Skip to content

J514s: SEP and Touch ID (T6030) - #2633

Open
RyanTheTide wants to merge 4 commits into
integration/aurorafrom
rescue/pr581-j514s-touchid-v2-rebase-20261121
Open

RyanTheTide wants to merge 4 commits into
integration/aurorafrom
rescue/pr581-j514s-touchid-v2-rebase-20261121

Conversation

@RyanTheTide

@RyanTheTide RyanTheTide commented Nov 21, 2026 •

Copy link
Copy Markdown
Contributor

Replaces #1691, rebased onto current integration/aurora (conflicts in drivers/soc/apple/profile.rs: the Touch ID Mac list in every_touch_id_mac_gets_a_sensor_the_shared_path_decodes grew on the base, so its length is now 19 with J514s added; all base entries kept).

Touch ID on the 14" M3 Pro MacBook Pro (J514s, T6030), rebased onto integration/aurora at 78ea426.

This is Adam Harris's series from #581, taken from the latest revision on his branch (adammharris/linux-aurora:feat/t6030-j514s-touchid @ fd840c9). Authorship and his sign-offs are unchanged. The code is his; the rebase applied without conflicts.

Changes

  • t6030-coprocessors.dtsi, t6030-peripheral-buses.dtsi: integration already describes the SEP, its DART and mailbox, and spi2 (all disabled). This adds the SEP DART's DMA window from the J514s ADT (vm-base 0x10000004000, vm-size 0xffff0000, the same BIT(40) aperture as T8122) and the Mesa sensor as spi2's child (CS 0, 8 MHz, CPHA). Everything stays disabled.
  • t6030-j514s.dts: enables the mailbox, spi2 and the sensor. Power is NUB GPIO 33, active high; the data-ready line (AP GPIO 200) is left out and the driver polls, as on J613. Calibration is apple/mesacal-j514s.bin from the Mac's own iBoot system container.
  • soc: apple: sep: a T6030 profile, which is T8122 with J514s's spi2 base (sensor 0x3352, SPI mode 1, variant-5 key store, xART guard level 1, wide DMA mask). 610b968 selects it by the apple,t6030 SoC compatible, and the board DT decides whether the SEP attaches. The J514s enables sep_mbox. The J516s keeps it disabled (its sensor wiring is unchecked), so with the mailbox check from soc: apple: sep: select the platform profile by SoC #2631 its SEP and SEP DART stay disabled.
  • tools/aurora-sep/tests/test_t6030_sep.py: checks the J514s tree against the ADT values, that J516s leaves the SEP and sensor off, and the profile match.

Testing

On this branch (no hardware):

  • Rebased onto integration/aurora c3d01cb (2026-11-21, arm64 LLVM=1 cross): test_t6030_sep.py 4/4 and test_t8122_sep.py 4/4; KUnit apple_sep* on arm64 QEMU (drivers/soc/apple/sep.kunitconfig) 306/306 passed; drivers/soc/apple/ builds with W=1, no warnings in profile.rs or xart_guard.rs; t6030-j514s.dtb and t6030-j516s.dtb with CHECK_DTBS=y W=1 (dtschema 2026.9): 129 warning lines on base and head, 0 new; run-host-tests.py --only t6030-pmp-dt 1/1, --only sep-suspend-errors 22/22.
  • test_t6030_sep.py 4/4 and test_t8122_sep.py 4/4 pass with the kernel's dtc.
  • t6030-j514s.dtb and t6030-j516s.dtb build with W=1 and add no warnings over integration (34 before, 34 after, identical).
  • drivers/soc/apple/ builds with W=1 on the Fedora-base + asahi.config release config (CONFIG_APPLE_SEP=m); no warnings in profile.rs or xart_guard.rs.

Hardware results on the same code before the rebase:

  • Adam Harris (@adammharris), J514s, system firmware 27.0, macOS on the same disk: read-only first load, keybag provisioning with the guard protecting all 10 macOS session records and refusing nothing, sensor up 18 ms after power-on, fprintd enrol and match, a match after reboot and after suspend, and macOS Touch ID still working afterwards.
  • @c4software, a second J514s on Omarchy 4.5.0 with the same series applied to linux-aurora 7.2.11.aurora14: read-only first load, provisioning (level 1, 10 records protected, 0 refusals), enrol and match, enrolment surviving a reboot. A suspend retest with no display attached is still open there.

