| 16 Jun 2022 |
 @elvishjerricco:matrix.org | Alright good. Not systemd stage 1's fault | 06:50:02 |
| @elvishjerricco:matrix.org | Still leaves me baffled... | 06:50:10 |
| @elvishjerricco:matrix.org | * Still leaves me baffled about my system... | 06:50:14 |
| @elvishjerricco:matrix.org | I only noticed it because I decided to try something really dumb and ran mount -o remount,rw /nix/store (notice I forgot bind, though I dunno if that matters). I have no idea if my system had a rw store before I did this, but now it does even after a reboot | 06:51:29 |
| 20 Jun 2022 |
|  @maxeaubrey:matrix.org joined the room. | 13:23:56 |
| 22 Jun 2022 |
| @elvishjerricco:matrix.org | Huh. This is a process running on my system: /nix/store/rb28rjj4whf1y6bhbxmg8zk8j7izi7i7-systemd-stage-1-250.4/bin/systemd-tty-ask-password-agent --watch --plymouth
That's the plymouth agent from stage 1. Is that... a problem?
| 00:19:56 |
|  @oxalica:matrix.org joined the room. | 04:30:36 |
 Paul Haerle | I've got a question regarding the "shutdown ramfs" https://github.com/NixOS/nixpkgs/blob/master/nixos/modules/system/boot/systemd/shutdown.nix. As I understand it, it's important (for some cases?) to run zpool sync with filesystems already unmounted before shutdown?
Anyway, the systemd service in question generates that ramfs at runtime, just before shutdown.
Is there a reason why that's preferred to generating the ramfs at build time? I might be missing something, but I believe it has got no dependencies, only known at runtime? | 12:31:42 |
 @janne.hess:helsinki-systems.de | In reply to @phaer:matrix.org
I've got a question regarding the "shutdown ramfs" https://github.com/NixOS/nixpkgs/blob/master/nixos/modules/system/boot/systemd/shutdown.nix. As I understand it, it's important (for some cases?) to run zpool sync with filesystems already unmounted before shutdown?
Anyway, the systemd service in question generates that ramfs at runtime, just before shutdown.
Is there a reason why that's preferred to generating the ramfs at build time? I might be missing something, but I believe it has got no dependencies, only known at runtime? I did the initial (broken) implementation this way because I stole the logic from arch… I think it'd be preferrable to do it as a nix build in the future | 12:32:42 |
| Paul Haerle | Janne Heß: Thanks for the quick answer! In what way do you consider it broken? Is there something besides the build time thing just discussed, that you like to have improved regarding shutdown ramfs? | 12:33:40 |
| @janne.hess:helsinki-systems.de | In reply to @phaer:matrix.org
Janne Heß: Thanks for the quick answer! In what way do you consider it broken? Is there something besides the build time thing just discussed, that you like to have improved regarding shutdown ramfs? ElvishJerricco already fixed everything before 22.05 was released ;) | 12:34:11 |
| @janne.hess:helsinki-systems.de | The only thing left to do (at some point) is to convert the ramdisk build to a nix build, I don't think there's anything left to improve (afaic) | 12:34:38 |
| Paul Haerle | Okay, so one more question. zpool sync could run on live systems as well, iiuc. I guess the reasoning behind doing it in a ramfs might be, that unmounting before will prevent everyone from writing after zpool sync? | 12:36:14 |
 @linus:schreibt.jetzt | yeah, things like the journal will still be written to after a sync if you do it in the booted system | 12:37:10 |
|  @sigmasquadron:matrix.org joined the room. | 23:49:53 |
| 23 Jun 2022 |
| @elvishjerricco:matrix.org | I would be very happy if someone who actually knows a thing or two about networkd would fix this up to be merge-worthy: https://github.com/NixOS/nixpkgs/pull/169116 | 18:30:12 |
| 25 Jun 2022 |
 kn | here's a little improvement to boot.initrd.systemd.emergencyAccess https://github.com/NixOS/nixpkgs/pull/179002 | 12:36:37 |
| kn | can someone explain why options such as this one are visible = false? I'd appreciate seeing them in configuration.nix(5) | 12:37:22 |
| kn | I booted my machine and went afk, so the zfs passphrase prompt timed out, stage-1 wet its pants and all I could do was hit enter, only to be told that a rescue shell was not available due to a locked root account -- hard to reboot via power button | 12:39:25 |
| kn | with this option visible, I would've immediately set it to config.users.users.root.hashedPassword before switching setting boot.initrd.systemd.enable = true | 12:40:25 |
 Arian | Because the systemd initrd is not finished and unsupported | 13:16:02 |
| Arian | We don't want people to think it's public API at this point | 13:16:13 |
| @janne.hess:helsinki-systems.de | Yeah, having them visible would mean all kinds of deprecation phases which would hinder the rapid development we need until it's finished | 15:00:41 |
| 26 Jun 2022 |
 flokli | kn: i saw your PR. I think the description of the option isn't up to date | 06:53:00 |
| kn | In reply to @flokli:matrix.org
kn: i saw your PR. I think the description of the option isn't up to date How so? | 08:02:50 |
| flokli | I think i got confused | 08:19:54 |
| flokli | It's using passwd only, not shadow? | 08:20:09 |
| Paul Haerle | I've been working on and off on a little flake which allows to build small-ish custom images. My main motivation is to create kexec-utable installer images for remote systems, possibly with custom kernels. I did it once with the old stage1, then with not-os and recently ported it over to systemd-in-stage1.
It's not much yet and size surely can be further reduced (e.g. with a smaller kernel and less user-land), but I am posting it here because I am eager for feedback and I think it can already be useful for others who aren't too deep into these things yet. It took me a while to get a decent flake structure, helpers, a basic test suite, ci and cache to run
https://github.com/dep-sys/nix-dabei
| 10:23:14 |
| Paul Haerle | And I hope this question is appropriate for this channel: I understand that utillinuxMinimal is built without systemd support for bootstrapping, but I wonder if there's a way to get only the systemd-enabled utillinux in my final closure with systemd in stage 1? Just overriding utillinuxMinimal to be indentical with utillinux breaks the builds, but maybe theres a more fine-grained solution? | 10:27:05 |
| 28 Jun 2022 |
| @janne.hess:helsinki-systems.de | This is not the right channel at all, no. This channel is about using systemd in the initrd, not about building util-linux | 11:50:45 |
