Yes, you missed a big something. It is the same problem as that with
fullscreen VGA: who 'owns' the device?
Drew Northup, N1XIM
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf
> Of Colin Davidson
> Sent: Thursday, December 14, 2000 7:26 PM
> To: Plex86
> Subject: Re: [plex86] brain fart regarding networking
>
>
> > On Wed, Dec 13, 2000 at 09:16:47PM -0600, <David Creswick wrote:
> > > handle the packet forwarding and it would just think it's on two
> networks.
> > > (the second one being one on a single machine with a host
> system and the
> > > guest system)
> > Hm. So this would be a point to point network? PPP sounds good to me.
> >
> >
> > Not as fast as possible with an emulated NE2k or somthing, but needs to
> > special treatment on either end. Hell, it's workable /right now/.
> >
> >
> > -=- James Mastros
>
> Hmm indeed. Is the idea to have an Ethernet driver interface in the host
> module, then on receipt of a packet, check that it's targetted to the
> "virtual" device in the guest (by checking the target address in the
> packet), or maybe not even checking, assuming that the host/guest
> OS will do
> any necessary checking, then pass the packet on to the guest.
> Similarly the
> guest "virtual" device would send ethernet packets to the host module
> directly. This would be very efficient, much more so than emulating an
> actual card. Consider: packets could be transferred through buffers with
> flags indicating when they contain data for forwarding. No need to signal
> through emulated IO, the host driver could just check the incomming buffer
> when it has control (or a single emulated IO access could flag the host).
> Outgoing data would require an emulated hardware interrupt, but only one.
> Sweet. And no need for a "virtual LAN".
>
> P.s., James, protocols would be handled by the host/guest OSes, this would
> be a device-level connection. Actually, not much work. Much simpler than
> emulating a card, unless I've missed something(s?).
>
>
