Description of problem: Ok so packagers/maintainers fail to see the advantages of using load module configuration file as opposed to loading the module in the unit when it gets started since the module does not get loaded at install time which forces users to either restart the systemd-modules-load.service manually or do it via modprobe themselves or having to reboot. So I'm wondering if there is something that is preventing us using a path unit to do (re)start the service once an user/package creates/drops an configuration file into the /etc/modules-load.d/ directory? Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Created attachment 541818 [details] Path unit I guess something like this should work...
Basically the issue here is that when you install a package that contains a load module the module does not get loaded thus users will have to reboot or manually load the module...
Hmm, so if at all, then the packaging scripts should load the modules after installation, but I feel very uncomfortable about that too. I am really not convinced this is really a good idea, so far none of the systemd configuration files are automatically reread and activated, and it sounds wrong to me to start now here for this, especially as modules loading is a bit risky in many ways... I really prefer if configuration files like this one are applied only when the user can assume they are. For example, when the user reboots the system, or when he installs packages, or when he explicitly invokes systemd-load-modules.service. But all of these are either already covered or can be handled from the respective packaging scripts explicitly (for example by invoking /usr/lib/systemd/systemd-modules-load from there), we don't need an automatism for that. And I am pretty sure we shouldn't act for these somewhat dangerous operations behind the users back, just because something starting placing files in the dir, and isn't even finished yet... I mean, we don't have transactions in the file systems in Linux. The next best thing we have is that user actions can imply them, such as booting/package installation/direct invocation. But just moving things around in the dir isn't that... Really not convinced of this one, sorry.