DOS2linux

DOS2linux

[Bart Vandewoestyne]

I have a routine from a DOS driver that looks like this:

static int getslotinfo( void )
{
  static char buff[320], *s=&buff[0]; int valid;

  inregs.h.ah=0xd8; inregs.h.al=0x1; inregs.h.cl=DiSC_Id.slot>>12;
inregs.h.ch=0;
  sregs.ds=FP_SEG(s); inregs.x.si=FP_OFF(s);
  int86x(0x15, &inregs, &outregs, &sregs);
  valid=outregs.h.ah;
  if(!valid) { DiSC_Id.it=buff[itconf]; DiSC_Id.dma=buff[dmachd]; }
  return(valid);
}

(full DOS-code is at http://mc303.ulyssis.org/heim/downloads/DISCDRV.C
)

Doing some research learned me that this piece of code does the
following things (according to http://www.ctyme.com/intr/rb-1641.htm
):

1) set AX register to 0xd800
2) set slot number to DiSC_Id.slot, (eg. 1 in my case -> base is
0x1000)
3) set function number to read
4) assign a 320-byte buffer for standard configuration data block
5) execute a software interrupt via the DOS specific int86x function,
this puts configuration data into the 320-byte buffer.
6) check if we get a valid return
7) if we have a valid situation, assign values from the configuration
block to DiSC_Id.it (it level) and DiSC_Id.dma (dma level)


So here's my question:

On http://www.ctyme.com/intr/rb-1641.htm I can see that this is all
about reading data from an EISA SYSTEM ROM.  I can't imagine there
doesn't exist some linux-API that allows me to do just the same.

What function calls and header files should I use in order to read
this 'EISA SYSTEM ROM' and assign the correct values to DiSC_Id.it and
DiSC_Id.dma ?

If there doesn't exist an API for this, what memory ranges should i
probe in order to get these values?


Thanks for answers,
mc303

-- 
Ing. Bart Vandewoestyne			 Bart.Vandewoestyne@pandora.be
Hugo Verrieststraat 48			       GSM: +32 (0)478 397 697
B-8550 Zwevegem			 http://users.pandora.be/vandewoestyne
----------------------------------------------------------------------
"Any fool can know, the point is to understand." - Albert Einstein

Re: DOS2linux

[Randy.Dunlap]

Bart Vandewoestyne wrote:
> 
> I have a routine from a DOS driver that looks like this:
> 
> static int getslotinfo( void )
> {
>   static char buff[320], *s=&buff[0]; int valid;
> 
>   inregs.h.ah=0xd8; inregs.h.al=0x1; inregs.h.cl=DiSC_Id.slot>>12;
> inregs.h.ch=0;
>   sregs.ds=FP_SEG(s); inregs.x.si=FP_OFF(s);
>   int86x(0x15, &inregs, &outregs, &sregs);
>   valid=outregs.h.ah;
>   if(!valid) { DiSC_Id.it=buff[itconf]; DiSC_Id.dma=buff[dmachd]; }
>   return(valid);
> }
> 
> (full DOS-code is at http://mc303.ulyssis.org/heim/downloads/DISCDRV.C
> )
> 
> Doing some research learned me that this piece of code does the
> following things (according to http://www.ctyme.com/intr/rb-1641.htm
> ):
> 
> 1) set AX register to 0xd800
                        ^^^^^^ actually to 0xd801

> 2) set slot number to DiSC_Id.slot, (eg. 1 in my case -> base is
> 0x1000)
> 3) set function number to read
> 4) assign a 320-byte buffer for standard configuration data block
> 5) execute a software interrupt via the DOS specific int86x function,
> this puts configuration data into the 320-byte buffer.
> 6) check if we get a valid return
> 7) if we have a valid situation, assign values from the configuration
> block to DiSC_Id.it (it level) and DiSC_Id.dma (dma level)
> 
> So here's my question:
> 
> On http://www.ctyme.com/intr/rb-1641.htm I can see that this is all
> about reading data from an EISA SYSTEM ROM.  I can't imagine there
> doesn't exist some linux-API that allows me to do just the same.

