shurr@cbnews.ATT.COM (Larry A. Shurr) (02/01/90)
In article <7713@pt.cs.cmu.edu> ralf@b.gp.cs.cmu.edu (Ralf Brown) writes: }In article <Jan.25.16.16.09.1990.13042@occlusal.rutgers.edu> koenig@occlusal.rutgers.edu (Evan M. Koenig) writes: }}After playing with 386 protected mode, I finally got real mode to }}work with a 4 gig linear address space. }}If you wan't the source code, I can email it. ^ | Look out! The apostrophe man's gonna get you! :-) }Count me in! All seriousness aside... I, too am interested in the code. This may be worth posting. In any case, I would like to have it. Please e-mail or post. regards, Larry -- Signed: Larry A. Shurr (cbema!las@att.ATT.COM or att!cbema!las) Clever signature, Wonderful wit, Outdo the others, Be a big hit! - Burma Shave (With apologies to the real thing. The above represents my views only.) (You may now R'eply. Forwarding from cbnews to my mail address now works!!!)
harlow@plains.UUCP (Jay B. Harlow) (02/02/90)
In article <40970040@hpindda.HP.COM> kmont@hpindda.HP.COM (Kevin Montgomery) writes: >> This is because your computer most probably does not decode address line >> A20 (at least in real mode). If you want to access the addresses beyond >> 1M you got to turn A20 ON implicitly. > >Ahhhhhg! I must be real confused- of what purpose is ignoring A20? Certain Companies when they released thier NEW 286 Intel machine, for a follow up to thier popular PC machine, decided they NEEDED to mimick it in EVERY aspect. Including the 1M limit wrap around ( i.e. address FFFF:0010 wraps to 0000:0000 ) at least being 'smart' and realizing that even 1M chunks (2nd M, 4th M, 6th M.... ) would be nice to use in any upcoming PROTECTED mode systems they set up a special port to toggle how the hardware sees A20, ( interesting enough the 486 has direct support for this 'feature' ) Of course this same company chose to ignore Intel warning about using INT 0 - INT 1F, any one try using the BOUND inst on a AT ;-). > What boxes do this? Nearly all of them, Must follow the leader ;-). who cares if it is brain-dead..... ( from a reliable system point of view ) > Of course ALL the clone makers, have to follow the leader to maintain compatibility ( or do they...... ;-) Jay -- Jay B. Harlow <harlow@plains.nodak.edu> uunet!plains!harlow (UUCP) harlow@plains (Bitnet) Of course the above is personal opinion, And has no bearing on reality...
Ralf.Brown@B.GP.CS.CMU.EDU (02/02/90)
In article <3288@plains.UUCP>, harlow@plains.UUCP (Jay B. Harlow) wrote: }In article <40970040@hpindda.HP.COM> kmont@hpindda.HP.COM (Kevin Montgomery) }writes: }>> This is because your computer most probably does not decode address line }>> A20 (at least in real mode). If you want to access the addresses beyond }>> 1M you got to turn A20 ON implicitly. }> } Certain Companies when they released thier NEW 286 Intel machine, for }a follow up to thier popular PC machine, decided they NEEDED to }mimick it in EVERY aspect. Including the 1M limit wrap around }( i.e. address FFFF:0010 wraps to 0000:0000 ) at least being 'smart' and Probably because a Certain OS Company actually *used* the 1M wraparound.... (the CP/M-compatible "CALL 5" entry point jumps to 000C0h, but is constrained on HOW it jumps there by the need for PSP:0006h to be the size of the first segment. That's how CP/M programs determined how much space they had, since the jump instruction at location 5 on a CP/M machine goes to the lowest byte of the OS, which is the next byte after the last byte available to the program) -- UUCP: {ucbvax,harvard}!cs.cmu.edu!ralf -=- 412-268-3053 (school) -=- FAX: ask ARPA: ralf@cs.cmu.edu BIT: ralf%cs.cmu.edu@CMUCCVMA FIDO: Ralf Brown 1:129/46 "How to Prove It" by Dana Angluin Disclaimer? I claimed something? 14. proof by importance: A large body of useful consequences all follow from the proposition in question.
feustel@well.UUCP (David Alan Feustel) (02/03/90)
Your problem is almost certainly that address line A20 is gated off.
A bios call is available to turn it on.
--
Phone: 219-482-9631 FAX: by arrangement
E-mail: feustel@well.sf.ca.us
{ucbvax,apple,hplabs,pacbell}!well!feustel kemp@umn-cs.cs.umn.edu (Stuart R. Kemp) (02/05/90)
I am posting this on behalf of Evan Koenig. Please direct
all replies to him.
From koenig@occlusal.rutgers.edu Sat Feb 3 13:14:39 1990
Hello to everyone and thanks for your responses. Since so many people
responded to question about the 386, I'm writing this rather generically.
In sum here's what has happened since I've posted my original question.
Addressing 4gigs in real mode works. I've been using it for over a week
now in my C programs. I load multiple images into expanded(extended?) memory
and do some edge detection work in the expanded memory. I am therefore
absolutely sure that my addressing scheme is good.
