| 6 Apr 2022 |
 @elvishjerricco:matrix.org | we do not | 19:37:01 |
| @elvishjerricco:matrix.org | I don't even think nixos jumps back to an initramfs for shutdown. | 19:40:32 |
| @elvishjerricco:matrix.org | It definitely should; we could do proper ZFS pool exporting for instance | 19:40:51 |
 Arian | During bootup | 19:41:17 |
| @elvishjerricco:matrix.org | that's weird | 19:41:26 |
| Arian | Yes... | 19:41:37 |
| @elvishjerricco:matrix.org | are you also using the systemd handover patch? | 19:41:53 |
| Arian | Yep | 19:42:01 |
| @elvishjerricco:matrix.org | I wonder if the problem goes away if you revert that patch | 19:42:17 |
| @elvishjerricco:matrix.org | obviously not a solution | 19:42:21 |
| @elvishjerricco:matrix.org | but it would be useful info | 19:42:28 |
| @elvishjerricco:matrix.org | Arian: I remember having issues with a SYSTEMD_READY tag getting unset as initrd transitioned into stage 2, which caused dev nodes to disappear. Maybe that's part of your problem? | 19:58:09 |
| @elvishjerricco:matrix.org | but that never affected nodes that were mounted in stage 1, I only saw problems when I had a LUKS device setup in stage 1 but not mounted until stage 2 | 19:59:20 |
| @elvishjerricco:matrix.org | And this was forever ago on the old PR... | 19:59:41 |
 @janne.hess:helsinki-systems.de | In reply to @elvishjerricco:matrix.org
obviously not a solution I will try | 23:36:44 |
| 7 Apr 2022 |
| @janne.hess:helsinki-systems.de | In reply to @elvishjerricco:matrix.org
I wonder if the problem goes away if you revert that patch I don't think that helps at all. Looking at the code, it's the same as setting x-initrd.mount in /etc/fstab, causing the mount to be extrinsic in systemd-analyze dump. Having an extrinsic mount seems to be the same as DefaultDependencies=no | 09:36:37 |
| @janne.hess:helsinki-systems.de | ah I think it's a udev issue (?) | 09:59:32 |
| Arian | we think it has to do with LVM now | 10:06:25 |
| Arian | trying out without LVM in between | 10:06:29 |
| @janne.hess:helsinki-systems.de | In reply to @elvishjerricco:matrix.org
Arian: I remember having issues with a SYSTEMD_READY tag getting unset as initrd transitioned into stage 2, which caused dev nodes to disappear. Maybe that's part of your problem? That sounds like exactly my problem | 11:02:00 |
| @elvishjerricco:matrix.org | FWIW, I tried using the udev patch with the handover patch and booting off an LVM drive (in qemu), and it worked fine for me. I didn't have LUKS though | 15:26:22 |
| @janne.hess:helsinki-systems.de | that's very interesting | 15:26:53 |
| @elvishjerricco:matrix.org | oh wait | 15:27:06 |
| @janne.hess:helsinki-systems.de | I think I might have found the issue but I'm not sure | 15:27:07 |
| @elvishjerricco:matrix.org | well, I guess that was qemu-vm, which doesn't have /nix on the / drive | 15:27:17 |
| @elvishjerricco:matrix.org | * well, I guess that was qemu-vm, which doesn't have /nix/store on the / drive | 15:27:27 |
| @elvishjerricco:matrix.org | Janne Heß: What do you think it is? | 15:27:39 |
| @janne.hess:helsinki-systems.de | I think /run/udev/data/ is lost in the handover and that‘s the root of the issue | 15:28:56 |
| @elvishjerricco:matrix.org | Hm. I dunno what that directory is | 15:29:12 |
| @janne.hess:helsinki-systems.de | it‘s the udev database where it stores its state | 15:29:45 |