I don't have any direct EISA experience on Linux, but the
Linux Device Drivers book (remember that one?) says:

"EISA devices are configured by software, but they don't need any
particular operating system
support. EISA drivers already exist in the Linux kernel for Ethernet
devices and SCSI controllers."

and

"As far as the driver is concerned, there is no special support for
EISA in the kernel, and the
programmer must deal with ISA extensions by himself. The driver uses
standard EISA I/O
operations to access the EISA registers. The drivers that are already
in the kernel can be used as
sample code."

See (and read) http://www.xml.com/ldd/chapter/book/ch15.html
and some drivers that already do this.


> What function calls and header files should I use in order to read
> this 'EISA SYSTEM ROM' and assign the correct values to DiSC_Id.it and
> DiSC_Id.dma ?
> 
> If there doesn't exist an API for this, what memory ranges should i
> probe in order to get these values?

~Randy

Re: DOS2linux

[Alan Cox]

> 1) set AX register to 0xd800
> 2) set slot number to DiSC_Id.slot, (eg. 1 in my case -> base is
> 0x1000)
> 3) set function number to read
> 4) assign a 320-byte buffer for standard configuration data block
> 5) execute a software interrupt via the DOS specific int86x function,
> this puts configuration data into the 320-byte buffer.
> 6) check if we get a valid return
> 7) if we have a valid situation, assign values from the configuration
> block to DiSC_Id.it (it level) and DiSC_Id.dma (dma level)

Oh god, thats the deep magic EISA weirdness department

> On http://www.ctyme.com/intr/rb-1641.htm I can see that this is all
> about reading data from an EISA SYSTEM ROM.  I can't imagine there
> doesn't exist some linux-API that allows me to do just the same.

Well actually its one of those things that needs writing cleanly but 
currently appears in its own form in some EISA drivers

EISA slots are I/O mapped at 0x1000, 0x2000, 0x3000, 0x4000 -> 0x8000
The ID port is at base+0xc80
Configuration data follows at base+0xc84, 0xc88 ...

I would assume the 320 byte buffer is providing this same data block, and
maybe more but I don't know the details. I thought EISA boards had gone
away

Alan

Re: DOS2linux

[Bart Vandewoestyne]

Alan Cox wrote:
> 
> Oh god, thats the deep magic EISA weirdness department

*gnarf* ;-)  Not good for a beginner like me I assume? :-(

> Well actually its one of those things that needs writing cleanly but
> currently appears in its own form in some EISA drivers

Hmm... with 'writing cleanly' you mean that there should become things
available like eisa_register_device() etc...?

> EISA slots are I/O mapped at 0x1000, 0x2000, 0x3000, 0x4000 -> 0x8000
> The ID port is at base+0xc80
> Configuration data follows at base+0xc84, 0xc88 ...

Yep, that was also what I figured out.

> I would assume the 320 byte buffer is providing this same data block, and
> maybe more but I don't know the details.

That is also what I think, but the problem is that I don't know at
which offset to look for that data...
If you look at the code:

static int getslotinfo( void )
{
  static char buff[320], *s=&buff[0]; int valid;

  inregs.h.ah=0xd8; inregs.h.al=0x1; inregs.h.cl=DiSC_Id.slot>>12;
inregs.h.ch=0;
  sregs.ds=FP_SEG(s); inregs.x.si=FP_OFF(s);
  int86x(0x15, &inregs, &outregs, &sregs);
  valid=outregs.h.ah;
  if(!valid) { DiSC_Id.it=buff[itconf]; DiSC_Id.dma=buff[dmachd]; }
  return(valid);
}

Would it help if i told you that itconf and dmachd are defined as (see
http://mc303.ulyssis.org/heim/downloads/DISCINC.H )

#define itconf                0xb2
#define dmachd                0xc0

So if my EISA board is at 0x1000, i should be able to read these
values from 0x1000+0xb2 and 0x1000+0xc0 ???  And if 'yes', any idea
about how to read them? (byte, word, long...? My guess would be as a
byte, but I'm not sure...)

