Skip to content

arm64: dts: exynos: google: zuma-husky: the main PMIC's interrupt, so the Pixel 8 Pro has an RTC alarm - #243

Merged
derekross merged 3 commits into
derekross:omamobilefrom
gitColinX:omamobile-zuma-husky-rtc-alarm
Nov 12, 2026
Merged

derekross merged 3 commits into
derekross:omamobilefrom
gitColinX:omamobile-zuma-husky-rtc-alarm

Conversation

@gitColinX

@gitColinX gitColinX commented Nov 11, 2026 •

Copy link
Copy Markdown
Contributor

On omamobile the Pixel 8 Pro's main PMIC has no interrupt: sec-pmic-acpm firmware:power-management:pmic-1: no interrupt, so no RTC alarm at boot, and rtc0 has no wakealarm. This is #236's fix for the 9a, on zuma.

Three commits, the way #236 has them:

  • the google,zuma-vgpio2ap-sysreg compatible in the sysreg binding (no clock, as zumapro's and gs201's);
  • the VGPIO2AP syscon at 0x15090000 in zuma.dtsi;
  • interrupts = <GIC_SPI 74 IRQ_TYPE_LEVEL_HIGH>, google,vgpio2ap-sysreg and wakeup-source on pmic-1 in zuma-husky.dts.

Where the two values come from

  • GIC SPI 74: stock's s2mpg14mfd node on my phone has interrupts = <0 0x4a 4> (its own dtbo, board id 0x40206).
  • 0x15090000: SYSREG_VGPIO2AP in Google's include/linux/mfd/samsung/s2mpg14-register.h (kernel/google-modules/soc/gs, android-gs-shusky-6.1-android16); drivers/mfd/s2mpg14-irq.c maps it with INTC0_IPEND at 0x290, which S2MPG14_VGPIO2AP_INTC0_IPEND already is here.

On my phone (G1MNW, 2026-11-11): the same changes on husky-test-20261109-t1 (with #233 and one more test commit that nothing on this boot ran), 7.3.0-rc6-omamobile-push1-g755b5d1c5532, its own modules installed.

  • No "no RTC alarm" line; rtc0 has wakealarm.
  • The s2mpg14-pmic line in /proc/interrupts stayed at 0 for the first two minutes. rtcwake -m no -s 60 armed the alarm and it fired on time: rtc-alarm0 0 to 1, s2mpg14-pmic 0 to 2.
  • No failed units; tools/phone-smoke passes every required check.

Not tested: suspend. It has never run on the 8 Pro, and I only armed the alarm with the phone awake. I have not booted this exact branch on today's omamobile; the device tree changes are the same lines. The 8a (zuma-akita.dts) gets the syscon node from zuma.dtsi but nothing there points at it; @aninmucek may want the same three lines on its pmic-1 once someone reads its stock interrupt.

Checks: tools/kernel-pr-check --build --merge passes: it merges cleanly onto fc7086a68, checkpatch --strict has no errors (its warnings are the Co-Authored-By trailer's spelling), 10 device trees build with W=1 and pass dtbs_check with no new warnings, and dt_binding_check passes on the one binding. No C files are touched.

gitColinX and others added 3 commits November 11, 2026 11:56
Tensor G3 (zuma) latches the main PMIC's in-band interrupt in a VGPIO2AP
system register block at 0x15090000, as zumapro does at the same address
and gs201 at 0x182f0000; the PMIC driver reads it through the
google,vgpio2ap-sysreg phandle of the samsung,s2mpg14-pmic binding. Add
the compatible, with no clock, as theirs.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Signed-off-by: Colin Lundholm <265166644+gitColinX@users.noreply.github.com>
The system registers that latch the main PMIC's in-band interrupt (GIC
SPI 74 on this SoC), the way zumapro.dtsi and gs201.dtsi describe
theirs. The address is SYSREG_VGPIO2AP in Google's s2mpg14-register.h
(kernel/google-modules/soc/gs, android-gs-shusky-6.1-android16), where
s2mpg14-irq.c maps it with INTC0_IPEND at 0x290. A PMIC node points at
it with google,vgpio2ap-sysreg; nothing uses it until a board does.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Signed-off-by: Colin Lundholm <265166644+gitColinX@users.noreply.github.com>
Without an interrupt the S2MPG14 driver supplies its rails but none of
the PMIC's events reach the host ("no interrupt, so no RTC alarm" at
boot), so rtc0 has no wake alarm on the Pixel 8 Pro. The interrupt is
GIC SPI 74, as stock's s2mpg14mfd node has it on this board
(interrupts = <0 0x4a 4>), latched by the VGPIO2AP sysreg, as the gs201
boards and the Pixel 9a have it; wakeup-source makes the alarm a wake
source for suspend, which has not been run on this phone yet.

On a Pixel 8 Pro (G1MNW) with these three commits on the display and
GPU test tree: no "no RTC alarm" line, rtc0 has wakealarm, the
s2mpg14-pmic interrupt line counts 0 until an alarm armed with
rtcwake -m no -s 60 fires on time (rtc-alarm0 0 to 1), no failed units.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Signed-off-by: Colin Lundholm <265166644+gitColinX@users.noreply.github.com>
@derekross

Copy link
Copy Markdown
Owner

Thank you, @gitColinX. This is a tidy port of #236: the same three steps in the same order, and the two values that had to be the Pixel 8 Pro's own are sourced rather than copied. I checked both against Google's tree on android-gs-shusky-6.1-android16: SYSREG_VGPIO2AP 0x15090000 and INTC0_IPEND 0x0290 are in s2mpg14-register.h (lines 613-614) and s2mpg14-irq.c maps them as you say (line 314). For the interrupt there is a public line as well as your phone's dtbo: dts/zuma-pmic.dtsi line 34 in kernel/devices/google/zuma has interrupts = <GIC_SPI IRQ_INTCOMB_VGPIO2AP_ALIVE IRQ_TYPE_LEVEL_HIGH 0>, and IRQ_INTCOMB_VGPIO2AP_ALIVE is 74 in soc/gs's dt-bindings/interrupt-controller/zuma.h. The same node has s2mpg14,wakeup = "enabled", so wakeup-source is stock's choice too.

