From mboxrd@z Thu Jan 1 00:00:00 1970 From: Scott Wood Subject: Re: [PATCH 1/8] kvm tools: Add initial SPAPR PPC64 architecture support Date: Wed, 7 Dec 2011 12:31:14 -0600 Message-ID: <4EDFB0F2.3030806@freescale.com> References: <4EDD94A4.2080003@ozlabs.org> <4EDE58FE.4040904@freescale.com> <4EDF1746.7090106@ozlabs.org> Mime-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-1" Content-Transfer-Encoding: 7bit Cc: , To: Matt Evans Return-path: In-Reply-To: <4EDF1746.7090106@ozlabs.org> Sender: kvm-ppc-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On 12/07/2011 01:35 AM, Matt Evans wrote: > Hi Scott, > > On 07/12/11 05:03, Scott Wood wrote: >> On 12/05/2011 10:05 PM, Matt Evans wrote: >>> This patch adds a new arch directory, powerpc, basic file structure, register >>> setup and where necessary stubs out arch-specific functions (e.g. interrupts, >>> runloop exits) that later patches will provide. The target is an >>> SPAPR-compliant PPC64 machine (i.e. pSeries); there is no support for PPC32 or >>> 'bare metal' PPC64 guests as yet. Subsequent patches implement the hcalls and >>> RTAS required to boot SPAPR pSeries kernels. >> >> You just sent out 28 patches removing "everything is x86" >> dependencies -- may I suggest that the PPC code be structured so that >> there isn't an "everything on PPC (or even PPC64) is SPAPR" assumption, >> even if SPAPR is initially the only sub-arch present? > > I had anticipated this comment (though not the "28 patches" remark, easy now). I was just using that to illustrate how it's easier to handle earlier than later -- no offense intended. :-) > It is a fair comment, but you hit the nail on the head with your other mail > (regarding configuring in i8042, presumably to emulate crappy dev boards) > asking whether kvmtool has a config system. It does not. > > Since we currently lack any kind of build-time configuration (or any fancy > run-time -M a la QEMU) it's a bit hard to cater for multiple > platforms. I'm aware that the PPC patches are painfully PPC64-with-SPAPR and I > don't present them as perfect, but I really think we need to attack the > configuration stuff before bifurcating. Is this something you'd like to see to? Just putting all SPAPR stuff in SPAPR-named files (or at least SPAPR-named functions), and likewise for book3s stuff, etc. would be an improvement. I see that you did this for some things, but not all. Try to make it obvious where the target-specific branching would take place, even if the actual branching mechanism is currently just a FIXME comment. -Scott