Skip to content

Device port: Pixel 6 (oriole, gs101 / Tensor G1) – tracking issue, contributors welcome! #2

Open

Description

@bilacc

Let's bring Omarchy to the Pixel 6! 🚀

Omarchy Mobile (formerly Omandroid) already shows what a Tensor phone can do with a mainline kernel and native Omarchy on the Pixel 7 Pro. I'd love to see the Pixel 6 (oriole, Google Tensor / gs101) join it, and I think it's one of the most promising next targets. Its siblings, the Pixel 6 Pro (raven) and the Pixel 6a (bluejay), are very welcome here too, since they share the SoC.

This is the tracking issue for the port, opened as docs/DEVICES.md asks, so that efforts don't overlap.

About me: I own a Pixel 6 and I'm kicking this off, but I can't work on it full time. I'll test, coordinate and contribute where I can. The real goal is a community effort: if you have a Tensor phone and some curiosity (XDA folks, this means you 👋), there's a piece of this for you. Pick a stage, say so in a comment, and let's build it together.

Why gs101 is a reasonable target

  • gs101.dtsi, gs101-oriole.dts and gs101-raven.dts have been in mainline Linux since about 6.8, with clocks, pinctrl, UFS and USB upstream.
  • CONTRIBUTING.md §5.2 notes that much of the gs201 (cheetah) work was ported from gs101, so a lot of it should port back.
  • postmarketOS keeps a close-to-mainline gs101 kernel (linux-postmarketos-gs101). What it enables on oriole is summarised, with links, in the comments.

What should carry over and what must be redone

Likely reusable:

  • the userspace packages (omamobile-shell, hyprland, mesa, omamobile-keyring);
  • the Omarchy patches;
  • the docker build environment;
  • phone-ssh, phone-install-package and the firmware-extraction tools;
  • the multi-device build: device/<codename>/device.conf plus DEVICE=<codename> bin/omamobile …;
  • the slot B flashing model.

Device-specific, needs new work:

  • a device/oriole/ directory with plans, notes and overlays;
  • packages/omamobile-device-oriole/;
  • device/oriole/device.conf, the board DTS and a dtbo carrying oriole's board ID.

Still cheetah-specific (the audit table in docs/RELEASES.md): mainly the initramfs init, including the DECON TRIG_CON write at 0x1c240030, which is cheetah-only and must never run on another phone.

Different hardware: the panel (S6E3FC3 on oriole, I believe; to verify), touch, PMIC rails, GPIOs, battery model and firmware names. gs101's interrupt numbers differ from gs201's, so never copy IRQ numbers.

Bring-up checklist

This follows CONTRIBUTING §5.4. The linked plans are cheetah's, to use as templates. Comment to claim a stage.

  • 0. Stock capture: mostly done. Board entry (0x20304/0x10000), partitions, boot image format and build confirmed from the factory image and a retail phone, merged in device-oriole: descriptor and notes from the factory image; images for devices without init_boot #5. Still open: soc_id, the live DT and /proc/iomem, which need an unlocked or rooted phone; run bin/omamobile capture and post facts, not files.
  • 0b. Per-device tooling: done upstream. Device values now come from device/<codename>/device.conf, and DEVICE=<codename> bin/omamobile … selects the device.
  • 1. First light: slot B boot, simple-framebuffer, USB NCM, SSH, watchdog handover, panic=10 (BRINGUP-M1)
  • 2. Clocks, pinctrl, UFS with the read-only guard, then the Arch root on userdata
  • 3. Touch and buttons
  • 4. GPU: Mali-G78 (Valhall Job Manager) on Panfrost plus Mesa; pmOS's gs101 fork has the GPU node and supplies (GPU-PLAN as a template)
  • 5. Display: simpledrm, then DSI, then the KMS driver and refresh rate (DISPLAY-PLAN)
  • 6. Fuel gauge and battery
  • 7. Wi-Fi (WIFI-BATTERY-PLAN)
  • 8. Bluetooth
  • 9. Haptics and torch
  • 10. ACPM
  • 11. CPU idle and UFS power
  • 12. USB-C and charging
  • 13. Suspend
  • 14. AoC, sensors and audio
  • 15. GNSS, NFC, modem and cameras
  • 16. Fingerprint

Stages 1-5 give a usable phone on a desk. After stage 2, most stages are independent.

Before you touch a phone

Please read CONTRIBUTING §3 (safety model) and AGENTS.md. In short:

  • Only slot B and userdata are ours. Slot A's stock Android is the way back.
  • Never flash bootloader, radio, devinfo, persist, efs*, modem* or vbmeta.
  • Back up efs, persist and modem_userdata off-device before any radio or PMIC work.
  • Don't post captures, serials, IMEIs or MACs in this thread.