What I ran: the gate passes on today's omamobile (clean merge onto fc7086a68fca, all three commits signed off, prefixes right, checkpatch with only the trailer-spelling warnings, the device trees build with W=1 and no new warnings). CI hasn't run on this PR yet because the run is waiting for my approval, so I ran the schema checks by hand in the meantime: the binding passes, zuma-husky.dtb and zuma-akita.dtb validate with nothing new, and your zuma-husky.dtb against the old schemas is rejected on google,zuma-vgpio2ap-sysreg, which shows the binding commit is doing its job. I also decompiled both boards before and after: the 8 Pro gains the node and the three properties, the 8a gains only the node, which nothing there uses and no driver binds, and no other board changes.

Your evidence is the right kind: the s2mpg14-pmic count staying at 0 for two minutes and then moving with the alarm shows both that the interrupt arrives and that it doesn't storm.

Nothing to change in the commits. I'll merge once CI is green. Four small things, none of which holds it up:

  • The run was on husky-test-20261109-t1 (g755b5d1c5532), which isn't on your fork, so I can't see what it had. Could you push that branch, or post git rev-parse HEAD:drivers/mfd/sec-irq.c HEAD:drivers/mfd/sec-acpm.c HEAD:drivers/mfd/sec-common.c HEAD:drivers/rtc/rtc-s5m.c from it? On omamobile they are 41975501d452, c644530c44b0, 6a40169c4f59 and e38d8b918a90. The driver path is the one the Pixel 9a has run on the published tree since arm64: dts: exynos: google: zumapro-tegu: the main PMIC's interrupt, so the RTC alarm wakes the Pixel 9a #236, so I'm not waiting for this, but a tested tree should exist somewhere.
  • Please paste the tools/phone-smoke report from that boot here.
  • On your first boot of omamobile with this in: the sec-pmic/rtc lines from dmesg, ls /sys/class/rtc/rtc0/wakealarm, and the s2mpg14-pmic count after a minute. If the PMIC ever failed to probe, touch would lose its supplies, so that's the quick tell.
  • A short entry in device/husky/NOTES.md on omarchy-mobile with this run, and ROADMAP's "Time" row, which still says "No RTC".

One correction to the description, not the commits: suspend has run on the model, on Somnius's phone (omarchy-mobile #640 and #642); "never on this unit" is the accurate form, as your commit message already has it.

#244 touches the same file in another place; the two merge cleanly in either order.

@derekross
derekross merged commit e082e84 into derekross:omamobile Nov 12, 2026
1 check passed
@gitColinX

gitColinX commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor Author

@derekross, I owe you answers to the four things you asked here on 11-12. Sorry for the delay.

The tested tree. g755b5d1c5532 was Brett's tag husky-test-20261109-t1 (on his fork, 54e10dcccd6b) plus three commits of mine. The four hashes from it:

drivers/mfd/sec-irq.c     41975501d452
drivers/mfd/sec-acpm.c    c644530c44b0
drivers/mfd/sec-common.c  6a40169c4f59
drivers/rtc/rtc-s5m.c     06cb3c191f81   (omamobile had e38d8b918a90)

The three MFD files are omamobile's. rtc-s5m.c differs, and I should have said so at the time: besides this PR's device-tree commit, that tree carried two power-off commits, ikegam's S2MPG12 disarm (now #233) and my S2MPG14 extension of it (now #262). Neither touches the alarm path, and no power off was run on that kernel. I can push the branch to my fork if you still want it.

The tools/phone-smoke report from that boot (2026-11-11T07:41Z, tools at 7f66078): every required check ok. The one "not ok" was new dmesg errors: brcmf_msgbuf_rxbuf_data_post: No PKTID available x435, the Bluetooth patch lines and the exynos-clkout probe line, which Brett's 11-13 notes say are gone on 82f31d3a7. Boot, storage guard, root, USB gadget, display (card2-DSI-1, 1344x2992), GPU (panthor), session, lock, touch (SALIXP), buttons, Wi-Fi, Bluetooth, battery and thermal (hottest zone 54 °C) ok; sound, sensors and cameras skipped. I can paste the full table if you want it here.

On published omamobile with this in (82f31d3a7, my unit, up 10 hours):

[    2.602853] s5m-rtc s2mpg14-rtc: registered as rtc0
[    2.603133] s5m-rtc s2mpg14-rtc: setting system clock to 2026-11-15T19:21:04 UTC

/sys/class/rtc/rtc0/wakealarm is there, and the s2mpg14-pmic and rtc-alarm0 interrupt counts read 0 after 10 hours idle (no storm). Touch works, so the PMIC probed. On the earlier boot of 5c8b9e542, rtcwake -m on -s 12 moved the s2mpg14-pmic count from 0 to 1.

NOTES and ROADMAP: in derekross/omarchy-mobile#835 (the 2026-11-15 NOTES entry has the RTC lines, and ROADMAP's Time row no longer says "No RTC").

And yes on the description: "never on this unit" is right. Somnius's runs were first on the model.

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