MacSCSI and HFS
Lance Keashly
my Mac. I went the route of the MacSCSI interface from Mr. Bass and hardware
wise it went fine. But software! What A Pain.
The code I got with the interface is pretty poor from the side of trying to
understand how the driver is made. The code that interfaces to the hardware
is OK. I could read it, understand it, and modify it. But the driver itself
is far from helpful, and if you want to make your MacSCSI work with HFS THIS
IS THE AREA of concern.
I understand that the original driver code is from the Aztec C distribution
so I thought I could find some more useful documentation. But NOOOO the
thing is written in this sparsly documented ASSEMBLY CODE (I bought a C
compiler package!) with no C code involved at all! I know C but assembler
takes longer.
Anyway to get the HFS working you need to worry about the control calls the
finder (5.0) will ask of the driver. Apple tech notes says somthing about it
somewhere. The driver MUST RESPOND either do it, pretend to do it, or say it
can't. The open routine MUST ALSO SET THE DRIVER QUEUE ENTRY CORRECTLY. the
Driver as given doesn't even bother. Here's the way I've modified my driver,
look thru it and see what's helpful, I've done crude partitioning that works
and my drive is an ST225 (20 meg). The partioning is done by multiple
drivers that are compiled for different areas of the disk. To get the HFS
on the drive, once the thing is mounted on finder 5.0 with the HD20 stuff
ERASE THE DISK. If things go right it will make it into HFS disk. It does
on mine.
Also, this stuff works great! The disk is as fast as the HYPER's plus I can
run MultiMac with a 10meg partition and finder 4.1 . This combination is GREAT,
I don't really have MultiMac (a bootlegged copy I use only to show what a Mac
can do) but when I find out where to get it I will.
THE CODE: Hacked from the John Bass release.
/*
* MacSCSI partitioned driver routine.
* This version was developed on the Mac under Aztec C.
*
*/
#asm
;:ts=8
;
;
Start
dc.w $4f00 ;locked, read, write, ctrl, status
dc.w 0 ;no delay
dc.w 0 ;no events
dc.w 0 ;no menu
dc.w open-Start ;open routine
dc.w rdwrt-Start ;prime routine
dc.w control-Start ;control routine
dc.w status-Start ;status routine
dc.w tst-Start ;close routine
dc.b 6
dc.b ".part1" ;name of driver
ds 0 ;for alignment
;
;
drvref dc.w 0 ;driver ref num
dc.w 0 ;file system ID
dc.b 0 ;sidedness of disk
dc.b 0 ;needs flush flag
dc.w 0 ;error count
;
diskbase_
dc.l 0 ;base of disk memory
;
openflg
dc.w 0 ;once only open flag
;
;
open
movem.l d0-d7/a0-a6,-(sp) ;save regs
jsr copen_
movem.l (sp)+,d0-d7/a0-a6 ;restore regs
lea openflg,a0
tst.w (a0)
bne skip
st (a0)
; first, fix the drive queue
move.l #18,d0 ;size of drive queue entry
dc.w $a51e ;NewPtr - system heap
move.l #$80000,(a0)+ ; set the hidden flags (NO EJECT)
clr.w 10(a0) ;local file system
movem.l d2-d7/a0-a6,-(sp) ;save regs
jsr get_DS_ ;get the size of the drive
move.l d0,-(sp)
jsr get_DN_ ;get the drive number
move.l (sp)+,d1
movem.l (sp)+,d2-d7/a0-a6 ;restore regs
move.w d1,12(a0) ; HFS WANTS THE SIZE OF THE DRIVE IN BLOCKS
move.w d0,d7 ;d7 has the drive number
swap d0
move.w 24(a1),d0 ;driver reference number
lea drvref,a2 ;get address of driver reference number
move.w d0,(a2) ;set up status buffer
dc.w $a04e ;AddDriver
move.l (a1),-(sp) ;push handle to resource
dc.w $a992 ;DetachResource
link a6,#-50
move.l sp,a0
move.w d7,22(a0)
dc.w $a00f ;DON"T KNOW WHAT THIS IS
unlk a6
skip
move.l #0,d0 ;good return
rts ;all done by INIT
;
rdwrt
movem.l d0-d7/a0-a6,-(sp) ;save regs