The problem as I stated was that I couldn't access the 1 to 2 meg range.
After some further investigation, I found I couldn't get to the 3 to 4
range. The problem? Address line A20 is gated off by the keyboard controller
as so many of you have told me.
R. Kirchner@informatik.uni-kl.de says that you have to write
d1h to port 64h
00h to port 60h
He says that there was a program in a german magazine in regards to this.
Can anyone send it to me? I can understand most of the german.
Any Way, here is the assembly code needed to access 4 gigs. Clip the
assembly portion and save to a file called 4gig.asm. Also included
is a small c program written for Turbo C 2.0 that will test the 32 bit
addressing. Save that file to peek32.c
If anyone gets the A20 addressing line problem solved post it to
comp.sys.ibm.pc and send me some mail.
Good Luck to all and enjoy!!
Evan Koenig koenig@occlusal.rutgers.edu
-------------------------Cut here 4gig.asm---------------------------------
; This program will go into protected mode, return but
; leave fs,gs pointing to a 32 bit descriptor.
; This will result in a 4 gigabyte real mode !!!!
; Just play back what's in memory
; 4 GIG .ASM
;*********************************************************
;Program by Neal Margulis -- Use Masm 5.0, OR TASM V1.0
; This program was taken the article
; 80386 Protected Mode Initialization
; Dr. Dobb's Journal, October 1988
;Note: Basically I deleted all the code that dumps some
; test patterns to the screen in protected mode and
; real mode. After returning from protected mode
; CS,DS,ES point to a real mode descriptor (64k segments)
; But, GS,FS are left pointing to the 4gig segment.
;
; #### DONT EVER PLAY WITH GS OR FS, LEAVE THEM ALONE ####
; #### IF YOU MESS WITH THEM, YOU WILL LOSE ACCESS TO ####
; #### THE 4 GIG RANGE. IF YOU WANT TO LEARN MORE ####
; #### ABOUT THE DESCRIPTORS, I SUGGEST YOU PICK UP A ####
; #### FEW BOOKS ON THE 80386!!! NO ONE PARTICULAR ####
; #### BOOK IS GOOD BY ITSELF. I SUGGEST YOU TRY TO ####
; #### FIGURE AS MUCH AS YOU CAN FROM THIS CODE AND ####
; #### THEN BUY A BOOK. ####
;
; THIS PROGRAM SHOULD BE RUN FIRST. I SUGGEST YOU PUT
; IT IN THE AUTOEXECFILE. AFTER THIS PROGRAM RUNS, YOUR
; ASSEMBLY OR C PROGRAM CAN ACCESS ALL 4GIGS
;
; -EVAN KOENIG koenig@occlusal.rutgers.edu
;
;*********************************************************
.386p
descriptor STRUC
limt_0_15 dw 0
base_0_15 dw 0
base_16_23 db 0
access db 0
gran db 0
base_24_31 db 0
descriptor ENDS
code_seg_access equ 09ah
data_seg_access equ 092h
CSEG segment word use16 'code'
assume cs:CSEG,ds:CSEG
mov ax,CSEG
mov ds,ax ; PUT Code Segment in DS
mov ax,seg PMODE ; Move PMode segment in ax
and eax,0FFFFh ; eax 0000 PMODE Let eax = a040
shl eax,4h ; eax a040 0000
mov ebx,eax ; ebx a040 0000
shr eax,16 ;
mov gdt_PM_1.base_0_15,bx
mov gdt_PM_2.base_0_15,bx
mov gdt_PM_1.base_16_23,al
mov gdt_PM_2.base_16_23,al
mov ax,seg C3
and eax,0ffffh
shl eax,04h
mov ebx,eax
shr eax,16
mov gdt_c3_5.base_0_15,bx
mov gdt_c3_5.base_16_23,al
mov ax,cs
and eax,0ffffh
shl eax,4h
add eax,offset gdttbl
mov dword ptr gdtaddr+2,eax
; NOW LOAD THE DESCRIPTOR TABLE
lgdt gdtaddr
A20_ON:
cld
cli
; Enter Protected Mode
mov eax,cr0
or eax,1
mov cr0,eax
DB 0eah,0h,0h,08h,0h ; jmp to PMODE use descriptor 1
gdtaddr label qword
dw 48
dd ?
