Jump to content

Compiling ACPI v2.0 driver for Windows XP SP3 and Windows 2003 SP2 (x32/x64)

Mov AX, 0xDEAD

By Mov AX, 0xDEAD
in Windows XP

 Share

Recommended Posts

Mov AX, 0xDEAD

Mov AX, 0xDEAD

Posted
  • Author
Posted
3 hours ago, Dietmar said: 
STACK_TEXT:  
ba551858 80a30d7b 00000003 ba551bb4 00000000 nt!RtlpBreakWithStatusInstruction
ba5518a4 80a319e6 00000003 897d7868 898bf000 nt!KiBugCheckDebugBreak+0x19
ba551c84 80a31f77 000000a5 00000003 8989dba4 nt!KeBugCheck2+0x574
ba551ca4 ba732b29 000000a5 00000003 8989dba4 nt!KeBugCheckEx+0x1b
ba551cc0 ba75c355 8989dba4 c0140008 00000000 ACPI!ACPIBuildCompleteMustSucceed+0x31 [e:\software\windowssourcecode\microsoft.leaked.source.code.archive_2020-10-04\nt5src\source\xpsp1\nt\base\busdrv\acpi\driver\nt\buildsrc.c @ 431]
ba551cf8 ba75a1b5 ba732af8 c0140008 00000000 ACPI!AsyncCallBack+0xe5 [e:\software\windowssourcecode\microsoft.leaked.source.code.archive_2020-10-04\nt5src\source\xpsp1\nt\base\busdrv\acpi\driver\amlinew\sync.c @ 82]
ba551d1c ba75d133 c0140008 4556414c 898bf000 ACPI!RunContext+0x1f5 [e:\software\windowssourcecode\microsoft.leaked.source.code.archive_2020-10-04\nt5src\source\xpsp1\nt\base\busdrv\acpi\driver\amlinew\ctxt.c @ 649]
ba551d44 ba75d25f 898bf000 00000000 ba77ab08 ACPI!InsertReadyQueue+0x15a [e:\software\windowssourcecode\microsoft.leaked.source.code.archive_2020-10-04\nt5src\source\xpsp1\nt\base\busdrv\acpi\driver\amlinew\sched.c @ 275]
ba551d64 ba74e137 897d7a48 00000000 89913020 ACPI!RestartCtxtPassive+0x54 [e:\software\windowssourcecode\microsoft.leaked.source.code.archive_2020-10-04\nt5src\source\xpsp1\nt\base\busdrv\acpi\driver\amlinew\sched.c @ 384]
ba551dac 80bd81ac 00000000 00000000 00000000 ACPI!ACPIWorker+0xbf [e:\software\windowssourcecode\microsoft.leaked.source.code.archive_2020-10-04\nt5src\source\xpsp1\nt\base\busdrv\acpi\driver\nt\worker.c @ 302]
ba551ddc 80ae4212 ba74e078 00000000 00000000 nt!PspSystemThreadStartup+0x34
00000000 00000000 00000000 00000000 00000000 nt!KiThreadStartup+0x16

Ok, this trace dont show usefull info

c0140008 = AMLIERR_UNEXPECTED_ARGTYPE, in all cases AMLIERR_UNEXPECTED_ARGTYPE  used with code like this

 rc = AMLI_LOGERR(AMLIERR_UNEXPECTED_ARGTYPE,
                                     ("ValidateArgTypes: expected Arg%d to be type Integer (Type=%s)..);
 

seems AMLI_LOGERR() has some requirement to enable text message, need to change it to show messsage in any case


Mov AX, 0xDEAD

Mov AX, 0xDEAD

Posted
  • Author
Posted (edited)

Hi!

1) Little patch to enable text output  of wanted function:

In function IsTraceOn() inside trace.c after line:
    BOOLEAN rc = FALSE;
 add few lines:
     if (
        !_stricmp(pszProcName, "VALIDATEARGTYPES") ||
        !_stricmp(pszProcName, "VALIDATETARGET") ||
        !_stricmp(pszProcName, "xxx") ||
        !_stricmp(pszProcName, "yyy")
        )
    {
            rc = TRUE;
    }
You can add/replace xxx/yyy to any wanted function, just look at begin of function for line like TRACENAME("FATAL"), word FATAL is what you need