jsr ScsiRdWr_ ;Call the C stuff
movem.l (sp)+,d0-d7/a0-a6 ;restore regs
move.l #0,d0
bra tst ;exit via IOdone
;
control
move.l #-17,d0
cmp.w #1,26(a0) ;is it KillIO
bne.s next ;no, exit
move.l #0,d0 ;okay return
rts ;must respond this way
next
cmp.w #5,26(a0) ;either check or erase
beq.s ok ; lie and say we did it
cmp.w #6,26(a0) ;definitly format
bne tst
movem.l d0-d7/a0-a6,-(sp) ;save regs
jsr cformat_ ; do it if you want I don't
movem.l (sp)+,d0-d7/a0-a6 ;restore regs
move.l #0,d0
ok
move.l #0,d0 ;say we did it OK
bra tst
;
status
move.l #-18,d0 ;we don't do status so say so
;
tst
movem.l d4-d7/a4-a6,-(sp) ;save regs going into IOdone
move.l $8fc,a0 ;get IOdone address
jsr (a0) ;call IOdone
movem.l (sp)+,d4-d7/a4-a6 ;restore them after IOdone
rts ;and jump to it
;
public _Uend_,_Dorg_,_Cend_
save_
lea Start+(_Uend_-_Dorg_)+(_Cend_-Start),a4 ; setup baseadr
move.l a0,Pbp_
move.l a1,Dp_ ;save arguments
rts
restore_
move.l Dp_,a1 ; pass DCEptr
rts
#endasm
#include <memory.h>
#include <resource.h>
#include <desk.h>
#include <pb.h>
/*
* varibles used by asm routines
* Dp and Pbp are device driver arguments handled by save/restore
* ptr and dvrarg needed by AddDriver call
*/
DCEPtr Dp;
ParmBlkPtr Pbp;
/*
* Local Stuff
*/
extern long DrvSize;
extern long FirstBlock;
/*
* ScsiRdWr -- do a driver read/write function.
*/
ScsiRdWr()
{
register struct ioParam *ip;
register long len,part;
register short blkno;
register char *addr;
save();
ip = & Pbp->u.iop;
part = Dp->dCtlPosition & 0xFFFFFE00;
/* this does my partitioning */
part = (part >> 9L) + FirstBlock;
len = (ip->ioReqCount + 511) & 0xFFFFFE00;
addr = ip->ioBuffer;
while(len >= 512) {
blkno = part;
if((Pbp->ioTrap & 0xff) == 2) {
ScsiRead(blkno, addr);
} else {
ScsiWrite(blkno, addr);
}
len -= 512;
addr += 512;
part += 1;
}
ip->ioActCount = ip->ioReqCount;
restore(); /* exit to I/O Done */
}
copen()
{
/* these are the c routines that are called on open */
save();
/* make sure the interface is reset, dangerous */
ScsiReset();
/* init the disk controller for the st225 */
init_st();
restore();
}
cformat()
{
/* these are the c routines that are called on format(initialize)*/
save();
/* do nothing right now */
/* Scsifmt(); */
restore();
}
long get_DS()
{
long x;
/* returns the size of the drive */
save();
x = DrvSize;
restore();
return(x);
}
int get_DN()
{
int x;
/* returns the drive number */
save();
x = DRVNUM ;
restore();
return (x);
}
--
Lance Keashly
New Media Technologies Ltd.
..decvax!microsoft!ubc-vision!winston!keashly
..ihnp4!alberta!ubc-vision!winston!keashly
#108 4664 Lougheed Highway
Burnaby, B.C., Canada,
V5C 5T5
(604) 291-7111
Rick Watson
Ok, here is a bit of my MacSCSI code. This will add a customizable icon
for your MacSCSI. A get info for the disk (and other places) will say
"SCSI" instead of "AppleTalk" for "Where". Note that I use Megamax C, so
you may have to do some hacking to get back to Aztec C. To re-customize
the icon you will have to create the icon, decompile it with REdit, and
edit the resulting hex back into the following code.
Rick Watson
University of Texas Computation Center
arpa: ri...@ngp.UTEXAS.EDU ri...@ngp.ARPA
uucp: ...seismo!ut-sally!ut-ngp!rick ri...@ut-ngp.UUCP
bitnet: ccaw001@utadnx
phone: 512/471-3241
goes into np_drv.c:
/* io parameter block */
#define io_VRefNum 22 /* volume reference number */
#define io_CSCode 26 /* code */
#define io_csParam 28 /* cs parameters */
#define io_lun 28 /* 2 cs logical unit number */
#define io_sect 30 /* 2 cs sector number */
#define io_addr 32 /* 4 cs buffer address */
.
.
