- Halfplement and export VdmMapFlat
- Make mention of the existence of VdmFlushCache and VdmUnmapFlat, which are not exported in x86 builds (macros are therefore defined to just return TRUE) but are exported in MIPS... builds...
See VDDSVC.H and NT_VDD.H in the DDK
(those headers don't exist in ROS; we'll have to create them later on).
svn path=/branches/ntvdm/; revision=61288
Studying the "vddtest" example mentioned in revision 61283 allowed me to see that valid VDD DLL "handles" start at 1 and cannot be equal to zero.
Add helper macros to convert from indices to handles. Fixes "vddtest" example.
Have fun at running it :)
svn path=/branches/ntvdm/; revision=61287
- Thanks Thomas for having pointed me towards the fact that the NTVDM exports were really STDCALL and not CDECL (as I've thought first, because originally I was checking whether the getXX() functions were stdcall or not, and distinguishing VOID stdcall or cdecl functions without the symbols was hopeless).
- Halfplement and export Sim32pGetVDMPointer and MGetVdmPointer, needed to run correctly the RageStorm "galaxy" sample mentioned yesterday (see revision 61283). Now it loads and the dispatch call works \o/
- VDD_INIT_PROC and VDD_DISPATCH_PROC have the same signature --> rename them to VDD_PROC.
Have fun at testing!
svn path=/branches/ntvdm/; revision=61286
Implement basic VDD support:
implement the (un)register & dispatch call BOP (called BOP_3RDPARTY): see "Undocumented DOS" by Schulman et al. page 267 (as always ;) ), as well as:
http://www.ragestorm.net/tutorial?id=27 - "Tutorial - Calling Win32 from DOS"
http://www.airborn.com.au/layout/vddtest.html - "DOS to 32 bit DLL under XP - an example" (almost the same explanations as the previous article)
http://sta.c64.org/blog/dosvddaccess.html - "Accessing Windows device drivers from DOS programs"
and...
... your lovely Win2k / Win2k3 DDK section: "Device Technologies/Other Devices/Devices Requiring VDDs" !!
Ready to test? :D
svn path=/branches/ntvdm/; revision=61283
- As already done for interrupts and I/O ports, add a registering system for BOPs.
- INT32: Move ControlBop to the 32-bit interrupts module where it is used only.
- DOS: Add (un)documented BOP_DOS (0x50) and BOP_CMD(0x54) used respectively by NTIO.SYS/NTDOS.SYS and by COMMAND.COM.
It appears that they take an extra parameter (so, skip 1 byte-instruction after the BOP instruction as we do for the Control BOP 0xFF).
See "Undocumented DOS 2nd edition" by Schulman et al., page 267.
svn path=/branches/ntvdm/; revision=61278
- Add/fix few comments;
- Don't forget to redisable access to AC registers after modifying them in int 10h. Usurp, please retest Rescue Rover 2 :)
- Make EmulatorGet/Set/ClearFlag internal functions only and use set/get<FLAG> instead.
svn path=/branches/ntvdm/; revision=61263
Fix the file reading subfunction (AH = 3Fh) of INT 21h.
When reading from the console, it always stops on a carriage return.
svn path=/branches/ntvdm/; revision=61256
Rewrite the VGA emulation.
Nah, it's a joke!
- Store VGA registers setup in VGA_REGISTERS structure.
- Improve BiosSetVideoMode and introduce a VgaSetRegisters function for setting VGA regs.
- Initialize Bda->CrtBasePort: it is used to determine the address of some I/O ports which depend on whether we are in color or monochrome mode.
- Add UnregisterIoPort in io.c (and fix some DPRINTs) (needed for VGA).
- When setting the misc. VGA register, check whether we're going to monochrome emulation, and reset some port addresses in that case.
svn path=/branches/ntvdm/; revision=61246
- Last race condition fix (see revision 61230).
- Add a comment why we do twice the same emptiness check (one before holding the mutex and one just after).
svn path=/branches/ntvdm/; revision=61239
The current "fast" method used in MOVS gives incorrect results. It has been temporarily
replaced with a slow method that works correctly.
The string functions do nothing when CX is 0 on entry, fix CMPS and SCAS.
svn path=/branches/ntvdm/; revision=61203
Revert r61158, and fix the prefix issue correctly, but resetting the flags
again after the call to Fast486InterruptInternal.
svn path=/branches/ntvdm/; revision=61160
Move the prefix reset below the interrupt check, because Fast486InterruptInternal
might set OPSIZE in some circumstances.
svn path=/branches/ntvdm/; revision=61158
- Remove some DPRINTs.
- Fix a bug in VgaEnterGraphicsMode (erroneous change of the real saved resolution when going into DoubleVision(TM) mode).
- Do not update the text cursor when being in graphics mode (otherwise I've already hit some ScanlineSize==0 problems) <-- [TheFlash], can you recheck that, please?
- Little optimisation: do not do anything in VgaRefreshDisplay if nothing has to be done.
svn path=/branches/ntvdm/; revision=61157
FPU instructions never generate #UD. If CR0.EM is set, they generate #NM.
If there is no FPU and CR0.EM is cleared, they do nothing.
Add a helper macro FPU_ST for accessing FPU registers.
Add a compile-time option to disable the FPU.
svn path=/branches/ntvdm/; revision=61150
- Move some PS/2 initialization from bios.c to ps2.c
- NULLify interrupt entries which are in fact pointers to data tables.
- Add a framework for adding I/O port handlers.
svn path=/branches/ntvdm/; revision=61107
Handle the case when the last page is also the first page in Fast486ReadLinearMemory ( when paging is enabled).
svn path=/branches/ntvdm/; revision=61104
The 486 doesn't have AC, VIF, VIP or ID.
Fix and simplify POPF.
[NTVDM]
Fix the count in the XMS copy function (INT 15h, AH = 87h).
svn path=/branches/ntvdm/; revision=61082
DAA, DAS: Update the flags.
AAA, AAS: Subtract from AX instead of AL. The specification says AL, but
so far in all known CPUs and emulators, AX is subtracted.
AAD: Clear AH.
svn path=/branches/ntvdm/; revision=61069
