disko

Sender	Message	Time
9 Mar 2024
lassulus	* yeah, generation of options manpages for disko is still a complicated topic, since everything is kinda recursive it's hard to use the old doc genrators for that. Also the old list based approach was kind of lacking in the referencing and updating part, like if you add a partition in between everything has to shift a bit. and yes, size is the new recommended way, start is usually 0, which means the next available space in sgdisk, which should be reproducible if you run the same script multiple times. so partitions usually only need size. you can use priority if you want certain partitions to be in front of others, but so far this didn;t made any difference in any bootloader. porting table partitions to gpt partitions is sadly not a straight forward thing, since the latter alwas have a label and before it was optional.	11:39:54
lassulus	this all ties in, into making the create part of every type idempotent, so running the createScript multiple times should just add stuff where it can and we should be able to run this on every nixos-rebuild switch	11:41:26
lassulus	opt-in in the beginning ofc	11:41:49
lassulus	but not every feature from the old table type got ported to the new gpt type, like msdos tables. but I have no usecase for them, if someone needs them (and is not happy with the weird gpt-hybrid thingie) please message me and we can implement a new type in tandem (meaning you do it and I tell you how :))	11:49:57
void	how do I set bootable = true; now?	11:52:47
void	or better, is there a migration guide ;P	11:52:59
lassulus	bootable = true didn't have any effect on gpt partitions before actually	11:55:24
lassulus	ah migration guide would be very cool, but sadly I haven't written one yet	11:55:54
lassulus	maybe we should start a pad and just get it done? but the knowledge is very scattered in my memory	11:57:19
void	I usually use a wiki when it happens, write everything I know atm, then restructure & fill the gaps.	12:00:31
lassulus	like my ideal idea was to generate a migration script, but I never finished it and then I realized that everybody was still using the legacy tables and so I wrote the deprecation notice first	12:02:58
void	is there a checkpoint/milestone in time where you tell users to say hey you can migrate now?	12:08:09
void	Im hesistating atm.	12:08:22
lassulus	Well you should have migrated sooner :D	12:11:22
lassulus	But it's not fully automatic yet. So the migration pain is yours until I build the migration script	12:12:11
lassulus	But not sure it will ever be done	12:12:21
@palo:xaos.space	https://pad.lassul.us/disko-migartion?view That's wat I remember I had to do for migrating.	12:43:04
10 Mar 2024
	SigmaSquadron joined the room.	14:50:47
SigmaSquadron	Hi all. Is there a way to create a partitionless LUKS device with Disko? All of the examples in the repo are putting the LUKS volume inside a GPT partition table, but since the LUKS volume will take up my entire disk, I would prefer to avoid partitioning the disk and only use LVM which takes the LUKS volumes as PVs.	14:53:42
SigmaSquadron	Redacted or Malformed Event	14:57:18
SigmaSquadron	I've tried using the following configuration, but it fails since it's missing the partition table type: { disko.devices.disk."nvme0" = { type = "disk"; device = "/dev/disk/by-id/nvme-ID"; content = { luks = { size = "100%"; content = { type = "luks2"; name = "luks-0"; passwordFile = "/tmp/disko/luks/nvme0.key"; askPassword = false; extraFormatArgs = [ "--cipher=aes-xts-plain64" "--hash=sha512" "--key-size=512" "--iter-time=5000" "--header=/dev/disk/by-id/usb-ID-part2" ]; extraOpenArgs = [ "--header=/dev/disk/by-id/usb-ID-part2" ]; settings = { header = "/dev/disk/by-id/usb-ID-part2"; preLVM = true; allowDiscards = true; }; initrdUnlock = true; content = { type = "lvm_pv"; vg = "NixOS"; }; }; }; }; }; }	15:01:08
