DOS ain't dead

> > In my opinion, this is because JEMM386 does not check whether memory
> size
> > is equal to the XBDA segment:
> > MemW[$413] shl 6 = MemW[$40E]
> > If not, he should refrain from moving the XBDA.
>
> In my opinion it is because AHCI expects some alignment of its structures
> which Jemm cannot easily guarantee, especially if it moves itself into an
> UMB first.
>
> You may try UMBPCI & UMBM, umbm.exe has got an /xbda option - it's "better"
> aligned, because umbm doesn't have to "move itself" ( although I've
> forgotten the details ).
>
> Or perhaps try jemm with option NOHI!
>
> Btw, what is the status of the "dark debugger"? Is that something that you
> created - or at least are involved with? I find it a very interesting
> project.

I added a few lines to the code in INIT.ASM (security is there anything below XBDA) and now it doesn't hang.

if ?MOVEXBDA and ?MOVEXBDAPM
    push ebx
    test bV86Flags,V86F_MOVEXBDA
    jz nomovexbda
    movzx eax,word ptr ds:[@XBDA]   ;does XBDA exist?
    and eax,eax
    jz nomovexbda
  if 1                       
    mov cx,ds:[@MEM_SIZE]
    shl cx,6
    cmp ax,cx
    jnz nomovexbda
  endif
...

I was also able to modify UMBM so that it transfers GRUB4DOS and XBDA data to UMB, thanks to which I have 22KB more conventional memory.

The dark debugger has been completed and so far I have no plans to change anything.

Hello Japheth, hello CandyMan,

> > In my opinion, this is because JEMM386 does not check whether memory
> size
> > is equal to the XBDA segment:
> > MemW[$413] shl 6 = MemW[$40E]
> > If not, he should refrain from moving the XBDA.
> In my opinion it is because AHCI expects some alignment of its structures
> which Jemm cannot easily guarantee, especially if it moves itself into an
> UMB first.

Not checking whether [@MEM_SIZE] shl 6 == [@XBDA] seems like a bad idea, if you ask me.

Also, I see that JEMM16.ASM does have this particular check. So I am mystified as to why INIT.ASM lacks the same check.

Thank you!

---
https://gitlab.com/tkchia · https://codeberg.org/tkchia · 😴 "MOV AX,0D500H+CMOS_REG_D+NMI"

> Enabling A20 line via port 92h also crash with JEMM386 loaded.

The NOA20 option should disable Jemm's port 92h trapping. However, IIRC, since Jemm resides in extended memory it needs A20 to be always on, so I'm not sure that NOA20 makes Jemm really ignore what happens with A20 - perhaps this option only controls what's mapped in at address 100000h-10FFFFh. Might be a good idea to look into the source code ...

Anyway, on the machines that are available for testing I couldn't reproduce your crash.

---
MS-DOS forever!