2) Configure WinDbg: for massive verbose output

  • build acpi.sys debug version
  • configure boot.ini to insta-break /BREAK
  • enable WinDbg output to file: Edit->Opem/Close log file (repeat at every session)
  • bu acpi!DriverEntry -  tell to stop at acpi.sys (windbg will save between sessions)
  • g - run kernel before first breakpoint
  • wait for break in acpi, you must see message Breakpoint 0 hit ACPI!DriverEntry:
  • ed Kd_ACPI_Mask 0xFFFFFFFF
  • !amli set spewon verboseon logon traceon
  • g - continue to load windows
  • type Ignore few times if asked, usually this is assertion check, so better to check in source files what condition was triggered
  • wait until desktop loaded, you must see a lot of text :)
  • commit log file: Edit->Opem/Close log file->Close
Edited by Mov AX, 0xDEAD
Dietmar

Dietmar

Posted
Posted (edited)

@Mov AX, 0xDEAD

You mean, that for the trace.c in base\busdrv\acpi\driver\amlinew

1.) I should add after line BOOLEAN rc = FALSE; in this trace.c

if (
        !_stricmp(pszProcName, "VALIDATEARGTYPES") ||
        !_stricmp(pszProcName, "VALIDATETARGET") ||
        !_stricmp(pszProcName, "xxx") ||
        !_stricmp(pszProcName, "yyy")
        )
    {
            rc = TRUE;
    }

and then compile acpi.sys debug version new with this new trace.c .

 

But what have I to write for "xxx" and "yyy" ?

Dietmar

PS: Maybe, that for this BSOD with c0140008 it is enough to add in trace.c only this one

if (
        !_stricmp(pszProcName, "VALIDATEARGTYPES") ||
        !_stricmp(pszProcName, "VALIDATETARGET")
        )
    {
            rc = TRUE;
    }

 

Edited by Dietmar
Mov AX, 0xDEAD

Mov AX, 0xDEAD

Posted
  • Author
Posted
2 hours ago, Dietmar said:

@Mov AX, 0xDEAD

if (   !_stricmp(pszProcName, "VALIDATEARGTYPES") ||

        !_stricmp(pszProcName, "VALIDATETARGET") ||
        !_stricmp(pszProcName, "xxx") ||
        !_stricmp(pszProcName, "yyy")
        )

and then compile acpi.sys debug version new with this new trace.c .

But what have I to write for "xxx" and "yyy" ?

@Dietmar

You can keep xxx/yyy as is, _stricmp() is string comparator, you can compare anything with anything, results are ORed

Dietmar

Dietmar

Posted
Posted (edited)

@Mov AX, 0xDEAD

 

Now, loong output stops with know Bsod containing 0xC0140008 .

Here is the reason,

thanks a lot

Dietmar 

8989dbf9: {
8989dbf9: CreateQWordField(NBUF=Buffer(0x30){
	0x8a,0x2b,0x00,0x00,0x01,0x10,0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x79,0x00},0xe,NBAS)AMLI:| | | | | | | | ValidateArgTypes(pArgs=897680fc,ExpectedTypes=BI)
AMLI:| | | | | | | | ValidateArgTypes=0

8989dc04: CreateQWordField(NBUF=Buffer(0x30){
	0x8a,0x2b,0x00,0x00,0x01,0x10,0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x79,0x00},0x16,NMAS)AMLI:| | | | | | | | ValidateArgTypes(pArgs=897680fc,ExpectedTypes=BI)
AMLI:| | | | | | | | ValidateArgTypes=0

8989dc0f: CreateQWordField(NBUF=Buffer(0x30){
	0x8a,0x2b,0x00,0x00,0x01,0x10,0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00
	0x00,0x00,0x00,0x00,0x79,0x00},0x26,NLEN)AMLI:| | | | | | | | ValidateArgTypes(pArgs=897680fc,ExpectedTypes=BI)
AMLI:| | | | | | | | ValidateArgTypes=0

