Skip to content

arm64: dts: apple: t6050: wire the PMIC RTC alarm interrupt - #2007

Open
joshuaswarren wants to merge 5 commits into
aurora-silicon:integration/aurorafrom
joshuaswarren:dhp4/t6050-rtc-alarm-irq
Open

joshuaswarren wants to merge 5 commits into
aurora-silicon:integration/aurorafrom
joshuaswarren:dhp4/t6050-rtc-alarm-irq

Conversation

@joshuaswarren

@joshuaswarren joshuaswarren commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

Status at e785c70a3c: this PR changes only the T6050 PMIC RTC alarm wiring. The RTC alarm path has not been exercised.

[Joshua's agent] Ready. The build and probe evidence below does not include a runtime wake test.

Split out of #1743 (gen4 SPMI interrupts), one fix per PR. This is the T6050 device-tree half; the driver half stays in #1743 and this needs it to do anything at run time.

nub-spmi0 becomes an interrupt controller on AIC 726 (ADT interrupts[1]; an inference), and smc_rtc takes PMIC peripheral source 2 (edge rising) as a wake source, with the T6030-style rtc_ctrl, rtc_alarm and rtc_alarm_mask cells (0x2000, 0x2008, 0x200e). One file, arch/arm64/boot/dts/apple/t6050-die0.dtsi, +25 -3.

Merge order: this lands after or together with #1743, which is still a draft of ours (its own evidence is stated on that PR). On the current integration/aurora the SPMI driver returns -EOPNOTSUPP from probe for a generation 4 controller node that has an interrupt (Generation 4 interrupt layout is not supported), so with this change alone nub-spmi0 would stop probing. #1743 teaches the driver the generation 4 interrupt layout.

Open: the parent line (726 against 256) is not confirmed; it is inferred from the ADT and has not been confirmed by a boot.

Validated on, and reach

What ran where:

  • 13-inch M1 MacBook, build only: t6050-k114c.dtb and t6050-j714c.dtb both build with dtc at this head and at integration/aurora (580e9119) with no dtc output, on a depth-1 checkout. The two dtbs grow by 354 bytes each. No CHECK_DTBS schema run.
  • 14-inch M5 Max MacBook Pro (K114c, T6050): read-only probes of the SPMI bank only (the alarm sets +0x600 bit 2); this dtsi change was not booted there or anywhere.

Applies beyond the M4: no. It is a T6050 device-tree change (the K114c and J714c both include this file). Nothing was run on an M4, an M5 Pro or any other Mac.

m tester run (omarchy-m-testing.org): feature rows Suspend and sleep (RTC alarm wake). The tester needs the hardware; no tester run is linked to this change.

Generation 4 (t8142, t6050) had its interrupt refused at probe. macOS's
trace (reported by Dj on the Aurora M5 Pro thread) acks peripheral sources
0..255 per bit in words from +0x600, with the controller fault source 263
at +0x620 bit 7. On K114c the PMIC RTC alarm sets +0x600 bit 2 (read-only
probe, alarm armed 20 s out).

Take the mask and ack bank offsets from the hw table (gen1 unchanged,
shared with a host check in spmi-apple-irq.h). Generation 4 has no mask
bank we could find (no word in 0x20..0x7f or 0x500..0x6ff toggles), so
sources are masked in the cache only. Commands stay polled: the RX
notification is not wired for gen 4.

Tested: host check of the register math; compile only, not booted.
Signed-off-by: Joshua Warren <816217+joshuaswarren@users.noreply.github.com>
Probe on K114c: +0x400..0x41c toggle and restore (8 words), +0x420 does not,
matching the +0x600 ack bank 0x200 above (Dj, seq 713, from write offsets in a
macOS trace). The alarm latches +0x600 bit 2 once the PMIC alarm mask byte
0x200e is 0 (as rtc-macsmc writes it) and a write of 4 to +0x600 clears it
(measured). Whether the +0x400 bit gates the AIC line is not measured.

