Skip to content

arm64: dts: exynos: google: zumapro-tegu: Add the haptics and the torch - #96

Merged
derekross merged 7 commits into
derekross:omamobilefrom
W4HO-ham:tegu-dt-haptics-torch
Nov 1, 2026
Merged

derekross merged 7 commits into
derekross:omamobilefrom
W4HO-ham:tegu-dt-haptics-torch

Conversation

@W4HO-ham

@W4HO-ham W4HO-ham commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor

Stacked on #90 and the DT and clock PRs under it. It uses #95's cirrus,vpbr-threshold-microvolt. Only the last commit is new.

The Pixel 9a's haptics and torch:

  • Haptics: a CS40L26 on USI9's I2C bus (0x43, interrupt gpa6-3 level low, reset gph2-3). It uses Google's limits for the 9a's LRA: 2.5 A boost peak, 10 V boost, 162 Hz, ReDC 14.1 ohm, and VP brownout prevention at 3.008 V.
  • Torch: a TI LM3644 on USI4's I2C bus (0x63, enable gpp24-1), both outputs on one white LED, limited to 75 mA. That's a dim flashlight, far below the part's 1.5 A flash and the LED's limits.

Against the palmtop: both nodes, their buses and pin groups match what it runs, property for property. The only differences are the PERIC0 clock numbers, renumbered by the clock series (same clocks), and the torch limit: the palmtop gets 75 mA from a runtime overlay, and this puts it in the DT.

CHECK_DTBS=y W=1 for zumapro-tegu.dtb is clean (with #95's binding), and checkpatch --strict is clean.

@derekross

Copy link
Copy Markdown
Owner

Thanks. Both nodes match Google's tegu sources: the CS40L26 at 0x43 on hsi2c_9 (interrupt gpa6-3, reset gph2-3, 2.5 A, 10 V, 162 Hz, VPBR at 3.008 V), the LM3644 at 0x63 on hsi2c_4 with FLASH_EN on gpp24-1, and the bus pin groups as in zumapro-pinctrl.dtsi. 14100007 is a nice touch: through the driver's conversion it gives exactly stock's 0x70CCD, where 14100000 would be one step lower. A short comment saying so would save the next reader the arithmetic.

The CI failure is the stack's, not this commit's: it validates against bindings that don't have #89's google,tg4c or #95's cirrus,vpbr-threshold-microvolt yet. With those two merged into the schemas, the 9a's DTB validates with no messages.

So this merges after the board stack, #89, #90 and #95. Once those land, please rebase it onto omamobile. The 100 kHz buses are a conservative choice (stock runs 950 kHz for haptics and 390 kHz for the torch), and that's fine with me.

@W4HO-ham
W4HO-ham force-pushed the tegu-dt-haptics-torch branch from d113ffd to 9254f9e Compare October 27, 2026 21:53
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Rebased onto the updated #90, which now carries #89's panel binding, so dtbs_check no longer reports google,tg4c as unmatched. Nothing in this commit changed.

@derekross

Copy link
Copy Markdown
Owner

Your commit is unchanged, and it's now the top of a stack that sits cleanly on today's omamobile. CI passes with dtbs_check, since #89's and #95's bindings are in.

This is the PR I'll rebase-merge to land the whole 9a stack, once the boot on #80 is in. The short comment on 14100007 would still be welcome in that push, but it's optional.

@derekross

Copy link
Copy Markdown
Owner

Thanks. The haptics and torch commit is the same patch as before, with an identical tree (8723ae7003af). The only change in the stack is the "Tested on" paragraph in the board commit. The short comment on 14100007 didn't make it in. It was optional, so that's fine.

This is still the PR I'll rebase-merge to land the 9a's board stack. Two things before that, both described on #80:

  1. The rebase. The google DTS Makefile conflicts with today's omamobile (the Pixel 7 board), and that's why no CI ran on the new heads. I did the rebase here with that one hunk resolved: it passes the gate (sign-offs, prefixes, checkpatch, W=1 builds of the device trees), every commit builds the 9a's DTB without warnings, dt_binding_check passes for the touched schemas, the 9a's DTB validates with no messages and is byte-identical to the one from your previous head, and the other seven Google DTBs are byte-identical to the base.

  2. The boot of this head after the rebase: uname -r, the commit, the tools/phone-smoke report (its display, touch and GPU rows), the uptime past 60 s, and the DTB's own /chosen/bootargs (or a list of what your bootargs overlay changes).

    A correction to what I wrote on arm64: dts: exynos: google: Add the Tensor G4 SoC and the Pixel 9a #80: tools/phone-smoke has no haptics section. It never vibrates the phone or lights the torch, so it can't cover this commit. For haptics and the torch, a read-only check is enough: the two probe lines from dmesg | grep -iE 'cs40l26|lm3644', the CS40L26 in /proc/bus/input/devices with its B: FF= line, and the LM3644's entry under /sys/class/leds/, with its max_brightness. Please don't run a vibration or torch test for this; that isn't needed to merge.

When those are in, I'll merge this and close #80, #81, #85, #87 and #90 as landed here. #105, #106, #119, #122 and #118 then follow on their own threads.

@derekross

Copy link
Copy Markdown
Owner

Thanks for the rebase. It's exactly the Makefile resolution from #80: gs201-panther.dtb kept, zumapro-tegu.dtb and its DTC_FLAGS line added. Nothing else moved. I compared every branch with the heads I looked at earlier today. The board commit's Makefile hunk is the only difference anywhere in the stack, every other commit is the same patch, and no file the stack touches differs apart from that Makefile.

With the conflict gone, CI ran on every head and passes, dtbs_check and dt_binding_check included. I ran the gate here as well, on this head and on the combined top of the stack (#118, #119 and #122 together), both merged onto today's omamobile. Both pass: sign-offs, prefixes, checkpatch (only the MAINTAINERS notes and the bootargs line), the W=1 builds of the four C files and the device trees. The other seven Google DTBs, the Pixel 7 Pro's included, are byte-identical to the base. The 9a's DTB has the same inputs as before the rebase, because the only change in what it includes is five clock IDs appended to the zumapro header, which it doesn't use. The commit-by-commit builds from this afternoon therefore still hold.

So one thing is left before I merge this, as asked on #80: the boot of this head. Please post:

  • uname -r and the commit it was built from (d8acd98c7d9e, or a tree with the same DTB);
  • the tools/phone-smoke report. I'm looking for the display row (1080x2424 on DSI-1), touch, and GPU (a render node, with panthor bound);
  • the uptime, past the bootloader watchdogs' 60 s;
  • the DTB's own /chosen/bootargs for that boot, or a list of what your bootargs overlay changes;
  • for the haptics and torch, the read-only check only: the probe lines from dmesg | grep -iE 'cs40l26|lm3644', the CS40L26's entry in /proc/bus/input/devices with its B: FF= line, and the LM3644 under /sys/class/leds/ with its max_brightness. No vibration or torch test, please.

When that's in, I'll rebase-merge this PR, which lands all twelve commits, and close #80, #81, #85, #87 and #90 as landed here.

Elsewhere in the stack, nothing new: #106's ramoops layout question and the three small changes on #122 are still open as I wrote them earlier today. #119 is ready and waits on #106. #105 waits on #100 and #104.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Booted this head on the palmtop: 7.3.0-rc6-omamobile-tegu-pr96-gd8acd98c7d9e, built from d8acd98 with the integration tree's config (olddefconfig), its own modules and this tree's zumapro-tegu.dtb as built. The only overlay was the one that sets /chosen/bootargs.

  • Uptime: 85 s at the smoke run, past the bootloader watchdogs' 60 s, on slot B. Then a push back to the integration build.

  • Bootargs (the DTB's /chosen/bootargs for this boot; the part before the second clk_ignore_unused pd_ignore_unused printk.devkmsg=on is our overlay, the rest is ABL's):

