Skip to content

arm64: dts: gs201-panther: the touchscreen, suspend-to-idle and LDO25M - #134

Merged
derekross merged 5 commits into
derekross:omamobilefrom
nesmeck:omamobile-gs201-panther-touch
Oct 30, 2026
Merged

derekross merged 5 commits into
derekross:omamobilefrom
nesmeck:omamobile-gs201-panther-touch

Conversation

@nesmeck

@nesmeck nesmeck commented Oct 29, 2026

Copy link
Copy Markdown
Contributor

The Pixel 7's (panther) touchscreen, plus the two board changes derekross suggested for it. Stacked: the first two commits are #129 (GPU) and #130 (haptics), because all of them edit the board file's header comment; the three new commits are on top. It also needs the driver fixes in #133 to be tested as below, though it builds and probes without them.

New commits:

  • default to suspend-to-idle: mem_sleep_default=s2idle in the bootargs, as the Pixel 7 Pro has. The firmware advertises PSCI SYSTEM_SUSPEND, so the kernel offered s2idle [deep], and gs201 never wakes from deep.
  • pin LDO25M at 1.8 V: the touch controller's VDD. Google's common gs201-pmic.dtsi allows 1.6-1.95 V and gs201-panther-pmic.dtsi doesn't narrow it; ABL leaves it at 1.8 V. Fixed at 1.8 V, as the Pixel Fold describes its own LDO25M.
  • add the touchscreen: the FocalTech FT3658 on USI0 SPI with the focaltech,ft3658 binding, wired as the Pixel 7 Pro's S3908 (chip select gpp20-3, interrupt gpa7-0 falling, reset gpp23-2), 10 MHz, 1080x2400; it enables the main PMIC, whose only regulators here are the controller's rails (LDO25M as VDD, LDO26M as AVDD). LDO26M's 3.0 V cites gs201-panther-pmic.dtsi in kernel/devices/google/pantah ("Panther touch controller requires 3.0v typ"). The node is the one tested during the driver's bring-up.

The omarchy-mobile side (CONFIG_TOUCHSCREEN_FOCALTECH_FT3658=m in panther's kernel config) follows in its own PR.

Tested on a Pixel 7 (slot B, 7.3.0-rc6), this branch merged with #133, modules installed before the boot:

  • /sys/power/mem_sleep reads [s2idle] deep;
  • vdd-ts reads 1.8 V and avdd-ts 3.0 V, each with spi0.0 as its only consumer;
  • focaltech_ft3658 loads at boot by itself and reads chip id 0x56, firmware 0x16; touches across the panel, the four corners included, drive Hyprland and unlock the PIN screen, with no SPI or CRC errors;
  • the GPU (panthor, CSF v1.5.0) and the haptics (cs40l26, a 300 ms rumble felt) keep working, and ACPM has 14 of 15 channels usable.

Checks: tools/kernel-pr-check --build OUT --merge origin/omamobile passes: a clean merge onto 2a318a9, sign-offs, and the seven device trees with W=1 and dtbs_check, with no new warnings. checkpatch --strict's one warning is the 144-column bootargs line, as in #108.

🤖 Generated with Claude Code

Enable the Mali-G710 that gs201.dtsi describes, as the Pixel 7 Pro does.
Its clock is the ACPM GPU clock, which this board already gets through
acpm_ipc, and as on the Pixel 7 Pro there is no mali-supply: the G3D rails
follow the PMU power domain and ACPM, and the bootloader leaves the GPU
powered.

Tested on a Pixel 7 from slot B (7.3.0-rc6), with the device package's
rule that keeps the GPU powered: panthor probes the Mali-G710 (id 0xa862,
shader_present 0x1110055), boots the CSF firmware (interface v1.5.0) and
registers renderD128, and Hyprland renders on it through Mesa's panfrost
("Mali-G710 MC7 (Panfrost)") with no MMU faults or GPU resets. As on the
Pixel 7 Pro, probe prints "no regulator (mali)" from the OPP core.