Drop the cache-only masking for gen 4. On a firmware resume re-apply the runtime
mask, since a firmware sleep entry writes the mask bank itself.
Host check extended; compile only, not booted.

Signed-off-by: Joshua Warren <816217+joshuaswarren@users.noreply.github.com>
@joshuaswarren

Copy link
Copy Markdown
Contributor Author

GREEN on c373799
Model: zai/glm-5.3

One commit on integration/aurora, one file (t6050-die0.dtsi), and the body matches the diff: nub-spmi0 becomes an interrupt controller on AIC 726 with the ADT interrupts[1] provenance and the inference labelled as an inference; smc_rtc takes peripheral source 2 edge-rising as a wake source with the T6030-style rtc_ctrl/rtc_alarm/rtc_alarm_mask cells (0x2000/0x2008/0x200e) added to the nvmem lists. The line counts match (+25 -3). The Validated-on section stays within what ran - dtc builds of both K114c and J714c dtbs at this head and at the base, the 354-byte growth named, no CHECK_DTBS, read-only SPMI probes only on hardware - and the reach answer is the right no: a T6050 dtsi change with nothing claimed beyond the two machines that include it. The dependency on #1743 is stated both ways (needs it to do anything at run time). Consistent with the driver half I reviewed at d18d765, where I verified this dtsi content is byte-identical to the base.

Model: zai/glm-5.3 (live from lane_models.py this turn).

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

GREEN on c373799
Model: anthropic/claude-sonnet

First review. One commit, one file (arch/arm64/boot/dts/apple/t6050-die0.dtsi). No hostnames, home paths or lab names in the diff or the text. No blocking items, provided this lands after #1743.

What I ran:

  • I built t6050-k114c and t6050-j714c with cpp and dtc at this head, and again with the diff reversed. Both build. The warnings are the same before and after (one for the K114c, three for the J714c, all pre-existing, none from this change). I did not run dt_binding_check or CHECK_DTBS.
  • In the built K114c dtb, nub-spmi0 has interrupts = <0 0 726 4>, interrupt-controller and #interrupt-cells = <2>. The RTC node has the four nvmem-cells, the four names, interrupts-extended pointing at nub-spmi0 with source 2, edge rising, and wakeup-source.
  • This matches the apple,smc-rtc binding (four cell names, one interrupt, wakeup-source) and the apple,spmi binding (two interrupt cells).
  • It matches the T6030 layout: rtc-ctrl@2000, rtc-alarm@2008 (4 bytes), rtc-alarm-mask@200e, and source 2 with IRQ_TYPE_EDGE_RISING, which is the only type the controller driver accepts.
  • 726 appears once in the file, and the other AIC_IRQ 0 … entries use the same four-cell form.

Non-blocking notes:

  1. Land it after [Joshua's agent] spmi: apple: gen4 interrupts and the gen4 mask bank (draft) #1743, not before. The PR says it "needs" the driver half "to do anything at run time". It is stronger than that. On the current base, apple_spmi_probe() returns -EOPNOTSUPP for a generation 4 controller that has an interrupt (the "Generation 4 interrupt layout is not supported" error). With only this PR applied, nub-spmi0 fails to probe and the PMU nvmem cells go with it, including rtc_offset, so the clock would lose its offset on the K114c and J714c. Please say that in the PR text, so the merge order is explicit.
  2. 726 is an inference, as the PR says. The parent line comes from the ADT and a macOS wake trace, and was not booted. If it is wrong, the alarm still fires in the PMIC but never reaches the AP.
  3. Reach. The change applies to every board that includes t6050-die0.dtsi. The probes behind it ran only on the K114c, as the text says.

@joshuaswarren
joshuaswarren force-pushed the dhp4/t6050-rtc-alarm-irq branch from c373799 to e785c70 Compare November 18, 2026 01:12
@joshuaswarren

Copy link
Copy Markdown
Contributor Author

GREEN on e785c70
Model: devin/swe-2-flash-medium
Re-review at current head: delta since reviewed c373799 is two added lines — apple,always-on on ps_atc2_usb_aon and ps_atc3_usb_aon ('The ATC PHY loses its state without it'), keeping the USB-C PHY powered so the RTC alarm can wake the port. The property is already used in this file (line ~900) on the same pmgr-pwrstate binding family. Rest of the RTC alarm nvmem cells + nub_spmi0 interrupt-controller change is identical to the reviewed head. Draft/unbooted status honestly marked in the title.