@adammharris, could you boot this branch on your J514s and confirm the SEP attaches and a match works (after a reboot and, if you can, a suspend)? A short result row here is enough. @c4software, a row from your J514s would be welcome too.

Thanks to Adam Harris for the bring-up, to c4software for the second-machine test, and to malik-na for the review on the earlier PR.

The T6030 profile is selected by the SoC compatible; each board device tree decides whether the SEP and the sensor are enabled.

Assisted-by: LLM (Claude:Opus 5.5)

adammharris and others added 4 commits November 21, 2026 11:01
…ouch ID sensor

The SEP, its DART and mailbox, and spi2 are already described (disabled)
in t6030-coprocessors.dtsi and t6030-peripheral-buses.dtsi. Touch ID
needs two things more, as the J514s ADT (system firmware 27.0) records
them:

- The SEP DART only accepts IOVAs in the window iBoot programmed: vm-base
  0x10000004000, vm-size 0xffff0000, the same BIT(40) aperture as on
  T8122. Describe it with apple,dma-range and apple,dma-offset.
- The Mesa sensor sits on spi2, chip select 0, at 8 MHz with CPHA. Add it
  as spi2's child.

Everything stays disabled: apple_sep enables the SEP and its DART when it
loads, and boards enable the mailbox, the bus and the sensor.

Assisted-by: LLM
Signed-off-by: Adam Harris <amh421@icloud.com>
Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
Enable the SEP mailbox, spi2 and the Mesa sensor (0x3352) on the 14" M3
Pro. Per the J514s ADT, the sensor is powered from NUB GPIO 33, active
high. As on J613, its data-ready line (AP GPIO 200) is left out and the
driver polls. The calibration comes from this Mac's own iBoot system
container as apple/mesacal-j514s.bin.

J516s shares the SoC, but its sensor wiring is unchecked, so it stays
off.

Assisted-by: LLM
Signed-off-by: Adam Harris <amh421@icloud.com>
Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
The J514s SEP is pre-booted by iBoot and speaks the variant-5 key store,
and its sensor is the same 0x3352 part T8122 carries, in SPI mode 1. So
J514s takes the T8122 profile with its own spi2 base, including the wide
DMA mask and xART guard level 1 from the first boot. Only J514s is
matched; the other T6030 Mac (J516s) has not had its ADT read.

Add a test that checks the J514s device tree against the ADT values and
that J516s leaves the SEP and the sensor off.

Tested on J514s, system firmware 27.0, with macOS on the same disk:
- the SEP attaches warm and advertises 12 endpoints
- the keybag is provisioned; the guard protects all 10 existing macOS
  session records and has refused nothing
- the sensor answers 18 ms after power-on and registers its serial
- enrolment and a match work, and a match still works after a reboot
  and after suspend
- macOS Touch ID still works afterwards

Assisted-by: LLM
Signed-off-by: Adam Harris <amh421@icloud.com>
Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
The T6030 profile was picked only when the root compatible also named
the J514s, so the 16-inch M3 Pro (J516s) and any other T6030 Mac had no
profile at all. The SEP, its DART and spi2 are SoC blocks at the same
addresses on every T6030 Mac, so the SoC compatible selects the profile.

Whether a board actually has the sensor wired is described by its device
tree: only the J514s enables sep_mbox, spi2 and the mesa node today, so
nothing binds on the J516s until its board file enables them.

Signed-off-by: Ryan <ryan@aurorasilicon.org>
@RyanTheTide

Copy link
Copy Markdown
Contributor Author

Blockers: 1.

  1. drivers/soc/apple/profile.rs:728 now returns the T6030 profile for any apple,t6030 root, and the module init at drivers/soc/apple/sep.rs:3646-3650 runs dt::enable_sep_and_dart() for every WarmRegister profile without checking the board. On a J516s (or any other T6030 board), that applies a changeset that sets the SEP DART's DMA window and status = "okay" on the DART and on the SEP node (drivers/soc/apple/dt.rs:224-256). The DART then probes over the DART iBoot left live for the pre-booted SEP, and the SEP platform device is created, even though the J516s tree leaves sep_mbox disabled. The commit message of 610b968 says "nothing binds on the J516s until its board file enables them". That is only true for spi2 and mesa; it is not true for the SEP and its DART, and test_t6030_sep.py checks only the static DTB, so it doesn't catch this. Selecting by SoC is the right direction. To make the board DT the switch, have the init path skip the changeset when the SEP's mboxes target is not available (the J514s enables it, the J516s doesn't), and log that the board leaves the SEP off. Add a KUnit or host test for the J516s case.
  2. The PR body still says "Only apple,t6030 + apple,j514s is matched" under Changes, which 610b968 reverses. Fix the bullet so it matches the head.

