Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.os.msdos.djgpp > #1726 > unrolled thread

resident PMODEDJ?

Back to article view | Back to comp.os.msdos.djgpp

Contents

  resident PMODEDJ? "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-02-15 04:57 -0500
    Re: resident PMODEDJ? rugxulo@gmail.com - 2014-02-15 19:13 -0800
      Re: resident PMODEDJ? "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-02-16 07:45 -0500
        Re: resident PMODEDJ? "Rod Pemberton" <dont_use_email@xnothavet.cqm> - 2014-03-03 19:00 -0500
    Re: resident PMODEDJ? "Rod Pemberton" <dont_use_email@xnohavenotit.cnm> - 2014-02-17 16:59 -0500
      Re: resident PMODEDJ? "Rod Pemberton" <dont_use_email@xnothavet.cqm> - 2014-03-03 19:00 -0500

#1726 — resident PMODEDJ?

As a part of some testing, I'm doing the following:

  CWSDPMI -p
  start assembly .com TSR which uses DPMI
  start a DJGPP program using DPMI afterwards

I was doing some testing of some custom DPMI assembly code
in conjuction with a DJGPP program.  I'm having a problems
with the assembly code.  This has nothing to do with
DJGPP.  That's not what I'm asking about.

I wanted to try to use PMODEDJ instead of CWSDPMI, since
AIUI it doesn't have paging enabled.  The CWSDPMI TSR can
be told to stay resident with the '-p' parameter.

PMODEDJ's documentation says it will only stay resident
one time.  Is there a switch, parameter, hidden switch,
tweak, patch, etc that can be used to make PMODEDJ
(pmodetsr.exe) stay resident longer?  I.e., permanently.


Rod Pemberton

[toc] | [next] | [standalone]

#1731

Hi,

On Saturday, February 15, 2014 3:57:39 AM UTC-6, Rod Pemberton wrote:
>
> I wanted to try to use PMODEDJ instead of CWSDPMI, since
> AIUI it doesn't have paging enabled.  The CWSDPMI TSR can
> be told to stay resident with the '-p' parameter.

You mean paging or swapping? I'm no systems developer myself, so I
don't really claim to understand it. I'm not sure you can disable
paging in all circumstances (e.g. if EMM386 loaded).

If you just want to disable swapping, try "-p -s-" or use CWSDPR0
instead. Or you could probably use something else (e.g. HDPMI32 -r).

[toc] | [prev] | [next] | [standalone]

#1732

On Sat, 15 Feb 2014 22:13:44 -0500, <rugxulo@gmail.com> wrote:

>> I wanted to try to use PMODEDJ instead of CWSDPMI, since
>> AIUI it doesn't have paging enabled.  The CWSDPMI TSR can
>> be told to stay resident with the '-p' parameter.
>
> You mean paging or swapping?

I meant paging.  Thanks for the info though.

The issue is I need to keep PMODEDJ resident (to execute both programs).

Keeping PMODEDJ resident would allow me to see if the TSR code is working
or failing the same way for both.  I.e., if it works for one but not the
other, then I likely have a paging or perhaps a memory locking problem.
If it fails for both, then I more likely have some other coding mistake.

I was going to look at PMODEJ's code when I get a chance to see what  
triggers
it to unload.  It stays resident for one use.  So, there is something that
causes it to unload, e.g., counter, pointer, etc.  It might be possible to
recompile it, or patch a byte to prevent it from unloading.  A small TSR or
two to modify interrupts might work too.

I don't know if any of those ideas will be easy or if such an attempt will
"break" something internally in PMODEDJ.  Usually, the DPMI host cleans up
when it exits.  So, forcing a DPMI host to stay resident when it wasn't
designed to do so may prevent it from cleaning up or initializing things
properly with multiple uses.  But, I won't know any of that for sure until
I look and try, unless someone more experienced with PMODEDJ posts some
in-depth information.


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#1759

On Sun, 16 Feb 2014 07:45:39 -0500, Rod Pemberton  
<dont_use_email@xnohavenotit.cnm> wrote:

> [...]
> If it fails for both, then I more likely have some other coding mistake.

I made some other coding mistakes, mentioned in the "resident service
providers" thread.


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#1735

On Sat, 15 Feb 2014 04:57:39 -0500, Rod Pemberton  
<dont_use_email@xnohavenotit.cnm> wrote:

> [...]
>
> PMODEDJ's documentation says it will only stay resident
> one time.  Is there a switch, parameter, hidden switch,
> tweak, patch, etc that can be used to make PMODEDJ
> (pmodetsr.exe) stay resident longer?  I.e., permanently.

