Skip to content

arm64: dts: exynos: google: gs201-panther: describe the GNSS receiver - #268

Open
angelrenefm wants to merge 1 commit into
derekross:omamobilefrom
angelrenefm:pr/panther-gnss
Open

angelrenefm wants to merge 1 commit into
derekross:omamobilefrom
angelrenefm:pr/panther-gnss

Conversation

@angelrenefm

@angelrenefm angelrenefm commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

The Pixel 7's GNSS receiver, described as the Pixel 7 Pro's. One commit, device tree only.

The stock panther dtbo (board 0x30406) has the same spi5 node, bcm4775@0 child and four GPS pin groups as cheetah's (0x30306); Google's source has one file for both, gs201-common-gps.dtsi in kernel/devices/google/gs201 (android-gs-pantah-6.1-android16), and the rails come from the shared gs201-pmic.dtsi. So this adds cheetah's gnss@0 on USI5's SPI, its pins, and LDO9S-LDO11S in panther's sub-PMIC node, next to the fingerprint's LDO15S. Everything stays disabled; the overlay that turns it on is omarchy-mobile#861.

tools/kernel-pr-check --build onto today's omamobile:

ok - 1 commits signed off by their authors, none by an AI
ok - no Co-Authored-By trailers
ok - subjects have subsystem prefixes
ok - checkpatch --strict: no errors (0 warnings and checks to read)
ok - policy (tools/pr-check): identifiers, blobs
ok - C builds: no C files touched # SKIP
ok - 10 device trees build with W=1 and dtbs_check, no new warnings

Tested on a Pixel 7. Its kernel has run this node since 4 Nov (a local build). This branch's gs201-panther.dtb against the phone's live tree: gnss@0, the four pin groups and the three LDOs have the same properties, apart from phandles and status.

  • 4 Nov, outdoors, cold start: first fix 178 s in, then 303 fixes, up to 16 satellites.
  • Today, indoors, omamobile-gnss force for 20 s: bcm4775x-gnss binds on spi0.0, the chip powers up, and /run/omamobile-gnss/nmea-tty gives 1089 sentences (GPS, GLONASS, Galileo, BeiDou and QZSS GSA), with no fix indoors. omamobile-gnss off powered it down again.

Not tested: this exact tree on the phone (the running kernel is a local build with the same node), and the overlay through omamobile-boot-update.

I used AI help (Claude) to work on this.

@derekross

Copy link
Copy Markdown
Owner

Thank you, Ángel. I compared the new blocks with the Pixel 7 Pro's in gs201-cheetah.dts and they are the same line for line, apart from the three LDOs being disabled, which is right for a phone whose sub-PMIC is switched on by an overlay. They also agree with Google's gs201-common-gps.dtsi on the pantah branch: the same four pins and pulls, 26 MHz, mode 3. Our gate passes on today's omamobile, CI is green with dtbs_check, and in the built DTB USI5, its SPI, the sub-PMIC and the three rails are all disabled, so a Pixel 7 without the overlay boots exactly as before. Nothing else in the tree changes.

One thing stops it from working as proposed. nstandby and mcu_resp are on gph2, and gph2's pin controller (pinctrl_hsi2, pinctrl@14440000) is status = "disabled" in gs201.dtsi. The Pixel 7 Pro and the Pixel 7a (lynx) turn it on in their board files; panther's board file doesn't, and neither does omarchy-mobile#861's overlay, which enables only the sub-PMIC, the rails, USI5 and the SPI. So on omamobile plus this commit plus the overlay, the GNSS node has no GPIO chip and no pin controller for two of its lines and the driver never binds. It works on your phone because your tree also has #195, which enables pinctrl_hsi2 for the Wi-Fi pins. The build and dtbs_check can't see this.

Could you pick one of these?

  1. Add &{/soc@0/pinctrl@14440000} { status = "okay"; }; to the overlay in omarchy-mobile#861. This is the one I'd prefer: this commit stays "everything disabled", and the overlay is applied when the image is built, so the controller probes with the pin groups you put in the base here.
  2. Or put status = "okay"; in the new &pinctrl_hsi2 block here, as lynx does, and say in the message that it turns the HSI2 pin controller on at every Pixel 7 boot.

Either way, please mention in the comment above &usi5 that gph2's controller has to be on, and name the overlay as it is called in #861 (gnss.dtso). And please boot this tree once (today's omamobile with this commit and without #195) with the overlay, and post the bcm4775x-gnss bind line and a short omamobile-gnss force run. Your description lists that as not tested, and it is the boot that would have shown this.