    clk_ignore_unused pd_ignore_unused rdinit=/init panic=10 bootconfig omamobile.noresize s3c2410_wdt.tmr_atboot=1 modprobe.blacklist=gs_decon,panthor systemd.unit=multi-user.target systemd.wants=tegu-curated-root-guard.service omamobile.tegu-curated-root systemd.mask=ModemManager.service systemd.mask=NetworkManager.service systemd.mask=autovt@tty1.service systemd.mask=bluetooth.service systemd.mask=display-manager.service systemd.mask=getty@tty1.service systemd.mask=omamobile-aoc.service systemd.mask=omamobile-boot-success.service systemd.mask=omamobile-firstboot.service systemd.mask=omamobile-fp.service systemd.mask=omamobile-gnss.service systemd.mask=omamobile-modem.service systemd.mask=omamobile-speaker-protection.service systemd.mask=omamobile-wifi.service systemd.mask=sddm.service systemd.mask=storageproxyd.service systemd.mask=tegu-battery-hold.service systemd.mask=tegu-battery.service systemd.mask=tegu-display-confirm.service systemd.mask=tegu-display.service systemd.mask=tegu-gpu.service systemd.mask=tegu-native-root-deadman.service systemd.mask=tegu-native-root-verify.service systemd.mask=tegu-normal-reboot.service systemd.mask=tegu-touch.service systemd.wants=tegu-curated-autostart.service omamobile.tegu-curated-autostart omamobile.tegu-standard-reboot clk_ignore_unused pd_ignore_unused printk.devkmsg=on bootconfig console=null exynos_drm.panel_name=google-tg4c.<panel id> tcpci_max77759.conf_sbu=0 arm64.nomte ufs_pixel_fips140.fips_first_lba=151942 ufs_pixel_fips140.fips_last_lba=152453 ufs_pixel_fips140.fips_lu=0 ufs_pixel_fips140.use_hw_keys=true id_aa64mmfr1.vh=0  bootconfig
    
  • Haptics and torch, read-only:

    • cs40l26-vibra 1-0043: Cirrus Logic CS40L26 ID 0x40a260, revision 0xa1
    • firmware 0x1800d4 rev 7.2.81 running, input: cs40l26_vibra as .../i2c-1/1-0043/input/input2
    • /proc/bus/input/devices: N: Name="cs40l26_vibra", B: FF=107030000 0
    • lm3644 0-0063: torch: 25 steps, 2550-72390 uA per channel, /sys/class/leds/white:torch with max_brightness 25

    No vibration and no torch test.

  • Display and GPU: both are deferred on this tree alone, so those rows read "not ok". /sys/kernel/debug/devices_deferred says 19470000.display-controller gs-decon: DSIM and panel and 1f000000.gpu panthor: get 'core' clock failed. The panel driver for google,tg4c is drm/panel: Add the Google TG4C panel (Pixel 9a) #89 and the GPU's clock is clk-acpm's zumapro variant (clk: samsung: acpm: Add Tensor G4 (zumapro) with a paired GPU clock #84), and neither is in this stack. The DT nodes are there and okay. gs_decon, gs_dsim_cmd and panthor load when modprobed, then wait on those two providers. On the integration tree, which has both, the same nodes give the 1080x2424 desktop at 120 Hz and panthor bound with a render node.

  • Touch: ok (Synaptics TouchComm Touchscreen).

tools/phone-smoke:

Check Result Detail
boot (required) ok kernel 7.3.0-rc6-omamobile-tegu-pr96-gd8acd98c7d9e, slot b, up 85s
storage guard (required) ok 62 partitions; only userdata and slot B's boot images writable
root (required) ok ext4 on userdata, 21G free
kernel health (required) ok no oops or BUG; 0 WARN, 10 error lines in dmesg
systemd units not ok failed: tegu-curated-autostart.service
USB gadget (required) ok bound to 11210000.usb
display not ok no connected DRM connector
GPU not ok no render node
session (required) not ok Hyprland isn't running, the shell isn't running
lock ok PAM file present (the shell didn't answer lock status)
touch ok Synaptics TouchComm Touchscreen
buttons ok gpio-keys
Wi-Fi skip no wireless interface
Bluetooth skip no controller
battery skip no battery power supply
sound cards skip none
sensors skip no sensor proxy
cameras not ok libcamera lists none (or a camera app holds them)

If you'd like the display and GPU rows from this tree too, I can boot it again with #84 and #89 merged on top.

