From mboxrd@z Thu Jan 1 00:00:00 1970 From: Cyrill Gorcunov Subject: Re: [PATCH] kvm tools: PCI -- Make PCI device numbers being unique Date: Tue, 10 May 2011 00:10:06 +0400 Message-ID: <4DC84A1E.3070600@gmail.com> References: <4DC6E111.805@gmail.com> <1304880489.12850.9.camel@lappy> <4DC6E6CE.70508@gmail.com> <4DC6E8F3.3050307@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Cc: Sasha Levin , Ingo Molnar , Asias He , Prasad Joshi , kvm-vger To: Pekka Enberg Return-path: Received: from mail-ew0-f46.google.com ([209.85.215.46]:62787 "EHLO mail-ew0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751726Ab1EIUKM (ORCPT ); Mon, 9 May 2011 16:10:12 -0400 Received: by ewy4 with SMTP id 4so1648471ewy.19 for ; Mon, 09 May 2011 13:10:10 -0700 (PDT) In-Reply-To: Sender: kvm-owner@vger.kernel.org List-ID: On 05/09/2011 11:53 PM, Pekka Enberg wrote: >> From: Cyrill Gorcunov >> Subject: [PATCH] kvm tools: PCI -- Make PCI device numbers being unique v2 >> >> PCI device numbers must be unique on a bus (as a part >> of Bus/Device/Function tuple).Make it so. Note the patch >> is rather a fast fix since we need a bit more smart pci device >> manager (in particular multiple virtio block devices most >> probably should lay on a separate pci bus). >> >> v2: Sasha spotted the nit in virtio_rng__init, ioport >> function was touched insted of irq__register_device. > > Hey, I don't like the new patch subject trend you're trying to start at all. You can make it > > kvm tools,pci: Make PCI device numbers unique > > but in this particular case "PCI" already appears in the title so > > kmv tools: Make PCI device numbers unique > > is the right thing to do. PCI stands for kvm tools subsystem, but if you like more the last -- I'm fine with it. > > In addition, the changelog doesn't really tell me much. Does it fix something? > Why would we need a "smart pci device manager" and why is that relevant for this > patch? Hmmh? > > Pekka The thing is that at moment the id's passed to MP table is incorrect, they are to be 5 bits long (mp spec). The smart manager we need -- it's because there could be multiple virtio block device and they _are_ to be separated pci devices, ie with own numbers and intx# assignents. As result we probably should have such virtio devices to lay on a separate pci bus, or if the number of pci devices exceed the width of address line then we should pass them to another pci bus. That is what I had in mind but I'm not sure all this should come to the changelog. -- Cyrill