NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/os/msdos/programm/8571

Text File | 1992-08-17 | 1.9 KB | 35 lines

Newsgroups: comp.os.msdos.programmer
Path: sparky!uunet!pipex!warwick!mrccrc!doc.ic.ac.uk!rhbnc!csqx.cs.rhbnc.ac.uk!adrian
From: [e-mail] (A Johnstone)
Subject: Virtual to physical address mapping
Message-ID: <[e-mail]>
Sender: [e-mail] (USENET News System)
Nntp-Posting-Host: csqx.cs.rhbnc.ac.uk
Reply-To: [e-mail] (A Johnstone)
Organization: Dept of Comp Sci, Royal Holloway & Bedford New College Uni London
Date: Mon, 17 Aug 1992 10:00:34 GMT
Lines: 22

I'm a hardware engineer with plenty of Unix and VMS systems programming experience
but rather ignorant about MS-DOS internals, so please pardon my naivity.

I have a homebrew SCSI card that is meant to perform DMA. It works fine on an
XT but not on an AT. On an XT, buffers live at the same physical addresses as
the compiler and linker specify, so it is easy to figure out what to put in the
DMA controller registers. On a 386 under MS-DOS 5 this is no longer true. I have
no way of figuring out what the true physical address of a particlular virtual
address is, and since I don't know where in the hardware address space my buffer
is living, I can't tell my SCSI chip where to DMA to. I must stress that even under
the simplest MS-DOS setup, with no high drivers or other cleverness, program
addresses no longer equal hardware addresses. I assume that under Windows or when
using memory extenders the situation becomes even murkier.

Any advice, hints or pointers to the literature on the subject of memory allocation
in an extended memory environment would be welcome. What I'm really hoping is
that one of you will be able to tell me about an MS-DOS interrupt that takes a
segment:offset pair and returns a 24-bit number containg the equivalent physical
main-memory address. However I've got this hunch that things in MS-DOS land are not
so simple...

                                   Adrian