Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!sun-barr!newstop!sun!imagen!qmsseq!pipkins
From: pipkins@qmsseq.imagen.com (Jeff Pipkins)
Newsgroups: comp.sys.ibm.pc
Subject: Re: Need Help with ROM/empty address distinction
Message-ID: <90@qmsseq.imagen.com>
Date: 16 Jan 90 17:16:28 GMT
References: <1990Jan15.162053.14109@ee.rochester.edu>
Reply-To: pipkins@qmsseq.UUCP (Jeff Pipkins)
Organization: QMS Inc., Mobile, Alabama
Lines: 66

In article <1990Jan15.162053.14109@ee.rochester.edu> jal@ee.rochester.edu writes:
>I am trying to write a program which can tell me what is going at
>at various memory locations.  I am trying to distinguish between 
>RAM,  ROM and nothing at all addresses.  Now I have no problem
>distinguishing between RAM and ROM,  but how do I distinguish between
>ROM and nothing.  In my experiments nothing sometimes is read as the
>same hex data and other times the data changes between reads.

The following code is an excerpt from a program similar to the one you
describe.  It works because of a convention set forth in the IBM AT Tech
Ref. Manual that says what add-on rom modules must look like in order
to be recognized by the BIOS.  Because add-on rom is only allowed from
C800:0000 to E000:0000, the following statement is used to call the
check_rom() function:

   puts("\nChecking for added ROM from C8000 to E0000...");
   check_rom((addr) 0xC0008000, (addr) 0xE0000000);

Here is the code for check_rom(), along with a function that it needs:

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
 *                                                                           *
 *   This routine checks for optional ROM code from "address" to "limit"     *
 *                                                                           *
 *   Each 2K block from "address" up to "limit" has the potential to contain *
 *   ROM.  If ROM is present, the first word will be AA55 and the next byte  *
 *   will contain the number of 512 byte blocks.                             *
 *                                                                           *
 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */

void check_rom(address, limit)
   unsigned int huge *address;
   unsigned int huge *limit;

{
   unsigned int blocks;

   while (address < limit) {
      if (*address == 0xAA55) {
         blocks = *(address+1) & 0x00FF;
         printf("   %lX: %uK bytes.\n", adr8086(address), blocks>>1);
         address += (blocks * 512)/2;
      }else {
         address += 2048/2;
      }
   }
}

/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
 *                                                                           *
 *   This function takes as input a 32-bit long and returns another 32-bit   *
 *   long that is the screwed-up 8086 address version.  In other words,      *
 *   the high word is shifted left 4 bits and then added to the low word     *
 *   to produce a 20-bit address.                                            *
 *                                                                           *
 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */

unsigned long adr8086( adr)
   unsigned long adr;

{
   return( ((adr>>12)&0x000FFFF0L) + (adr&0x0000FFFFL) );
}

Any opinions expressed here (including the expression "screwed-up") are
my own, and not necessarily shared by my employer.  So there.