| 23 Apr 2023 |
 @elvishjerricco:matrix.org | The initrd systemd unit that imports my ZFS pool is basically instant, taking only as long as it takes to enter the password. But that import service takes about 6 seconds longer with plymouth, and it seems to take those 6 seconds for the password prompt to appear at all | 07:50:56 |
| @elvishjerricco:matrix.org | * The initrd systemd unit that imports my ZFS pool is basically instant, taking only as long as it takes to enter the password. But that import service takes about 6 seconds longer with plymouth, and it seems to spend those 6 seconds waiting for the password prompt to appear at all | 07:51:24 |
 @janne.hess:helsinki-systems.de | Wouldn't be surprised after looking at some Plymouth code. Their design doesn't seem to be optimal or even good | 09:08:33 |
| @elvishjerricco:matrix.org | Yea, but adding 6 superfluous seconds to the boot time seems like something the Canonical or RedHat people would have gotten upset about and fixed | 09:12:19 |
 @lily:lily.flowers | In reply to @elvishjerricco:matrix.org
Yea, but adding 6 superfluous seconds to the boot time seems like something the Canonical or RedHat people would have gotten upset about and fixed I mean our plymouth hasn't been updated in a while, so it could also just already be fixed upstream (I had a branch I mentioned at https://github.com/NixOS/nixpkgs/pull/217728#discussion_r1136088299 which seemed to incidentally fix several other plymouth issues along the way but I never got a response) | 11:17:33 |
| @lily:lily.flowers | I probably should just clean up the commits/messages on that branch and PR it to get more review on it | 11:23:05 |
| @elvishjerricco:matrix.org | Huh. TIL gpt-auto-generator doesn't work with pkgs.systemdStage1. I mean, it doesn't work at all without a patch to put the root fs on the cmdline instead of fstab, but with that fixed, I still had to set boot.initrd.systemd.package = pkgs.systemd | 21:32:04 |
| @elvishjerricco:matrix.org | Oh, it's probably because of withEfi = false in systemdMinimal | 21:33:26 |
| @elvishjerricco:matrix.org | (I am becoming increasingly convinced that we should default boot.initrd.systemd.package = config.systemd.package) | 21:34:29 |
 @aktaboot:tchncs.de | speaking of that, is there a way to let the luks decryption not show asterisks for the passphrase ? | 21:36:10 |
| @elvishjerricco:matrix.org | I believe so... | 21:36:28 |
| @aktaboot:tchncs.de | I will look into it tommorrow if I remember | 21:36:39 |
| @elvishjerricco:matrix.org | boot.initrd.luks.devices.foo.crypttabExtraOpts = ["password-echo=no"] I think? | 21:37:02 |
| @elvishjerricco:matrix.org | Hm. For a simple nixosTest, the initrd is only 0.036M | 21:39:26 |
| @elvishjerricco:matrix.org | 35864 bytes to be specific | 21:39:49 |
| @elvishjerricco:matrix.org | I'll do some more testing on this, but I'm really thinking we should just use the regular systemd in initrd | 21:40:06 |
| 24 Apr 2023 |
 oddlama | I'd like to get a quick debug shell going with systemd stage1 for some networking stuff. Do I have to create a new unit and insert it into the dependency chain or is there a better way? | 15:16:03 |
| @elvishjerricco:matrix.org | oddlama: rd.systemd.unit=rescue.target will boot you into rescue mode in initrd instead of actually booting | 15:16:46 |
| @elvishjerricco:matrix.org | as a kernel param, that is | 15:17:00 |
| @elvishjerricco:matrix.org | though networking won't be started in rescue mode, and annoyingly network.target is one of those ones that systemctl won't let you manually start | 15:17:37 |
| oddlama | that will run instead of switch root, right? | 15:17:40 |
| oddlama | So if I want a shell earlier then this doesn't work? | 15:17:44 |
| @elvishjerricco:matrix.org | So I guess you might have to add a wants dependency between rescue.target and network.target or something? | 15:18:01 |
| oddlama | In reply to @elvishjerricco:matrix.org
though networking won't be started in rescue mode, and annoyingly network.target is one of those ones that systemctl won't let you manually start hmm that's unfortunate | 15:18:06 |
| @elvishjerricco:matrix.org | In reply to @oddlama:matrix.org
that will run instead of switch root, right? initrd will approach rescue.target instead of initrd.target | 15:18:21 |
| @elvishjerricco:matrix.org | the earliest possible shell would be rd.systemd.unit=emergency.target, since emergency mode literally only starts a shell by default | 15:19:06 |
| oddlama | Maybe by using emergency.target | 15:20:24 |
| oddlama | Nvm you beat me to it | 15:20:33 |
| oddlama | I guess I could just manually start the things I want to debug from the emergency shell then | 15:21:34 |
| @elvishjerricco:matrix.org | yea, rescue or emergency mode in initrd will start very nearly nothing | 15:22:19 |
