Add configration option for partition UUIDs #951
Conversation
| @@ -74,6 +79,16 @@ in | |||
| description = "Name of the partition"; | |||
| default = name; | |||
| }; | |||
| uuid = lib.mkOption { | |||
There was a problem hiding this comment.
Maybe call the option guid so it's harder to mistake for the filesystem's id?
There was a problem hiding this comment.
I called it uuid because, within the Linux context, it's always referred as that. For example, the name of /dev/disk/by-partuuid, blkid calls it PARTUUID etc. I could have named it partuuid, but I figured that if there is ever a separate filesystem UUID option, that would be distinguished as partition.content.uuid vs partition.uuid.
| default = null; | ||
| example = "809b3a2b-828a-4730-95e1-75b6343e415a"; | ||
| description = '' | ||
| The UUID (also known as GUID) of the partition. Note that this is distinct from the UUID of the filesystem. |
There was a problem hiding this comment.
Document which types of partition table support the option. Maybe also add a corresponding assertion to the module, to give a clear error rather than unexpected behaviour.
There was a problem hiding this comment.
Does disko support anything other than GPT? At least, I included it in the gpt.nix module assuming that it would only be available for GPT.
There was a problem hiding this comment.
currently we still support DOS with the legacy table, but that is deprecated and it's fine to do it here for gpt only
| args = '' | ||
| --partition-guid="${toString partition._index}:${if partition.uuid == null then "R" else partition.uuid}" \ | ||
| --change-name="${toString partition._index}:${partition.label}" \ | ||
| --typecode=${toString partition._index}:${partition.type} \ | ||
| "${config.device}" \ | ||
| ''; | ||
| createArgs = '' | ||
| --align-end ${lib.optionalString (partition.alignment != 0) ''--set-alignment=${builtins.toString partition.alignment}''} \ | ||
| --new=${toString partition._index}:${partition.start}:${partition.end} \ | ||
| ''; |
There was a problem hiding this comment.
not sure why this was split, but isn' the device more of a createArg?
There was a problem hiding this comment.
Just to avoid repeating itself, nothing more. It seemed nicer this way. I can revert it to the original style if you want.
Closes #948. This is my first ever Nix-related PR, so I'd appreciate feedback if you have any!
I wasn't sure if mdadm does expose
partuuid, I assumed that it did but that may be wrong. I did run the tests, but I wasn't sure if it was wise to include UUIDs in example files. It's probably not a good idea for users to copy the UUID, they should generate their own.