From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:41067) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1eKHpp-0007Nu-KQ for qemu-devel@nongnu.org; Thu, 30 Nov 2017 00:56:34 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1eKHpm-0003K2-JE for qemu-devel@nongnu.org; Thu, 30 Nov 2017 00:56:33 -0500 Received: from mx1.redhat.com ([209.132.183.28]:45320) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1eKHpm-0003IR-Bk for qemu-devel@nongnu.org; Thu, 30 Nov 2017 00:56:30 -0500 Date: Thu, 30 Nov 2017 13:56:18 +0800 From: Peter Xu Message-ID: <20171130055618.GA22901@xz-mi> References: <20171129084628.12336-1-david@gibson.dropbear.id.au> <20171129084628.12336-4-david@gibson.dropbear.id.au> <20171130040248.GU3023@umbus.fritz.box> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <20171130040248.GU3023@umbus.fritz.box> Subject: Re: [Qemu-devel] [for-2.12 3/7] pci: Fold pci_bus.h into pci.h List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: David Gibson Cc: Marcel Apfelbaum , qemu-devel@nongnu.org, Igor Mammedov , Eduardo Habkost , "Michael S . Tsirkin" On Thu, Nov 30, 2017 at 03:02:48PM +1100, David Gibson wrote: > On Wed, Nov 29, 2017 at 12:38:00PM +0200, Marcel Apfelbaum wrote: > > On 29/11/2017 10:46, David Gibson wrote: > > > include/hw/pci/pci_bus.h is now very small and can only safely be included > > > after hw/pci/pci.h. So, just fold it into pci.h. > > > > > > > I don't get the benefit from merging the header files. > > I would go the other way around and find stuff specific > > to pci_bus and add it there, like the pci_bus_new* > > you touched in the prev patch. > > Hrm. Except the point of the earlier patch was that those are > actually spoecific to root buses, so would really belong in say > pci-host.h, rather than pci-bus.h. > > A log of PCI stuff deals with interaction between the device and bus > though, so it just seemed like more trouble than it was worth to go > disentangling them properly. > > > Maybe if *all* code files requiring pci.h would automatically > > need pci_bus.h would make sense, but I didn't check that. > > Yeah, I don't think every user of pci.h needs pci_bus.h, although a > fair few do as you can see from the diff. Well, I guess it's up to > Michael. I'll be tolerably content either way - I'd say this is the > least important patch of the series. I would tend to keep pci_bus.h too considering that it's a separate header file already (though I see that Linux doesn't have that specific pci_bus.h, but everything is in pci.h). Or, if we really want to do that, IMHO we can also consider to include pci_bus.h directly in pci.h to avoid movement within headers. Thanks, -- Peter Xu