dw 0
; global descriptor table
gdttbl label dword
gdt_null descriptor <,,,,,> ; descrip 0
gdt_PM_1 descriptor <0ffffh,,,code_seg_access,0c0h,0> ; descrip 1
gdt_PM_2 descriptor <0ffffh,,,data_seg_access,08fh,0> ; descrip 2
gtd_3 descriptor <0ffffh,0,0,data_seg_access,08fh,0> ; descrip 3
gdt_rm_4 descriptor <0ffffh,0,0,data_seg_access,08fh,0> ; descrip 4
gdt_c3_5 descriptor <0ffffh,,,code_seg_access,080h,0> ; descrip 5
CSEG ends
;//////////////////////////////////////////////////////////////////////////
; HERE IS OUR PROTECTED MODE PROGRAM
;/////////////////////////////////////////////////////////////////////////
PMODE segment para public use32 'code'
assume cs:PMODE
mov ax,18h ; SELECTOR 3 (18h/8h = 3) IS A 4 GIG DATA SEGMENT
mov es,ax ; WITH BASE AT 0
MOV fs,ax ; Let ES and FS point to 4 GiG data segment
mov gs,ax
mov ax,10h ; Set DS to selector 1, but we don't care mov ds,ax ;
; It isn't my code
; JUMP BACK TO REAL MODE
; LEAVE GS POINTING TO A 32 BIT DESCRIPTOR
;
; MAKE A SHORT JUMP. CLEARS OUT PREFETCH QUEUE. JUMP BACK TO REAL MODE
db 0eah ,0h,0h,0h,0h,28h,0h;
align 16
pdat db 0ach
lastpm label dword
PMODE ends
c3 segment para public use16 'code'
assume cs:c3
mov ax,20h ; Selector #4 real mode
mov es,ax ; Set all segment registers back to the real
mov ds,ax ; mode descriptor
; Notice!!! Leave Gs and FS pointing to the 4 gig
; address space. NEVER NEVER NEVER CHANGE IN REAL
; MODE, ELSE 4 GIGS GO BYE BYE
mov eax,cr0 ; Clear protection mode bit
and eax,07ffffffeh
mov cr0,eax
jmp far ptr flushr1
flushr1:
; WELCOME BACK TO REAL MODE
; YOU CAN CALL DOS TO PRINT OUT A MESSAGE HERE IF YOU WANT
; DON'T DO IT IN PROTECTED MODE. THE IDT TABLES AREN'T SET UP!
A20_off:
mov ah,04ch
mov al,01h
int 21h
c3 ends
.8086
end
---------------------------Cut here peek32.c--------------------------------
/* 32 Bit peek and poke */
/* by Evan Koenig */
/* Assumptions: gs and fs point to a 32 bit descriptor */
/* */
/* This code was written for Turbo C V2.0. This program*/
/* should be compiled with TCC and TASM. TCC will */
/* automatically invoke TASM. I guess for other */
/* compilers it might be similar. */
#pragma inline
#include <dos.h>
main()
{
unsigned int segment,offset,ret,ret2;
unsigned long address;
unsigned long ret32,real32;
printf("Enter a 32 bit address to at. Enter it in hex\n");
scanf ("%lx",&address);
asm .386;
asm push esi;
asm mov esi,dword ptr address;
asm mov edx,gs:[esi];
asm mov dword ptr ret32,edx;
asm pop esi;
asm .8086;
printf("32 bit peek returned %lx\n",ret32);
}tonyg@batserver.cs.uq.oz.au (Tony Gedge) (02/06/90)
Has anyone ever thought of doing a similar thing with a 286 processor ? (Dives for cover to avoid the falme onslaught :-) But seriously, would it be possible, or is the 286 too retarded ? ant (Believe it or not, this is not tonyg) -- ------------------------------------------------------------------------- | Computer Science Department, | tonyg@batserver.cs.uq.oz.au | | University of Queensland, Australia.| "CC stands for Cryptic Crossword" | -------------------------------------------------------------------------
yip@ztivax.UUCP (Dr Yeung-Cho Ip) (02/12/90)
In article <1990Feb5.153408.15670@umn-cs.cs.umn.edu> kemp@umn-cs.cs.umn.edu (Stuart R. Kemp) writes: > >If anyone gets the A20 addressing line problem solved post it to >comp.sys.ibm.pc and send me some mail. > Now here is a small program written in TURBO C, which toggels the state of the A20 bit every time it is called. I can't test it with Koenigs program, since I don't have a 386 box. I hope it meets your needs. If you want to set A20 to a constant value, you only have to change the line before the last one in main(). ( put_dat((ea^0x02) | 0xD1); /* toggel a20 bit */) ( ^ ) ( | ) ( This is the bit value of the a20 bit ) Hope that helps Michael Franzen, c/o yip@ztivax.uucp, Siemens AG, Munich ------------------------ cut here --------------------------- #include <dos.h> #define COMMAND 0x64 #define STATUS 0x64 #define DATA 0x60 put_com(c) unsigned char c; { if (inportb(STATUS)&0x1) inportb(DATA); /* Datenport loeschen */ do ; while(inportb(STATUS)&0x3); outportb(COMMAND, c); } put_dat(d) unsigned char d; { do ; while(inportb(STATUS)&0x2); outportb(DATA, d); } unsigned char get_dat() { while(!(inportb(STATUS)&0x1)); return(inportb(DATA)); } main() { unsigned char ea; disable(); put_com(0xD0); /* Read EA Port */ ea = get_dat(); put_com(0xD1); /* Write EA Port */ put_dat((ea^0x02) | 0xD1); /* toggel a20 bit */ enable(); }