Re: KVM and/or Qemu ?

[Avi Kivity <avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>]

James Jacobsson wrote:
> Is KVM supposed to be a generic VT-x/SVM abstraction layer, or is it
> supposed to be yet-another Qemu backend?
>
>   

kvm is intended to be a generic hardware virtualization abstraction 
layer (not limited to VT-x/SVM -- if the ppc folk want in, they're welcome).

That said, I find it hard to imagine someone will ignore all the amazing 
work by Fabrice & Co. and start from scratch.  Qemu's feature set will 
be hard to match.


> The reason I'm asking is that I'm a bit confused. When I'm looking at
> the kernel module sources and the interface (kvmctl.h), it looks like
> an VT-x/SVM abstraction layer, but when trying to build KVM, it
> insists on compiling Qemu, as well as many of the data structures
> seems very Qemu influenced.
>   

Certainly the kvm module can be built standalone (and is, in upstream 
kernels).  So can libkvm.a.

The top-level Makefile does not support it, but user/Makefile just a 
chdir away.

> If KVM isn't supposed to be for Qemu only, I'd like to start working
> on getting a better separation between KVM and Qemu, 

Okay.  Please coordinate work on this list.

> as well as
> documenting the KVM interface.
>
>   

This is a most welcome addition.

Is there a web site where you can nominate people for sainthood?


PS.  I have some plans for changing the userspace interface for guest 
smp support as well as general cleanups, so if you want to work on 
mutating the userspace interface first towards its final form, it will 
save some rewriting of the documentation.

-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV