<sewn>
the context of the issue I linked, which is hosting screenshots of kiss linux user's desktops and such
angerisagift has quit [Ping timeout: 268 seconds]
angerisagift has joined #kisslinux
phinxy has quit [Quit: WeeChat 3.5-dev]
phinxy has joined #kisslinux
dazzelpansy has joined #kisslinux
shokara has quit [Read error: Connection reset by peer]
shokara has joined #kisslinux
dilyn has quit [Quit: Connection closed]
sad_plan has joined #kisslinux
<sad_plan>
hi
dilyn has joined #kisslinux
<dilyn>
illiliti: I finally have the capacity to actually rigorously use ZFS but I'm hitting an issue. What's your /etc/tinyramfs/config and kernel command line? mine:
<dilyn>
useful, thank you; I assume based on the way the zfs hooks work you *have* to mark pools that need to be mounted as legacy mount, otherwise they'll all mount at /mnt/root instead of /
<illiliti>
yes
<illiliti>
iirc there is a workaround to this issue but i didn't implement it
<illiliti>
which kernel and zfs version you use?
<dilyn>
6.2.12+ (ubuntu 23.04 kernel), whatever latest zfs is
<dilyn>
have to be bleeding edge on ubuntu because I got an arc GPU xD
<illiliti>
intel arc?
<dilyn>
yessir
<dilyn>
a750
<illiliti>
cool
<dilyn>
got it for encoding/decoding once I build my NAS; i'll switch my main machine to like a 7900xtx or something when that happens. but that's a long-term strategy
<illiliti>
well your only way to understand what's wrong is to boot into initramfs and type commands by hand until you get an error
<illiliti>
there are some breakpoints you can use to drop into shell before specified stage is ran
<illiliti>
grep init.sh for rdbreak
<dilyn>
for sure, that was the last thing I was going to do (arguably maybe it should be the first thing I did but oh well)
<illiliti>
it's worth to check if /dev/disk/by-uuid/<your uuid> gets created in initramfs at boot
<illiliti>
if not, then it's device manager issue
<illiliti>
in your case it's systemd obviously
<illiliti>
if uuid does not exist in /dev/disk/by-uuid, you can try to specify path to devnode in zfs_root instead as a quick workaround
<dilyn>
definitely
<illiliti>
can you send me output of zfs list if possible?
<dilyn>
yeah that's the change I made and why I asked for yours lmfao
<dilyn>
but *technically* it SHOULD be supported:P
<illiliti>
yeah
<dilyn>
I'm about to test the changes i'll lyk
<illiliti>
the problem is that it requires major changes to tinyramfs
<illiliti>
like event-based system where hooks= order won't matter
<dilyn>
aha aha
<illiliti>
that's something i want to implement
<dilyn>
I never used tinyramfs enough to notice when I was messing around with it last time
<illiliti>
but have no time for this yet
<illiliti>
i don't use tinyramfs often too
<illiliti>
i just generated it once and forgot
<illiliti>
btw i don't know how zfs mounts pools with nested pools
<illiliti>
i never tested such configuration
<dilyn>
wdym?
<dilyn>
it's just a single pool
<dilyn>
how mounting datasets is handled is a tricky bitch tho
<illiliti>
i mean if you mount rpool/ubuntu, does it mean that rpool/ubuntu/usr will get mounted too?
<illiliti>
i don't know how mount works in this case
<illiliti>
mine configuration is simple as you see
<dilyn>
rpool/ubuntu/usr won't get mounted because it's canmount=off. it only exists to create rpool/ubuntu/usr/local
<dilyn>
you can't have rpool/ubuntu/usr mount to, say, /usr, because it won't exist for initrd to handoff to (because if you set it to automount it will get mounted to /mnt/root instead of /)
<dilyn>
you'd have to extend tinyramfs' zfs hook to mount datasets before switch_root
<dilyn>
I think the way would just be a zfs.init.late that checks if a var is set in config and mounts datasets - trivial with a zfs mount -a BUT if you're doing a dual boot off a single pool and have datasets setup like I do you'd need to set a different passphrase on the root datasets so that you don't get a collision -- because rpool/kiss was made after
<dilyn>
rpool/ubuntu, rpool/kiss would get mounted over rpool/ubuntu
<dilyn>
if you provide separate keys you just fail on the key for the dataset you don't care about and I think zfs mount -a only mounts unlocked rpools
<illiliti>
i think that's something host system should do, e.g via fstab
<dilyn>
I think I agree
<illiliti>
initramfs task is to mount root and switch root to it
<dilyn>
100%
<dilyn>
you can use the zfs mount generator to create a bunch of systemd services to properly map out the mount order and do the mounting during init after you've switch rooted but... why would I let systemd do that when I can just use fstab do the one thing it was designed for lmfao
<illiliti>
yes
<illiliti>
if you manage to get it to work, lmk. i'm going to sleep
dilyn has quit [Quit: Connection closed]
<sewn>
what's it with zig devs trying to blame kiss linux for zig builds being segfault
<sad_plan>
maybe theyre drunk?
dilyn has joined #kisslinux
<dilyn>
illiliti: figured out *one* thing: was using the wrong ID for the pool
<dilyn>
zpool wants the ID of the pool itself, not the disk. I guess -- hard to debug. My keyboard doesn't work in initrd
<dilyn>
the keyboard is seen by the kernel, the module is loaded. so why it doesn't function, I'm not certain of
dilyn has quit [Quit: Connection closed]
sad_plan has quit [Quit: nyaa~]
dilyn has joined #kisslinux
dilyn has quit [Quit: Connection closed]
dilyn has joined #kisslinux
<dilyn>
setting hostonly in config results in success. Will investigate why soon enough