On 10/06/2022 00:38, Heiko Thiery wrote: > Hi Tim, Hi Simon, > > [SNIP] > >> >> Heiko, >> >> You can add multi-dtb support to this so that it's usable by the other >> imx8mn boards with the following: >> >> [...] >> >> I don't mind sending this as a follow-up to your patch here. > > Since this patch moves the parts from the 2 imx8mn-evk boards to one > "common" file it would be better to do more changes on that in a > separate patch. > >> It looks like there are only the following boards in mainline that >> would benefit from using this shared include: >> imx8mn-beacon-kit-u-boot.dtsi >> imx8mn-var-som-symphony-u-boot.dtsi >> imx8mn-venice-u-boot.dtsi
There's also 'imx8mn-bsh-smm-s2-u-boot-common.dtsi'. Slightly different because only half of the blobs are there with IMX8M_DDR3L. >> Have you compared the binman portions of imx8m{m,n,p}-u-boot.dtsi? > > No not yet. I looked a bit and they look very much alike. I suspect it's possible to eventually unify everything into a shared 'imx8m-u-boot.dtsi', but I didn't actually try. >> There are a lot of differences due to different property ordering and >> label/node naming conventions. I would like to see these normalized >> but i'm not clear which is the best example to normalize to. >> Specifically I don't know: >> 1. what is the convention for property ordering in dt... is it simply >> alphabetical order? AFAIK property order doesn't matter for binman. Node order is very significant though. For names, I think we should: - Prefer hyphens to underscores - Prefer lowercase to uppercase - Prefer meaningful names to things like 'blob-ext@1' - Avoid first char being a number (because labels can't have that) >> 2. have we settled on a convention for the blob naming, if so what is >> the best example? I had suggested 'ddr-1d-imem-fw' etc. on a series from Peng, see latest version of it [1]. [1] arm: dts: imx8m: update binman ddr firmware node name https://lore.kernel.org/u-boot/20220603071715.15212-4-peng....@oss.nxp.com/ > I am not aware that there is a conventional here. But maybe simon can > give some hints here. >