Skip to content

Pixel 8 Pro CPU DVFS: zuma's CPU levels from the ECT and CPU thermal zones (stacked on #157) - #158

Merged
derekross merged 13 commits into
derekross:omamobilefrom
greenbrettmichael:omamobile-zuma-husky-cpufreq
Nov 13, 2026
Merged

derekross merged 13 commits into
derekross:omamobilefrom
greenbrettmichael:omamobile-zuma-husky-cpufreq

Conversation

@greenbrettmichael

@greenbrettmichael greenbrettmichael commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

Stacked on #157 (the GPU, for zuma's ACPM clocks and TMU). Only the last commit is new. Updated after review: turbo-mode above stock's battery-limiter caps, and the energy model from stock's per-level power.

This gives the Pixel 8 Pro CPU frequency scaling and CPU thermal zones. Until now every core ran at the bootloader's rate: the X3 at 1.16 GHz of 2.91, the A715s at 1.07 of 2.37 and the A510s at 1.43 of 1.70.

What it adds to zuma.dtsi:

  • Clocks: the three clusters on their ACPM DVFS clocks (ZUMA_CLK_ACPM_DVFS_CPUCL0-2, from Pixel 8 Pro GPU: zuma's CMU_G3D, ACPM GPU clock, panthor and a GPU thermal zone #157's clk-acpm zuma variant), through cpufreq-dt.
  • Operating points: the ACPM firmware's own levels, read from the ECT the bootloader leaves at 0x90000000 (Google's ect_parser.c format), up to each cluster's max-freq in stock's exynos-acme domains (Google's zuma-cpu.dtsi):
    • little: 324–1704 MHz (10 levels);
    • mid: 402–2367 MHz (15);
    • big: 500–2914 MHz (15).
      The ECT's levels above those maximums are listed in comments and left out, as stock leaves them out. The ECT's G3D list is the GPU table Pixel 8 Pro GPU: zuma's CMU_G3D, ACPM GPU clock, panthor and a GPU thermal zone #157 took from stock's DT, and its little list matches stock's cpufreq time_in_state, which checks both the parser and the source.
  • The energy model: each level's opp-microwatt is its power in stock's default pixel-em profile (Google's zuma-b0-ipop.dts, the same in the Pixel 8 Pro's own device tree), with #cooling-cells. A power coefficient can't be used: there is no opp-microvolt, since the firmware sets the voltages, and with one the energy model fails to register (EM: invalid perf. state: -22, which the first version had).
  • turbo-mode on the levels above stock's battery current limiter caps (zuma-shusky-bcl.dtsi: 1945 MHz mid, 2294 MHz big), as on gs201 (arm64: dts: exynos: google: gs201: CPU frequency scaling and thermal zones (Pixel 7 Pro) #169): with no BCL driver, they are used only with boost on.
  • Thermal zones on the ACPM TMU's big, mid and little sensors, as arm64: dts: exynos: google: zumapro: CPU DVFS, thermal zones and the GPU's top levels #87 proposes for zumapro: throttle from 80 C, force the lowest level from 95 C, critical at 105 C.

Tested on husky (tag husky-test-20261031-c1: this branch merged with #156, built and installed with modules-push, a normal boot):

  • At boot: clk-acpm, gs101_acpm_tmu and panthor load at about 4.0 s, cpufreq-dt brings up three schedutil policies (0-3, 4-7, 8), and each policy is bound as the cooling device of its zone. The zones read 52-55 C idle.
  • The firmware follows: in every sample below, the policy's rate and ACPM's readback (cpucl0-2 in clk_summary) agreed.
  • All nine cores loaded for 90 s:
    • the zones reached 80-85 C within 3 s;
    • throttling then held them at 79-88 C, below the 95 C trip, and none reached critical;
    • the mid cluster ran 1.9-2.2 GHz (from a fixed 1.07), and the little cluster 0.95-1.55 GHz;
    • step_wise held the X3 at 0.5-0.9 GHz, because the shared die heat keeps its sensor at 80 C;
    • 4 s after the load, all three zones read 68 C.
  • Bursts: each cluster reaches its top level (time_in_state shows 1704, 2367 and 2914 MHz).
  • The rest of the phone: Wi-Fi, touch and the GPU are unaffected. With panthor loaded at boot, Hyprland's aquamarine now picks panthor's render node for the simple-framebuffer output ("falling back to sole renderD", renderer "Mali-G715 MC7 (Panfrost)"), so the desktop draws on the GPU.