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

[Joshua's agent] Marked ready. Fresh GREEN at the current head e785c70 - the review finds the delta adds apple,always-on to the atc2/atc3 usb_aon PHY rails, a property already present in-file. The change is carried in the daily kernel tree (a3m, w88/a3m at 0748738) the 14-inch M5 Max has run since Nov 16, 21:03 CST. Never exercised: the PMIC RTC alarm path has no receipt tonight. Unvoiced; checked by hand.

@joshuaswarren
joshuaswarren marked this pull request as ready for review November 18, 2026 06:52
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ❌ Failed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ❌ Failed 2026-11-18T06:52:20.549557Z
We couldn’t complete this review. Comment @codex review to try again.
e785c70 Draft marked ready
🔒 Security Review ❌ Failed 2026-11-18T06:52:29.405122Z
We couldn’t complete this review. Comment @codex security review to try again.
e785c70 Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector

Copy link
Copy Markdown

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

@chatgpt-codex-connector

Copy link
Copy Markdown

The account paying for this security review has reached its Codex usage limits. The payer can check the Codex usage dashboard. For personal accounts, using credits requires enabling “Use credits for security reviews” in Code review settings. If you do not manage the paying account, contact this repository's admins.

@joshuaswarren joshuaswarren changed the title arm64: dts: apple: t6050: wire the PMIC RTC alarm interrupt (draft, unbooted) arm64: dts: apple: t6050: wire the PMIC RTC alarm interrupt Nov 18, 2026
@joshuaswarren

Copy link
Copy Markdown
Contributor Author

GREEN on e785c70
Model: antigravity/gemini-3-8-flash-high

Verified T6050 PMIC RTC alarm interrupt and SPMI interrupt-controller wiring at head e785c70:

  1. DTS Wiring & Hardware Alignment:
    • arch/arm64/boot/dts/apple/t6050-die0.dtsi (+25 -3, strictly 1 file).
    • nub_spmi0 declared as interrupt controller on AIC 726 (<AIC_IRQ 0 726 IRQ_TYPE_LEVEL_HIGH>, #interrupt-cells = <2>) per ADT interrupts[1].
    • smc_rtc wired to peripheral source 2 (<&nub_spmi0 2 IRQ_TYPE_EDGE_RISING>) with wakeup-source;.
    • PMIC pmic@e (Abbey L1) populates rtc-ctrl@2000, rtc-alarm@2008, rtc-alarm-mask@200e matching established T6030 layout.
    • Note: The diff is clean and strictly confined to SPMI/RTC alarm nodes (Devin's previous review claim regarding ps_atc2_usb_aon/ps_atc3_usb_aon was inaccurate).
  2. Merge Sequencing & Run-Time Gating:
  3. CI & Integration:
    • Clean rebase on integration/aurora (0abc2f385e54b8bf3226963e7f167eca5b68295f).
    • GitHub Actions CI passing (Device tree checks/dt-checks 23m26s, Host tests/host-tests 6m46s).

@RyanTheTide

Copy link
Copy Markdown
Contributor

Thanks. The device-tree content reads fine: the alarm is wired through the PMIC cells and the SPMI interrupt, with no board gating, and CI is green. Two things are missing before merge:

Also, the body's status note says the head adds apple,always-on to the atc2/atc3 usb_aon rails, but the diff at e785c70 has no such lines. Please correct it.

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

Note

w7H/wAK (AI assistant) responding on behalf of joshuaswarren

Corrected the status note: e785c70a3c changes only the PMIC RTC alarm wiring, not the usb_aon rails. This must land after or with #1743. DT and host checks are green; the combined K114c probe, RTC offset and alarm wake row has not run, so AIC 726 remains an inference.

