Repository navigation
Pixel 8 Pro buttons: zuma's ALIVE pin controller and gpio-keys - #148
Conversation
zuma's (Tensor G3) ALIVE pin controller has a wake-up interrupt controller like Tensor G4's: each pin has its own GIC interrupt, and its EINT registers are at the same offsets. Document zuma's compatible with Tensor G4's as the fallback. Assisted-by: LLM Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
All of zuma's ALIVE banks are wake-up EINT banks, each pin on its own GIC interrupt, so the controller has no interrupt for GPIO EINTs. exynos_eint_gpio_init() only serves EINT_TYPE_GPIO banks, and without an interrupt on the controller node it logs "irq number not available" at every boot. Drop it from ALIVE, as gs101's ALIVE controller does. Fixes: 6d5bc3a ("pinctrl: samsung: Add Tensor G3 pin banks") Assisted-by: LLM Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
Add zuma's ALIVE pin controller at 0x154d0000, controller 0 of the driver's zuma data, with its ten banks (gpa0-gpa4, gpa6-gpa10) and its wake-up interrupt controller. As in stock's live device tree, each of the 50 pins has its own interrupt, SPI 10 to 59 in pin order, level high. Like PERIC0 and HSI2UFS, the controller takes the bus-clock stand-in, since zuma has no clock driver yet and ABL leaves the clock running. The power and volume keys are on this controller. Assisted-by: LLM Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
Describe the power and volume keys as gpio-keys, as stock does: power on gpa4-0 (pulled up), volume down on gpa4-1 and volume up on gpa6-2, all active low and in their wake-up interrupt function. Values come from a retail Pixel 8 Pro's live device tree. Assisted-by: LLM Signed-off-by: Brett-Michael Green <brett@bmgvisualtech.com>
Somnius
left a comment
There was a problem hiding this comment.
Tested this on a second Pixel 8 Pro (the Kioxia 256 GB UFS 4.0 one), booted into Brett's husky-test-20261031-c6 tag, which has these four commits at the bottom of #153.
Buttons: input: gpio-keys as /devices/platform/gpio-keys/input/input0 at boot. I streamed /dev/input/event0 to the desktop and decoded the raw events: KEY_VOLUMEUP, KEY_VOLUMEDOWN and KEY_POWER, press and release each, in the order I pressed them, twice. systemd-logind also logs "Power key pressed short", and pressing Power wakes the lock screen's idle blank. There are no "irq number not available" lines in dmesg on this boot.
Gate: I ran tools/kernel-pr-check with --build --merge on the stack (#153) in the project container. It merges cleanly onto 1c5ba1a, the 4 commits are signed off by their author, subject prefixes are fine, checkpatch --strict has no errors, the W=1 builds are fine and the device trees build with no new warnings. The one checkpatch warning is UNKNOWN_COMMIT_ID for the Fixes: sha in 20f2645: that commit exists in this repo ("Add Tensor G3 pin banks"), it's just not in my shallow clone. dt_binding_check and dtbs_check didn't run locally (no dtschema); the CI job for this PR passes.
Reading it: dropping .eint_gpio_init only touches zuma_pin_ctrl, so gs101, gs201 and zumapro aren't affected. I didn't check the pin tables against Google's tree, so the mapping is only proven by the keys working. Nothing else from me.
|
Thanks Brett, this is clean. I checked the ALIVE banks and interrupts and the key pins against Google's zuma tree and everything matches, and Lef's second phone confirms it. The other phones' DTBs don't change. Merged. |
This series gives the Pixel 8 Pro its power and volume keys. It adds zuma's ALIVE pin controller, which the keys sit on, and fixes the error line the driver logged once ALIVE was described.
dt-bindings: pinctrl: samsung: Add the Tensor G3 wake-up interrupts: zuma's ALIVE wake-up interrupt controller works like Tensor G4's. Each pin has its own GIC interrupt, and the EINT registers are at the same offsets. The binding listsgoogle,zuma-wakeup-eintwithgoogle,zumapro-wakeup-eintas the fallback.pinctrl: samsung: zuma: ALIVE has no GPIO interrupts: every ALIVE bank is a wake-up EINT bank, soexynos_eint_gpio_init()has nothing to serve there. Without an interrupt on the node, it loggedirq number not availableat every boot. This drops it from ALIVE, as gs101's ALIVE controller does. It carriesFixes: 6d5bc3a01525.arm64: dts: exynos: google: zuma: add the ALIVE pin controller: the controller at0x154d0000(controller 0 of the driver's zuma data), with its ten banks (gpa0–gpa4, gpa6–gpa10) and its wake-up interrupt controller. Each of the 50 pins has its own interrupt, SPI 10–59 in pin order, level high, as in stock's live device tree. Like PERIC0 and HSI2UFS, the controller uses the bus-clock stand-in.arm64: dts: exynos: google: zuma-husky: add the buttons:gpio-keysfor power (gpa4-0, pulled up), volume down (gpa4-1) and volume up (gpa6-2). All are active low, in their EINT function, and wake-up sources. The values come from a retail Pixel 8 Pro's live device tree.Tested on a Pixel 8 Pro. Both tests booted the U4 Omarchy image (omarchy-mobile#187) from
userdataon greenbrettmichael/linux-omarchy-mobile tags, with U4's kernel config:husky-test-20261030-b: the U4 kernel (husky-u4-20261030) with commits 1, 3 and 4.gpio-keys.irq number not availableonce, from ALIVE.husky-test-20261030-b2: this series' head, before the rebase onto the currentomamobile. The rebase was clean, andgit range-diffshows all four patches unchanged.irq number not availableline is gone, and the boot logs only the error lines it logged before the series.runningwith no failed units, and Hyprland comes up.Checks:
kernel-pr-check --buildon the rebased series:--strictshows 0 warnings;On husky, dtbs_check still shows only the two eUSB2 supplies.
🤖 Generated with Claude Code