DFB1 Support thread

[Badwolf]

This is what MAPROM2.2 does:-

Code: Select all

TTRAM_Install:
; inputs:
;	d0: size
;--------------------------------------------------------------
	move.l	d0,-(sp)			; size
	move.l	#TTRL_MEMBOTTOM,-(sp)		; start
	move.w	#20,-(sp)			; maddalt(size, start)
	trap	#1				; 
	add.l	#6,sp				; keep size on stack
	tst.l	d0				; tos supports altram?
	bne		.fail

	move.w	#0,-(sp)			; allocate st-ram only
	move.l	#$10000,-(sp)			; 64Kb			
	move.w	#68,-(sp)			; Mxalloc
	trap	#1
	addq.l	#8,sp
	tst.l	d0				; failed to allocate?
	beq		.fail
	cmp.l	#-32,d0				; mxalloc not available?
	beq		.fail
	
	move.l	d0,d1				; 64Kb fastram buffer
	move.l	#$5f465242,d0			; '_FRB'
	bsr 	CK_WriteJar			; write cookie

	move.l	(sp)+,d0
	move.l	#TTRL_MEMBOTTOM,d1	
	add.l	d0,d1				; d1 = memtop
	move.l	d1,$5a4.w			; write fastram top
	move.l	#TTRL_FASTRAMMAGIC,$5a8.w	; write fastram magic
	rts
.fail:
	addq.l	#4,sp
	moveq.l	#0,d0
	rts

Which, assuming cookie writing succeeds -- and it extends the jar if not -- looks ok to me.

Crucially, which is what I forgot when I wrote my first AltRAM declaration program, is to TSR.

Code: Select all

SYS_QuitKeepResident:
;--------------------------------------------------------------
	move.w	#0,-(sp)		; Ptermres()
	lea		gSYS_TsrSize(pc),a0
	move.l	(a0),-(sp)
	move.w	#49,-(sp)
	trap	#1
	addq.l	#8,sp

gSYS_TsrSize is initialsied here:

Code: Select all

_entrypoint
;--------------------------------------------------------------
  	move.l	4(sp),a0		; a0 = basepage
  	lea		gSYS_Stack(pc),sp	; initalize stack
	move.l	#100,d0			; basepage size
	add.l	#$1000,d0		; 
  	add.l	$c(a0),d0		; text size
  	add.l	$14(a0),d0		; data size
  	add.l	$1c(a0),d0		; bss size
	lea		gSYS_TsrSize(pc),a1
	move.l	d0,(a1)

Now, I don't know enough about TSRs, nor assembly, nor TOS to know a) what's happening here and b) what happens to the Mxalloc()'d memory for the FRB at this point. Does Ptermres() stop the release of all memory Mxalloc()'d by a process?

'Cos if not, I can see a problem.

BW

[Badwolf]

From what I gather, HDDriver may even install its own (this could be confirmed by booting clean with it and inspecting the cookie list with the CPX). Assuming it does that, subsequently ignoring its own _FRB would be a bug.

...

And my own TK040 driver checks for 'XFRB', which is a sized version of _FRB. I don't remember what software defines that one.

I stumbled across that one the other day and I think XFRB is an HDDriver construct.

Which implies it ought to know better.

I wonder what happens if HDDriver is loaded from floppy *after* TT-RAM is declared?



I did try to buy a SCSI adapter to check this out myself but couldn't find anything on EBay to plug into the Falcon and give a 50pin IDC port. Anyone any pointers?

BW

[stephen_usher]

I did try to buy a SCSI adapter to check this out myself but couldn't find anything on EBay to plug into the Falcon and give a 50pin IDC port. Anyone any pointers?

This would do it. You'd need a terminator.

https://www.ebay.co.uk/itm/134126945703

The name of the connector you're looking for is "HD50" by the way.

You could buy one of these HD50 to DB50 cables and find a female IDC connector and make up the rest of the cable.

https://www.ebay.co.uk/itm/304136985036

[Badwolf]

This would do it. You'd need a terminator.
https://www.ebay.co.uk/itm/134126945703

Eeek! Iw as actually thinking something more like this:- https://www.ebay.co.uk/itm/392236581097

But we need Male rather than Female, IIRC.

BW

[mfro]

... Does Ptermres() stop the release of all memory Mxalloc()'d by a process?...

'Cos if not, I can see a problem.

BW

No worries, that's what Ptermres() does. It's not only "stopping the release" but it completely unhooks allocated blocks from memory management. They simply do not exist anymore from the GEMDOS point of view.

[stephen_usher]

Eeek! Iw as actually thinking something more like this:- https://www.ebay.co.uk/itm/392236581097

But we need Male rather than Female, IIRC.

BW

See my later edit to my post...

[stephen_usher]

OK, a DB50 female IDC socket is not cheap at £23! https://uk.rs-online.com/web/p/d-sub-connectors/1210157

P.S.

Alternatively... https://www.ebay.co.uk/itm/230460442947

So the whole set of two cables would cost just over £10

[Badwolf]

No worries, that's what Ptermres() does. It's not only "stopping the release" but it completely unhooks allocated blocks from memory management. They simply do not exist anymore from the GEMDOS point of view.

Grand, thanks, mfro. So it looks to me like MAPROM/FASTRAM is doing the right thing.

I think the only check left is to see if _FRB is actually set.

Although I would be interested in knowing if there's any change to what happens if HDDriver is run *after* MAPROM (eg from floppy).

The more information to report to Uwe the better.

BW

[Badwolf]

Alternatively... https://www.ebay.co.uk/itm/230460442947
So the whole set of two cables would cost just over £10

I was staring wistfully at that one but it's female again.

What are the two combined you referred to? The best combination I could find was an HD50 to HD68 and then an HD68 to IDC50.

But then I thought life's too short! :lol:

BW

[stephen_usher]

Alternatively... https://www.ebay.co.uk/itm/230460442947
So the whole set of two cables would cost just over £10

I was staring wistfully at that one but it's female again.

What are the two combined you referred to? The best combination I could find was an HD50 to HD68 and then an HD68 to IDC50.

But then I thought life's too short! :lol:

BW

This: https://www.ebay.co.uk/itm/304136985036 Plus this: https://www.ebay.co.uk/itm/230460442947