On order: this and #181 don't conflict with each other. #195 will need a rebase after whichever of them lands first.

The Pixel 7's stock dtbo for board 0x30406 (the one ABL picks on this
phone) describes the GNSS receiver exactly as the Pixel 7 Pro's (board
0x30306): the spi5 node with its bcm4775@0 child at 26 MHz in mode 3,
host_req on gpa6-4, mcu_req on gpp4-3, mcu_resp on gph2-2 and nstandby
on gph2-3, with the same pin settings; the two fragments compare equal.
Google's source has one file for both, gs201-common-gps.dtsi in
kernel/devices/google/gs201 (android-gs-pantah-6.1-android16). Its
supplies are the S2MPG13's L9S_GNSS_CORE, L10S_GNSS_RF and
L11S_GNSS_AUX, with the same ranges as cheetah's (0.725-1.3 V,
1.6-1.95 V, 1.6-1.95 V), from the shared gs201-pmic.dtsi.

So add cheetah's gnss@0 node on USI5's SPI, its pin groups, and the
three LDOs in the sub-PMIC node. The LDOs are disabled, as LDO15S is,
because the fingerprint overlay enables the sub-PMIC on this phone;
USI5 and its SPI stay disabled as in gs201.dtsi. A bring-up overlay
(omarchy-mobile's device/panther/overlays/gnss.dtso) turns them on,
together with gph2's pin controller (pinctrl_hsi2, disabled in
gs201.dtsi), which mcu_resp and nstandby need.

Assisted-by: LLM
Signed-off-by: Ángel René <17342133+angelrenefm@users.noreply.github.com>
@angelrenefm

Copy link
Copy Markdown
Contributor Author

Thank you, Derek, good catch: my phone's tree has #195, which turns pinctrl_hsi2 on. I took your first option: #861's overlay now turns on pinctrl@14440000, and this commit stays "everything disabled". The comment above &usi5 says gph2's controller has to be on and names device/panther/overlays/gnss.dtso.

Checked here: kernel-pr-check passes (10 device trees with W=1 and dtbs_check, no new warnings); the overlay applies to this commit's gs201-panther.dtb and changes seven status lines (the six, plus the pin controller), and applies together with fingerprint-sensor.dtso.

Not done yet: the boot you asked for (today's omamobile with this commit, without #195, with the overlay). It needs the phone in fastboot; I'll post the bcm4775x-gnss bind line and an omamobile-gnss force run here when it's done. The commit message now names the overlay and the pin controller too.

I used AI help (Claude) to work on this.

@angelrenefm

Copy link
Copy Markdown
Contributor Author

The boot you asked for, on 2026-11-18: today's omamobile (26867fc) with this commit and without #195, kernel 7.3.0-rc6-omamobile-push1-g9468af10c2b7, and its gs201-panther.dtb with gnss.dtso (plus fingerprint-trusty, fingerprint-sensor and torch) applied by fdtoverlay.

How: a kexec from the phone's usual kernel with that DTB (kexec -c --dtb=…; the default kexec_file_load silently keeps the running DTB, which I caught on the first try by the FDT size). The only thing added to the DTB was the bootloader's four memory@ nodes, copied from the running one. Not a cold boot from fastboot.

After the boot, before anything else:

omamobile-gnss force:

bcm4775x-gnss spi0.0: BCM4775x GNSS on SPI, irq 114; supplies off until /dev/ttyBCM opens
bcm4775x-gnss spi0.0: power on (supplies enabled, standby)
bcm4775x-gnss spi0.0: ttyBCM opened by lhd[5257]

spi0.0 is bound to bcm4775x-gnss (under 109400c0.usi/10940000.spi), the three rails go to enabled, and /run/omamobile-gnss/nmea-tty gave 996 lines in 20 s ($PSSGR, $PGLOR, $GPGSA, $QZGSA, $BDGSA, $GNGSA…), no fix indoors. Then unforce and off: receiver stopped, chip unpowered.

Not checked: the fingerprint sensor on this boot. Trusty doesn't come back after a kexec (trusty_init_msg_buf: SMC_FCZ_FFA_RXTX_MAP failed, its FF-A buffers are still mapped from the previous kernel), so the fingerprint stack can't start; that's the kexec, not this tree. A cold boot would settle it. After a normal reboot into the phone's usual kernel, fingerprint, Wi-Fi and the rest came back as before.

I used AI help (Claude) to work on this.

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