.
control:
cmpi.w #21, io_CSCode(a0)
beq.s cntrl1 /* if Finder 4.1 get icon call */
.
.
.
cntrl1:
clr.l d0
move.l a1, -(sp)
lea icnlst, a1 /* pass icon list address */
move.l a1, io_csParam(a0)
move.l (sp)+, a1
bra tst
.
.
.
icnlst:
dc.w 0x0000,0x0000,0x0000,0x0000,0x03FF,0xFFF8
dc.w 0x0600,0x0018,0x0C00,0x0038,0x1800,0x0068
dc.w 0x3000,0x00C8,0x6000,0x0188,0xC000,0x0308
dc.w 0xFFFF,0xFE08,0xFFFF,0xFE08,0xC000,0x0608
dc.w 0xC000,0x0608,0xC000,0x0608,0xC000,0x0608
dc.w 0xC000,0x0608,0xC000,0x0608,0xC001,0xC608
dc.w 0xC001,0xC608,0xC000,0x0608,0xC000,0x0608
dc.w 0xC000,0x0618,0xC000,0x0630,0xC000,0x0660
dc.w 0xC000,0x06C0,0xFFFF,0xFF80,0xFFFF,0xFE00
dc.w 0x0000,0x0000,0x0000,0x0000,0x0000,0x0000
dc.w 0x0000,0x0000,0x0000,0x0000
dc.w 0x0000,0x0000,0x0000,0x0000,0x03FF,0xFFF8
dc.w 0x07FF,0xFFF8,0x0FFF,0xFFF8,0x1FFF,0xFFF8
dc.w 0x3FFF,0xFFF8,0x7FFF,0xFFF8,0xFFFF,0xFFF8
dc.w 0xFFFF,0xFFF8,0xFFFF,0xFFF8,0xFFFF,0xFFF8
dc.w 0xFFFF,0xFFF8,0xFFFF,0xFFF8,0xFFFF,0xFFF8
dc.w 0xFFFF,0xFFF8,0xFFFF,0xFFF8,0xFFFF,0xFFF8
dc.w 0xFFFF,0xFFF8,0xFFFF,0xFFF8,0xFFFF,0xFFF8
dc.w 0xFFFF,0xFFF8,0xFFFF,0xFFF0,0xFFFF,0xFFE0
dc.w 0xFFFF,0xFFC0,0xFFFF,0xFF80,0xFFFF,0xFE00
dc.w 0x0000,0x0000,0x0000,0x0000,0x0000,0x0000
dc.w 0x0000,0x0000,0x0000,0x0000
dc.b 4 /* length of name */
dc.b "SCSI" /* name */
dc.b 0 /* alignment */
John Bass
> my Mac. I went the route of the MacSCSI interface from Mr. Bass and hardware
> wise it went fine. But software! What A Pain.
>
> THE CODE: Hacked from the John Bass release.
>
> /*
> * MacSCSI partitioned driver routine.
> * This version was developed on the Mac under Aztec C.
> *
> */
>
> #asm
> /*
> ..ihnp4!alberta!ubc-vision!winston!keashly
>
> #108 4664 Lougheed Highway
> Burnaby, B.C., Canada,
> V5C 5T5
> (604) 291-7111
The above are segments from Lance Keashly's posting of our Copyrighted
MacSCSI np_dvr.c routine. While I support Lance's desire to share with
others -- his respect for others intellectual property is greatly lacking.
Lance violated the copyright in two ways:
1) removing the copyright notice which SPECIFICALLY prohibits
making or maintaining ANY copies without the copyright notice
included.
2) the removed copyright provided for unrestricted personal
non-comercial use --- any other reproduction requires
our written permission -- including posting to the net or
BBS's.
Copyrights are useful for protection of any written work from a few line
poem or jingle to a full length novel or programming system. We published
the MacSCSI article in DDJ to help the general hobbiest on a tight budget,
not to put others into our business -- thus the need to protect the copyright.
We depend on a much improved version of the software for part of our income.
It has been licensed at a nominal fee to several firms including the
Warp 9 which are now shipping their $895 internal drive via mail order.
Lastly Lance violated the rights of Manx in that the assembly language
segments were derived from the Aztec C Ram Disk -- we obtained their written
permission for publication in the Sept Doctor Dobbs issue and our use.
Thanks to Lance for his willingness to share, but please remove all copies
of the referenced article -- I have requested that Lance repost the article
with the proper copyright notices included.
Overall it is highly recommended to publish changes in a diff format with
minor context ...
John Bass, MacSCSI author and copyright holder.