8989dc1a: Store(NHLAACPIAsyncAcquireGlobalLock: Entered with context 897680ac
ACPIAsyncAcquireGlobalLock: Got lock immediately, Context 897680ac
ACPIReleaseGlobalLock: Lock released by context 897680ac
=Buffer(0x8){
	0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00},NBAS)AMLI:| | | | | | | ValidateTarget(pdataTarget=89768110,ExpectedType=DataObject,ppdata=ba55bcd4)
AMLI:| | | | | | | ValidateTarget=0 (pdataTarget=89768160)
=Buffer(0x8){
	0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00}
8989dc23: Add(NHLAACPIAsyncAcquireGlobalLock: Entered with context 897680ac
ACPIAsyncAcquireGlobalLock: Got lock immediately, Context 897680ac
ACPIReleaseGlobalLock: Lock released by context 897680ac
=Buffer(0x8){
	0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00},Subtract(NHLLACPIAsyncAcquireGlobalLock: Entered with context 897680ac
ACPIAsyncAcquireGlobalLock: Got lock immediately, Context 897680ac
ACPIReleaseGlobalLock: Lock released by context 897680ac
=0x0,One,)AMLI:| | | | | | | ValidateArgTypes(pArgs=8976827c,ExpectedTypes=II)
AMLI:| | | | | | | ValidateArgTypes=0
AMLI:| | | | | | | ValidateTarget(pdataTarget=897682a4,ExpectedType=DataObject,ppdata=ba55bcd4)
AMLI:| | | | | | | ValidateTarget=0 (pdataTarget=897682a4)
=0xffffffff,NMAS)AMLI:| | | | | | | ValidateArgTypes(pArgs=897680fc,ExpectedTypes=II)
AMLI_ERROR(c0140008): Unexpected argument type
ValidateArgTypes: expected Arg0 to be type Integer (Type=Buffer)
AMLI:| | | | | | | ValidateArgTypes=c0140008

8989dc33: }
*** Fatal System Error: 0x000000a5
                       (0x00000003,0x8989DBA4,0xC0140008,0x494E495F)

Break instruction exception - code 80000003 (first chance)

A fatal system error has occurred.
Debugger entered on first try; Bugcheck callbacks have not been invoked.

A fatal system error has occurred.

Connected to Windows XP 2600 x86 compatible target at (Wed Apr  6 13:29:01.796 2022 (UTC + 2:00)), ptr64 FALSE
Loading Kernel Symbols
.........................
Loading User Symbols

*******************************************************************************
*                                                                             *
*                        Bugcheck Analysis                                    *
*                                                                             *
*******************************************************************************

Use !analyze -v to get detailed debugging information.

BugCheck A5, {3, 8989dba4, c0140008, 494e495f}

Probably caused by : ACPI.sys ( ACPI!ACPIBuildCompleteMustSucceed+31 )

Followup: MachineOwner

 

Edited by Dietmar
Dietmar

Dietmar

Posted
Posted (edited)

@Mov AX, 0xDEAD

NMAS appears in DSDT only in this Method.

Here is the whole Windbg output with name otto1

https://ufile.io/9zatgp5y

Dietmar

            

 Method (_INI, 0, NotSerialized)  // _INI: Initialize
            {
                CreateQWordField (NBUF, \_SB.PC00.HDAS._Y35._MIN, NBAS)  // _MIN: Minimum Base Address
                CreateQWordField (NBUF, \_SB.PC00.HDAS._Y35._MAX, NMAS)  // _MAX: Maximum Base Address
                CreateQWordField (NBUF, \_SB.PC00.HDAS._Y35._LEN, NLEN)  // _LEN: Length
                NBAS = NHLA /* \NHLA */
                NMAS = (NHLA + (NHLL - One))
                NLEN = NHLL /* \NHLL */
            }

 

Edited by Dietmar
Dietmar

Dietmar

Posted
Posted (edited)

@Mov AX, 0xDEAD

 

I think, that in the translation of the Acpi2 word CreateQWordField ---> CreateDWordField

something went wrong. The first argument in this function is from type buffer and I think,

CreateQWordField also returns a value from type buffer.

And the Method _INI  is parsed from XP, other Method with also CreateQWordField in it may be just not parsed from XP

and so no error. This would mean, that any DSDT, that contains a Method _INI and the Acpi2 word CreateQWordField in it

would give until now Bsod with BugCheck A5, {3, xxx, c0140008, yyy}

Dietmar

 

PS: Maybe, that the same error for the Acpi2 word CreateQWordField also happens in Vista (Longhorn) 5048 acpi.sys .

Edited by Dietmar
WinWord2000

WinWord2000

Posted
Posted
On 4/4/2022 at 10:33 PM, Mov AX, 0xDEAD said:

Hi Winword2000, Club

This is possible only if someone leak win2000 acpi sources from MS. Second option is binary patching without sources, but you need to find some Bro who interested with this job

I think it is also possible if we change the ntoskrnl and hal  of Windows 2000 with the ntoskrnl and hal of the leaked XP sp1 , Is that correct ?

 WinWord2000 (Mox ax,bx) Grazie a tutti !

George King

George King

Posted
Posted
3 minutes ago, WinWord2000 said:

I think it is also possible if we change the ntoskrnl and hal  of Windows 2000 with the ntoskrnl and hal of the leaked XP sp1 , Is that correct ?

