Skip to content

Give a Mac's disk one password, with no recovery key - #657

Merged
maralcbr merged 2 commits into
quattro-upstreamfrom
mac/norecovery-owner-only
Sep 28, 2026
Merged

maralcbr merged 2 commits into
quattro-upstreamfrom
mac/norecovery-owner-only

Conversation

@maralcbr

Copy link
Copy Markdown
Member

Owner decision (2026-09-28): drop the disk recovery key and match upstream's single-password default (omacom#2005). Counterpart of the upstream runtime change on omacom#13362.

  • Boot check (omarchy-mac-boot): accepts an owner-only disk (one keyslot, owner_slot recorded) and still a Mac set up before with owner and recovery slots recorded; the header must hold exactly the recorded slots. luks-slots already treated an empty recovery= as none and a missing one as keep.
  • Owner setup: no recovery slot, no recovery key screen, no omarchy-drive-recover arming; the re-key keeps the owner's slot and runs luks-slots owner=<slot> recovery=. In-place encryption and the owner password unlocking the disk are unchanged. An older runtime's interrupted setup journal (any phase, including done) finishes with the owner's slot alone.
  • Existing Macs keep their recovery slot as a plain second passphrase; nothing removes it. omarchy-drive-recover, its units, luks-recovery.sh and the arming migration go; a new migration removes armed units (sudo only when present). omarchy-drive-password still refuses a recovery-key-form password, but only on a disk with more than one slot.
  • Mac manual, lifecycle-dispatch doc, convergence spec and the mac-image acceptance harness (conversion, password-change and factory-reset now expect one slot; recovery-reset removed) follow.

Release order: the omarchy-mac-boot from this commit must publish with or before any runtime that drops the key; an older boot check fails an owner-only disk after setup. Noted on omacom/omarchy-pkgs#673.

Tested in an Arch container: packages/omarchy-mac/boot/test/all passes (new owner-only and two-slot boot-check fixtures, owner-only luks-slots); test/shell has the same 15 environment failures as the base commit and nothing else (rewritten provision-owner and re-key journal tests, older-journal phases, legacy drive-password guard, retirement migration). Not run on hardware (demo freeze). Design debate and a second review: fixed the done-phase journal and stale docs; accepted that a resumed older setup takes the old recovery key if the owner types it as the password.

Not here: the installer's encryption help still promises a recovery key (PlainLanguage.encryptLinuxDiskRecovery in omacom/omarchy-mac-installer).

Owner setup no longer adds, shows or arms a disk recovery key: the re-key keeps
the owner's slot alone and records it with an empty recovery=. The Mac boot
check accepts that owner-only disk, and still a Mac set up before with an
owner and a recovery slot recorded. That recovery slot stays as a plain second
passphrase: nothing removes it, and omarchy-drive-recover, its units and the
migration that armed them go, with a migration that disarms the units.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant