From mboxrd@z Thu Jan 1 00:00:00 1970 From: ezequiel.garcia@free-electrons.com (Ezequiel Garcia) Date: Mon, 29 Jul 2013 11:51:30 -0300 Subject: [PATCH 3/5] ARM: dove: add MBus DT node In-Reply-To: <51F67AD8.1070904@gmail.com> References: <1375101114-28858-1-git-send-email-sebastian.hesselbarth@gmail.com> <1375101114-28858-4-git-send-email-sebastian.hesselbarth@gmail.com> <51F661DE.1070106@gmail.com> <20130729135227.GA2417@localhost> <51F67AD8.1070904@gmail.com> Message-ID: <20130729145129.GD2417@localhost> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Mon, Jul 29, 2013 at 04:23:20PM +0200, Sebastian Hesselbarth wrote: > On 07/29/2013 03:52 PM, Ezequiel Garcia wrote: > > Hi Sebastian, > > > > (Ccing devicetree ML) > > > > On Mon, Jul 29, 2013 at 02:36:46PM +0200, Sebastian Hesselbarth wrote: > >> On 07/29/2013 02:31 PM, Sebastian Hesselbarth wrote: > >>> This adds a MBus node including ranges and pcie apertures required later. > >>> > >>> Signed-off-by: Sebastian Hesselbarth > >>> --- > >>> arch/arm/boot/dts/dove.dtsi | 19 +++++++++++++++++++ > >>> 1 file changed, 19 insertions(+) > >>> > >>> diff --git a/arch/arm/boot/dts/dove.dtsi b/arch/arm/boot/dts/dove.dtsi > >>> index 397674c..bdda016 100644 > >>> --- a/arch/arm/boot/dts/dove.dtsi > >>> +++ b/arch/arm/boot/dts/dove.dtsi > >>> @@ -29,6 +29,20 @@ > >>> marvell,tauros2-cache-features = <0>; > >>> }; > >>> > >>> + mbus { > >>> + compatible = "marvell,dove-mbus", "marvell,mbus", "simple-bus"; > >>> + #address-cells = <2>; > >>> + #size-cells = <1>; > >>> + pcie-mem-aperture = <0xe0000000 0x10000000>; /* 256M MEM space */ > >>> + pcie-io-aperture = <0xf2000000 0x00200000>; /* 2M I/O space */ > >> > >> Actually, current v9 of the mbus patch set still requires "controller" > >> property to match the corresponding controller node. I had a short > >> discussion with Ezequiel to possibly just use of_find_compatible_node > >> and blindly assumed post-v8 will already use it. > > > > Ah, regarding this: despite your good arguin against the 'controller' property approach, > > I still feel a bit inclined for it, as I like the way it tightly-binds the two nodes. > > I understand that the phandle property *shows* you that both are > related. But with DT you should always ask for every property, if > (a) it is really required to do the job and (b) does it really > describe the HW or just your SW needs/wishes. > I see and I understand your point. For some reason it still feels a bit dirty to set this kind of compatible nodes requirement (one node simply requiring another node). I would hate to set a precedent for a dirty solution... Maybe the DT maintainers can shed some light on this? -- Ezequiel Garc?a, Free Electrons Embedded Linux, Kernel and Android Engineering http://free-electrons.com