Re-tested on husky-test-20261101-e1 (this branch as pushed, merged with #160 and #156): all three perf domains register an energy model with no error, the policies come up at 1704/1945/2294 MHz with boost off, and boost on gives 2367/2914 MHz.

tools/kernel-pr-check --build on the commit: checkpatch --strict has nothing to read, and 8 device trees build with W=1 and dtbs_check with no new warnings.

Not here:

  • a smarter governor for the X3 (power_allocator with a sustainable power) instead of step_wise. Under an all-core load it gives up the big core first;
  • the ECT's levels above stock's maximums;
  • suspend.

🤖 Generated with Claude Code

@derekross

Copy link
Copy Markdown
Owner

Thanks, cpufreq on all three clusters with throttling that holds under 95 C is great to see. The limits and power coefficients match Google's zuma-cpu.dtsi, and the cooling states fit each table. Same request as on #157: push the c1 tag and paste uname -r and a smoke report. Please also make the "as on zumapro" wording point at #87. This merges after #157.

@derekross

Copy link
Copy Markdown
Owner

Thanks Brett. Since this went up, my Pixel 7 Pro CPU scaling landed as #169, and I compared the two: same mechanism, and your limits and coefficients match Google's zuma-cpu.dtsi. The one place I'd like them to agree: on gs201 I marked the levels above stock's battery-current-limiter caps as turbo-mode, because with no BCL driver nothing pulls the clusters down when the battery voltage droops, and a burst at the top level on a weak battery risks a brownout. Husky's stock has the same caps (zuma-shusky-bcl.dtsi, vdroop: mid 1945 MHz, big 2294 MHz), so please mark mid 2130/2245/2367 and big 2363 to 2914 as turbo-mode. They stay available through boost, as on the 7 Pro. Please also rebase onto #157 once that's updated, and the round-1 points still stand: the c1 tag on your fork, uname -r and a smoke report, and the #87 wording.

@greenbrettmichael
greenbrettmichael force-pushed the omamobile-zuma-husky-cpufreq branch 2 times, most recently from 6e4fa69 to 4cc3108 Compare November 1, 2026 14:04
@greenbrettmichael

Copy link
Copy Markdown
Contributor Author

Thanks Derek. Rebased onto the updated #157 and force-pushed (4cc310893).

  • turbo-mode: mid 2130, 2245 and 2367 MHz and big 2363 to 2914 MHz are turbo-mode. I checked the caps in zuma-shusky-bcl.dtsi first: vdroop1, vdroop2 and smpl_warn all cap cpucl1 at 1945000 and cpucl2 at 2294000. The DT comment explains why, as gs201's does. On e1, the policies come up at 1704/1945/2294 MHz with boost off, and boost on gives 2367/2914 MHz.
    • One thing for both phones: switching boost off again leaves cpuinfo_max_freq and scaling_max_freq at the turbo level until a reboot. cpufreq_frequency_table_cpuinfo() only ever raises cpuinfo.max_freq ("If the driver has set its own cpuinfo.max_freq above max_freq, leave it as is"). So on husky and on gs201, boost is one-way per boot. Boot is safe, since boost starts off. I put the caps back by hand on this boot. I haven't looked for an upstream fix yet.
  • A fix the smoke run turned up: the energy model never registered. Each cluster logged EM: invalid perf. state: -22, because dynamic-power-coefficient needs opp-microvolt, and the firmware sets the voltages. So the coefficients are gone. Each level now has opp-microwatt from stock's default energy model: the pixel-em profile in Google's zuma-b0-ipop.dts, which the Pixel 8 Pro's own device tree has too. With that, all three perf domains register, with no error. gs201 has the same coefficient-without-voltage setup, so its energy model may be failing the same way. Its stock pixel-em profile would fix it there too.
  • The tag and the wording: husky-test-20261031-c1 is on my fork now, and the zone wording points at arm64: dts: exynos: google: zumapro: CPU DVFS, thermal zones and the GPU's top levels #87.

tools/kernel-pr-check --build on the commit: checkpatch --strict is clean, and 8 trees build and pass dtbs_check with no new warnings.

tools/phone-smoke on husky-test-20261101-e1

uname -r: 7.3.0-rc6-omamobile-husky-wifi (the test kernels pin the release name; this is build #11 of tag husky-test-20261101-e1).