@derekross

Copy link
Copy Markdown
Owner

Thanks for the boot report, and for keeping to the read-only checks. The kernel release has no -dirty and its hash is this head, and the uptime and touch rows are in. The haptics and torch lines all match what the drivers print. B: FF=107030000 0 is exactly the rumble bit plus the periodic and gain bits that the memless layer adds on top of it, and 25 steps from 2550 to 72390 uA is what leds-lm3644.c reports for the node's limit. I rebuilt this head on today's omamobile. It still merges cleanly, the gate passes, and the 9a's DTB is byte-identical to the one I checked yesterday.

One thing still stands between this PR and a merge: the display and GPU rows. I don't think the reason you give holds, though. The TG4C panel driver (#89, cf17900a3178 drm/panel: Add the Google TG4C panel) and clk-acpm's zumapro variant (#84, f171fb4ccbf6 clk: samsung: acpm: Add Tensor G4 (zumapro) with a paired GPU clock) both merged before this stack was rebased. They're in the base under it, so they're already in the tree at d8acd98c7d9e. Merging #84 and #89 on top would give you the same tree again.

So I suspect the config. You built with your integration tree's config and olddefconfig. DRM_PANEL_GOOGLE_TG4C has no default, so it comes out off unless that config already had it. If the panel module isn't built, gs-decon waits on "DSIM and panel", which is the message you saw. Panthor's "get 'core' clock failed" would follow in the same way if EXYNOS_ACPM_CLK is off or clk_acpm isn't loaded, because the 9a's GPU node takes its core clock from &acpm_ipc. The test boot's command line also had modprobe.blacklist=gs_decon,panthor and masked your display and GPU units, so even with the modules built, neither would have come up on its own.

Could you check and boot once more?

  1. On the build you just ran: zcat /proc/config.gz | grep -E 'DRM_PANEL_GOOGLE_TG4C|EXYNOS_ACPM_CLK|DRM_GS_DECON|DRM_GS_DSIM_CMD|DRM_PANTHOR' (or the same lines from the build's .config), and lsmod | grep -E 'clk_acpm|panel_google_tg4c'. That tells us whether it was the config.
  2. Rebuild d8acd98c7d9e with those symbols set to m (or y). Boot it with a bootargs overlay that doesn't blacklist gs_decon or panthor and doesn't mask the display and GPU units, so they load the way they will for everyone.
  3. Post uname -r, the sha256sum of the zumapro-tegu.dtb you built (before the bootargs overlay; mine is b5ea1c7d42c4…), and the tools/phone-smoke display and GPU rows. I'm looking for 1080x2424 on DSI-1, and a render node with panthor bound. Please also post the error lines that the kernel health row counts (dmesg --level=err, with identifiers removed).

No haptics or torch re-check is needed. The probe lines you posted cover them, and step 2 doesn't change those drivers.

For the record, this boot's bootargs aren't the DTB's own. The DTB built from this head carries console=tty0 clk_ignore_unused pd_ignore_unused rdinit=/init panic=10 mem_sleep_default=s2idle. Your overlay drops console=tty0 and mem_sleep_default=s2idle, and adds s3c2410_wdt.tmr_atboot=1, the blacklist and the unit masks. None of that matters for the haptics and torch result. The watchdog option keeps the timers running and fed by the kernel rather than stopped, so the 60 s result holds either way.

Once the display and GPU rows are in from this commit, I'll rebase-merge this PR, which lands all twelve commits, and close #80, #81, #85, #87 and #90 as landed here. After that, #55 needs a rebase, and it should bring the 9a's pmic-1 interrupt and google,vgpio2ap-sysreg with it, because its binding makes both required. The rest of the stack rebases onto the landed commits.

@derekross

Copy link
Copy Markdown
Owner

Thanks for your patience with this one. A status from my side, and an offer.

I rebuilt this head on today's omamobile, which is fourteen commits further on (the Pixel 7's touchscreen and haptics, the MAX77779 charger, the raven haptics). None of them touch a file the 9a's tree includes. The stack still merges cleanly and the gate passes. The 9a's DTB is byte-identical to the one I checked on the 28th (b5ea1c7d42c4…) and validates with no messages, and the other seven Google DTBs are byte-identical to the base. So nothing has moved under you, and nothing needs a rebase.

The one thing still open is what I asked for on the 29th: the display and GPU rows from a boot of this commit, built with DRM_PANEL_GOOGLE_TG4C and EXYNOS_ACPM_CLK on, with no blacklist or masks on gs_decon, panthor and their units, plus the DTB's sha256 and the dmesg --level=err lines.

While checking that, I found a gap on my side too. Our shared config fragment in omarchy-mobile doesn't build the TG4C panel driver either, and it doesn't build the ACPM thermal driver your thermal zones need (GS101_ACPM_THERMAL). So a kernel built with bin/omamobile kernel today couldn't light the 9a's panel, whatever the DT says. I'm adding both to kernel/configs/omamobile.config as modules now; until that lands, please set both symbols to m yourself for this boot.

So that the boot and touchscreen aren't held up, I've landed them now by rebase-merging #81 (its five commits had all run in the form that lands: the board in your 10-28 boot, touch in your 10-29 boot of this head). This branch and the ones above it only need a git rebase omamobile, which drops those five commits on its own; I checked that it gives exactly the same tree as this head merged today.

@W4HO-ham
W4HO-ham force-pushed the tegu-dt-haptics-torch branch from d8acd98 to 532a6f3 Compare October 31, 2026 02:39
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Rebased onto #90's new head (one commit, unchanged). Compiles; checkpatch clean.

@derekross

Copy link
Copy Markdown
Owner

Thanks, the rebase is clean and the 9a's DTB is still b5ea1c7d42c4…. Our config fragment now builds DRM_PANEL_GOOGLE_TG4C and GS101_ACPM_THERMAL (omarchy-mobile #191), so a bin/omamobile kernel build of this commit, booted without the gs_decon/panthor blacklist and masks, should give the display and GPU rows. Those rows, the DTB's sha256 and the dmesg --level=err lines are the last thing before I merge #85, #87, #90 and this together.

Turn on CMU_G3D and the Mali-G715 for the Pixel 9a, with its core clock
on ACPM's paired "gpu" clock (index 17), which sets the shader (G3D) and
L2 (G3DL2) domains together, and coregroup and stacks on CMU_G3D's gates.

The operating points are the shader rates of the 9a's GPU DVFS table up
to 649 MHz. The levels above it, to 940 MHz, wait for a GPU thermal
zone, since nothing throttles the GPU yet.

Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
Give the three CPU clusters their ACPM DVFS clocks (cpucl0-2, by the
zumapro ACPM clock ids) and operating points for cpufreq-dt, with
capacities and power coefficients for the energy model from Google's
zumapro-cpu.dtsi, and add the ACPM TMU sensor and the big, mid, little
and GPU thermal zones.

The CPU levels are the APM firmware's own (its DVFS table, read on the
Pixel 9a), up to the 9a's thermal headroom: 1950 MHz little (Google's
max-freq for that cluster), 2367 MHz mid and 2687 MHz big. The little
cluster's 324 and 610 MHz lie below stock's 820 MHz floor. Each CPU zone
throttles its cluster from 80 C, forces the cluster's lowest levels
from 95 C and shuts down at 105 C, all below stock's 112 C (CPU) and
115 C (GPU) hot trips. gpu-thermal throttles panthor's devfreq from
85 C.

Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
… MHz

With gpu-thermal throttling the GPU, add the 9a's GPU levels above 649
MHz: 723, 807, 850 and 890 MHz. The table's top, 940 MHz, stays out
until it has run on the 9a.

Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
The Tensor G4 (zumapro) display pipeline has the gs201 DSIM, at the same
register layout and with the same three register windows. Add
google,zumapro-dsim with google,gs201-dsim as its fallback.

Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
Add DSIM0, the DSI master between DECON0 and the panel: its link and
D-PHY register windows and its interrupt (GIC SPI 252), as in Google's
GPL zumapro device tree, disabled until a board names its panel.

Link: https://android.googlesource.com/kernel/devices/google/zumapro/+/8831630eee2ad095b85ed452944e6353940966a5/dts/zumapro.dtsi
Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
Turn on the Pixel 9a's display pipeline: DECON0 with a 64 MiB reusable
scanout pool below 4 GiB and the bootloader's splash as its boot
framebuffer, scanning out to DSIM0, and the TG4C panel on DSIM0's
virtual channel 0. With DECON's output on DSIM0, gs-decon switches the
link and the panel between 60 and 120 Hz.

Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
Add the Pixel 9a's haptics, a Cirrus Logic CS40L26 on USI9's I2C bus
(0x43, interrupt on gpa6-3, reset on gph2-3), with Google's limits for
its LRA (2.5 A boost peak, 10 V boost, 162 Hz, ReDC 14.1 ohm) and VP
brownout prevention at 3.008 V, and its torch, a TI LM3644 on USI4's
I2C bus (0x63, enable on gpp24-1) driving one white LED from both
outputs, limited to 75 mA.

Signed-off-by: W4HO <87742609+W4HO-ham@users.noreply.github.com>
@W4HO-ham
W4HO-ham force-pushed the tegu-dt-haptics-torch branch from 532a6f3 to 5628f18 Compare October 31, 2026 15:59
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Rebased onto today's omamobile (b3a7833, with #106's ramoops in it); nothing dropped since ramoops was never in this branch, seven commits as before. Device tree compiles, checkpatch clean.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

The boot you asked for, from this head (5628f18) on the palmtop.

Build. defconfig, then kernel/configs/omamobile.config, then olddefconfig (there's no 9a fragment), make Image.gz dtbs modules. That's bin/omamobile kernel's steps, run natively on an arm64 build VM with Ubuntu clang 21.1.8 because the container's Android clang isn't on that machine. The config has:

CONFIG_DRM_GS_DSIM_CMD=m
CONFIG_DRM_GS_DECON=m
CONFIG_DRM_PANEL_GOOGLE_TG4C=m
CONFIG_DRM_PANTHOR=m
CONFIG_EXYNOS_ACPM_CLK=m

One thing had to change first. Built from the fragment as it stands, the 9a boots with no USB at all: it reached the lock screen, but no gadget ever appeared on the host. The eUSB2 PHY, and the ACPM/mailbox/PMIC path its rails come through, are modules in defconfig, and the initramfs loads none. omarchy-mobile#227 builds those four in. The boot below is with that change. Nothing else differs.

Boot. uname -r: 7.3.0-rc6-omamobile-push1-g5628f1888904, no -dirty. The bootargs overlay keeps the DTB's own arguments, adds s3c2410_wdt.tmr_atboot=1, omamobile.noresize and cs40l26_vibra.max_buzz_ms=100, and masks only units that must not run unasked (omamobile-boot-success, storageproxyd, omamobile-fp, omamobile-modem, omamobile-aoc, omamobile-firstboot) plus bring-up-only units. Nothing blacklists or masks gs_decon, panthor or the graphics unit. The package's omamobile-tegu-graphics-start loaded them the way it will for everyone: gs_decon gs_dsim_cmd panel_google_tg4c panthor clk_acpm in lsmod.

DTB sha256 (before the bootargs overlay): 44857e681743d4d71339f592bf1ddf34c6de331b139724e5bc1de2265aad3f95. It isn't your b5ea1c7d42c4…, and I think I know why: this head is the rebase onto b3a7833, and git diff d8acd98c7d9e 5628f1888904 -- arch/arm64/boot/dts/exynos/google/ changes zumapro-reserved-memory.dtsi (ramoops), which the 9a's DTB includes. If you built yours from d8acd98, that would explain the difference.

tools/phone-smoke:

| display | ok | card1-DSI-1 connected, 1080x2424 |
| GPU | ok | render node on panthor |
| kernel health (required) | ok | no oops or BUG; 0 WARN, 18 error lines in dmesg |

dmesg --level=err (no identifiers in it). Eight of the 18 are the palmtop's bring-up units trying to load out-of-tree modules built for the integration kernel, refused for version magic (tegu_nfc_clkpad, goodix_fp, tegu_gpu_dvfs_activate), plus one systemd generator line. The kernel's own:

samsung-pinctrl 10840000.pinctrl: irq number not available   (also 10c40000, 12040000, 13040000, 13060000, 154d0000)
exynos-clkout exynos-clkout: cannot match parent device
exynos-clkout exynos-clkout: probe with driver exynos-clkout failed with error -22
Warning: unable to open an initial console.
cpu cpu0: EM: invalid perf. state: -22   (also cpu4, cpu7)
panthor 1f000000.gpu: error -ENODEV: _opp_set_regulators: no regulator (mali) found

The graphics unit itself reported "failed: gauge", only because max77779_battery isn't in this tree. It logged "up: GPU, display, haptics, torch" before that. The palmtop is back on its daily kernel.