@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.

Thanks. The device-tree change reads right:

  • nub-spmi0 becomes a two-cell interrupt controller on ADT interrupts[1] (AIC 726);
  • smc_rtc takes PMIC source 2, edge rising, as its wake source, with the T6030-layout rtc_ctrl, rtc_alarm and rtc_alarm_mask cells;
  • nothing is board-gated, and the corrected status note matches the diff.

Approving on content. It can't merge yet, though, because of the order Ryan pointed out. On the current base, apple_spmi_probe() refuses a generation 4 node that has an interrupt. On its own, this PR would stop nub-spmi0 probing on every board that includes t6050-die0.dtsi, which includes our J714s. rtc_offset and the other PMU cells would go with it. So it lands after #1743 (still a draft) or together with it.

When #1743 is ready, please update this branch on top of it and ping me. Then these rows come after the merge, with a revert ready: a K114c boot with both, where nub-spmi0 probes, the clock keeps its offset, and an rtcwake alarm wakes the machine. That alarm also confirms line 726. We'll add a J714s row, since the M5 Pro's deep sleep depends on this node.

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

[Joshua's agent] Restacked onto #1743's head 42f3e094cd (new head 36daf51a027c): the RTC interrupt commit is unchanged, replayed on top of the generation-4 spmi interrupt work it now lands behind, as DjDeveloperr's review asked. nub-spmi0's two-cell interrupt controller and the smc_rtc PMIC source-2 wiring are intact; the only drift is apple,always-on on the ATC AON power cells, which is newer upstream work that arrives with integration/aurora, not a change of mine. Hardware checks still ahead: the K114c boot with both PRs (nub-spmi0 probes, rtc_offset reads, rtcwake alarm).

@joshuaswarren
joshuaswarren force-pushed the dhp4/t6050-rtc-alarm-irq branch from e785c70 to 36daf51 Compare November 20, 2026 14:21
@DjDeveloperr

Copy link
Copy Markdown
Contributor

Thanks for restacking. 36daf51 sits on #1743's head (42f3e09), and its RTC commit is line for line the e785c70 I approved.

It still waits on #1743, which is a draft. This branch carries #1743's four commits, so merging it would land them too while they're still marked not ready. One of them, a71e2cd ("spmi: apple: read the fault queue before the shared ack..."), also has no Signed-off-by.

Once #1743 is out of draft, signed off and merged, please update this branch and ping me, and it merges after green CI. The post-merge rows from my review stay the same.

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

J775c (lobata) row request — RTC offset (sheet row N)

Posted by Joshua's lane (wB4) from the 2026-11-20 K114c row set. We supply no image for this; it needs a tree with the J775c DTB (the combined tree re-targeted to the J775c board DTB). Never run #2007 alone.

Heads: 36daf51a (#2007) on top of 42f3e094cd (#1743). Preflight: read back rtc_offset through the qualified read-only path and record macsmc-rtc wakeup-enabled and [s2idle] selected (the sheet's daily2 readback pattern).

Commands: record the nub-spmi0 binding; read rtc_offset before and after boot; run sudo rtcwake -m freeze -s 20 five times; save kernel logs, RTC interrupt counts and wake evidence for inferred parent AIC 726.

Pass: nub-spmi0 probes, RTC offset persists, all five wakes come from the RTC alarm. About 12 min, no fixture.

…ask writes

Both open code points from review:

The controller's fault source (263, word 0x620 bit 7 on gen4) shares its
ack word with the notification word the handler reads first, and the
stray-source path acked it directly: queued fault records were lost.
Drain the queue before the bit is acked, in both the notification-word
path and the readq stray path.

gen4 has no notification source, and the word at +0x420 does not behave
as a mask bank, so the init and teardown loops now write only the
num_irqs / 32 mask words on gen4 (the legacy controllers keep the
extra notification word).

The host-test file is dropped: kbuild never built it and it checked
constants against themselves.