 WinWord2000 (Mox ax,bx) Grazie a tutti !

You can try to replace Windows 2000 acpi.sys with one that is linked on first page to WinCert. Or try explore that file on running Windows with Depency Walker to see if there are missing import. If no missing import, then you can try to replace it on running system

1
WinWord2000

WinWord2000

Posted
Posted (edited)
1 hour ago, George King said:

You can try to replace Windows 2000 acpi.sys with one that is linked on first page to WinCert. Or try explore that file on running Windows with Depency Walker to see if there are missing import. If no missing import, then you can try to replace it on running system

 

I know someone in another Forum who I think is Damnation here, who started doing this work, but he encountered problems in debugging :   https://forum.eclectic4un.me/viewtopic.php?f=25&t=89 .

@Damnationdo you know how to handle this error ?

If anyone @Dietmar or @Mov AX, 0xDEAD or @George King know of  debugging knowledge then help Damnation please 

Windows 2000 is the only classic system that can run this because of its similarity to XP

I don't know about this stuff ,my specialist is Security and Network 

WinWord2000 (Mox ax,bx in EclecBoard) Grazie a tutti !

Edited by WinWord2000
George King

George King

Posted
Posted (edited)
On 4/4/2022 at 10:59 PM, Mov AX, 0xDEAD said:

Use official MS way:

  • create separate .asm files in amd64 folder, let's name as amd64_helpers.asm
  • add AMD64_SOURCES=  amd64\amd64_helpers.asm to file "sources" of project
  • place inlined __asm {...} code with preprocessor #ifdef _X86_ ...... #endif, it will processed only for x32 builds

example of x64 asm:
  

PUBLIC  OSNotifyDeviceCheck

EXTRN       g_AmliHookEnabled:DWORD
EXTRN       memcpy:PROC
EXTRN       memset:PROC
  
OSNotifyDeviceCheck PROC
  ...
  call memcpy
  ...
OSNotifyDeviceCheck ENDP

 

Can you please prepare x64 ASM for this project? I have no idea how I can do it. I understand these steps except ASM creation.. Or does it mean only to copy current ASM code to standalone file?

I would like to compile it for x64 target to give it a try on my machine.

 

EDIT: I have added #ifdef _X86_ to osnotify.c and devpower.c.

But now I fail with these warnings, how I can solve variable conversion type? Is there a way to bypass it?

image.thumb.png.9d8bbbcd6371be0aa939e2a4f86befa0.png

 

Edited by George King
Dietmar

Dietmar

Posted
Posted

@Mov AX, 0xDEAD

 

Can the trace.c function also be used for the Acpi Bsod

0x000000A5 (0x00000002, xxx, 0x00000001(0), yyy)

The "1" in this BSOD means: 1 : ACPI cannot convert the BIOS' resource list into the proper
format. This probably represents a flaw in the BIOS' list
encoding procedure.

Which functions have to be set at this places, for 0x000000A5 (0x00000002, xxx, 0x00000001, yyy)

and then later for 0x000000A5 (0x00000002, xxx, 0x00000000, yyy)

via

if (   !_stricmp(pszProcName, "VALIDATEARGTYPES") ||

        !_stricmp(pszProcName, "VALIDATETARGET") ||
        !_stricmp(pszProcName, "xxx") ||
        !_stricmp(pszProcName, "yyy")
        )

 

Dietmar

Mov AX, 0xDEAD

Mov AX, 0xDEAD

Posted
  • Author
Posted (edited)
10 hours ago, Dietmar said:

