Skip to content

Offer Touch ID on the 16-inch M4 Pro (J616s) - #140

Open
scottjones wants to merge 3 commits into
mainfrom
touchid-j616s
Open

scottjones wants to merge 3 commits into
mainfrom
touchid-j616s

Conversation

@scottjones

@scottjones scottjones commented Nov 14, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

The 16-inch M4 Pro (J616s) read Touch ID none. Its sensor sits at the root of the device tree, not under /soc. Like the Neo, it loads a sealed root key, and its calibration is sealed too.

 touchid-eligibility
-  sensor under /soc only; calibration = firmware-name entries run together
+  sensor under /soc or at the root; each firmware-name entry read on its own
+  admits apple,j616s (T6040); the 14-inch J614s stays unsupported
 touchid-root-key
-  MSRk only, seal must name the Neo (00008140)
+  --record MSRk|FSC2, --chip 00008140|00006040 (default: the Neo's key)
 touchid-enable
+  J616s: reseals mesacal-j616s.bin (FSC2) and msrk-j616s.bin (MSRk) from the factory records
+  refuses a firmware name outside apple/
 disable-fingerprint
+  deletes msrk-j616s.bin and mesacal-j616s.bin too; the next enable rebuilds them

Evidence

  • Before: touchid-eligibility on the J616s prints none.
    After: it prints available, with the sensor at /proc/device-tree/spi@429108000/fingerprint@0.

  • On the J616s, touchid-root-key --chip 00006040 [--record FSC2] rebuilds both files byte for byte as aurora-silicon's aurora-touchid-setup does. With them, fprintd enrols, verifies (verify-match), and matches again after a cold boot (the enclave restores the enrolment).

  • test/touchid-consent-test.sh gains a J616s section:

    • eligibility, including the cases where the factory records are missing;
    • both files rebuilt root-only;
    • a seal for another SoC refused;
    • an unsafe firmware name refused;
    • a calibration rebuilt by a run whose key then fails, rolled back;
    • an unlisted T6040 board with a root-level sensor refused;
    • disable deleting the key and the calibration.

    The section fails on the old scripts. All Touch ID tests and the scope, architecture and fixture checks pass locally.

  • 7.3 (SEP interface v5), 2026-11-18: the same J616s on the edgeB2 7.3 kernel (Move edge Macs to 7.3: linux-aurora, Mesa and aura (edgeB2), with aura as the boot loader omarchy-pkgs#3025) with libfprint-git from libfprint-git: Touch ID on SEP interface v4 and v5 omarchy-pkgs#2737: fprintd-verify matches, and the enclave logs interrupt-driven sensor capture.

  • Merged with current main; the omarchy-mac suite passes (123 ok).

Merge Danger

Door: two-way for the package files, not for the enclave: once Linux has advanced the enclave's anti-replay state, neither turning Touch ID off nor a revert moves it back.

Lands with the 7.3 edge move. The linux-aurora commit omarchy-pkgs pins today for 7.2 (bed544b, 7.2.11) ships no T6040 device tree at all, so a J616s gets no sensor node from it and reads none. That is read from the source, not run on a J616s with that kernel. On 7.3 it needs:

The M4 Pro as an install target is separate installer work. #160 (16-inch M4 Max) builds on this commit: merge this one with a merge commit, not a squash, so #160 needs no rebase.

Blast Radius: J616s

Other listed Macs keep their behaviour. An unlisted board whose sensor sits at the root of the device tree now reads unsupported:board instead of none, so fingerprint setup says "Touch ID can't be used on this Mac yet" where it said nothing, and setup goes on. The Neo's live node (7.3) names only apple/mesacal-j700.bin, so its key keeps the default apple/msrk-j700.bin. The Neo now passes its default chip and record explicitly, with the same rebuild and checks as before.

Open for maintainers:

🤖 Generated with Claude Code

The J616s read "none": its sensor sits on an SPI controller at the root
of the device tree, not under /soc, and its SoC wasn't on the list.
Like the Neo, its sensor loads a sealed root key, and its calibration is
sealed too. Its node names both (mesacal-j616s.bin, then
msrk-j616s.bin), and the calibration name was read as both entries run
together.

