| 28 Jan 2023 |
 @lily:lily.flowers | About initrd resume:
It is vital that this be done prior to remounting any filesystems (even as read-only) otherwise data may be corrupted.
At https://www.kernel.org/doc/html/latest/power/swsusp.html | 22:27:45 |
| @lily:lily.flowers | In reply to @elvishjerricco:matrix.org
does just having stuff in the page cache pre-resume fuck with it? Maybe. I'm not quite sure, it doesn't elaborate | 22:28:03 |
 @elvishjerricco:matrix.org | Yea, that's annoying. If you're going to warn about it, you should explain it | 22:28:45 |
| @elvishjerricco:matrix.org | * Yea, that's annoying. If they're going to warn about it, they should explain it | 22:28:52 |
| @lily:lily.flowers | (I would hope mounting an unrelated disk that wasn't present at hibernation would be fine, but idk why it's a problem to begin with so maybe not) | 22:29:15 |
| @elvishjerricco:matrix.org | An EFI variable should work, right? lol | 22:29:44 |
| @lily:lily.flowers | You know, that's an even better idea | 22:30:16 |
| @lily:lily.flowers | Cursed, but better | 22:30:23 |
| @elvishjerricco:matrix.org | lol great | 22:30:38 |
| @elvishjerricco:matrix.org | I did not mean for that to be taken seriously | 22:30:44 |
| @elvishjerricco:matrix.org | but yea it might be better | 22:30:49 |
| @elvishjerricco:matrix.org | Well, the issue author created a version of the hibernate test that exhibits the issue | 22:31:33 |
| @elvishjerricco:matrix.org | they just hard coded the resume offset | 22:31:38 |
| @elvishjerricco:matrix.org | so that's fragile and we can't actually have that in nixpkgs :P | 22:31:52 |
| @elvishjerricco:matrix.org | but it helps diagnose for now | 22:31:57 |
| @lily:lily.flowers | I wonder if one of the more esoteric filesystems in the kernel allows you to either deterministically know or set a specific offset for a file | 22:32:58 |
| @elvishjerricco:matrix.org | oh | 22:33:59 |
| @elvishjerricco:matrix.org | oh no | 22:34:00 |
| @elvishjerricco:matrix.org | best bad idea yet | 22:34:03 |
| @elvishjerricco:matrix.org | networking | 22:34:06 |
| @elvishjerricco:matrix.org | the test machine tells a server machine the offset, which remembers it until initrd (using initrd networking) asks for it | 22:34:41 |
| @elvishjerricco:matrix.org | absolutely cursed | 22:34:45 |
| @lily:lily.flowers | Very cursed yeah 😆 | 22:35:17 |
| @elvishjerricco:matrix.org | Ok, I used their config and I see this:
resume # [ 0.948879] systemd[1]: Found device /dev/vda1.
resume # [ 0.949179] systemd[1]: Mounting /sysroot...
resume # [ 0.949847] systemd[1]: Starting Resume from hibernation using device /dev/vda1...
resume # [ 0.951809] systemd-hibernate-resume[115]: Could not resume from '/dev/vda1' (254:1).
| 22:37:33 |
| @elvishjerricco:matrix.org | it doesn't say why it couldn't resume, so maybe I need to enable some debug output | 22:37:48 |
| @elvishjerricco:matrix.org | but | 22:37:49 |
| @lily:lily.flowers | Wait I've got another one. What if you made a swap partition but nuked that partition from the table and set an offset relative to the previous partition | 22:37:54 |
| @elvishjerricco:matrix.org | the Mounting /sysroot thing is scary | 22:37:57 |
| @lily:lily.flowers | In reply to @elvishjerricco:matrix.org
the Mounting /sysroot thing is scary Yes | 22:38:02 |
| @elvishjerricco:matrix.org | I mean, I've known about that ordering issue for a while | 22:38:13 |