@Mov AX, 0xDEAD 

8989dc1a: Store(NHLAACPIAsyncAcquireGlobalLock: Entered with context 897680ac
ACPIAsyncAcquireGlobalLock: Got lock immediately, Context 897680ac
ACPIReleaseGlobalLock: Lock released by context 897680ac
=Buffer(0x8){
	0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00},NBAS)AMLI:| | | | | | | ValidateTarget(pdataTarget=89768110,ExpectedType=DataObject,ppdata=ba55bcd4)
AMLI:| | | | | | | ValidateTarget=0 (pdataTarget=89768160)
=Buffer(0x8){
	0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00}
8989dc23: Add(NHLAACPIAsyncAcquireGlobalLock: Entered with context 897680ac
ACPIAsyncAcquireGlobalLock: Got lock immediately, Context 897680ac
ACPIReleaseGlobalLock: Lock released by context 897680ac
=Buffer(0x8){
	0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00},Subtract(NHLLACPIAsyncAcquireGlobalLock: Entered with context 897680ac
ACPIAsyncAcquireGlobalLock: Got lock immediately, Context 897680ac
ACPIReleaseGlobalLock: Lock released by context 897680ac
=0x0,One,)AMLI:| | | | | | | ValidateArgTypes(pArgs=8976827c,ExpectedTypes=II)
AMLI:| | | | | | | ValidateArgTypes=0
AMLI:| | | | | | | ValidateTarget(pdataTarget=897682a4,ExpectedType=DataObject,ppdata=ba55bcd4)
AMLI:| | | | | | | ValidateTarget=0 (pdataTarget=897682a4)
=0xffffffff,NMAS)AMLI:| | | | | | | ValidateArgTypes(pArgs=897680fc,ExpectedTypes=II)
AMLI_ERROR(c0140008): Unexpected argument type
ValidateArgTypes: expected Arg0 to be type Integer (Type=Buffer)
AMLI:| | | | | | | ValidateArgTypes=c0140008

 

@Dietmar

Quote

8989dbf9: CreateQWordField(NBUF=Buffer(0x30){...NBAS)
ValidateArgTypes(ExpectedTypes=BI)
ValidateArgTypes=0

8989dc04: CreateQWordField(NBUF=Buffer(0x30){...NMAS)
ValidateArgTypes(ExpectedTypes=BI)
ValidateArgTypes=0

8989dc0f: CreateQWordField(NBUF=Buffer(0x30){...NLEN)
ValidateArgTypes(ExpectedTypes=BI)
ValidateArgTypes=0

Creating fields was OK

Quote

8989dc1a: Store(
=Buffer(0x8){0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00},NBAS)
ValidateTarget(ExpectedType=DataObject)
ValidateTarget=0
=Buffer(0x8){0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00}

NBAS = NHLA shows as OK, but NHLA was evaluated as Buffer, must be Integer   

Quote

8989dc23: Add(
=Buffer(0x8){0x00,0xc0,0x26,0x65,0x00,0x00,0x00,0x00}
    ,Subtract(=0x0,One,)
    ValidateArgTypes(ExpectedTypes=II)
    ValidateArgTypes=0
    ValidateTarget(ExpectedType=DataObject)
    ValidateTarget=0

NHLL - One is OK, NHLL evaluated as 0x0 Integer, result = 0 - 1 = -1 = 0xffffffff

Quote

=0xffffffff,NMAS)
ValidateArgTypes(ExpectedTypes=II)
AMLI_ERROR(c0140008): Unexpected argument type
ValidateArgTypes: expected Arg0 to be type Integer (Type=Buffer)
ValidateArgTypes=c0140008

NMAS = (NHLA + (NHLL - One)) is Failed
Add() expect two integers, but first argument (Arg0=NHLA) evaluated as Buffer/Field

How it was declared (on my bios):

Quote

    OperationRegion (PNVA, SystemMemory, PNVB, PNVL)

    Field (PNVA, AnyAcc, Lock, Preserve)
    {
        ...
        NHLA,   64,
        NHLL,   32,
        ....
        }

NHLL is OK, 32bit field, at Subtract() evaluated as 0x0 Integer
NHLA is BAD, seems XP ACPI don't handle 64bit field declaration as Integer and alias it as Buffer type

This declaration is not CreateQWordField opcode, it handled in other place(OperationRegion/Buffer opcode i guess )

On my Skylake H110 board this code has OS check, so DSDT authors know about ACPI compatibility

Quote

Method (_INI, 0, NotSerialized)  // _INI: Initialize
{
    If (OSYS >= 0x07D6)
    {
        CreateQWordField (NBUF, \_SB.PCI0.HDAS._Y25._MIN, NBAS)  // _MIN: Minimum Base Address
        CreateQWordField (NBUF, \_SB.PCI0.HDAS._Y25._MAX, NMAS)  // _MAX: Maximum Base Address
        CreateQWordField (NBUF, \_SB.PCI0.HDAS._Y25._LEN, NLEN)  // _LEN: Length
        NBAS = NHLA /* \NHLA */
        NMAS = (NHLA + (NHLL - One))
        NLEN = NHLL /* \NHLL */
        If (VDID != 0xFFFFFFFF)
        {
            VMMH (Zero, One)
        }
    }
}

 

Edited by Mov AX, 0xDEAD
2

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
 Share
Go to topic listing

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...