On Mon, Mar 21, 2016 at 11:00 PM, Alexandre Courbot <acourbot@xxxxxxxxxx> wrote: > On 03/22/2016 10:41 AM, Rob Herring wrote: >> >> On Sun, Mar 20, 2016 at 1:55 AM, Alexandre Courbot <gnurou@xxxxxxxxx> >> wrote: >>> >>> On Sat, Mar 19, 2016 at 5:47 AM, Rob Herring <robh@xxxxxxxxxx> wrote: >>>> >>>> On Tue, Mar 15, 2016 at 11:58:42AM +0900, Alexandre Courbot wrote: >>>>> >>>>> GM20B's definition is mostly similar to GK20A's, but requires an >>>>> additional clock. >> >> >> [...] >> >>>>> gpu@0,57000000 { >>>>> compatible = "nvidia,gk20a"; >>>>> @@ -45,3 +49,22 @@ Example: >>>>> iommus = <&mc TEGRA_SWGROUP_GPU>; >>>>> status = "disabled"; >>>>> }; >>>>> + >>>>> +Example for GM20B: >>>>> + >>>>> + gpu@0,57000000 { >>>> >>>> >>>> Drop the comma and leading zero. >>> >>> >>> Even though this is how it appears in the actual DT? >> >> >> Yes, those will need to get fixed, too. > > > Sorry, I just want to confirm that I understand why this needs to be fixed. > The parent node has #address-cells = <2>, and the practice of specifying two > cells in the node name is consistent with what I see in > http://www.devicetree.org/Device_Tree_Usage. > > However in the device tree usage example one can interpret the two cells as > being two different components of the address, whereas in our case we are > using two cells because the address is 64-bit - hence we should specify it > in the name as a single entity. Is this correct? Exactly, commas are for separating distinct fields like chip select and offset in the wiki example. A 64-bit address is a single field. The other reason to fix it is dtc is going to start warning for this. Rob -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html