Check Result Detail
boot (required) ok kernel 7.3.0-rc6-omamobile-husky-wifi, slot b, up 289s
storage guard (required) ok 62 partitions; only userdata and slot B's boot images writable
root (required) ok ext4 on userdata, 97G free
kernel health (required) ok no oops or BUG; 0 WARN, 0 error lines in dmesg
systemd units ok none failed
USB gadget (required) ok bound to 11210000.usb
display ok card2-DSI-1 connected, 1344x2992
GPU ok render node on panthor
session (required) ok Hyprland and the Omarchy shell running
lock ok PAM file present, the shell can lock (passwordPam true, locked true)
touch ok SALIXP Touchscreen
buttons ok gpio-keys
Wi-Fi ok wlp1s0 on brcmfmac, link up
Bluetooth skip no controller
battery skip no battery power supply
sound cards skip none
sensors skip no sensor proxy
cameras skip no libcamera tools
thermal ok hottest zone 50 °C

The "0 error lines" is only because I cleared the ring buffer with dmesg -C during the screen-off test on this boot. From journalctl -k -b -p err, the boot's error lines, leaving out brcmfmac's, are:

  • panthor: _opp_set_regulators: no regulator (mali) found, expected: ACPM sets the rail;
  • the HK3's one refused brightness write before allow_tx (sending dcs data 51 02 bc failed: -1), whose level is kept and sent again;
  • exynos-clkout: cannot match parent device, from the zuma PMU already on omamobile, and on the d3 boot too;
  • systemd's gpt-auto generator, and "unable to open an initial console".

@derekross

Copy link
Copy Markdown
Owner

Thanks Brett, and thanks for checking the caps in zuma-shusky-bcl.dtsi first.

  • turbo-mode: set on mid 2130/2245/2367 and big 2363-2914, as asked, and the boost-off rates on e1 match the caps.
  • The energy model: good catch. I compared every opp-microwatt with the default profile of pixel_em in Google's zuma-b0-ipop.dts, and all 40 levels match (10 little, 15 mid, 15 big). The kernel side agrees with you: dev_pm_opp_of_register_em() prefers opp-microwatt, and the coefficient path needs a voltage the firmware never gives us.
  • Tags and wording: the c1 tag is on your fork, and the arm64: dts: exynos: google: zumapro: CPU DVFS, thermal zones and the GPU's top levels #87 wording is in.

CI is green, with dtbs_check, and every commit's husky device tree builds.

Thanks too for the two notes about the Pixel 7 Pro side. You're right that cpufreq_frequency_table_cpuinfo() never lowers cpuinfo.max_freq, so turning boost off doesn't take the turbo levels away again until a reboot. And gs201 has the same coefficient-without-voltage setup, so its energy model most likely doesn't register either. Neither holds up this PR. I'll follow up on both for the 7 Pro.

The merge order: Lef's #163 first, then #160, then this one, which needs one rebase over both. The conflicts are in clk-zuma.c, the clock binding, zuma.dtsi and zuma-husky.dts; the clock-id header merges without one. Once that rebase is green, I'll merge this PR, which lands #157's commits with it.

@greenbrettmichael

Copy link
Copy Markdown
Contributor Author

Thanks Derek. Rebased onto today's omamobile (1a882ca), which now carries #163 and #200, and force-pushed (e00ab31e1). #157's branch now holds the same first 12 commits.

  • CMU_TOP IDs: the G3D ids start at 28 now (28-40, after arm64: dts: exynos: google: zuma-husky: haptics and the rear torch #200's PERIC0 22-27), and TOP's nr_clk_ids is still CLK_DOUT_G3D_SWITCH + 1. Nothing else in the header moved.
  • Conflicts, each resolved by keeping both sides:
    • clk-zuma.c: the PERIC0 and PERIC1 muxes, gates and dividers in CMU_TOP sit next to G3D's, and CMU_G3D follows CMU_PERIC0. I rewrote the file comment to cover both, and the "firmware-owned" comment above the TOP gates lists HSI0, PERIC0, PERIC1 and G3D;
    • the clock binding: google,zuma-cmu-g3d in the enum, and its clock block before PERIC0's;
    • zuma.dtsi: the google,zuma-acpm.h and gpio.h includes;
    • zuma-husky.dts: opp-table-gpu after the battery node, and &cmu_g3d and &gpu before Pixel 9a (zumapro): the modem link on the shared gs201-cp host and zuma PHY, with CPIF (stacked on #186) #187's &pcie_1.
  • git range-diff against the old head: commits 3-7 and 9-11 are unchanged; 1, 2, 8, 12 and 13 differ only by the resolutions above.

tools/kernel-pr-check --build --merge passes on the 13 commits: it merges cleanly onto 1a882ca, checkpatch --strict has nothing to read, W=1 builds of the 4 touched C files pass, 10 device trees build and pass dtbs_check with no new warnings, and dt_binding_check passes on 3 bindings.

I haven't booted this exact head. The code is the same as e1's except the new clock ids and the merged neighbours.

#160 next: the two conflict only in zuma.dtsi, where both add nodes at the same place. I'll rebase this one again once #160 lands, as you planned.

@greenbrettmichael

Copy link
Copy Markdown
Contributor Author

A boot of the exact heads now, as promised: omamobile (1a882ca) with #160 (d5c35c5fa) and then #158 (e00ab31e1) merged. The one zuma.dtsi conflict between them is resolved by keeping both sides in address order. Test tag: husky-test-20261109-t1.

  • Build: the pinned clang 22.0.2 (clang-r596125, tools/fetch-toolchain), main's tooling, the modules installed with modules-push, and slot B flashed with omamobile flash. The initramfs is the e1 one, unchanged (husky packs no modules into it), so the kernel was the only change on this boot.
  • Device package: omamobile-device-husky 0.1.8 from main. I installed it on e1 first and booted it once there (clean: the rescue stood aside with no hsi2c_11). The point was that its MAX77759 blocklist is in place before this kernel turns the bus on.
  • uname -r: 7.3.0-rc6-omamobile-push1-g54e10dcccd6b

What the two PRs do on this boot:

tools/phone-smoke on husky-test-20261109-t1
Check Result Detail
boot (required) ok kernel 7.3.0-rc6-omamobile-push1-g54e10dcccd6b, slot b, up 57s
storage guard (required) ok 62 partitions; only userdata and slot B's boot images writable
root (required) ok ext4 on userdata, 97G free
kernel health (required) ok no oops or BUG; 0 WARN, 46 error lines in dmesg (34 known, 12 new)
new dmesg errors not ok N N: brcmf_msgbuf_rxbuf_data_post: No PKTID available !! x2; Bluetooth: N: BCM: '…' x2; exynos-clkout exynos-clkout: probe with driver exynos-clkout failed with error N x1
systemd units ok none failed
USB gadget (required) ok bound to 11210000.usb
display ok card2-DSI-1 connected, 1344x2992
GPU ok render node on panthor
session (required) ok Hyprland and the Omarchy shell running
lock ok PAM file present, the shell can lock (passwordPam true, locked true)
touch ok SALIXP Touchscreen
buttons ok gpio-keys
Wi-Fi ok wlp1s0 on brcmfmac, link up
Bluetooth ok hci0, powered true
battery ok 20% Charging, 22 °C
sound cards skip none
sensors skip no sensor proxy
cameras skip no libcamera tools
thermal ok hottest zone 49 °C

The "new" dmesg lines, none from these PRs:

@derekross

Copy link
Copy Markdown
Owner

Thank you, Brett, and thanks for the boot of the exact pair; I fetched t1 and it is this head with
#160 and the zuma.dtsi resolution in address order, nothing else. The rebase reads well too: the
CMU_TOP ids at 28-40 sit after #200's with no duplicates in any block, and on the Pixel 7 Pro side
its device tree is byte-identical and clk-acpm, panthor and the TMU driver compile to the same
instructions.

One thing has changed since my 11-01 review, and it isn't something you could have seen then: the
Pixel 8a (akita) now has a board file on zuma.dtsi. With this PR its CPUs get your
operating-points-v2 too, and cpufreq-dt (built in) would start scaling the 8a's clusters through
ACPM at boot. Your turbo marks follow the 8 Pro's battery-limiter caps from
zuma-shusky-bcl.dtsi (1704/1945/2294 MHz), but Google's zuma-akita-bcl.dtsi caps the 8a at
1475/1549/2147 MHz in every event, so by your own reasoning the 8a's non-turbo tops would be
1425/1418/2147, and with the shared table it would run above them with boost off. Nobody has run
cpufreq on the 8a yet either; it still lives in the initramfs.

The simplest fix, I think: keep the OPP tables in zuma.dtsi, move the CPUs'
operating-points-v2 into zuma-husky.dts, and say in the table's comment that the turbo marks
are the 8 Pro's caps. The 8a can then add its own with its own caps. The TMU sensor and the thermal
zones are fine to share. If
you'd rather keep it shared, then @aninmucek would need to agree and boot it on the 8a first, with
its own caps. Since this needs the rebase over #160 anyway, both can go in one push.

The plan otherwise stands: #160 merges first (its CI run failed to start on GitHub's side, as did
this one's, so I'll re-run them), then this rebased over it, and I'll close #157 when this lands.
Small and optional while you're in there: #87's zones landed through #96, so the "as #87 proposes"
in two messages can say "as zumapro.dtsi does".

Document zuma's (Tensor G3) CMU_G3D, which clocks the GPU, and the
CMU_TOP clocks it takes: the G3D GLB, NOCD and SWITCH paths and the
shared PLL dividers behind them. The inputs and their order are those of
zumapro's CMU_G3D, which has the same layout.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
Describe the GPU's clocks on zuma (Tensor G3): CMU_TOP's G3D GLB, NOCD
and SWITCH paths, with the shared PLL dividers they select from, and
CMU_G3D at 0x1ee00000, with its two PLLs, muxes, dividers and the GPU's
core, coregroup and stacks gates.

As on zumapro, firmware sets these up and keeps G3D on, and the GPU's
rate changes go through ACPM, so everything here is read-only. Google's
zuma CMUCAL tables give the same offsets, field widths and parents as
zumapro's CMU_G3D: the PLLs have PLL0822x's M/P/S layout, and the
L2_GLB and STACKS dividers have no ratio field.

Link: https://android.googlesource.com/kernel/google-modules/soc/gs/+/refs/heads/android-gs-shusky-6.1-android16/drivers/soc/google/cal-if/zuma/cmucal-sfr.c
Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
… ids

The Tensor G3 (zuma) ACPM firmware numbers its DVFS domains as Google's
enum acpm_dvfs_id for zuma: MIF, INT, CPUCL0-2, DSU, BCI, G3D, G3DL2,
TPU, INTCAM, TNR, CAM, MFC, DISP, AOC and BW, the same 17 as the
Tensor G4's. Add their clock ids, and the paired GPU clock after them
(17), as the gs201 and zumapro headers do, so device trees can name
them.

Link: https://android.googlesource.com/kernel/google-modules/soc/gs/+/refs/heads/android-gs-shusky-6.1-android16/drivers/soc/google/cal-if/zuma/acpm_dvfs_zuma.h
Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
The ACPM protocol driver creates zuma-acpm-clk for the Tensor G3, but
clk-acpm had no match for it. Add the zuma variant, built like the
zumapro one:

- the DVFS domains as clocks. zuma's are zumapro's 17, in the same
  order (Google's enum acpm_dvfs_id is zuma's), so G3D is 7 and G3DL2
  is 8, as the Pixel 8 Pro's stock GPU node names them
  (gpu1_cmu_cal_id 0xb040007, gpu0_cmu_cal_id 0xb040008);
- a paired gpu clock (17) that moves G3D and G3DL2 together. The
  Pixel 8 Pro's stock gpu_dvfs_table_v2 is the Pixel 9a's, pair for
  pair, so zumapro's table is reused, without its 940 MHz level: zuma's
  gpu_dvfs_max_freq is 890 MHz;
- zuma_gpu_max_khz, 0 (read-only) by default, as on gs201 and zumapro.

The firmware's DVFS channel is 0, as stock's cal_if node has it.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
The Tensor G3 (zuma) has a Mali-G715, arch 11.8.10 r1p2 on the
Pixel 8 Pro, like the Tensor G4's. Add its compatible, so that panthor
can match its integration-specific cache hash, as it does zumapro's.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
The Tensor G3 (zuma) integrates the same Mali-G715 as the Tensor G4,
with the same ASN hash values: the Pixel 8 Pro's stock device tree has
l2-hash-values = <0x36db6d 0x5b6db6 0x00>, from Google's zuma-gpu.dtsi,
the source zumapro's match already cites. On zumapro, CSF startup fails
with a GPU bus fault without them.

Match google,zuma-mali with zumapro's SoC data.

