At 10:17 PM 12/20/99 +0100, you wrote:
>Nelson Rush wrote:
>>
>> Looks like someone has the Linux kernel running on top of itself as a user
>> mode virtual machine.
>
>Found this one through their forum:
>
>http://www.nocrew.org/software/a386/.
>
>I like it even better :).
>
>Unfortunately, both these projects rely on availability of the
>source code, something which is often not the case.
>So no windows-on-linux this way...
>
>-- Ramon
>
>
No, but what it does do is grant the possibility of a Linux-on-top method.
Ie. Use Linux as a sub-VM OS and run it on top of other OS's. This sort of
thing would allow you to have nested OSs running. Say for instance I run NT
and then run Linux on top of NT, now we don't have to deal with writing the
rest of the software for NT! We can then write a VM engine with this
user-mode linux kernel that runs other OSs while running on top of other
OSs. No more need to port the VM engine to work with different platforms,
it only needs to work with Linux (but backwards, the VM being on top of
Linux instead of under). Quick note, we may need to run a pure Linux with
the VM running on it to run the first OS so we have more control from the
beginning of the chain.
Here's a clearer picture of what I'm thinking of.
Linux(w/VM)->BeOS->LinuxVM->NT->LinuxVM->Linux->LinuxVM->Win98
The Linux sub-VM is used as the glue between the OSs, there could even be
intercommunication between them along some sort of mapped memory. This
would allow switching directly to any of the OSs running in the chain. A
switch to BeOS from Win98 would entail: the LinuxVM->Win98 to communicate
to the Linux(w/VM)->BeOS to expect the chain to collapse.
What would happen then is that the LinuxVM->Win98 would go idle/drop to the
previous VM, the LinuxVM->Linux would activate and see the message to
collapse and then idle/drop to the previous VM, the LinuxVM->NT would
activate and see the message to collapse and then idle/drop to the previous
VM, and then the Linux(w/VM)->BeOS would activate see the message that it
is now the master and then go on with its business. To go forward in the
chain we need only to do the same in the opposite direction. Motion
throughout this chain would never even stir up the hosted OSs, since the
LinuxVM's would never fully activate, they'd see the message that they're
in the middle of a transfer to another master. I think it'd be a fast,
simple, and efficient method.
Sincerely,
Nelson Rush
"Guns don't kill people. It's those damn bullets. Guns just make them go
really really fast." -- Jake Johanson
