Skip to content

arm64: dts: apple: t8132: enable the Secure Enclave on every M4 board - #2247

Draft
RyanTheTide wants to merge 3 commits into
integration/aurorafrom
spec-m3m4/t8132-sep-all-boards
Draft

RyanTheTide wants to merge 3 commits into
integration/aurorafrom
spec-m3m4/t8132-sep-all-boards

Conversation

@RyanTheTide

@RyanTheTide RyanTheTide commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

Stacked on #2245 (itself on #2244); merge those first.

Problem. Every T8132 has the SEP, and its driver must bind on every board: the enclave takes the deep-sleep request, and Touch ID and the secure storage paths go through it. The DT enabled the SEP and its DART only for the J713/J715, so the J604 (fingerprint sensor enabled), the iMacs and the Mac mini never bind it.

Change. Enable sep and sep_dart in t8132.dtsi, give the SEP power-domains = <&ps_sep>, <&ps_gpio> there, and mark ps_sep firmware-owned in t8132-pmgr.dtsi. The J713/J715 power ownership file (also used by the legacy-loader adapter) still disables ps_sep, so that board file keeps a one-line re-enable.

Testing. cpp + dtc on all six T8132 boards: same warning counts. J713 and J715 decompile to the same trees except for dropped explicit status = "okay" lines. On J604, J623, J624 and J773g the SEP and its DART are enabled with their domains.

Still to run. A boot on the J604, an iMac and the J773g with the SEP bound and a clean s2idle; a Touch ID enroll and match on the J604.

Assisted-by: LLM (Claude:Opus 5.5)

RyanTheTide and others added 3 commits November 19, 2026 22:14
Every T8132 has the ANE; the DT enabled it, with its three DARTs, on the
J713 only, the board it was first run on (#1952). The node's resources
and power domains are the SoC's, and the driver binds only when the
loader hands off the firmware preload (aura#260 does so for any
apple,t8132-ane node), so enable it in the SoC file and drop the J713
copy. The J713's DT is unchanged.

Not tested on the J604, J715, J623/J624 or J773g (desk audit, lane
SPEC-M3M4).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…nclave and display power wiring

The J715 (MacBook Air 15", M4) is the J713 board with a larger panel and
six amps; its ADT matches the J713's apart from the radio island and the
ATC PHY trims (t8132-j715.dts). Yet the deep-sleep controller, both
cluster power gates, the SEP (which Touch ID's match path needs, and the
J715 DT already enables the Mesa sensor), the DCP and external-DCP
power gating, the USB4/DCP DART clock-protection slices, the radio
port's CLKREQ and the ACIO PHY links sat in t8132-j713.dts only, with
"tested on this machine only" as the reason.

Move them unchanged into the shared J713/J715 file. The J713 DT compiles
to the same tree (dtc -s output identical before and after).

Not tested on a J715: needs a J715 boot,
Touch ID enroll, and an s2r cycle with and without a dock.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every T8132 has the SEP, and its driver must bind on every board: the
enclave takes the deep-sleep request, and Touch ID (the J604 and the Airs
have the sensor enabled) and the secure storage paths go through it. The
DT enabled the SEP and its DART only in the J713/J715 board file, so the
J604, the iMacs and the Mac mini never bound it.

Enable both in the SoC file, give the SEP its power domains there, and
mark the enclave's power state firmware-owned in the SoC power tree. The
J713/J715 ownership file still leaves ps_sep off for the legacy-loader
adapter, so the board file keeps one line turning it back on.

J713 and J715 compile to the same tree apart from dropping explicit
status = "okay" lines (decompiled and compared). Not tested on the J604,
J623/J624 or J773g: needs a boot with the SEP bound on each, and a Touch
ID enroll on the J604.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@RyanTheTide

Copy link
Copy Markdown
Contributor Author

Blockers: 3.

  1. The SEP driver refuses every T8132 Mac this PR targets. profile::select() returns the T8132 profile only for apple,j713 or apple,j715 (drivers/soc/apple/profile.rs:712-714). For J604, J623, J624 and J773g it falls through to Err(ENODEV) (:721), and the KUnit case at :785-791 asserts exactly that for J604. With this PR those boards get an enabled sep node whose probe fails, so none of the stated goals happens there: no sleep request, no Touch ID on the J604, no secure-storage path. The driver side has to come first: pick the profile by SoC, and describe the sensor bus in the DT so the desktops can say they have none. Prove it with a bind row. Until then this DT change only adds a failed probe on four boards.

  2. None of the three commits has a Signed-off-by (5df0283, cb64329, e2e9109), and all three end with Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>. Add the human sign-off and drop the trailer.

  3. The PR conflicts with integration in t8132-j713-j715.dtsi, inherited from arm64: dts: apple: t8132-j713-j715: give the J715 the J713's sleep, enclave and display power wiring #2245 (see there).

The DT side is otherwise consistent. sep_mbox is enabled for all six boards through t8132-jxxx.dtsi. ps_sep stays apple,always-on, and the new apple,firmware-owned sits next to it in t8132-pmgr.dtsi. J713/J715 still turn ps_sep back on after t8132-j713-pmgr.dtsi disables it.

HW row: once the driver selects by SoC, run this head on the J604, one iMac (J623/J624) and the J773g: apple-sep bound with no ENODEV, one clean s2idle, and Touch ID enroll and match on the J604. Also a J713 regression with Touch ID and deep sleep. 0 AVC, 0 failed units, no new kernel errors on each.

Assisted-by: LLM (Claude:Opus 5.5)

@RyanTheTide

Copy link
Copy Markdown
Contributor Author

Closing for now: the enclave driver can't bind on these boards until the profile is picked by SoC (#2631); the DT enable can come back on top of that with boot rows.

@RyanTheTide RyanTheTide reopened this Nov 22, 2026
@RyanTheTide

Copy link
Copy Markdown
Contributor Author

Reopened and kept as a draft; the comment above lists what it still needs before review.

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