Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!wuarchive!texbell!vector!attctc!chasm
From: chasm@attctc.Dallas.TX.US (Charles Marslett)
Newsgroups: comp.os.minix
Subject: Re: ATI VGAWonder
Summary: "Traditional INT-2 problem"
Keywords: interrupt hang incompatibility
Message-ID: <11042@attctc.Dallas.TX.US>
Date: 18 Jan 90 06:35:06 GMT
References: <1990Jan16.044748.18071@ux1.cso.uiuc.edu> <2275@kodak.UUCP>
Organization: The Unix(R) Connection, Dallas, Texas
Lines: 63

In article <2275@kodak.UUCP>, hysell@kodak.UUCP (John Hysell) writes:
> In article <1990Jan16.045403.18941@ux1.cso.uiuc.edu>, bc@ux1.cso.uiuc.edu (B.C.) writes:
> > I got the following message:
> > 
> > Unexpected interrupt or trap: vector = 13.  pc = 0x147  text+data+bss = 0x11E
> > 
> > My Norton Programmer's Guide tells me that interrupt 13 (decimal) is invoked
> > during the vertical retrace period on a color display.  I am using an ATI
> > VGAWonder 512k VGA card.  It seems to work OK, but I have gotten this error
> > twice.  Paul Allen tells me that the kernel ought to charge right on after it
> > gets this error, but it doesn't -- it hangs.  Completely.
> > 

The problem is that the ATI card is generating an interrupt on a (BIOS) mode
set -- and there is no routine to handle it.  Some motherboard BIOSes clear
the interrupt (being designed after the VGA came out), others do not.

A (really) permanent fix is to cut the INT-2 line on the VGA Wonder card (not
nice, really).  Another, more esthetic, and less reliable, fix is to execute
the following code fragment before starting MINIX (say in the boot image
initialzation code):

    MOV   DX,03D4h    ; Get the CRTC address
    MOV   AL,11h      ; Select register 0x11, Vertical-Sync-End
    OUT   DX,AL
    INC   DX
    IN    AL,DX       ; Read the old value
    OR    AL,20h      ; Set bit 5 (disable vertical interrupts)
    AND   AL,0FFH-10h ; Clear any pending vertical interupt
    OUT   DX,AL       ; Issue the new state

    JMP   $+2         ; For those with latent paranoia
    JMP   $+2         ; (or who have programmed 3780s)
    JMP   $+2
    OUT   DX,AL       ; We clear the interrupts twice!

If this fixes the problem, it's a general one, not unique to ATI -- and it's
the reason IBM cut the interrupt pin loose on the true blue VGA.

Charles Marslett
chasm@attctc.dallas.tx.us
[of course, it could also be unique to ATI, and then this is all non-sequitur]

>    problems.  I finally found that if I booted DOS and used the ATI utility
>    program VSETUP and forced the card into CGA; SHUT OFF THE XT; and then booted
>   MINIX from a cold start, everything would work fine.
>      Later, while running MINIX 1.2 on the AT, I did not have these problems.
>    The chief difference (aside from obvious machine change) was that I have
>    reconfigured the ATI card for 16-bit I/O.  Under MINIX 1.3, it remains very
>    stable.  (Next; to try MINIX 1.5)

It should be more stable, the EGA is (rather) well supported with 1.3, but
did not work to well with 1.1 or 1.2.

> Good luck;
> -John Hysell
> hysell@kodak.com
> hysell@kodak.uucp

Ditto,

Charles Marslett
chasm@attctc.dallas.tx.us