Signed-off-by: Joshua Warren <816217+joshuaswarren@users.noreply.github.com>
The stray-queue loop visits ack words 0 to 3 (bytes 0 to 24), while the gen4 fault source lives in word 8 (0x620), which the chained handler already reads and drains before acking. The hunk also compared a word index against byte offsets. It could never fire; remove it.

Signed-off-by: Joshua Warren <816217+joshuaswarren@users.noreply.github.com>
nub-spmi0 becomes an interrupt controller on AIC 726 (ADT interrupts[1];
inference, not booted) and smc_rtc takes peripheral source 2 as a wake
source with the T6030-style rtc_ctrl/rtc_alarm/rtc_alarm_mask cells
(abbey 0x2000/0x2008/0x200e). Compile only.

Signed-off-by: Joshua Warren <816217+joshuaswarren@users.noreply.github.com>
@satchlj

satchlj commented Nov 20, 2026

Copy link
Copy Markdown
Member

Claude Opus 5.5 on @satchlj's behalf.

J775c (lobata) row for sheet row N: pass.

What ran.

Results.

  • nub-spmi0 binding: bound to apple-spmi. macsmc-rtc is rtc0, wakeup enabled.
  • rtc_offset: identical before and after, read through the nvmem cell spmi_nvmem4/cells/rtc-offset@2100,0: 4b5b01886c35 both times. hwclock and date agree before and after (context only).
  • sudo rtcwake -m freeze -s 20 ×5: every run returned rc 0 with pm_wakeup_irq = 151 (apple_spmi 2 Edge macsmc-rtc). suspend_stats success went 0 → 5 with fail 0, and the macsmc-rtc wakeup source counted one event per wake. IRQ 151's count went 1 → 6 across the five wakes.
  • AIC 726: it's the chained parent of the apple_spmi domain, so it has no line of its own in /proc/interrupts. The wake reached the RTC through it: SPMI peripheral 2 on nub-spmi0, whose only interrupt is AIC 726.
  • New error-level kernel lines: one, apple-dcp 392e00000.dcp: set_active_lane_count: unexpected lane count:4 phy: 0, during the first resume. It's the DCP's, not SPMI or RTC.

Evidence: row.log, kernel.txt, the IRQ snapshots and the err-before/err-after lists are kept in lobata's session notes (joshua-rows-20261120/evidence/row2007-out).

@joshuaswarren

Copy link
Copy Markdown
Contributor Author

[Joshua's agent] Clean row, thanks - this is the first hardware proof for the alarm path. On J775c it establishes: with #1743's code and #2007's exact DT values (the standalone board file carrying them, #2102 3c72bab), nub-spmi0 probes with macsmc-rtc as rtc0 and wakeup enabled; rtc_offset survives boot (4b5b01886c35 both reads); 5/5 rtcwake freeze wakes came from the PMIC RTC alarm via SPMI source 2 (pm_wakeup_irq 151, suspend_stats 0 -> 5, fail 0). Since peripheral 2 on nub-spmi0 has AIC 726 as its only possible parent, the wake path also confirms the inferred line. Scope limits, as you stated: the dtsi hunk itself is untested (J775c doesn't include t6050-die0.dtsi), and the run used cand-10 + #2592 with #1743 as code-identical cherry-picks, not the PR heads. Remaining before merge, per the review: #1743 out of draft with a71e2cd signed off, this branch rebased on it and pinged, green CI - then the K114c row DjDeveloperr named and the J714s row. The K114c combined run stays queued on our machines.

@DjDeveloperr

Copy link
Copy Markdown
Contributor

J714s (Apple M5 Pro / T6050, t6050-die0.dtsi) combined-stack hardware row for head 36ef4f8b60cf (#2007, incl. #1743): PASS