@derekross

Copy link
Copy Markdown
Owner

Thank you, @W4HO-ham, for this boot. It's exactly the run I asked for, and thanks for chasing down the missing USB along the way instead of working around it.

What I checked:

  • The kernel is this head. 7.3.0-rc6-omamobile-push1-g5628f1888904 has no -dirty, and its hash is this PR's head.
  • The DTB hash. I rebuilt this head's zumapro-tegu.dtb and got exactly your 44857e681743…. Your explanation of why it differs from my b5ea1c7d42c4… is right: the rebase onto b3a783356 brought ramoops into zumapro-reserved-memory.dtsi. The seven patches are the same ones I reviewed at d8acd98c7d9e (git range-diff shows all seven as unchanged).
  • Display and GPU came up the way they will for everyone: card1-DSI-1 connected, 1080x2424, a render node on panthor, gs_decon gs_dsim_cmd panel_google_tg4c panthor clk_acpm loaded by the package's start script, and nothing blacklisted or masked for them. The extra boot arguments (tmr_atboot, noresize, max_buzz_ms) and the masked units don't touch what this PR tests.
  • The error lines. None of them is caused by this stack:
    • exynos-clkout … cannot match parent device comes from the base. The PMU registers a clock-out child on every SoC, and the driver has no Tensor G4 entry yet. Your soc, nfc: the Pixel 9a's NFC clock pad as a PMU clock #104 rework adds one, so this line goes away when soc, nfc: the Pixel 9a's NFC clock pad as a PMU clock #104 lands.
    • EM: invalid perf. state: -22 on the three clusters and panthor's no regulator (mali) found both come from OPP tables that carry no voltages. The energy model needs voltages, and panthor runs on its clock alone without them. The thermal zones use the default step-wise governor, which doesn't need an energy model, so both are harmless for now. Worth knowing if anyone adds a power-allocator zone later.
    • The pinctrl irq number not available lines are the base's controllers without interrupts.
  • The gate on today's omamobile (caf5f8a2b): it merges cleanly, the sign-offs and prefixes pass, checkpatch has no warnings, and the eight Google device trees build with W=1 and no new warnings. The 9a's DTB passes dt-validate with no messages, and the other seven Google DTBs, the Pixel 7 Pro's among them, are byte-identical to the base. CI is green on this head.

The USB problem is real but separate from this stack: the eUSB2 PHY and its ACPM/PMIC path being modules that the initramfs never loads is a question for our config fragment. I'll take it up on omarchy-mobile#227.

So this is ready. I'll rebase-merge it, which lands the seven commits, and close #85, #87 and #90 as landed here. Their patches are all in this branch.

After that, here's the order for the rest:

  1. soc, nfc: the Pixel 9a's NFC clock pad as a PMU clock #104 is independent of this stack and merges onto today's base as it is.
  2. arm64: dts: exynos: google: zumapro-tegu: Add the NFC controller #105, arm64: dts: exynos: google: zumapro-tegu: Add the fuel gauge and USB-C port controller #119, arm64: dts: exynos: google: zumapro: Add USI10 and its I2C controller #136 and arm64: dts: exynos: google: zumapro-tegu: Add Bluetooth #122 then need a git rebase omamobile. That drops this PR's commits, and for arm64: dts: exynos: google: zumapro-tegu: Add the NFC controller #105 also soc, nfc: the Pixel 9a's NFC clock pad as a PMU clock #104's, on its own.
  3. arm64: dts: exynos: google: zumapro-tegu: Add the NFC controller #105 and arm64: dts: exynos: google: zumapro-tegu: Add the fuel gauge and USB-C port controller #119 touch neighbouring lines (&gpa7 and &gpa8), so whichever lands second gets a small context conflict. The resolution keeps both blocks, &gpa7 first. I've put the exact hunk on arm64: dts: exynos: google: zumapro-tegu: Add the fuel gauge and USB-C port controller #119.

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