Assisted-by: Claude Opus 5.5
Signed-off-by: Alejandro Garcia <skipper.pelygro@gmail.com>
Describe the CS40L26A haptics driver at 0x43 on USI8 I2C, as on the
Pixel 7 Pro: the same bus (gpp16-0/1 at 400 kHz), interrupt (gpa8-1)
and reset (gpp24-3) as the Pixel 7 Pro, and the same stock boost and
actuator values (peak current 2.5 A, boost 11 V, f0 145 Hz, ReDC
13 ohm), all read from the Pixel 7's live device tree. Enable the
PERIC0 pin controller, USI8 and its I2C bus with it.

The driver loads its RAM firmware (cs40l26.wmfw) and, if present, the
wavetable (cs40l26.bin), from the phone's own vendor files; without the
firmware it does not probe.

Tested on a Pixel 7 from slot B (7.3.0-rc6), with the GPU enabled as
well: cs40l26-vibra probes the CS40L26 (ID 0x40a260, revision A1), loads
cs40l26.wmfw (v7.2.91) and cs40l26.bin, and registers cs40l26_vibra;
FF_RUMBLE effects of 100-500 ms at 25-80 % play through tools/ff-rumble
and are felt, with no errors from the driver.

Assisted-by: Claude Opus 5.5
Signed-off-by: Alejandro Garcia <skipper.pelygro@gmail.com>
The firmware advertises PSCI SYSTEM_SUSPEND, so the kernel offers deep
sleep and makes it the default: /sys/power/mem_sleep reads
"s2idle [deep]" on the Pixel 7. Linux has no gs201 sleep sequence yet,
and the phone never wakes from deep. Pass mem_sleep_default=s2idle, as
the Pixel 7 Pro does, so a suspend that names no state stays in s2idle.

Tested on a Pixel 7 from slot B (7.3.0-rc6): /sys/power/mem_sleep reads
"[s2idle] deep".

Suggested-by: Derek Ross <derekross@gmail.com>
Assisted-by: Claude Opus 5.5
Signed-off-by: Alejandro Garcia <skipper.pelygro@gmail.com>
LDO25M is the touch controller's VDD. It has no voltage here, so the
regulator core leaves it at whatever the PMIC holds. Google's common
gs201-pmic.dtsi constrains it to 1.6-1.95 V, panther's
gs201-panther-pmic.dtsi doesn't narrow that, and ABL leaves it at
1.8 V, which vdd-ts reads on the phone. Fix it at 1.8 V, as the Pixel
Fold describes its own LDO25M, so the touch controller gets stock's
voltage whatever the PMIC held before.

Tested on a Pixel 7 from slot B (7.3.0-rc6): vdd-ts reads 1.8 V, with the
touch controller as its only user, and the controller probes as before.

Suggested-by: Derek Ross <derekross@gmail.com>
Assisted-by: Claude Opus 5.5
Signed-off-by: Alejandro Garcia <skipper.pelygro@gmail.com>
Describe the FocalTech FT3658 on USI0 SPI, with the focaltech,ft3658
binding: the Pixel 7 Pro's S3908 wiring (chip select gpp20-3, interrupt
gpa7-0 falling, reset gpp23-2, gs201-pinctrl.dtsi's ts_irq and
ts_reset), 10 MHz, and the 1080x2400 panel. Enable the main PMIC, whose
only regulators here are the controller's rails: LDO25M as VDD and LDO26M
as AVDD. LDO26M's 3.0 V is Google's own value for this phone:
gs201-panther-pmic.dtsi in kernel/devices/google/pantah sets
regulator-max-microvolt = <3000000> on m_ldo26_reg ("Panther touch
controller requires 3.0v typ").

Tested on a Pixel 7 from slot B (7.3.0-rc6), with the GPU, the haptics
and the driver's pending DMA-safe buffer and fsleep() fixes: the ft3658
module loads at boot and reads chip id 0x56, firmware 0x16; vdd-ts reads
1.8 V and avdd-ts 3.0 V; touches across the panel, the four corners
included, drive Hyprland and unlock the PIN screen, with no SPI or CRC
errors. The GPU and the haptics keep working.