Link: https://android.googlesource.com/kernel/devices/google/zuma/+/42395e8adcb880aa336f0da6bb2a19ed53929c40/dts/zuma-gpu.dtsi
Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
Describe CMU_G3D and the Mali-G715 at 0x1f000000, both disabled, as
zumapro does: the GPU's registers, its JOB, MMU and GPU interrupts
(SPI 347, 348 and 346) and its size come from the Pixel 8 Pro's stock
mali node, and its core, coregroup and stacks clocks from CMU_G3D.

There is no power domain: stock's own power-domain driver reports
pd-g3d on at probe, so the bootloader leaves G3D powered.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
Turn on CMU_G3D and the Mali-G715 on the Pixel 8 Pro:

- core is ACPM's paired gpu clock (ZUMA_CLK_ACPM_GPU), which sets G3D
  and G3DL2 together through the GPU DVFS pairs;
- coregroup and stacks stay on CMU_G3D's gates;
- an OPP table with the shader rates 150-649 MHz.

Stock goes up to 890 MHz. The levels from 723 MHz on are left out until
a GPU thermal zone is in the device tree, as on the Pixel 9a, because
nothing upstream throttles the GPU yet.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
The Tensor G3 (zuma) reads its SoC temperatures through the same ACPM
TMU service as gs101 and zumapro, with zumapro's seven sensors in the
same order: the Pixel 8 Pro's stock device tree names them BIG, MID,
LITTLE, G3D, ISP, TPU and AUR (ids 0-6) on ACPM channel 9. Add its
compatible.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
The Tensor G3 (zuma) TMU service is the Tensor G4's: ACPM channel 9 and
seven sensors in the same order, BIG, MID, LITTLE, G3D, ISP, TPU and
AUR, as the Pixel 8 Pro's stock device tree has them (acpm_tmu on
channel 9, tmuctrl_0 to tmuctrl_6). Match google,zuma-acpm-tmu with
zumapro's data.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
…zone

Add zuma's TMU sensor under acpm_ipc and a gpu-thermal zone on its G3D
sensor (3). It throttles panthor's devfreq from 85 C and shuts down at
105 C, as zumapro.dtsi does, below the stock Pixel 8 Pro's own GPU
trips (passive from 100 C, hot at 115 C). For that, the GPU node
gets #cooling-cells.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
With gpu-thermal throttling panthor's devfreq, add the Pixel 8 Pro's
GPU levels from 723 to 890 MHz, stock's top (gpu_dvfs_max_freq).

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
Run the three CPU clusters on their ACPM DVFS clocks (cpucl0-2) through
cpufreq-dt, and add thermal zones for them on the ACPM TMU's big, mid
and little sensors (0-2):

- the operating points are the ACPM firmware's own levels, read from
  the ECT the bootloader leaves at 0x90000000, up to each cluster's
  max-freq in stock's exynos-acme domains: 1704 MHz little, 2367 MHz
  mid and 2914 MHz big. The ECT's little levels up to 1704 MHz are the
  ones stock's cpufreq statistics list;
- the energy model is stock's: each level's opp-microwatt is its power
  in the default pixel-em profile of Google's zuma-b0-ipop.dts, which
  the Pixel 8 Pro's own device tree has too. A dynamic power
  coefficient can't be used, since the firmware sets the voltages and
  there is no opp-microvolt: with one, the energy model fails to
  register ("EM: invalid perf. state: -22"). The CPUs get #cooling-cells
  for cpufreq cooling;
- the levels above stock's battery current limiter caps (1945 MHz mid,
  2294 MHz big, from Google's zuma-shusky-bcl.dtsi) are turbo-mode, as
  on gs201: with no BCL driver nothing pulls the clusters down when the
  battery voltage droops, so they are only used with boost enabled.
  Those caps are the Pixel 8 Pro's, so the tables stay in zuma.dtsi but
  only zuma-husky.dts points its CPUs at them: the Pixel 8a's caps in
  Google's zuma-akita-bcl.dtsi are lower, and it needs its own;
- a CPU zone throttles its cluster from 80 C, forces its lowest level
  from 95 C and is critical at 105 C, as zumapro.dtsi does. The TMU
  sensor and the zones are shared by all zuma boards.

The ACPM clock ids header moves from the board to zuma.dtsi.