The first three commits carry Assisted-by: LLM in their messages. They are Adam's commits, carried unchanged, so leave them as they are.

HW row: J514s on this head (fleet kernel, default cmdline): SEP attaches, fprintd-verify matches after a reboot and after s2idle, 0 AVC, 0 failed units, 0 new kernel errors. Also a J516s boot of the same kernel showing apple_sep leaves the SEP DART and the SEP disabled (no "SEP DART changeset applied" line) and 0 new DART errors.

Assisted-by: LLM (Claude:Opus 5.5)

@RyanTheTide

Copy link
Copy Markdown
Contributor Author

On the two points:

  1. soc: apple: sep: select the platform profile by SoC #2631 covers this. Its dt::enable_sep_and_dart() returns before touching the SEP or its DART unless the SEP's mboxes target is available, and logs "the device tree leaves the SEP mailbox disabled on this board; the SEP stays disabled". The J516s tree keeps sep_mbox disabled, so on that board neither the changeset nor the DART probe happens. soc: apple: sep: select the platform profile by SoC #2631 lands first and this PR is rebased on it, as the two already overlap in the Touch ID KUnit list. The J516s boot row you asked for (no "SEP DART changeset applied" line, 0 new DART errors) still holds: that rebased head needs it, and it needs a J516s owner because we have none in the lab.
  2. The Changes bullet now matches 610b968: the profile is chosen by the SoC, and whether the SEP attaches comes from the board DT through the mailbox.

Assisted-by: LLM (Claude:Opus 5.5)

@RyanTheTide

Copy link
Copy Markdown
Contributor Author

Merge-train adversarial review (MERGETRAIN-LINUX2) at 610b968: BLOCKERS.

Not merging until the blockers below are fixed or answered; the details give file:line and a failure case.

Full review

#2633 J514s: SEP and Touch ID (T6030) @610b9683

VERDICT: BLOCKERS

Blockers

  1. drivers/soc/apple/profile.rs:728 now returns the T6030 WarmRegister profile for any apple,t6030 root. At module init, drivers/soc/apple/sep.rs:3646-3651 then calls dt::enable_sep_and_dart() without checking the board, and drivers/soc/apple/dt.rs:188-256 sets status okay on the SEP DART (with the J514s-ADT DMA window from t6030-coprocessors.dtsi) and on the SEP node. At this head nothing checks whether sep_mbox is available. The 11-21 reply says soc: apple: sep: select the platform profile by SoC #2631 adds that check and that "this PR is rebased on it", but soc: apple: sep: select the platform profile by SoC #2631 is still OPEN, is not in base b241e35, and is not in this PR's history (dt.rs at head is byte-identical to base around line 188). Failure: on a J516s (t6030-j516s.dts leaves sep_mbox/spi2/mesa disabled), loading apple_sep probes the SEP DART over the DART iBoot left live for the pre-booted SEP, with an aperture read only from the J514s ADT, and creates the SEP platform device with a disabled mailbox. 610b968's message ("nothing binds on the J516s") is wrong for this head. Fix: merge soc: apple: sep: select the platform profile by SoC #2631 first and rebase onto it, or drop 610b968 so the profile stays J514s-matched until then. Add a host or KUnit case for the J516s init path, not only the static DTB check in tools/aurora-sep/tests/test_t6030_sep.py:113-118.
  2. Commit hygiene: b710d8c, b9e39ab and 0571993 carry Assisted-by: LLM trailers, which rule 4 forbids. The existing review chose to leave Adam's commits unchanged. Under the binding rule this still blocks, so ask Adam to reword them, or reword them with his consent and keep his Signed-off-by.

Nits

  • 610b968's sign-off is Ryan <ryan@aurorasilicon.org> but the other commits use Ryan Murray. Use the full name for DCO consistency.
  • profile.rs:447 hardcodes controller_base: 0x2_8910_8000, which duplicates spi2's reg in t6030-peripheral-buses.dtsi. This is the existing profile pattern, not new here, but it is a per-SoC constant the DT already carries.
  • 0571993's message still says "Only J514s is matched", and the next commit reverses that. Acceptable in a series, but the profile.rs:433-443 doc comment should not claim the hardware results for "T6030" in general: they are J514s results.