How to help

Every contribution counts, big or small:

  • Have a Pixel 6, 6 Pro or 6a? Comment with your model and how much time you have. A safe first contribution is a stock capture, keeping only the facts you share. I'll do one on my oriole as well.
  • No phone? Kernel and DT work against upstream gs101-oriole.dts can start without one. See the postmarketOS summary in the comments for what already exists.
  • Discussion: questions and chat are welcome in the XDA thread. Claim stages here, so they're tracked in one place.

Plans go in device/oriole/<SUBSYSTEM>-PLAN.md (template in CONTRIBUTING §5.5). Kernel work: fork linux-omarchy-mobile, branch off omamobile with one subsystem per branch (e.g. omamobile-gs101-oriole-dts), and open PRs against omamobile. To match the current packages and overlays, build omamobile-int (what the Pixel 7 Pro runs), but base your own commits on omamobile. See CONTRIBUTING §6.1.

Thanks for reading, and I'm looking forward to seeing Omarchy Mobile boot on a Pixel 6! 🎉

Activity

  1. bilacc commented on Oct 13, 2026

    @bilacc
    ContributorAuthor

    Hi @derekross! 👋 A quick heads-up so the extra traffic doesn't come as a surprise: I've started an XDA thread that points contributors here, to this issue, to claim stages, following CONTRIBUTING.md and the safety model:
    https://xdaforums.com/t/dev-wip-contributors-wanted-native-omarchy-arch-linux-arm-hyprland-on-pixel-6-6-pro-6a-mainline-kernel-no-android.4804933/

    If you're happy with the effort, a Pixel 6 | oriole | gs101 | WIP (see #2) row in docs/DEVICES.md would help people find it. And of course, if you'd like anything run differently (scope, where discussion happens, branch layout for a gs101 port), just say so and I'll update the issue and the thread.

    Status: no stock capture from my oriole yet. I'll post the facts from it (not the files) here when I can, so stage 0 is still open if someone else gets there first.

    Thanks for Omandroid, it's a great base to build on!

  2. derekross commented on Oct 13, 2026

    @derekross
    Owner

    Thanks for organizing this and for the XDA thread. The issue reads the safety model exactly right, including the DECON write that must never run on another phone.

    • DEVICES.md: done. The Pixel 6 is listed as a WIP community port pointing here, and the 6 Pro and 6a point here too (Wanted).
    • Second device: the Pixel 6 is our preferred second device. gs101 already has mainline clocks, pinctrl, UFS and USB, and it will prove the multi-device build and release work.
    • Stage 0b is done: as of today, device values come from device/<codename>/device.conf, and DEVICE=<codename> bin/omandroid … selects the device. Cheetah's images are byte-identical to before. See device/cheetah/device.conf and docs/RELEASES.md, whose audit table lists what is still cheetah-specific (mainly in the initramfs init). So stage 0, the stock capture, is the real first step. A first PR would be device/oriole/device.conf plus device/oriole/NOTES.md, with facts only.
    • Kernel work: outside contributors can't push branches to linux-omandroid, so fork it, branch off omandroid (one subsystem per branch, e.g. omandroid-gs101-oriole-dts), and open PRs against omandroid. CONTRIBUTING §6.1 now says so. To match the current packages and overlays, build the new omandroid-int branch (what the Pixel 7 Pro runs today, as one snapshot commit); base your own commits on omandroid, the curated upstream-ready branch.
    • Discussion: XDA for chat and this issue for claiming stages works for me.

    On postmarketOS's gs101 kernel: a summary of what it supports on oriole would be a great starting point, as you suggest. It's a citable source, so please link it.

    Looking forward to Omarchy on a Pixel 6.

  3. derekross commented on Oct 13, 2026

    @derekross
    Owner

    Heads-up: the project is now called Omarchy Mobile.

    Omandroid was never Android, and the other Omarchy Mobile effort is merging into this one, so the project now carries that name. Everything moved today:

    • Repos: derekross/omarchy-mobile and derekross/linux-omarchy-mobile. Old URLs redirect, but please update your remotes.
    • Kernel branches: omamobile (curated, upstream-ready) and omamobile-int (what the Pixel 7 Pro runs, as a snapshot). The old omandroid and omandroid-int branches stay a short while for anyone mid-clone, then go away.
    • Inside the code, the short name is omamobile, like Omarchy's omacalc and omawrite: bin/omamobile, omamobile-* packages and commands, the [omamobile] package repository and omamobile-keyring, and device/<codename>/device.conf as before. The release suffix is -omamobile.
    • CONTRIBUTING, DEVICES and the release design are updated. Nothing else about the process changed: fork linux-omarchy-mobile and open PRs against omamobile; device notes and device.conf go to this repo.

