From mboxrd@z Thu Jan  1 00:00:00 1970
From: =?utf-8?q?Rafa=C5=82_Cygnarowski?= <zswi@pers.pl>
Subject: Re: documentation for fs redirector
Date: Thu, 15 Nov 2007 19:52:19 +0100
Message-ID: <200711151952.25905.zswi@pers.pl>
References: <200711122300.05659.zswi@pers.pl> <200711151726.57487.zswi@pers.pl> <473C833F.7000408@sbcglobal.net>
Mime-Version: 1.0
Content-Type: multipart/signed;
  boundary="nextPart6872242.dYNCbhaFVq";
  protocol="application/pgp-signature";
  micalg=pgp-sha1
Content-Transfer-Encoding: 7bit
Return-path: <linux-msdos-owner@vger.kernel.org>
In-Reply-To: <473C833F.7000408@sbcglobal.net>
Sender: linux-msdos-owner@vger.kernel.org
List-Id: <linux-msdos.vger.kernel.org>
To: linux-msdos@vger.kernel.org

--nextPart6872242.dYNCbhaFVq
Content-Type: text/plain;
  charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

Hi!

Dnia czwartek, 15 listopada 2007, Mike McCarty napisa=C5=82:
> Without actually seeing the defective code from DOS and your
> attempted workaround, it's impossible to tell. I am an old hand
> at MSDOS and assembler, though not much with the undocumented
> I/Fs. However, if you want you can shoot me some code and I'll
> have a look. Perhaps we can work this out via e-mail.
>
> A TSR may not be able to do what you want, unless it is rather
> "smart". CX may be clobbered so far up that there's no way
> for you to fix it, except in a DOS version specific manner,
> if at all. One possibility is to patch DOS at install. IOW,
> find the defective code, and patch over it, possibly jumping
> out into your own code in a couple of places. It may not be
> possible to fix this w/o clobbering something else that DOS needs,
> so the patches might be significant. IOW, you might need to
> trap multiple interrupt vectors and save information for more
> than one level of DOS, along with flags indicating what was
> taking place at the time CX got clobbered, and restore the
> values the various levels of DOS need.

This is fragment of dosemu code called after int2f AX=3D112E (just before=20
quitting):

if (LOW(state->eax) =3D=3D MULTIPURPOSE_OPEN /* 2E */) {
	u_char *stack_ch, *stack_cl;

	// original CX value when int21 ah=3D6c00 is called
        u_short in2e_fileattr =3D sda_ext_attr(sda);=20

	// normally this should be enought
	SETWORD(&(state->ecx), out2e_result);

	// but I try to modify pushed CX on stack
	stack_ch =3D (u_char *) (Addr(state, ss, esp) - (784 + 6));
	stack_cl =3D (u_char *) (Addr(state, ss, esp) - (784 + 7));
	if (
        	((in2e_fileattr & 0x00ff) !=3D *stack_cl) &&
		((in2e_fileattr & 0xff00) !=3D *stack_ch)
	) {
                // sanity check
		error("Wrong stack offset for MULTIPURPOSE_OPEN!"		=09
			"Expected value: 0x%04x. Value on stack 0x%02x%02x\n",
                        in2e_fileattr, *stack_ch, *stack_cl);
		leavedos(1);
	}

        *stack_ch =3D (out2e_result & 0xFF00) >> 16;
        *stack_cl =3D out2e_result & 0x00FF;
}

I use this program to test if CX was set correctly:

#include <stdio.h>
#include <io.h>
#include <dos.h>
#include <conio.h>
#include <errno.h>

#define u_short unsigned int
#define u_char unsigned char

int open2e(u_short openmode, u_short fileattr, u_short fci, char *filename,=
=20
int *respond)
{
	int __fd =3D -1, res;

	asm {
		mov ax, 0x6c00
		mov bx, openmode
		mov cx, fileattr
		mov dx, fci
		mov si, filename
		int 0x21
		jc blad
	}

	asm	mov res, cx;
	asm	mov __fd, ax;
	*respond =3D res;
	return __fd;

blad:
	asm	mov errno, ax;

	return -1;
}

int main(void)
{
	int fd, res;
	char *filename =3D "F:\\TEST.TXT";

	clrscr();

	fd =3D open2e(0x0042, 0x002a, 0x0001, filename, &res);

	if (fd =3D=3D -1) {
		perror("open2e");
	} else {
		printf("2e respond: 0x%04x\n", res);
		printf("fd =3D=3D %d\n", fd);
		write(fd, "write test", 11);
		close(fd);
	}

	getch();

	return 0;
}

If you need more do not hesitate to ask.

Best regards,
=2D-=20
Rafa=C5=82 Cygnarowski
rafi@pers.pl

--nextPart6872242.dYNCbhaFVq
Content-Type: application/pgp-signature; name=signature.asc 
Content-Description: This is a digitally signed message part.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.5 (GNU/Linux)

iD8DBQBHPJVpiTbqoF5x7W4RAlFxAJ0UPf+QDS9rMHMT009Re9ltPZ4ABQCfcrDb
P4ssysWsjKiZKOu/HG2I40Q=
=+Psz
-----END PGP SIGNATURE-----

--nextPart6872242.dYNCbhaFVq--