Missing rows

  • No hardware row on this head. The J514s results (Adam, c4software) predate the rebase. Needed: J514s on this head with the fleet kernel and default cmdline: SEP attaches, fprintd-verify matches after a reboot and after s2idle, 0 AVC, 0 failed units, 0 new kernel errors. The c4software suspend retest is still open.
  • J516s: one boot of the final head (after soc: apple: sep: select the platform profile by SoC #2631) showing that apple_sep leaves the SEP DART and SEP disabled (no "SEP DART changeset applied" line) and 0 new DART errors. There is no J516s in the lab, so this needs an owner.
  • If the J514s's live DT comes from its Stage 2, the dts commits reach hardware only after a Stage 2 DTB refresh. Say which DT the row booted.

@RyanTheTide

Copy link
Copy Markdown
Contributor Author

Clarification to my merge-train review above: the commit-trailer point is a nit, not a blocker (already-pushed history stays as it is, no force push). The other blockers listed there still stand.

@DjDeveloperr DjDeveloperr left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Pre-PR checklist review (§1–§6) at 610b9683bbeb: REQUEST_CHANGES (2 blockers + missing §5 hardware rows).

Blockers

  1. Code / safety blocker — J516s probes SEP DART and binds apple,sep at this head (§1/§3):

    • drivers/soc/apple/profile.rs:728 (commit 610b9683) now returns &T6030 (Bootstrap::WarmRegister) for every apple,t6030 root compatible, including arch/arm64/boot/dts/apple/t6030-j516s.dts (which leaves sep_mbox, spi2, and mesa disabled).
    • At 610b9683, drivers/soc/apple/sep.rs:3647-3651 calls dt::enable_sep_and_dart(p.dart_window) for every WarmRegister profile, and drivers/soc/apple/dt.rs:188-260 has no sep_mbox availability check (#2631 is still open and is neither in base 024cec8c9347 nor in this PR's history).
    • Failure mode: Loading apple_sep on a J516s applies the OF changeset to set status = "okay" on sep_dart (probing Linux's DART driver over the DART iBoot left live for the pre-booted SEP, with an aperture taken only from the J514s ADT) and sets status = "okay" on sep, creating the platform device with a disabled mailbox. Commit 610b9683's message ("nothing binds on the J516s until its board file enables them") is inaccurate at this head, and tools/aurora-sep/tests/test_t6030_sep.py:113-118 only inspects the static DTB.
    • Fix: Land #2631 first and rebase this PR onto it (or carry the mailbox check here), and add a test covering the disabled-mailbox init path.
  2. §4 Attribution — AI trailers in commit messages:

    • Commits b710d8c7f27a, b9e39ab31d07, and 0571993ee239 carry Assisted-by: LLM trailers in the commit messages. Per pre-pr-checklist.md §4, commits must carry human identity and human Signed-off-by: only (no Assisted-by: or AI trailers in commits; disclosure belongs at the end of the PR body, where it already appears).
    • Since these commits were already rebased onto this branch with an added Signed-off-by:, please drop the Assisted-by: LLM commit trailers (with Adam's consent or via Adam) while preserving Adam's authorship and Signed-off-by:.

Missing §5 Evidence (at the live head)

  • J514s on the rebased head (default cmdline, noting which DTB booted): SEP warm attach, fprintd-verify after reboot and after s2idle, 0 SELinux denials (AVC), 0 failed units, 0 new kernel errors. (Prior Adam Harris / c4software rows predate the rebase and 610b9683.)
  • J516s after rebasing onto #2631: boot confirming apple_sep leaves sep_dart and sep disabled (no "SEP DART changeset applied" line) and 0 new DART errors.

Nits

  • Commit 610b9683 uses Signed-off-by: Ryan <ryan@aurorasilicon.org> whereas the other commits use Ryan Murray <ryan@aurorasilicon.org>.
  • drivers/soc/apple/profile.rs:433-443 doc comment on T6030 states the hardware verification results; worth clarifying they are J514s results rather than T6030-wide.

CI: 8/8 checks green (dt-checks, host-tests, source, rust-build, kunit, 3x DWC3/ATC).

Assisted-by: LLM (Gemini 4 Neon)

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.

3 participants