> I thought EISA boards had gone away

Unfortunately... the device I am trying to write a driver for is
especially designed for our university.  Not many of those boards
exist in the world I guess :-(
(I hope that last sentence didn't take away your interest in my
project ;-)

Greetzzz,
mc303

-- 
Ing. Bart Vandewoestyne			 Bart.Vandewoestyne@pandora.be
Hugo Verrieststraat 48			       GSM: +32 (0)478 397 697
B-8550 Zwevegem			 http://users.pandora.be/vandewoestyne
----------------------------------------------------------------------
"Any fool can know, the point is to understand." - Albert Einstein

Re: DOS2linux

[Alan Cox]

> Hmm... with 'writing cleanly' you mean that there should become things
> available like eisa_register_device() etc...?

Yeah

> > EISA slots are I/O mapped at 0x1000, 0x2000, 0x3000, 0x4000 -> 0x8000
> > The ID port is at base+0xc80
> > Configuration data follows at base+0xc84, 0xc88 ...
> 
> Yep, that was also what I figured out.
> 
> > I would assume the 320 byte buffer is providing this same data block, and
> > maybe more but I don't know the details.
> 
> That is also what I think, but the problem is that I don't know at
> which offset to look for that data...
> If you look at the code:
> 
> Would it help if i told you that itconf and dmachd are defined as (see
> http://mc303.ulyssis.org/heim/downloads/DISCINC.H )
> 
> #define itconf                0xb2
> #define dmachd                0xc0
> 
> So if my EISA board is at 0x1000, i should be able to read these
> values from 0x1000+0xb2 and 0x1000+0xc0 ???  And if 'yes', any idea
> about how to read them? (byte, word, long...? My guess would be as a
> byte, but I'm not sure...)
> 

Wild guess. Try iobase + 0xC80 to check if the card is in that slot, then
check inb(iobase+0xC80+0xb2) and see if that gives sane answers

RE: DOS2linux

[Camiel Vanderhoeven]

Bart Vandewoesteyne wrote:
> That is also what I think, but the problem is that I don't know at
> which offset to look for that data...
> If you look at the code:
> 
> static int getslotinfo( void )
> {
>   static char buff[320], *s=&buff[0]; int valid;
> 
>   inregs.h.ah=0xd8; inregs.h.al=0x1; inregs.h.cl=DiSC_Id.slot>>12;
> inregs.h.ch=0;
>   sregs.ds=FP_SEG(s); inregs.x.si=FP_OFF(s);
>   int86x(0x15, &inregs, &outregs, &sregs);
>   valid=outregs.h.ah;
>   if(!valid) { DiSC_Id.it=buff[itconf]; DiSC_Id.dma=buff[dmachd]; }
>   return(valid);
> }
> 
> Would it help if i told you that itconf and dmachd are defined as (see
> http://mc303.ulyssis.org/heim/downloads/DISCINC.H )
> 
> #define itconf                0xb2
> #define dmachd                0xc0
> 
> So if my EISA board is at 0x1000, i should be able to read these
> values from 0x1000+0xb2 and 0x1000+0xc0 ???  And if 'yes', any idea
> about how to read them? (byte, word, long...? My guess would be as a
> byte, but I'm not sure...)

Looking at the above piece of sourcecode, I would say your guess is
correct. "buff" is declared as an array of chars (bytes), so I would try
to read a byte at 0x10b2 and at 0x10c0. Just try it out & see what you
get back...

Camiel.

Re: DOS2linux

[Bart Vandewoestyne]

Camiel Vanderhoeven wrote:
> 
> > static int getslotinfo( void )
> > {
> >   static char buff[320], *s=&buff[0]; int valid;
> >
> >   inregs.h.ah=0xd8; inregs.h.al=0x1; inregs.h.cl=DiSC_Id.slot>>12;
> > inregs.h.ch=0;
> >   sregs.ds=FP_SEG(s); inregs.x.si=FP_OFF(s);
> >   int86x(0x15, &inregs, &outregs, &sregs);
> >   valid=outregs.h.ah;
> >   if(!valid) { DiSC_Id.it=buff[itconf]; DiSC_Id.dma=buff[dmachd]; }
> >   return(valid);
> > }
> >
> > Would it help if i told you that itconf and dmachd are defined as (see
> > http://mc303.ulyssis.org/heim/downloads/DISCINC.H )
> >
> > #define itconf                0xb2
> > #define dmachd                0xc0
> >
> > So if my EISA board is at 0x1000, i should be able to read these
> > values from 0x1000+0xb2 and 0x1000+0xc0 ???  And if 'yes', any idea
> > about how to read them? (byte, word, long...? My guess would be as a
> > byte, but I'm not sure...)
> 
> Looking at the above piece of sourcecode, I would say your guess is
> correct. "buff" is declared as an array of chars (bytes), so I would try
> to read a byte at 0x10b2 and at 0x10c0. Just try it out & see what you
> get back...

