12212 2017-10-10 09:41:50 +0000 Path for modprobe.d and modconf.d should be set through variables. 2023-07-10 16:40:44 +0000 1 1 7 Build System, Metadata & Runtime OE-Core configuration unspecified All Multiple RESOLVED FIXED Medium enhancement Future 1 ola.x.nilsson ola.x.nilsson quaresma.jose randy.macleod --- --- Don't know oldest_to_newest 77483 0 ola.x.nilsson 2017-10-10 09:41:50 +0000 The paths used for modprobe.d and modconf.d in kernel-modules-split.bbclass are hardcoded. They should use variables like all other paths. This matters both for files generated by the recipe/class and by the package makefile. 77545 1 richard.purdie 2017-10-12 14:59:19 +0000 I have no objection to this in principle however I do wonder what assumptions the underlying software has about these paths. Its one thing to make it configurable, its another whether all of the software using the paths responds to the configuration options. I won't want to do this if it meant carrying a lot of patches to software with no hope of upstreaming them as I don't see this as an important enough issue based on the information here. 77565 2 ola.x.nilsson 2017-10-13 08:02:04 +0000 systemd loads modules according to files in /etc/modules-load.d/*.conf /run/modules-load.d/*.conf /usr/lib/modules-load.d/*.conf according to 'man modules-load.d' modprobe uses conf files in /lib/modprobe.d/*.conf /etc/modprobe.d/*.conf /run/modprobe.d/*.conf according to 'man modprobe.d'. I'm not sure what other software you are thinking about. We put our files in /usr and /lib (read-only on our products) and it is possible to handle that with the current kernel.bbclass and kernel-modules.bbclass. But when we want to use the usrmerge feature it requires a rather ugly hack to split_module_packages. 85305 3 randy.macleod 2019-10-10 15:10:12 +0000 Ola, Are you able to send a patch? 85320 4 ola.x.nilsson 2019-10-11 10:38:04 +0000 I guess it's too late for the zeus release, so there is no hurry? Then yes, I can prepare a patch. Eventually :) 85322 5 randy.macleod 2019-10-11 13:52:26 +0000 Thanks Ola. It's possible that the change could be backported to zeus if it's not too disruptive or risky. 91758 6 randy.macleod 2021-10-14 15:12:59 +0000 Any update Ola? 91765 7 ola.x.nilsson 2021-10-15 10:57:35 +0000 Sorry, there is always something more pressing to do... I still want to add this functionality. 95750 8 quaresma.jose 2023-06-20 08:41:58 +0000 I have submit a patch for this [2] but because it looks like busybox don't support other path than /etc [1] so the implementation only relocate the default /etc location when systemd is enabled [3]. [1] https://git.busybox.net/busybox/tree/modutils/modprobe.c?id=669c40ed8ebf480c95ce36135104e474e361a7e6#n658 [2] https://lists.openembedded.org/g/openembedded-core/message/177186 [3] https://lists.openembedded.org/g/openembedded-core/message/183123 95881 9 quaresma.jose 2023-07-04 10:43:01 +0000 This bug should be fixed with [1] and [2] also adds systemd configurations. [1] https://git.yoctoproject.org/poky/commit/?id=303421dce02fec67daab0dc86d85e11963d5741f [2] https://lists.openembedded.org/g/openembedded-core/message/183823 95896 10 quaresma.jose 2023-07-10 16:40:44 +0000 This bug should be fixed with [1] and [2] also adds systemd configurations. [1] https://git.yoctoproject.org/poky/commit/?id=303421dce02fec67daab0dc86d85e11963d5741f [2] https://git.yoctoproject.org/poky/commit/?id=6704606b351c4d570cce96ee9afb38c565829790