This is the only way to assign devices rather than fixed gpt partitions.
Without reading the code it's not very obvious how disko actually
assigns devices to zpools.
A config like
```nix
{
vdev = [
{
mode = "mirror";
members = [ "data1" "data2" ];
}
{
members = [ "data3" ];
}
];
}
```
would result in the following command:
```shell
zpool create -f <name> mirror /dev/data1 /dev/data2 /dev/data3
```
which would result in a single vdev with a 3-way mirror, rather than a
vdev with a 2-way mirror and a second vdev with a single disk. By
reordering the vdevs to handle those with an empty mode first we
transform this into:
```shell
zpool create -f <name> /dev/data3 mirror /dev/data1 /dev/data2
```
which does have the desired outcome.
The type keyword was included before every vdev:
```shell
zpool create -f <name> /dev/sda log mirror /dev/sdb /dev/sdc log mirror /dev/sdd /dev/sde
```
but this is incorrect and should instead be:
```shell
zpool create -f <name> /dev/sda log mirror /dev/sdb /dev/sdc mirror /dev/sdd /dev/sde
```
lvcreate -l does not accept a '100%' parameter which currently leads to a crash. THis change automatically changes `100%` to `100%FREE` leading to the intended behavior.
Fixes#130
This should fix pretty much all cases where spaces or other special
characters would break disko due to improper quoting. I searched for all
instances of '.label', '.device' and '.name', so I believe I caught
whatever I could.
In some cases I changed single quotes to double quotes for consistency.
I know we don't usually fix bugs in the legacy table type, but it was so
easy I couldn't resist.
For disk type, option name used in disk partlabel
naming should be short. While setting a specialized
option imageName allow us to create image with long name
without side-effects.
Usually this is added by nixos-generate-config.
However now that we run nixos-generate-config before disko or we even
run nixos-facter, we likely not see the LVM block devices anymore.
Use `zfs set -u` to update the `mountpoint` flag without mounting or
unmounting the filesystem. This flag was added in OpenZFS 2.2.0, which
was released October 2023.
The previous logic would not update `mountpoint` if `config.options` was
empty or only contained `mountpoint`.
As `makeDiskImages` always requires a NixOS configuration, we can
simplify the code by convering it into a NixOS module. Then we can make
it responsible for populating `system.build.diskoImages` and
`system.build.diskoImagesScript`.
This is not an issue when initially creating the partition because
device mapper will just map on top of the filesystem, but it breaks
incremental updates because the check for `TYPE=` will fail, and the
subsequent attempt to run `mkswap` also fails:
mkswap: cannot open /dev/disk/by-partlabel/disk-nix-store-swap: Device or resource busy
Instead, just don't do anything, NixOS will take care of it.
Running `zfs set mountpoint=/mnt/my-ds tank/my-ds`, ZFS may try to
unmount the dataset even if the mountpoint didn't change.
To avoid the confusing error message, this command is now only run when
the mountpoint actually changes.