I tried reading bytes from the following locations:

0x1000+0xb2		-> gives me 255
0x1000+0xc80+0xb2	-> gives me 255
0x1000+0xc84+0xb2	-> gives me 255

same for 0xc0

I guess these aren't the values I should be expecting... :-(

Greetzzz,
mc303

-- 
Ing. Bart Vandewoestyne			 Bart.Vandewoestyne@pandora.be
Hugo Verrieststraat 48			       GSM: +32 (0)478 397 697
B-8550 Zwevegem			 http://users.pandora.be/vandewoestyne
----------------------------------------------------------------------
"Any fool can know, the point is to understand." - Albert Einstein

RE: DOS2linux

[Camiel Vanderhoeven]

Bart Vandewoestyne wrote:
>
> I tried reading bytes from the following locations:
> 
> 0x1000+0xb2		-> gives me 255
> 0x1000+0xc80+0xb2	-> gives me 255
> 0x1000+0xc84+0xb2	-> gives me 255
> 
> same for 0xc0
> 
> I guess these aren't the values I should be expecting... :-(

Did you do a normal read from memory, or did you use port-i/o? I think
you should use the latter. Of course, your card could be at 2000h in
stead of 1000h, or at X000h for that matter. I'm not very familiar with
the EISA architecture, but I do know that each card can use the
following I/O ranges:
X000h-X0FFh; X400h-X4FFh; X800h-X8FFh; XC00-XCFF, where X is the slot
number. There is a book on the EISA architecture available free of
charge as a PDF file from www.mindshare.com/pdf/eisabook.pdf. Perhaps
you'll find what you need to know from studying that. I hope this helps.

Veel success ermee,

Camiel.

Re: DOS2linux

[Bart Vandewoestyne]

Camiel Vanderhoeven wrote:
> 
> Did you do a normal read from memory, or did you use port-i/o?

I used the inb() function.

> you should use the latter. Of course, your card could be at 2000h in
> stead of 1000h, or at X000h for that matter.

My card is at 1000h, that's something i know for sure, because I can
probe the EISA ID at 0x1000+0xc80.

> I'm not very familiar with
> the EISA architecture, but I do know that each card can use the
> following I/O ranges:
> X000h-X0FFh; X400h-X4FFh; X800h-X8FFh; XC00-XCFF, where X is the slot
> number.

I guess you mean X0000h-X0FFF; X1000-X1FFF; ...

> There is a book on the EISA architecture available free of
> charge as a PDF file from www.mindshare.com/pdf/eisabook.pdf. Perhaps
> you'll find what you need to know from studying that. I hope this helps.

Tnx for the tip!  I will check it out!

I also found this URL: http://uw7doc.sco.com/cgi-bin/man/man?eisa+D4

It comes from UnixWare 7 documentation and there they have the kind of
translation that I want to do (that is: translate INT 15h call "Read
Function" (AH=D8h, AL=01h)) to linux.  As i understood there isn't
such thing available for linux?  Meaning I'll have to try and
implement that stuff myself?  But then the problem remains: how do i
get to the data that is in the 320 byte buffer returned from an INT
15h call "Read Function" (AH=D8h, AL=01h)
 

Greetzzz,
mc303

-- 
Ing. Bart Vandewoestyne			 Bart.Vandewoestyne@pandora.be
Hugo Verrieststraat 48			       GSM: +32 (0)478 397 697
B-8550 Zwevegem			 http://users.pandora.be/vandewoestyne
----------------------------------------------------------------------
"Any fool can know, the point is to understand." - Albert Einstein

