Baslinux mailing list

[Ken Martwick <kenm AT surfbest.net>]

James Miller wrote:

> On Thu, 4 Nov 2004 kenm AT surfbest.net wrote:
>
>  
>
> > When a PC first boots, the BIOS provides (real-mode) interrupt service
> > via an array of 256 "interrupt vectors" at the bottom of addressable
> > memory.  These do not normally concern Linux, but they do affect the
> > booting process if loadlin is used and have a major impact on the
> > compact Xvesa server.  Assuming the BIOS interrupt vectors are good
> > (many are buggy), the problems begin when DOS boots and adds its
> > changes to the vectors; these won't be usable after Linux boots.
> > My first idea to correct the Xvesa problems was to write a "clean"
> > set of BIOS interrupt vectors into memory after BL3 boots.  This
> > is still a good idea, but I haven't figured out exactly how to
> > do it, especially at physical address zero!
> > It is easier to make such a correction while still in DOS, but
> > DOS must still be able to allow loadlin to boot Linux, so my
> > idea was to only "correct" some of the altered vectors.  In
> > fact, I wrote a quick program to change the first 32.
> > On the basis of tests with a 486, a Pentium, and a Pentium II,
> > this seems to work flawlessly and to allow booting from a
> > "normal" DOS environment with XMS ("himem.sys" or equal).  I
> > do not use emm386, so I cannot say if it would work.  Tests
> > were conducted using PTS-DOS and FreeDOS.
> > In practice, the messiest part of this procedure is getting a
> > clean set of interrupt vectors.  I have found a "multiboot"
> > program which comes with utilities to do this with the use
> > of a temporary boot floppy.  After that, all that is needed is
> > to run my "fix1f.exe" program with the vectors "vector.bin" in
> > the same directory.  If someone can provide a site for this
> > (<200K) package, I believe I could email it as an attachment.
> >    
>
> This sounds interesting.  Could it be the solution to the DOS memory mgmt
> programs' fouling up of Xvesa?  That would be a great interim
> solution--maybe somehow making up a batch file that would run it, then
> boot.bat.  I'd be willing to host it at my add-ons site.  But I'm not
> quite clear on how it all works--especially the role of the boot floppy
> you mention.  We're talking about one or more programs that execute within
> a DOS session here, aren't we?  A DOS session started from a hard drive
> install of DOS?  And where the baslin directory is located with its
> relevant files?  So where does the "temporary boot floppy" fit into this?
> Is it just where some program was located that you transferred to your DOS
> hard drive install?  And from that hard drive install you run it, then
> your fix1f.exe, before running boot.bat?  Clarification would be
> appreciated.  And the 200K: is this fix1f.exe and another program that you
> got off the boot floppy, or just one or the other?
>
> Thanks, James
> _______________________________________________
> BasLinux mailing list
> BasLinux AT lists.ibiblio.org
> http://lists.ibiblio.org/mailman/listinfo/baslinux
>
>  
I'm sorry I was confusing.  The boot floppy I mentioned is only used once
to get the clean BIOS interrupt vectors for a particular PC.  The 200k 
size includes all the programs and source code.  My code is GPL and the
other programs are also open source.
Ken Martwick