On 10/20/17 05:11, Alexander.Steffen@xxxxxxxxxxxx wrote: >> Hi Alexander, >> >> On 10/19/17 10:06, Alexander.Steffen@xxxxxxxxxxxx wrote: >>> Hi Frank, Rob, >>> >>> I ran some tests with kernel v4.14-rc2, where I came across an issue >>> with your commit "of: overlay: add overlay symbols to live device >>> tree" (d1651b03c2df75db8eda3fbcd3a07adb337ee8b0, >>> >> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/d >> rivers/of?h=v4.14-rc5&id=d1651b03c2df75db8eda3fbcd3a07adb337ee8b0). >>> I am not sure whether this is a problem with your change or whether I >>> made a mistake somewhere. >>> >>> Without this commit, the following device tree overlay works fine on >>> a Raspberry Pi: >>> >>> /dts-v1/; >>> /plugin/; >>> >>> / { >>> fragment@0 { >>> target-path = "/soc/spi@7e204000"; >>> __overlay__ { >>> #address-cells = <1>; >>> #size-cells = <0>; >>> status = "okay"; >>> >>> spidev1: spi@1 { >>> compatible = "spidev"; >>> reg = <1>; >>> spi-max-frequency = <5000000>; >>> }; >>> }; >>> }; >>> }; >>> >>> But with the commit, it is rejected: >>> >>> OF: overlay: no symbols in root of device tree. >>> OF: overlay: of_build_overlay_info() failed for tree@/ >>> create_overlay: Failed to create overlay (err=-22) >>> >>> Only if I remove the spidev1 label, it continues to work with the >>> commit (but this is not always possible in more complex overlays). >>> >>> I hope this does not make a difference, but I use the configfs >>> interface to load the overlay, that is not part of mainline, but >>> shipped by several distributions nonetheless (for example in the >>> Raspberry Pi kernel, >>> >> https://github.com/raspberrypi/linux/commit/cccc24635da69799335566eb46 >> 4a4c9e1fb4a8ad). >>> >>> >>> Could you tell me whether this change in behavior is intentional and >>> whether I have to fix something in my usage? >> >> Yes, the change is intentional. >> >> If you are using overlays, the expectation is that the device tree >> that was used to boot contains symbols (the paths of nodes that >> have phandles) so that phandle references in the overlay can be >> fixed up to match the values in the base devicetree. Symbols >> will be added to the device tree if if is compiled with the >> "@" option. This is the first method to avoid the overlay >> load error. > > Ah, thanks, this does solve the problem. I use nothing special to > generate the base devicetree, just "make dtbs". Would it make sense > to add the "@" option to the kernel Makefile, so that symbols are > added by default? Or does this have any downside? I'd suspect that Yes, there is a large downside. Compiling with the "@" option will increase the size of the DTB. Most boards will not use overlays, and in those cases this overhead has no value. So "@" should be opt-in, not default. > many more people will be hit by this change, since most tutorials for > overlays that I have seen include the "@" option for the dtc calls.> >> If you compile the overlay with the "@" option then symbols >> will be added to the overlay DTB. Previous to commit >> d1651b03c2df, these symbols would be ignored when an overlay >> was loaded. If you remove the "@" from the compile of >> overlays, then everything should work the same way it did >> before the commit. This is the second way to avoid the >> load overlay error. >> >> >>> Thanks >>> Alexander >>> -- >>> 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 > -- 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