    Thanks for bearing with the churn. Better now than after a release.

  4. bilacc commented on Oct 14, 2026

    @bilacc
    ContributorAuthor

    postmarketOS on gs101: what its kernel and device packages enable for oriole

    I read pmOS's packages and the kernel source they build. Nothing here has been run on a phone. It shows what is enabled, not what is confirmed working.

    Sources

    What the fork adds to upstream v7.0's device tree

    • The S2MPG11 sub-PMIC. Upstream already has the S2MPG10 main PMIC.
    • A GPU node, google,gs101-mali / arm,mali-valhall-jm, with its supplies. The config uses Panfrost (DRM_PANFROST=y, Panthor off).
    • Touch on oriole: an ST FTS controller on SPI11 (compatible "st,spi"), driven by an out-of-tree driver (TOUCHSCREEN_ST_FTS_V521_SPI). This would need porting or rewriting before it could go upstream.

    What upstream already had

    • simple-framebuffer at 1080x2400;
    • gpio-keys;
    • UFS;
    • USB DRD (otg, role switch);
    • the MAX77759 with its TCPCI Type-C, GPIO and nvmem;
    • the cluster 0 watchdog.

    Boot and userspace details

    • deviceinfo_drm="false" (2026-05-18 commit: "We dont have drm"), so there's no display or KMS driver, only simplefb.
    • The kernel command line is clk_ignore_unused regulator_ignore_unused.
    • The initramfs modules are ufs-exynos, dm-mod and simplefb.
    • The boot image uses header v4. pmOS also flashes vbmeta, which Omarchy Mobile's model never touches.
    • soc-google-gs101 has an initramfs hook that writes 0x3061 to DECON at 0x1c300030 "to enable framebuffer refreshing". This is the gs101 counterpart of cheetah's TRIG_CON write at 0x1c240030. It's a useful lead for stage 1, but it should be verified with a read first, and cheetah's address must never be used.

    Not enabled

    • Wi-Fi: PCI_EXYNOS is off.
    • Audio: SND_SOC_SAMSUNG is off, and there's no AoC.
    • Panel driver and brightness: none.
    • Fuel gauge: BATTERY_MAX1720X=y is set, but the device tree has no fuel-gauge node.
    • Bluetooth, haptics, sensors, modem and cameras.

    What this means for our checklist

    • Stages 1-2 (boot, UFS, USB) have an upstream base, and pmOS adds some boot-time experience on top.
    • Stage 3 (touch, buttons): the buttons are upstream. Touch has a working out-of-tree reference in pmOS's fork.
    • Stage 4 (GPU): the Mali-G78 is a Valhall Job Manager GPU, so it uses Panfrost, not Panthor as I wrote in the checklist. The pmOS fork has the DT node and supplies.
    • Stage 5 and later (display KMS, battery, Wi-Fi, Bluetooth, audio and the rest) are new work. That is where Omarchy Mobile's gs201 work is the main reference.

    The pmOS wiki page for oriole redirects to another domain, so I haven't cited it.

  5. bilacc commented on Oct 14, 2026

    @bilacc
    ContributorAuthor

    Heads-up: the project is now called Omarchy Mobile.

    Omandroid was never Android, and the other Omarchy Mobile effort is merging into this one, so the project now carries that name. Everything moved today:

    • Repos: derekross/omarchy-mobile and derekross/linux-omarchy-mobile. Old URLs redirect, but please update your remotes.
    • Kernel branches: omamobile (curated, upstream-ready) and omamobile-int (what the Pixel 7 Pro runs, as a snapshot). The old omandroid and omandroid-int branches stay a short while for anyone mid-clone, then go away.
    • Inside the code, the short name is omamobile, like Omarchy's omacalc and omawrite: bin/omamobile, omamobile-* packages and commands, the [omamobile] package repository and omamobile-keyring, and device/<codename>/device.conf as before. The release suffix is -omamobile.
    • CONTRIBUTING, DEVICES and the release design are updated. Nothing else about the process changed: fork linux-omarchy-mobile and open PRs against omamobile; device notes and device.conf go to this repo.

    Thanks for bearing with the churn. Better now than after a release.

    Thanks Derek, and congrats on the merge. Omarchy Mobile is a better name. I've updated this issue for the rename, the device.conf work (0b ticked; stage 0 is now the first step) and the fork-and-PR flow against omamobile, and posted the postmarketOS gs101 summary below, with links. One correction from that research: the Mali-G78 is Valhall Job Manager, so stage 4 is Panfrost, not Panthor.