- touchid-eligibility finds the sensor at the root as well, admits the
  J616s, and reads each firmware-name entry on its own.
- touchid-root-key takes the SoC the seal must name (--chip) and the
  record kind (--record MSRk or FSC2). The default stays the Neo's key.
- touchid-enable rebuilds the J616s calibration and root key from its
  factory records on the ESP, checked against its iBoot System
  Container, and refuses a firmware name outside apple/.
- disable-fingerprint deletes either sensor key.

Tested:
- On a J616s, touchid-root-key rebuilds both files byte for byte as
  aurora-silicon's aurora-touchid-setup does. With those files Touch ID
  enrols and verifies, also after a cold boot.
- The consent test gains a J616s section, which fails on the old scripts.

Assisted-by: Claude:claude-opus-5-5
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@scottjones

Copy link
Copy Markdown
Collaborator Author

End-to-end on a J616s (7.3 test kernel), with this change cherry-picked onto the shipped pin (ea0097a) and installed as omarchy-mac 0.2.3-3:

  • From a fresh state (no agreement, driver not loaded, no sensor files), the boot status read available.
  • omarchy-setup-security-fingerprint: touchid-enable rebuilt mesacal-j616s.bin and msrk-j616s.bin (0600) and the activation read ready.
  • Enrolment completed.
  • sudo and the lock screen both unlocked with a finger.

One thing found along the way, outside this PR: the setup's pacman -S --needed libfprint-git fprintd usbutils replaced a newer local libfprint-git (-6, SEP interface v4 and v5) with the repo's -5 (v4 only). fprintd then refused the reader ("interface version 5, expected 4"). On the 7.3 kernel Touch ID needs libfprint-git -6 (omacom/omarchy-pkgs#2737) published with it.

@scottjones
scottjones marked this pull request as ready for review November 18, 2026 13:16
@scottjones
scottjones requested a review from maralcbr as a code owner November 18, 2026 13:16

@maralcbr maralcbr left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @scottjones, the J616s path reuses the Neo's rebuild cleanly: the seal record has to be named for the T6040, every record has to be present in this Mac's iBoot System Container, and the new firmware-name guard is a good addition. A few things before merge:

  1. The manual still describes the Neo only. mac-manual/content/04-security.md:42 should cover the 16-inch M4 Pro, including its calibration being rebuilt from the factory records (AGENTS.md: the manual changes with the behaviour). Suggest: "On a MacBook Neo or a 16-inch MacBook Pro with M4 Pro, setup also reads the sensor's key (and on the M4 Pro its calibration) from the factory records the installer left on the boot partition, into firmware files only root can read. Turning Touch ID off deletes the key." Adjust the last sentence to match point 4.
  2. Version. CONTRIBUTING.md:64 still asks for a bump in a behaviour-changing PR, and omarchy-mac/version is still 0.2.6. #30 would change that rule, but until it lands, please bump the version or get an explicit waiver.
  3. The calibration rollback has no test. rebuilt_calibration (touchid-enable:40-42,124) is never exercised: the other-chip case fails before anything is written. Could you add a case where the calibration rebuild succeeds and the key step fails, and check that mesacal-j616s.bin and touchid-on are gone? A container holding the FSC2 record, the seal and the trust bundle but not the MSRk record would do it.
  4. The calibration on disable. disable-fingerprint:20-27 deletes msrk-j616s.bin but keeps mesacal-j616s.bin, which touchid-root-key:8 now calls device-private and which is written root-only. Please either delete it with the key (the next yes rebuilds it, touchid-enable:119-126) or say in the README and manual why it stays.

One question: does the Neo's real firmware-name list only its calibration, or name apple/msrk-j700.bin second? The parse now reads a second entry on every board (touchid-eligibility:71-75). Any other name would rename the Neo's key, and disable-fingerprint:20 would miss it.

Nits, fine as follow-ups: the touchid-enable:2-4 header and a few Neo-only sentences in the README's fingerprint-readers paragraph now apply to the J616s too. That covers the FSC2 record, the device-private note, and unsupported:no-calibration rather than no-root-key. The removal error at disable-fingerprint:28 names both keys on any Mac. And the J614s refusal test uses a /soc sensor, so the root-level search isn't tested on an unlisted board.

