On Wed, Aug 8, 2018 at 3:29 PM Ian Lepore <ian@xxxxxxxxxxx> wrote: > > On Wed, 2018-08-08 at 17:01 -0400, Robert P. J. Day wrote: > > On Wed, 8 Aug 2018, David Gibson wrote: > > > > > > > > On Tue, Aug 07, 2018 at 04:43:03PM -0400, Robert P. J. Day wrote: > > > > > > > > > > > > i went looking for examples of that property in the current > > > > linux > > > > kernel code base, and was a bit puzzled to see that *all* > > > > references > > > > to that property -- both the setting and the processing -- is for > > > > powerpc exclusively. searching from the very top of the kernel > > > > source: > > > > > > > > $ grep -rl virtual-reg * > > > > arch/powerpc/boot/dts/wii.dts > > > > arch/powerpc/boot/dts/rainier.dts > > > > arch/powerpc/boot/dts/sequoia.dts > > > > arch/powerpc/boot/dts/ep405.dts > > > > ... snip ... > > > > $ > > > > > > > > so if the device tree processing code under drivers/of doesn't > > > > even recognize that property, how is it officially part of the > > > > spec? or am i misreading something? > > > virtual-reg is kind of a hack, I think we want to discourage its > > > use > > > as much as possible. > > discouraging its use is one thing, but the issue is whether it's > > even an *official* property under the spec. if it is completely > > defined and processed under only powerpc and is not even recognized > > by > > the basic kernel drivers/of code, how does it merit inclusion in the > > spec? > > > > i realize there may not be a perfect equivalence here, but when > > checking out parts of the spec, i like to check how those parts of > > the > > spec are processed by the kernel code under drivers/of, and it just > > seems odd if there is no mention of such a property in the kernel. or > > am i misunderstanding the correspondence between the DTSpec and what > > is implemented in the kernel in that they don't have to match? > > > > rday > > > > I hate to keep hearing "the kernel" as if linux is the only possible > consumer of devicetree data, and is thus the definitive source on what > the data should be. U-Boot, FreeBSD, and NetBSD are also consumers. I remind folks of this almost daily. Came up today with Risc-V bindings in fact. > The virtual-reg property is documented for PowerPC in the ePAPR > document. I don't know what relationship there is between that document > and the devicetree spec. ePAPR was the starting point for the DTSpec. We removed parts that were obviously PPC specific. Maybe virtual-reg should be too. Not sure if it would be useful for any other arch or not. I don't think it is for ARM. Rob -- To unsubscribe from this list: send the line "unsubscribe devicetree-spec" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
