I think that the top two were failing because the relevant /proc entries hadn't been created
Well, I don't get the error here anymore, so I can only assume it was operator error when I was first playing with this. That said, I don't think moving sysctl to S01 is particularly safe, because of Bryan's observation that the RTC entry won't exist until the rtc module is loaded.
As we don't dictate to folks how to configure their kernel (though we do warn them that a non-modular kernel is easier ATM wrt. udev), I believe the safest place to put the bootscript is some time after S05modules. Because the modules are loaded asynchronously though, I'm not sure we'll ever know what the "right" time is to attempt to set sysctl related stuff - slower machines, or those loading lots of modules, will take a while before sysctl can do its thing successfully.
So, with all that said, I think someone with some time on their hands needs to investigate this whole 'dev.d' setup (or just hit me with the relevant cluebat), and we'll put the relevant machinery in the bootscripts and/or the book. This, I believe, would completely deprecate 'sysctl', in favour of configuring devices immediately after their creation.
Regards,
Matt. -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
