From mboxrd@z Thu Jan 1 00:00:00 1970 From: Scott Wood Subject: Re: [PATCH 15/17] KVM: PPC: Support irq routing and irqfd for in-kernel MPIC Date: Mon, 22 Apr 2013 18:31:35 -0500 Message-ID: <1366673495.10399.10@snotra> References: <1366321186.9938.13@snotra> <67A4FF0D-3BCF-42DB-9143-C87F1EF9CD0D@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; delsp=Yes; format=Flowed Content-Transfer-Encoding: 8BIT Cc: , "kvm@vger.kernel.org mailing list" , Marcelo Tosatti , Gleb Natapov To: Alexander Graf Return-path: In-Reply-To: <67A4FF0D-3BCF-42DB-9143-C87F1EF9CD0D@suse.de> (from agraf@suse.de on Thu Apr 18 19:15:46 2013) Content-Disposition: inline Sender: kvm-ppc-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On 04/18/2013 07:15:46 PM, Alexander Graf wrote: > > On 18.04.2013, at 23:39, Scott Wood wrote: > > > Do we really want any default routes? There's no platform notion > of GSI > > here, so how is userspace to know how the kernel set it up (or what > GSIs > > are free to be used for new routes) -- other than the "read the > code" > > answer I got when I asked about x86? :-P > > The "default routes" really are just "expose all pins 1:1 as GSI". I > think it makes sense to have a simple default for user space that > doesn't want to mess with irq routing. > > What GSIs are free for new routes doesn't matter. Routes are always > completely rewritten as a while from user space. So when user space > goes in and wants to change only a single line it has to lay out the > full map itself anyway. It looks like you already write the routes in your QEMU patches, so I'd like to avoid adding MPIC default routes in KVM to keep things simple. It's legacy baggage from day one. With default routes, what happens if we later support instantiating multiple interrupt controllers? -Scott