On 7 May 2015 at 09:28, Pantelis Antoniou <pantelis.antoniou@xxxxxxxxxxxx> wrote: > Hi Geert, > >> On May 7, 2015, at 10:17 , Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote: >> >> On Tue, May 5, 2015 at 6:51 PM, Michal Suchanek <hramrach@xxxxxxxxx> wrote: >>> On 5 May 2015 at 17:18, Michal Suchanek <hramrach@xxxxxxxxx> wrote: >>>> where is a device tree compiler that can compile an overlay? >>>> >>>> The compiler at git://git.kernel.org/pub/scm/utils/dtc/dtc.git shrieks >>>> there is a syntax error as soon as the /plugin/ tag is present. >>> >>> ok, so I found a dtc patch here https://lkml.org/lkml/2013/1/4/295 >>> >>> This does not apply cleanly to current DTC but the parts that fail are >>> due to cosmetic changes in the code so I went ahead and applied it. >> >> Thanks for the link! That patch indeed requires some manual rework. >> >>> However, the created blob is not what the kernel can accept into >>> of_overlay_create so what do I do with the blob to load it? >> >> Use "OF: DT-Overlay configfs interface" from Pantelis? >> >> That depends on "configfs: Implement binary attributes", also from Pantelis, >> which requires reverting commit 28444a2bde8d1695i ("configfs_add_file: >> fold into its sole caller"). >> >> All of the above require manual fixups to apply. >> >> Still have to try the actual loading, new to the overlay game on mainline >> kernels (I did use it on boneblack with its original kernel and its overlay >> manager). >> > > Posted the patches recently, but for your convenience here’s the git tree. > > https://github.com/pantoniou/dtc/tree/dt-overlays5 > > FWIW, I’ll post the capemgr patch today too. FWIW here is a separate branch with the dt overlay stuff I am running. https://github.com/hramrach/linux-sunxi/commits/dt-overlay You might want to look at the small fixes there. It is worth noting that with your version of the patch adding -@ dtc option the built board dtbs do not contain the symbols needed for overlays unless overlays are enabled. So far dtbs were static wrt build options AFAIK. That is they were either built or not but unless the sources in question changed the dtbs would be the same. Introducing this option dynamically based on CONFIG_OF_OVERLAY makes dtbs content config dependent which does not seem like a very good idea. In fact I think -@ should be the default on compilers that support it and might be optionally turned off. Maybe some statistic of the size of the symbol table so generated and relative size increase for various boards might be nice thing to have. Thanks Michal -- 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