Conversation

Notices

1. Embed this notice
   Rich Felker (dalias@hachyderm.io)'s status on Thursday, 07-Mar-2024 16:32:53 JST Rich Felker

   • jade

   @ariadne @leftpaddotpy @dysfun Is it actually possible to run systemd the service manager without systemd the udev controller that decides how it wants to rename all your devices?

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Thursday, 07-Mar-2024 16:32:50 JST Rich Felker

     in reply to

     • Lennart Poettering
     • jade

     @pid_eins @ariadne @leftpaddotpy @dysfun None of that is part of any systems I boot. The kernel waits for rootfs to come up if it's mounting it directly; otherwise the mount command in the initramfs does. By the time the real init (pid 1) has control, that's all done.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Lennart Poettering (pid_eins@mastodon.social)'s status on Thursday, 07-Mar-2024 16:32:51 JST Lennart Poettering

     in reply to

     • jade

     @dalias @ariadne @leftpaddotpy @dysfun if you boot a physical system you need to wait for hw to show up, and then fsck it, mount it and so on. For that you need events for devices popping up. You cannot really build a realistic hw system without having hotplug events. It just doesn't make any sense whatsoever. Hence no, if you want systemd to to boot your system, it waits for the rootfs dev, and it waits for auxiliary devices, and it does that by watching udev, because that#s the API for that.

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Thursday, 07-Mar-2024 16:32:52 JST Rich Felker

     in reply to

     • Lennart Poettering
     • jade

     @pid_eins @ariadne @leftpaddotpy @dysfun I mean if you just want the default kernel handling of them, with no automatic hotplug reactions or hooking them up to something else, can you do that?

   • Embed this notice
     Lennart Poettering (pid_eins@mastodon.social)'s status on Thursday, 07-Mar-2024 16:32:53 JST Lennart Poettering

     in reply to

     • jade

     @dalias @ariadne @leftpaddotpy @dysfun In containers, where you do not have devices (because Linux doesn't virtualize the driver model/sysfs via namespaces or something equivalent) systemd runs without udev without you even having to do anything. In a well behaved container manager systemd's unit file conditions make sure don't start udev because device mgmt is not available anyway.

     On physical hw you need something to manager your devices in the general case, so yes, udev is used for that.

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Thursday, 07-Mar-2024 22:22:18 JST Rich Felker

     in reply to

     • Lennart Poettering
     • jade

     @pid_eins @ariadne @leftpaddotpy @dysfun I mean ultimately if it runs in containers I guess the answer has to be yes - you containerize the whole system to the extent systemd thinks it's in a container and leaves that stuff alone.

     But what I'm asking is if there's a non-hack way to just say "please leave my /dev alone".

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Thursday, 07-Mar-2024 22:22:19 JST Rich Felker

     in reply to

     • Lennart Poettering
     • jade

     @pid_eins @ariadne @leftpaddotpy @dysfun Like, I asked an honest question relevant to the topic (adoption of systemd by distro(s?) I care about) - whether it's possible to run only as service supervision & init without it taking responsibility for anything hotplug/device related - and you gave me insults, hyperbole, and a half answer to the opposite question (udev without systemd) which everyone already knows how to do. 😖

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Lennart Poettering (pid_eins@mastodon.social)'s status on Thursday, 07-Mar-2024 22:22:21 JST Lennart Poettering

     in reply to

     • jade

     @dalias @ariadne @leftpaddotpy @dysfun Well, systemd focusses on real-life systems that have potentially complex storage.Not having device events in userspace is not going to work for this.Sure in specialist cases you can make your system work like a 1980s PC with no hotplug,but this is generally not in focus for us,hence no,a system with udev but without systemd will require some hacking around,it's not a scenario systemd will be particularly friendly towards,or that we as maintainers care for.

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Thursday, 07-Mar-2024 22:22:21 JST Rich Felker

     in reply to

     • Lennart Poettering
     • jade

     @pid_eins @ariadne @leftpaddotpy @dysfun 🙄

   • Embed this notice
     Lennart Poettering (pid_eins@mastodon.social)'s status on Friday, 08-Mar-2024 02:21:09 JST Lennart Poettering

     in reply to

     • jade

     @dalias @ariadne @leftpaddotpy @dysfun there is not.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Friday, 08-Mar-2024 02:25:23 JST Rich Felker

     in reply to

     • Lennart Poettering
     • Neal Gompa (ニール・ゴンパ) :fedora:
     • jade

     @Conan_Kudo @pid_eins @ariadne @leftpaddotpy @dysfun I don't even have udev on my laptop. The kernel's builtin devtmpfs behavior is plenty advanced for almost anything you could need. Simple mdev rules can set additional permissions etc. in the few cases they're needed.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Neal Gompa (ニール・ゴンパ) :fedora: (conan_kudo@fosstodon.org)'s status on Friday, 08-Mar-2024 02:25:24 JST Neal Gompa (ニール・ゴンパ) :fedora:

     in reply to

     • Lennart Poettering
     • jade

     @pid_eins @dalias @ariadne @leftpaddotpy @dysfun Technically you could have simple program emit udev events back to systemd and not have udev, but ehh, why? Even embedded systems I worked on several years ago were complex enough to need some of udev's capabilities anyway.