Re: DOS2linux

[Brian Gerst]

Bart Vandewoestyne wrote:
> I also found this URL: http://uw7doc.sco.com/cgi-bin/man/man?eisa+D4
> 
> It comes from UnixWare 7 documentation and there they have the kind of
> translation that I want to do (that is: translate INT 15h call "Read
> Function" (AH=D8h, AL=01h)) to linux.  As i understood there isn't
> such thing available for linux?  Meaning I'll have to try and
> implement that stuff myself?  But then the problem remains: how do i
> get to the data that is in the 320 byte buffer returned from an INT
> 15h call "Read Function" (AH=D8h, AL=01h)

Basically, nobody ever implemented an EISA bus layer because it is rare
to see EISA hardware anymore.  The few EISA aware drivers in Linux
handle configuration themselves.  I had thought about doing it at one
time for the sake of completeness, but I don't have access to any EISA
hardware to test.

--

				Brian Gerst

Re: DOS2linux

[Dr. Kelsey Hudson]

On Mon, 27 Aug 2001, Alan Cox wrote:

> I would assume the 320 byte buffer is providing this same data block, and
> maybe more but I don't know the details. I thought EISA boards had gone
> away

oh come on alan, you know full well nothing ever goes away :) eisa should,
by all means (as should mca) but as long as the boards and machines still
exist you can expect them to rear their ugly heads and bite, especially
when you least expect it.

eisa is evil!

 Kelsey Hudson                                           khudson@ctica.com
 Software Engineer
 Compendium Technologies, Inc                               (619) 725-0771
---------------------------------------------------------------------------

Re: DOS2linux

[Camiel Vanderhoeven]

From: Bart Vandewoestyne <Bart.Vandewoestyne@pandora.be>
> > Did you do a normal read from memory, or did you use port-i/o?
>
>I used the inb() function.

Sorry for asking, I just wanted to make sure it wasn't something simple...

>My card is at 1000h, that's something i know for sure, because I can
>probe the EISA ID at 0x1000+0xc80.

Good.

> > I'm not very familiar with
> > the EISA architecture, but I do know that each card can use the
> > following I/O ranges:
> > X000h-X0FFh; X400h-X4FFh; X800h-X8FFh; XC00-XCFF, where X is the slot
> > number.
>
>I guess you mean X0000h-X0FFF; X1000-X1FFF; ...

No, I don't. Each EISA slot occupies four 256-byte ranges. That is because 
any address that looks like X100h-X3ffh; X500h-X7ffh etc. is misinterpreted 
by older ISA hardware.

Camiel.

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp

RE: DOS2linux

[Camiel Vanderhoeven]

One additional note: it looks like the 320 bytes you are looking for are
part of the NVRAM on the system board. Where this is stored exactly may
very well be system-specific, I'm afraid...

Camiel.

Dos2Linux

[Bart Vandewoestyne]

How do I translate the following piece of DOS-code to linux?

static union {
  unsigned int *a;      // One 32 bits address
  unsigned long  l;     // One 32 bits long
  unsigned int w[2];    // Two 16 bits words
  unsigned char  b[4];  // Four 8 bits bytes
} DMAaddr;

static union {
  signed int w;         // One 16 bits words
  signed char  b[2];    // Two 8 bits bytes
} DMAcntr;

...

static void setadr( unsigned int far *buff, unsigned int length )
{
  unsigned int lw;

  lw = FP_SEG( buff );                // Segment address of buffer
  DMAaddr.w[1] = ( lw >> 12 ) & 0xf;    // Makes real 32bit address
  DMAaddr.w[0] = ( lw << 4 ) & 0xfff0;
  DMAaddr.l += ( unsigned long )FP_OFF( buff );
  DMAcntr.w = length;
}


Thanks,
mc303

-- 
Ing. Bart Vandewoestyne			 Bart.Vandewoestyne@pandora.be
Hugo Verrieststraat 48			       GSM: +32 (0)478 397 697
B-8550 Zwevegem			 http://users.pandora.be/vandewoestyne
----------------------------------------------------------------------
"Any fool can know, the point is to understand." - Albert Einstein