* mason grumbles at the wild convolutions of the initramfs system. | 03:02 | |
mason | Or, I might actually just be angry at the zfs-initramfs package being fragile. | 03:04 |
---|---|---|
Xenguy | .oO( Just turn UEFI off ) | 03:12 |
mason | Xenguy: UEFI is pleasant, though. | 03:14 |
mason | It's confused handling of switchroot in the initramfs that makes things painful. | 03:14 |
Xenguy | Suit yourself mason, but it's not my cup of tea | 03:24 |
Xenguy | Off to soak | 03:25 |
mason | Xenguy: But what's that got to do with my initramfs commentary? | 03:25 |
mason | Have a good one. | 03:25 |
Xenguy | You said UEFI is pleasant, and I've never found it anything other than a hassle when installing a new OS, that's really my only point | 03:26 |
mason | Xenguy: I mean, how did we get there? | 03:26 |
Xenguy | I don't know that much about it actually | 03:26 |
mason | Xenguy: Maybe I misread and you weren't responding to me at all. | 03:26 |
Xenguy | I think maybe there was miscommuncation | 03:27 |
Xenguy | no worries | 03:27 |
mason | I'm good at that, so probably. | 03:27 |
Xenguy | hah | 03:27 |
Xenguy | Well it is just scrolling text, so there are the usual limitations, but of course | 03:27 |
Xenguy | Have a good evening/morning etc. | 03:27 |
mason | zfs-initramfs fails in weird ways. I gutted it and have my system booting automatically off of an encrypted ZFS root, but with any luck I can figure out how it's breaking and submit a patch. | 04:16 |
mason | FWIW, I've also tossed grub because I prefer elilo. Maybe we should ship elilo also - bootloader freedom, not just init freedom. :) | 04:20 |
mason | I'll add it to my "want to package" list. | 04:21 |
* mason stares at https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=720340 | 04:26 | |
mason | This is something that could use attention. Also on my list since it bites me everywhere. | 04:27 |
fsmithred | mason, I use a patch for that | 04:27 |
fsmithred | it might be in that bug report | 04:27 |
mason | fsmithred: Is it a reasonable patch? | 04:28 |
fsmithred | nfc, but it works | 04:28 |
fsmithred | what constitutes reasonable? | 04:28 |
mason | fsmithred: There appear to maybe be a couple similar reports, including a more extensive one from 2010. (This one's 2013.) | 04:28 |
mason | fsmithred: Reasonable, as in we're not embarassed to ship it by default. :) | 04:28 |
fsmithred | there's also a discussion about it (or two) on the forum | 04:29 |
mason | I don't see a patch in that one, but if you turn up a source I'd love to look at it. | 04:29 |
fsmithred | there are actually two fixes - one for encrypted lvm and one for encrypted not-lvm | 04:29 |
mason | The haltramfs idea seems interesting. | 04:29 |
mason | Hrm. | 04:29 |
mason | This one's encrypted/ZFS, but it'd be similar to encrypted/LVM. | 04:29 |
fsmithred | https://dev1galaxy.org/viewtopic.php?id=2690 | 04:31 |
fsmithred | that's the most recent | 04:31 |
fsmithred | https://dev1galaxy.org/viewtopic.php?pid=16284#p16284 | 04:33 |
fsmithred | same thread, later post | 04:33 |
mason | fsmithred: Looks like we're just mucking with timeouts, if I'm reading that correctly...? | 04:34 |
mason | fsmithred: Might be worth investing the time in a proper haltramfs so we can get a clean luksClose etc. | 04:35 |
mason | I'd have to read more, but I think the general notion is that you unmount everything possible, pivot away, and then unmount the remainder. | 04:35 |
mason | That's it from five miles up anyway. | 04:35 |
mason | fsmithred: This discussion makes me a little nervous about just going faster but leaving the key in RAM: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=587222 | 04:36 |
mason | Consider, for instance, someone booting into a "rescue" environment and then just digging through RAM. They key would still be there, because there wasn't a proper luksClose. As I understand it. | 04:37 |
mason | Anyway, FWIW, this is on my tentative task list. It's not something I'm hoping or expecting anyone else to do for me. :) | 04:37 |
fsmithred | the second piece does shut down the volume groups | 04:40 |
fsmithred | but that doesn't help if you have a plain encrypted partition | 04:40 |
mason | Even with root on one of them? I'll read it again. | 04:40 |
fsmithred | not sure | 04:41 |
mason | I'll reread it. Thank you for the pointer. | 04:41 |
fsmithred | looks like it would try to get them all | 04:43 |
fsmithred | nobody I've talked to about this knows enough to come up with a good solution | 04:46 |
fsmithred | so if you can do it, that would be great | 04:46 |
mason | I'll try. | 04:50 |
mason | Alright, the zfs-initramfs internal docs are wrong, and walking through the code has made everything fairly straightforward. | 18:05 |
mason | Which is to say, it's possible to use the code unmodified, which is preferable. | 18:06 |
golinux | :D | 18:08 |
onefang | Ah, but now you have to change the docs. B-) | 18:30 |
mason | onefang: FWIW, what it looks like if you trust the docs is that setting rpool= is enough, if you have a bootfs in your pool. In practise it seems to blow up without root= specified. | 18:32 |
amesser | are there any pre-built beowulf netinst images around. sry for stupid question, but i don't follow ml regulary | 19:35 |
sicelo | amesser: seems you need to install ascii, change your sources.list to reflect beowulf, etc .. | 21:31 |
amesser | tx. i used grml linux fpr testing - i just wanted to compare power consumption | 21:32 |
golinux | Pad for this week's meet is here: https://pad.dyne.org/code/#/2/code/edit/INGNizES8naax8HPKsfYR-el/ | 23:07 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!