Assisted-by: LLM
Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
@greenbrettmichael
greenbrettmichael force-pushed the omamobile-zuma-husky-cpufreq branch from e00ab31 to 6abc13c Compare November 12, 2026 17:59
@greenbrettmichael

Copy link
Copy Markdown
Contributor Author

Thanks Derek, good catch on the 8a. Rebased over #160 and onto today's omamobile (2db2bc3), and force-pushed (6abc13cef), with your fix in the same push:

  • CPU levels: the three OPP tables stay in zuma.dtsi, and the CPUs' operating-points-v2 moved to zuma-husky.dts (&cpu0 to &cpu8). The tables' comment now says the turbo-mode marks are the Pixel 8 Pro's caps and that a board with lower BCL caps needs tables of its own. The TMU sensor and the CPU and GPU zones stay shared. The commit message says the same.
  • arm64: dts: exynos: google: zumapro: CPU DVFS, thermal zones and the GPU's top levels #87: both messages now say "as zumapro.dtsi does". I checked its trips first: 80/95/105 C for the CPU zones and 85/105 C for the GPU, the same as ours.
  • In the built device trees: none of zuma-akita.dtb's nine CPUs has operating-points-v2 now (all nine had it on t1), so cpufreq-dt leaves the 8a alone. zuma-husky.dtb's /cpus and the three CPU tables are identical to t1's, phandles included.
  • Against t1: on 3b13c2d (right after Pixel 8 Pro display: gs-decon, DSIM0 and the HK3 panel at 120 Hz #160 landed) the rebased tree was exactly t1's a0d5f81ee17d plus the three other commits omamobile had gained and the move above. Moving on to 2db2bc3 needed no resolution: every commit is the same patch as on 3b13c2d. In zuma.dtsi both sides are kept in address order as in t1, and in zuma-husky.dts &cmu_g3d, the CPUs and &gpu sit before &decon0 and &dsi0, also as in t1.
  • git range-diff against e00ab31e1: commits 1-6, 9, 10 and 12 unchanged; 7 and 8 only in their context (the display nodes next to them); 11 only in its message; 13 by the move above.

tools/kernel-pr-check --build --merge passes: it merges cleanly onto 2db2bc3, checkpatch --strict has nothing to read, W=1 builds of the 4 touched C files pass, 10 device trees build and pass dtbs_check with no new warnings, and dt_binding_check passes on 3 bindings.

I haven't booted this head. Its code and the 8 Pro's CPU, GPU and display nodes are what t1 ran; what's new around them is omamobile's own since then.

@derekross

Copy link
Copy Markdown
Owner

Thank you, Brett, that is exactly the fix I had in mind, and the push is easy to check. The three tables stay in zuma.dtsi, the CPUs get operating-points-v2 in zuma-husky.dts, and the comment and the message say whose caps the turbo marks are. I built both boards from this head: none of the Pixel 8a's nine CPUs has the property now, so cpufreq leaves it alone, and each of the 8 Pro's nine points at the same table as before.

I also compared this head with your t1 tag, since you say plainly that the head itself hasn't booted: every driver file, binding and clock header is the same blob as in t1, and the two device-tree files differ from t1 only by what omamobile itself gained since (the PMIC interrupt, the s2idle argument, the GNSS and NFC nodes) and by the move. That is enough for me. On the Pixel 7 Pro side nothing changes: its device tree is byte-identical with and without this, and the drivers are the ones I checked last time. Our gate passes on the merge with today's omamobile, CI is green with dtbs_check and the binding check, and each device-tree commit builds on its own.

So this goes in as it is, and I'll close #157 when it lands, since its commits come with this one. If you get a chance to boot the merged omamobile on your 8 Pro afterwards, a line here with uname -r and the three policies would round it off.

One note for the 8a, not for you to do anything about: it shares the TMU sensor and the thermal zones from zuma.dtsi, so its next kernel will read temperatures and get the 105 C critical trip for the first time. I'll let its owner know.

@derekross
derekross merged commit 82f31d3 into derekross:omamobile Nov 13, 2026
1 check passed
@greenbrettmichael

Copy link
Copy Markdown
Contributor Author

Booted merged omamobile on my 8 Pro, at 82f31d3a7 (this PR's last commit): 7.3.0-rc6-omamobile-push1-g82f31d3a7a35, the kernel as the only change (modules by modules-push, slot B by omamobile flash, the e1 initramfs and omamobile-device-husky 0.1.8 unchanged).

Recorded in the 8 Pro's notes in derekross/omarchy-mobile#741.

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