From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ingo Molnar Subject: Re: [PATCH 1/38] KVM: Create kvm-intel.ko module Date: Thu, 30 Nov 2006 21:19:35 +0100 Message-ID: <20061130201935.GA14696@elte.hu> References: <456AD5C6.1090406@qumranet.com> <20061127121136.DC69A25015E@cleopatra.q> <20061127123606.GA11825@elte.hu> <20061130142435.GA13372@infradead.org> <20061130154425.GB28507@elte.hu> <20061130115957.c3761331.akpm@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Christoph Hellwig , Avi Kivity , kvm-devel@lists.sourceforge.net, linux-kernel@vger.kernel.org Return-path: To: Andrew Morton Content-Disposition: inline In-Reply-To: <20061130115957.c3761331.akpm@osdl.org> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org * Andrew Morton wrote: > It's a fat, complex, presumably arch-specific, presumably > frequently-changing API. So whatever we do will be unpleasant - > that's unavoidable in this case, I suspect. > > (hmm, the interface isn't versioned at present - should it be?) > > Maybe, perhaps, one day it _should_ be a syscall API. But right now > if we did that it would become a versioned syscall API with obsolete > slots and various other warts. yeah, very much agreed. For example the paravirtualization/accelerator downcalls/upcalls in KVM dont exist yet, so there's little to standardize. Once we see it from lhype & KVM how these things look like we can design a sane kernel interface around it. But i'm against the notion that KVM is 'just' a device. It's not, and it /will/ grow into something fundamental. > I get the feeling we'd be best off if we were to revisit this in a > year or so. yeah. I'd suggest merging it as-is into v2.6.20. In a year we'll have some real APIs to think about. Ingo