Mail Archives: djgpp/2008/07/23/15:01:57
"Eph" <t DOT ephraim AT gmail DOT com> wrote in message
news:ca0df607-1be3-4358-b84f-f0ccddde3197 AT d77g2000hsb DOT googlegroups DOT com...
> I've problems to get a call to
> __dpmi_simulate_real_mode_procedure_retf_stack working correctly.
> My problem is, that I need to call a 16Bit real Mode function placed
> in Memory below 1MB, which expects three parameters to be placed on
> the RM Stack.(see comments in the code).
>
> Below the code, how I tried to use the function:
>
> ---------------
> __dpmi_regs regs;
> struct {
> unsigned long u1;
> unsigned long u2;
> unsigned long u3;
> } stack;
>
> memset(&stack, 0, sizeof(stack));
>
> stack.u1 = parm1; // expected at bp[14]
> stack.u2 = parm2; // expected at bp[10]
> stack.u3 = parm3; // expected at bp[06]
>
> memset(®s, 0, sizeof(__dpmi_regs));
>
> regs.x.cs = seg; // segment of the function entry
> regs.x.ip = ofs; // offset of the function entry
DPMI 0.9 says ss:sp must be zero for the DPMI host to provide a real mode
stack, or ss:sp must point to a valid real mode stack. DJGPP doc's
(libc.info) say ss:sp and flags must be set to valid values or zero.
regs.x.ss=0;
regs.x.sp=0;
The DJGPP doc's don't state which "flags" must be set (DPMI or __dpmi_regs):
regs.x.flags=0; /* __dpmi_regs flags */
regs.h.bh=0; /* DPMI flags */
But, it's likely the doc's were referring to the DPMI flags... (BH bit 0 =
1 resets A20 and PIC) You'll also need to set any other registers that the
function uses since the call copies _all_ registers.
> __dpmi_simulate_real_mode_procedure_retf_stack(®s, sizeof(stack) /
> 2, &stack);
> ---------------
>
> What am I doing wrong?
Not sure.
> Can anyone please point it out!
Try the above. Let us know.
Rod Pemberton
- Raw text -