Behaviour changes to note in the PR body:

  • An unlisted board whose sensor sits at the root of the device tree now reads unsupported:board instead of none, so fingerprint setup prints "Touch ID can't be used on this Mac yet" where it used to say nothing (enable-fingerprint:54-59). It's harmless, because setup goes on, but it isn't quite "Other Macs keep their behaviour".
  • "Two-way" holds for the package files but not for the enclave: once Linux has advanced the anti-replay state, neither disabling Touch ID nor a revert moves it back.

CI: everything passes, and the package suites ran touchid-consent-test.sh, including the J616s section.

Not code issues: admitting the J616s, and its xART writes to the enclave state shared with macOS, needs an explicit maintainer decision. So does the merge-commit request for #160.

For release: once this is on main, the next omarchy-mac release ships J616s eligibility whatever the kernel. The "nothing changes on 7.2" claim therefore carries weight, and it depends on 7.2's device tree having no live sensor node for the J616s, which I can't confirm from here. omacom/omarchy-pkgs#2737 looks like it covers the libfprint v5 build, and it has to be published before the invitation can reach a J616s on 7.3, since your run showed --needed installing -5. What's still missing is suspend and resume on the J616s, a check that macOS still unlocks with working Touch ID after a Linux enrolment (ideally with FileVault on), and a run at the final head.

Reviewed at b6d63e85692a.

— Marcelo's agent


key=/usr/lib/firmware/updates/apple/msrk-j700.bin
if [[ -e /etc/omarchy-mac/touchid-on || -e /etc/modprobe.d/omarchy-mac-touchid.conf || -e $key || -e $key.new ]]; then
keys=(/usr/lib/firmware/updates/apple/msrk-j700.bin /usr/lib/firmware/updates/apple/msrk-j616s.bin)

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This deletes the J616s key but keeps mesacal-j616s.bin, which is written root-only and described as device-private. Could it go too, or could the README say why it stays?

— Marcelo's agent

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed in 6b5d94a: disable-fingerprint now deletes mesacal-j616s.bin (and its .new copy) from /usr/lib/firmware/updates along with the key; the next enable rebuilds both. The README and the manual say so. The consent test checks that turning Touch ID off leaves no J616s file.

# checked against this Mac's iBoot System Container, like the root key.
echo "Reading this Mac's Touch ID calibration..."
mkdir -p -m 0755 "$firmware/updates/${calibration%/*}"
rebuilt_calibration=1

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nothing tests this rollback yet: please add a case where the calibration rebuild succeeds and the key step fails.

— Marcelo's agent

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed in 6b5d94a: the consent test adds a J616s run with a container holding the FSC2 record, the seal and the trust bundle but not the MSRk record. The calibration rebuilds, the key step fails, and the test checks that mesacal-j616s.bin (and any .new) and touchid-on are gone. With the rollback removed (lines 40-42 and rebuilt_calibration=1), that case fails.

calibration=$(tr -d '\0' <"$sensor/firmware-name")
mapfile -t names < <(tr '\0' '\n' <"$sensor/firmware-name")
calibration=${names[0]:-$calibration}
root_key=${names[1]:-$root_key}

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This now reads a second firmware-name entry on every board. Could you confirm the Neo's node names nothing other than apple/msrk-j700.bin there?

— Marcelo's agent

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Checked on a MacBook Neo (7.3.0-0test4), read-only: /proc/device-tree/soc/spi@385108000/fingerprint@0/firmware-name holds one entry, apple/mesacal-j700.bin, with no second entry. So root_key keeps the default apple/msrk-j700.bin, which is the name disable-fingerprint deletes. No change needed here.

if ! rm -f /etc/omarchy-mac/touchid-on /etc/modprobe.d/omarchy-mac-touchid.conf "$key" "$key.new"; then
echo "Touch ID could not be turned off: remove /etc/omarchy-mac/touchid-on, /etc/modprobe.d/omarchy-mac-touchid.conf and $key" >&2
if ! rm -f /etc/omarchy-mac/touchid-on /etc/modprobe.d/omarchy-mac-touchid.conf "${keys[@]}" "${keys[@]/%/.new}"; then
echo "Touch ID could not be turned off: remove /etc/omarchy-mac/touchid-on, /etc/modprobe.d/omarchy-mac-touchid.conf and ${keys[*]}" >&2

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: this names both keys on any Mac.

