v86 monitors (was Re: Re: DRL updated! -- please stop that INT discuss

On Wed, 23 Oct 2002, tom ehlert wrote:

> Eric Auer wrote:
> > The only interrupt that causes troubles for emm386 type v86 monitors is
> > int 67h ($67, 0x67, as you like :-)), and for this, there is the elegant
> > solution mentioned above already implemented in almost all emm386
> drivers.
> 'almost all' -> it worked in MS-EMM386, so it was used by *some* guy,
> so it had to be supported by the other EMM's, too.

It's a fairly widely used technique. DOSEMU uses an interrupt
vector "trampoline" like this too, for *all* interrupts that it services.
Alternatively, the vector at 0:4*67h can point to a HLT instruction, so
that the v86 monitor just needs to recognize the cs:ip where GP(0)
occurred.

On a related subject, newer CPUs (Pentiums) have special virtual mode
extensions in which an

 INT-n

can be *really* equivalent (using a controlling 32-byte bitmap in the TSS)
to

 PUSHF
 CLI
 CALL FAR 0:n*4
(assuming that the trap flag is clear)

see this series of articles by Robert Collins

http://www.ddj.com/documents/ddj9801o/
http://www.ddj.com/documents/ddj9803p/
http://www.ddj.com/documents/ddj9805q/

Bart