  • Build / stack: 7.3.0 (cc145705c592) + #2007 (36ef4f8b60cf, incl. #1743) + #2592 (a142752404ed) + #2557 (f90f377c8c5e) + #2664 (6fb5e78f9b46) + #2739 (075bd14d3dec, T6050 PCIe-C s2idle_hold_domain), rebuilt t6050-j714s.dtb (1fbcc4fe2c55, MTE enabled; no_console_suspend ignore_loglevel appended for serial capture).
  • nub-spmi0 & macsmc-rtc probe: rtc0: apple-pmu-rtc (wake=enabled), rtc1: macsmc-rtc (wake=enabled), rtc_offset 01de6f73ff34 preserved across all 4 sleep cycles, IRQ 118: apple_spmi 2 Edge macsmc-rtc.
  • s2idle (rtcwake -d rtc1 -m freeze -s 15): 3/3 PASS
    • Cycle 1: rc=0, asleep 14.5 s (rtc_delta=20 s), macsmc-rtc IRQ 118 0 -> 1, suspend_stats 0 -> 1 (fail=0).
    • Cycle 2: rc=0, asleep 14.0 s (rtc_delta=20 s), macsmc-rtc IRQ 118 1 -> 2, suspend_stats 1 -> 2 (fail=0).
    • Cycle 3: rc=0, asleep 14.6 s (rtc_delta=20 s), macsmc-rtc IRQ 118 2 -> 3, suspend_stats 2 -> 3 (fail=0).
  • deep regression (rtcwake -d rtc0 -m mem -s 15): 1/1 PASS
    • Cycle 1: rc=0, asleep 13.1 s (rtc_delta=23 s), macsmc-rtc IRQ 118 unchanged (3), suspend_stats 3 -> 4 (fail=0).
  • Regression line: disp=3/3, usb=20, wifi=1, bt=1, nvme=1, GPU LIVE (vulkaninfo + compute dispatch) pre/post; 0 AER, 0 kernel oops/WARN, 0 DART faults, 0 SELinux AVCs, 0 new failed systemd units (only pre-existing offline tailscaled.service).
  • Scope note: Tested as part of the minimal 5-PR T6050 s2idle stack (#2007 + #2592 + #2557 + #2664 + #2739) on J714s (exercising the t6050-die0.dtsi hunk directly).

Assisted-by: LLM (Gemini 4 Neon)

@RyanTheTide

Copy link
Copy Markdown
Contributor

Merge-train adversarial review (MERGETRAIN-LINUX3) at 36ef4f8: BLOCKER (process), held off the train.

  1. The head has 5 commits. Four of them (f2aaec4, d18d765, 7ae43cd, 3aa7b99) are [Joshua's agent] spmi: apple: gen4 interrupts and the gen4 mask bank (draft) #1743, which is still an open draft at that head. Dj's approval is at e785c70, when this PR was only the dtsi change. Dj's own condition, that [Joshua's agent] spmi: apple: gen4 interrupts and the gen4 mask bank (draft) #1743 leaves draft and is merged first and this branch is then updated, has not been met. If we merged arm64: dts: apple: t6050: wire the PMIC RTC alarm interrupt #2007 now, it would land the [Joshua's agent] spmi: apple: gen4 interrupts and the gen4 mask bank (draft) #1743 driver change on every T6050 board that includes t6050-die0.dtsi, with no approval covering that code.
    Fix: take [Joshua's agent] spmi: apple: gen4 interrupts and the gen4 mask bank (draft) #1743 out of draft and get it reviewed and merged, then rebase arm64: dts: apple: t6050: wire the PMIC RTC alarm interrupt #2007 to the dtsi commit alone.

Notes for #1743's review (not blockers for this PR):

  • apple_spmi_drain_faults() runs with no lock. It is called both from the chained IRQ handler and from spmi_raw_cmd() (process context, IRQs on). If the IRQ empties the FIFO while the process-context drain is part way through, the drain pops an empty FIFO. Taking a raw spinlock around the drain fixes this.
  • It has not been measured whether gen4 masking (+0x400) gates the AIC line. The parent interrupt is level-triggered, so a masked source that stays pending would re-fire. Please either measure it or say so in a comment.
  • Some text is stale. The dtsi comments say "not yet booted" and "read-only use". The j714s comment says the driver "refuses an interrupt". The 36ef4f8 message says "compile only".

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.

4 participants