— Marcelo's agent

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed in 6b5d94a: disable-fingerprint removes only the files that exist, and if a removal fails the error names only those still there. The test now leaves a removable key beside a record that can't be removed and checks that the error names only the record.

…e manual

Turning Touch ID off now deletes mesacal-j616s.bin with the sensor key,
since the next yes rebuilds both from the factory records; the removal
error names only the files still there. The manual's Touch ID page and
the README's fingerprint-readers paragraph cover the 16-inch M4 Pro
where they described the Neo alone.

Tested:
- The consent test gains a J616s run whose calibration rebuilds and whose
  key can't (a container without the MSRk record): the calibration and
  the agreement are gone. It fails with the rollback removed.
- An unlisted T6040 board whose sensor sits at the root reads
  unsupported:board; it fails if the board check skips root sensors.
- Turning Touch ID off deletes the J616s calibration and its temporary
  copy, and a failed removal names only what is left.
- omarchy-mac/test/all passes (123 ok), as do the scope, architecture
  and fixture checks and the manual build.

Assisted-by: Claude:claude-opus-5-5
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@maralcbr maralcbr left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@scottjones re-reviewed at 6b5d94a: every code point from the last round is fixed, and I found no blockers in the update.

  • The manual (04-security.md:42) and the README now cover the 16-inch M4 Pro, its calibration included.
  • Turning Touch ID off deletes mesacal-j616s.bin and its .new with the key (disable-fingerprint:21-25). The error names only what's left. Both are tested (touchid-consent-test.sh:662-667,727-734).
  • The calibration rollback is now exercised. j616s-no-key-isc.img lets the calibration rebuild and the key step fail, and the test checks that nothing J616s and no touchid-on is left (:718-723).
  • The unlisted-board refusal now uses a sensor at the root of the tree (:699-701).
  • Thanks for checking the Neo's live firmware-name. With one entry, msrk-j700.bin stays the key name, which disable-fingerprint deletes.

Nits, fine as follow-ups: disable-fingerprint:20 says "rebuilds it" over three files, so "them" would fit. The README's "the MSRk record's IM4P" now describes the FSC2 calibration too, so "the MSRk or FSC2 record's" would be accurate.

On the version: the body says #30 has merged, but this head's CONTRIBUTING.md:64 still has the old bump rule, so I can't confirm it from the branch. If #30 is on main, no bump belongs here, and #161 looks like it covers the release.

CI: everything passes, and the package suites ran touchid-consent-test.sh, including the J616s section.

Still open, and not code issues:

  • Admitting the J616s, with its xART writes to the enclave state it shares with macOS, needs an explicit maintainer decision.
  • So does the merge-commit request for #160.
  • So does whether this needs to land before #161 cuts 0.2.7, the version omacom/omarchy-pkgs#3040 is meant to pin for the 7.3 move (for now it's a 0.2.6 placeholder).

For release (not the source merge):

  • omacom/omarchy-pkgs#2737 has to be published before the invitation can reach a J616s on 7.3.
  • The "reads none on 7.2" claim still comes from reading the source, not from a run.
  • Still missing: suspend and resume on the J616s, a check that macOS still unlocks with Touch ID after a Linux enrolment (ideally with FileVault on), and a run at the final head.

Reviewed at 6b5d94a20bcc.

— Marcelo's agent


key=/usr/lib/firmware/updates/apple/msrk-j700.bin
if [[ -e /etc/omarchy-mac/touchid-on || -e /etc/modprobe.d/omarchy-mac-touchid.conf || -e $key || -e $key.new ]]; then
# What Touch ID setup rebuilt from the factory records; the next setup rebuilds it.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: "the next setup rebuilds them", now that the list covers three files.

— Marcelo's agent

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.

2 participants