Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!purdue!decwrl!shelby!portia!dhinds From: dhinds@portia.Stanford.EDU (David Hinds) Newsgroups: comp.sys.ibm.pc.programmer Subject: Re: Virtual 8086 on an 80386 Summary: This isn't all that hard Message-ID: <10467@portia.Stanford.EDU> Date: 23 Mar 90 01:02:48 GMT References: Sender: David Hinds Distribution: na Organization: Stanford University Lines: 32 In article , blk@mitre.org (Brian L. Kahn) writes: > > I am interested in running DOS and standard programs on a Virtual DOS > system using a 386 base. The goal is a sort of augmented and paranoid > access control over the disk (and maybe network); rather than go into > that let me say for argument that I want detailed statistics on any > and all peripherals while running standard programs such as Lotus, > Word, WordPerfect, dBase, etc. I want to tally all writes to screen > memory, all disk accesses, any bypass of BIOS such as controller or > timer reprogramming, and so on. > > Can anyone offer insight into such an enterprise? Yes I know that > performance will suffer, and a drop of 50% is OK! I have read about > one or two virtual DOS's whose main goal is to switch between > environs, but I want to stick in a review mechanism between the > software (DOS + application) and the hardware. > It sounds like you would be better off writing your own virtual 86 supervisor. This is easier than it sounds, actually. I have a book on the 80386 ("80386 Technical Reference", I don't remember the author) that gives the skeleton of a supervisor for DOS in an appendix. There isn't much to this - it just sets up all the protected-mode interrupt handlers to bounce a virtual-mode interrupt back to the appropriate virtual-mode DOS or BIOS handler. You would need to set up a bunch of counters, and just add a little bit of code to these handlers to keep track of each call before passing control on to DOS. You can also trap I/O instructions in a similar fashion - I don't know if the example does this, though. The review mechanism should be absolutely, positively transparent. -David Hinds dhinds@popserver.stanford.edu