Assisted-by: Claude Opus 5.5
Signed-off-by: Alejandro Garcia <skipper.pelygro@gmail.com>
@derekross

Copy link
Copy Markdown
Owner

Thank you, @nesmeck. This completes everything from the #108 thread, mem_sleep_default=s2idle and the LDO25M pin, and the touchscreen node is careful work: the wiring and the 3.0 V AVDD both trace to Google's panther files.

Here's what I checked:

  • It merges cleanly onto today's omamobile, alone and together with Input: ft3658 - review fixes: DMA-safe buffers, fsleep(), the Google source #133. CI's dtbs_check is clean. Each of the five commits builds on its own. The only checkpatch warning is the long bootargs line, as on arm64: dts: exynos: google: Add Pixel 7 (panther) #108, so leave it.
  • The Pixel 7 Pro, Pixel Fold and Pixel 7a device trees are byte-identical with and without the stack.
  • LDO25M at 1.8 V sits inside Google's m_ldo25_reg range (1.6-1.95 V), and panther doesn't narrow it. The rail is already at 1.8 V, so registering it writes nothing.
  • The touch node matches gs201-panther-touch.dtsi: CS gpp20-3, reset gpp23-2, INT gpa7-0, 10 MHz, LDO25M/LDO26M. The falling edge matches Google's driver (IRQF_TRIGGER_FALLING).

Nothing blocks. A few small things, in any order:

  1. Two stock touch details the node doesn't carry. Could you add them, or say in the message why not?
    • Google's panther file sets a pull-up on the interrupt pin (&ts_irq { samsung,pin-pud = <EXYNOS_PIN_PULL_UP>; }), while the common gs201 group, which we share with the Pixel 7 Pro, has no pull. Touch works, so the controller is evidently driving the line, but matching stock is cheap. Please put the override in gs201-panther.dts (&ts_irq { samsung,pin-pud = <GS101_PIN_PULL_UP>; };), not in gs201-pinctrl.dtsi, because the Pixel 7 Pro uses that group.
    • Stock's controller-data has cs-clock-delay = <1>, a 1 µs wait from chip select to the first clock, which upstream spells spi-cs-setup-delay-ns = <1000>.
  2. The LDO25M commit's "Tested" paragraph says the touch controller is the rail's only user and probes as before. At that commit pmic-1 is still disabled and there's no touch node yet, so that test is really of the last commit. "Tested with the touchscreen commit on top" would be accurate.
  3. Enabling pmic-1 also binds the S2MPG12 RTC, not only the two regulators, as on the Pixel 7 Pro's suspend setup. A line in the touchscreen commit's message saying so would help anyone bisecting later.
  4. The smoke report. Could you paste a tools/phone-smoke report from the T6 boot (this branch with Input: ft3658 - review fixes: DMA-safe buffers, fsleep(), the Google source #133), with the full uname -r? CONTRIBUTING asks for one with every change that runs on a phone, and that one boot covers arm64: dts: exynos: google: gs201-panther: enable the GPU #129, arm64: dts: exynos: google: gs201-panther: add the haptics #130, Input: ft3658 - review fixes: DMA-safe buffers, fsleep(), the Google source #133 and this PR. The NOTES in device/panther: the touchscreen (FT3658 kernel option, T6) and the touch notes omarchy-mobile#149 could carry the release string too.

Merge plan: derekross/omarchy-mobile#145 first (the GPU rule and the s2idle drop-in), then #133, then this PR with "Rebase and merge". That lands #129's and #130's commits along with your three, and I'll close #129 and #130 as merged through this one. derekross/omarchy-mobile#149, with the kernel option, follows. If you push again for points 1-3, keep the stack as it is and I'll take the new head.

@derekross
derekross merged commit fbbcac1 into derekross:omamobile Oct 30, 2026
1 check passed
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