X-Authentication-Warning:	delorie.com: mail set sender to djgpp-bounces using -f
Date:	Wed, 30 Jul 2014 14:51:30 +0200
From:	Mateusz Viste <mateusz DOT viste AT delorie DOT com>
User-Agent:	Mozilla/5.0 (X11; Linux x86_64; rv:24.0) Gecko/20100101 Thunderbird/24.6.0
MIME-Version:	1.0
Newsgroups:	comp.os.msdos.djgpp
Subject:	Re: Using outportb() with djgpp
References:	<almarsoft DOT 2508132549819646607 AT news DOT free DOT fr>
In-Reply-To:	<almarsoft.2508132549819646607@news.free.fr>
Lines:	64
Message-ID:	<53d8ea52$0$2077$426a74cc@news.free.fr>
Organization:	Guest of ProXad - France
NNTP-Posting-Date:	30 Jul 2014 14:51:31 CEST
NNTP-Posting-Host:	82.225.72.113
X-Trace:	1406724691 news-2.free.fr 2077 82.225.72.113:61529
X-Complaints-To:	abuse AT proxad DOT net
Bytes:	3688
To:	djgpp AT delorie DOT com
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp
Reply-To:	djgpp AT delorie DOT com

Hi all,

My problem is somewhat solved (or at least I know exactly what is going 
wrong, and how to jump over it), so I felt it would be good to share my 
conclusions in this thread.

First, I quickly dropped DJGPP for this project. Not because it's not 
working fine, or anything, but simply because doing such low-level stuff 
in real mode is a bit clearer, because it's closer to the machine (and 
having it running on a hypothetical 8086 is always sounding 'cool').

These of you who follow comp.os.msdos.programmer already know the whole 
story.

For the others:

I have written before that my problem was appearing only in DJGPP, but 
not in real mode. This was true... but only until I started using XMS 
memory from within my real mode program. Then, the exact same problem I 
had on DJGPP appeared (ie. wild reboots). I installed JemmEx hoping to 
catch the problem without triggering an immediate reboot, and it did 
show me the problem: an "exception 0D" was occuring when I was writing a 
midi note to port 0x330, and trying to move XMS blocks immediately after 
that.

I am using a SoundBlaster 64 AWE card on an ISA slot. These cards 
require a magic TSR called "AWEUTIL" that emulates a standard MPU-401 
interface on port 0x330 by catching notes that are sent to it via some 
obscure NMI magic. What I understood is that AWEUTIL is catching the 
note, and then performs some work to translate the MIDI message to a 
native command of the SB64 "EMU8000" synth chip. As it appears, if I 
happen to try using XMS memory when AWEUTIL is 'busy', a crash happens. 
If I don't load AWEUTIL, then my program works fine. Also, if I load 
AWEUTIL, but use low memory instead of XMS, again, it works.

The workaround I found is simply to wait for 2ms after each time I write 
to port 0x330 (giving AWEUTIL enough time to do its work, I presume), 
then my program works fine, no crashes, everyone happy.

So I think that the problem I had in DJGPP was exactly the same. But 
there, I couldn't control how (and when) my program used XMS, since it's 
handled automatically, so it would probably be much harder to locate the 
problem.

cheers,
Mateusz





On 07/10/2014 10:20 PM, Mateusz Viste wrote:
> Hi,
>
> I am trying to write to a hardware port from within djgpp, using
> outportb(). It does work from time to time, but often outportb() will
> make the whole PC reboot. Specifically I am writing to port 0x330 (to a
> MPU401 adapter). If using the same code with turbo C, it works
> perfectly. Hence I am wondering: are there any special "rules" I should
> be aware of when trying to outportb() from within protected mode?
>
> regards,
> Mateusz

- Raw text -