Lily Foster	In reply to @sigmasquadron:matrix.org I've tried using the following configuration, but it fails since it's missing the partition table type: { disko.devices.disk."nvme0" = { type = "disk"; device = "/dev/disk/by-id/nvme-ID"; content = { luks = { size = "100%"; content = { type = "luks2"; name = "luks-0"; passwordFile = "/tmp/disko/luks/nvme0.key"; askPassword = false; extraFormatArgs = [ "--cipher=aes-xts-plain64" "--hash=sha512" "--key-size=512" "--iter-time=5000" "--header=/dev/disk/by-id/usb-ID-part2" ]; extraOpenArgs = [ "--header=/dev/disk/by-id/usb-ID-part2" ]; settings = { header = "/dev/disk/by-id/usb-ID-part2"; preLVM = true; allowDiscards = true; }; initrdUnlock = true; content = { type = "lvm_pv"; vg = "NixOS"; }; }; }; }; }; } is `luks2` a valid `type`? shouldn't it be just `luks`?	15:05:09
SigmaSquadron	Right, but even switching to just luks, the same error occurs, since it’s searching for the type in lib/types/table.nix	15:06:28
Lily Foster	In reply to @sigmasquadron:matrix.org Right, but even switching to just luks, the same error occurs, since it’s searching for the type in lib/types/table.nix hmmm it looks like disko supports luks as a content type for disk. one sec, let me poke at it	15:08:04
Lily Foster	In reply to @sigmasquadron:matrix.org I've tried using the following configuration, but it fails since it's missing the partition table type: { disko.devices.disk."nvme0" = { type = "disk"; device = "/dev/disk/by-id/nvme-ID"; content = { luks = { size = "100%"; content = { type = "luks2"; name = "luks-0"; passwordFile = "/tmp/disko/luks/nvme0.key"; askPassword = false; extraFormatArgs = [ "--cipher=aes-xts-plain64" "--hash=sha512" "--key-size=512" "--iter-time=5000" "--header=/dev/disk/by-id/usb-ID-part2" ]; extraOpenArgs = [ "--header=/dev/disk/by-id/usb-ID-part2" ]; settings = { header = "/dev/disk/by-id/usb-ID-part2"; preLVM = true; allowDiscards = true; }; initrdUnlock = true; content = { type = "lvm_pv"; vg = "NixOS"; }; }; }; }; }; } oh you're just missing type entirely. that's what's happening here (the disko test i just made for bare luks worked fine). try this: { disko.devices.disk."nvme0" = { type = "disk"; device = "/dev/disk/by-id/nvme-ID"; content = { type = "luks"; name = "luks-0"; passwordFile = "/tmp/disko/luks/nvme0.key"; askPassword = false; extraFormatArgs = [ "--cipher=aes-xts-plain64" "--hash=sha512" "--key-size=512" "--iter-time=5000" "--header=/dev/disk/by-id/usb-ID-part2" ]; extraOpenArgs = [ "--header=/dev/disk/by-id/usb-ID-part2" ]; settings = { header = "/dev/disk/by-id/usb-ID-part2"; preLVM = true; allowDiscards = true; }; initrdUnlock = true; content = { type = "lvm_pv"; vg = "NixOS"; }; }; }; }	15:19:45
SigmaSquadron	That did indeed work. Thank you!	15:23:20
11 Mar 2024
	$@fractivore:cyberia.club$ @fractivore:cyberia.club left the room.	06:53:07
	w joined the room.	14:21:25
	Vilsol joined the room.	19:14:16

* yeah, generation of options manpages for disko is still a complicated topic, since everything is kinda recursive it's hard to use the old doc genrators for that. Also the old list based approach was kind of lacking in the referencing and updating part, like if you add a partition in between everything has to shift a bit. and yes, size is the new recommended way, start is usually 0, which means the next available space in sgdisk, which should be reproducible if you run the same script multiple times. so partitions usually only need size. you can use priority if you want certain partitions to be in front of others, but so far this didn;t made any difference in any bootloader. porting table partitions to gpt partitions is sadly not a straight forward thing, since the latter alwas have a label and before it was optional.