The patch below allow you to build a resident version of
PMODEDJ.  PMODEDJ is a.k.a. PMODETSR.EXE.

This version is experimental.  Use at your own discretion.

Be warned that I could've introduced any number of unknown
issues, or more likely failed to fix many potential problems
that the program was never coded to handle.

However, preliminary tests seem to indicate that the few
trivial changes below make PMODEDJ resident.

The diff's are against pmode13s.zip available here:
http://www.delorie.com/pub/djgpp/current/v2misc/

I also had an issue with the MAKEFILE failing when TLINK
was executed.  The error complained about 16-bit DPMI ...
So, I made a batch file to build it.  That's below also.

Please don't ask me for additional fixes.


diff -Naur PMODE/pmode.asm PMODE_NEW/pmode.asm
--- PMODE/pmode.asm	2000-10-13 22:09:00.000000000 -0500
+++ PMODE_NEW/pmode.asm	2014-02-17 15:37:58.000000000 -0500
@@ -933,11 +933,11 @@
  ;°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°
  ; [fold]  [
  pm_init:
-	push ax
-	call pm_info
-	pop ax
-	jnc short @@initf0		; error?
-	ret				; yup, abort
+;	push ax
+;	call pm_info
+;	pop ax
+;	jnc short @@initf0		; error?
+;	ret				; yup, abort

  @@initf0:				; no error, init protected mode
  	test ax,2
@@ -2902,10 +2902,10 @@
  	add bx,bx
  	call [exitrouttbl+bx]		; deinit

-	mov es,[privatedataseg]
-	mov ah,49h
-	int 21h				; free private data
-	call [pm_exit]			; user defined cleanup routine
+;	mov es,[privatedataseg]
+;	mov ah,49h
+;	int 21h				; free private data
+;	call [pm_exit]			; user defined cleanup routine
  	pop ax
  	int 21h				; quit

diff -Naur PMODE/pmodetsr.asm PMODE_NEW/pmodetsr.asm
--- PMODE/pmodetsr.asm	1997-02-22 22:18:24.000000000 -0500
+++ PMODE_NEW/pmodetsr.asm	2014-02-17 13:52:50.000000000 -0500
@@ -43,8 +43,8 @@
  ; restore old int 2f vector since pmode is not reentrable and looks at  
int 2f
  	xor	ax,ax
  	mov	ds,ax
-	mov	eax, dword ptr cs:oldint2f
-	mov	ds:[2fh*4], eax		; atomic, cli not needed
+;	mov	eax, dword ptr cs:oldint2f
+;	mov	ds:[2fh*4], eax		; atomic, cli not needed

  	push	DGROUP
  	pop	ds


----------BUILD.BAT

del *.obj
del *.map
del *.exe

gcc -O2 -s padsec.c -o padsec.exe
gcc -O2 -s ehdrfix.c -o ehdrfix.exe

tasm -ml -zi -zn pmode.asm
tasm -ml -zi -zn pmodstub.asm
tasm -ml -zi -zn pmodetsr.asm

tlink -n -c -m -s pmodstub.obj pmode.obj
padsec pmodstub.exe

tlink -n -c -m -s pmodetsr.obj pmode.obj
ehdrfix pmodetsr.exe

----------

HTH,


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#1758

On Mon, 17 Feb 2014 16:59:38 -0500, Rod Pemberton  
<dont_use_email@xnohavenotit.cnm> wrote:
> On Sat, 15 Feb 2014 04:57:39 -0500, Rod Pemberton  
> <dont_use_email@xnohavenotit.cnm> wrote:

>> [...]
>>
>> PMODEDJ's documentation says it will only stay resident
>> one time.  Is there a switch, parameter, hidden switch,
>> tweak, patch, etc that can be used to make PMODEDJ
>> (pmodetsr.exe) stay resident longer?  I.e., permanently.
>
> The patch below allow you to build a resident version of
> PMODEDJ.  PMODEDJ is a.k.a. PMODETSR.EXE.
>
> This version is experimental.  Use at your own discretion.
>

Update:

After a trivial amount of use, it seems this modified version
of PMODETSR stays resident and can be used repeatedly for
single-use execution of DPMI applications, i.e., one DPMI application
after another.  *But*, it *seems* it's not reentrant.  When a PM
TSR is loaded, executing another DPMI application causes a lockup.
I haven't checked to see if there are any memory leaks etc.
I probably won't be doing much more with this.  I think it has
served it's purpose.  Perhaps, someone with more familiarity with
PMODETSR can add a '-r' resident or '-p' permanent flag in the future.


Rod Pemberton

[toc] | [prev] | [standalone]

Back to top | Article view | comp.os.msdos.djgpp

csiph-web