also with XP SP1 for nvme boot and storahci

Dietmar

Edited December 4, 2021 by Dietmar

[infuscomus]

@Dietmar

I went and did a full fresh install of XP with my kernel on a vanilla SP3 ISO - I got all the way to the OOBE and after exiting that - BSOD 0xC000021A

It seems I have missed something with my kernel - I got close to SP3 compatibility though.

[infuscomus]

@Dietmar

Tomorrow I will try to re-compile the decompiled XP winlogon that showed up on 4chan a while back - hopefully that can help get this working.

edit: I've also made a new build environment VM - old one got corrupted.

Edited December 5, 2021 by infuscomus

[infuscomus]

I'm not having much luck compiling winlogon.

Edited December 5, 2021 by infuscomus

[infuscomus]

@Dietmar

Had some success compiling winlogon with the server 2003 compiler, let see how this one goes.

https://ufile.io/ljjg9x9w

[ExtremeGrief]

@infuscomusis that for 149C fix?