  6. bilacc commented on Oct 14, 2026

    @bilacc
    ContributorAuthor

    @derekross Stage 0, part 1: no phone needed. Everything below comes from Google's public factory image.

    Stage 0, part 1: facts from Google's public oriole factory image (no phone needed)

    Source: Google's factory image oriole-cp3a.261005.002.a1. I read android-info.txt, fastboot-info.txt, vendor_boot.img and dtbo.img and only the header of boot.img, by fetching just the byte ranges needed (HTTP range reads). Below are facts only, no files. Nothing here comes from a phone, so nothing here identifies one.

    Partitions and images

    • The image ships no init_boot.img and no vendor_kernel_boot.img. fastboot-info.txt flashes boot, dtbo, pvmfw, vendor_boot and the vbmetas, then the dynamic partitions. That confirms the [I] in RELEASES.md: oriole's descriptor would be BOOT_IMAGES="boot vendor_boot dtbo" and UFS_RW="boot_b vendor_boot_b dtbo_b userdata". A real fastboot getvar all still has to confirm the partition table.
    • boot.img and vendor_boot.img both use header v4. vendor_boot has a page size of 0x800, two vendor ramdisks and a bootconfig section.
    • android-info.txt: board=slider|whitefin|oriole|raven. One image serves the Pixel 6, the Pixel 6 Pro and two gs101 development boards.

    Base DTBs (in vendor_boot; ABL picks one by soc_id)

    • There are two base DTBs: soc_id = <0x9845000 0x00> and <0x9845001 0x00>. Both have the root compatible google,gs101. These are the gs101 counterparts of cheetah's 0x9855000/0x9855001. Which value is A0 and which B0, and which one a given retail phone has, still needs the phone.
    • The root has #address-cells = <2> and #size-cells = <1>, as on gs201. So the 12-byte reg gotcha for reserved memory from AGENTS.md applies here too.
    • An /ect node and /chosen/bootargs are present. The stock bootargs use earlycon=exynos4210,mmio32,0x10A00000, the same UART address as cheetah's VENDOR_CMDLINE.

    Board dtbo (dtbo.img, 18 entries, page size 4096)

    board_id board_rev values panel strings touch reading
    0x20302, 0x20303, 0x20304 0x10000, 0x10100 (0x20303 also 0x10002) S6E3FC3 ST FTS oriole
    0x20202, 0x20203, 0x20204 0x10000, 0x10100 (0x20203 also 0x10001) S6E3HC3 Samsung sec_ts raven
    0x201, 0x20001 several S6E3HC2 both dev boards [I]
    • The panel and touch strings match pmOS's oriole DT (ST FTS on SPI11) and settle the panel question: oriole's panel is the S6E3FC3.
    • "goodix" appears in every entry. That's the under-display fingerprint sensor.
    • Which board_id/board_rev a retail oriole boots with is still [I]: probably the newest one, 0x20304/0x10000, since cheetah's is 0x30306/0x10000. getprop ro.boot.dtbo_idx on the phone answers it.

    Proposed device/oriole/device.conf values, so far

    CODENAME=oriole
    NAME="Pixel 6"
    SOC=gs101
    COMPATIBLE="google,gs101-oriole"
    PACKAGE=omamobile-device-oriole
    DTB=exynos/google/gs101-oriole.dtb
    SOC_IDS="<label>:0x9845001 <label>:0x9845000"   # labels to confirm
    DTBO_BOARD_ID=0x20304                            # [I] confirm with ro.boot.dtbo_idx
    DTBO_BOARD_REV=0x10000                           # [I]
    VENDOR_PAGESIZE=0x800
    BOOT_IMAGES="boot vendor_boot dtbo"
    UFS_RW="boot_b vendor_boot_b dtbo_b userdata"
    PANEL_SIZE=1080x2400
    

    Still needs a phone

    • The live DT (/sys/firmware/fdt, needs root);
    • ro.boot.dtbo_idx;
    • which soc_id the phone has;
    • fastboot getvar all (partition sizes, slot state);
    • /proc/iomem, to check reserved memory.
  7. derekross commented on Oct 14, 2026

    @derekross
    Owner

    This is excellent work, and exactly the kind of sourced, facts-only groundwork that makes a port go smoothly. Thanks for the Panfrost correction too.

    A few notes:

    • device.conf: your values look right. Please open the first PR with device/oriole/device.conf and device/oriole/NOTES.md, with the [I] items marked as unconfirmed, as you have them. BOOT_IMAGES="boot vendor_boot dtbo" is the first device without init_boot/vendor_kernel_boot, which is what the descriptor was built for. If bin/omamobile images assumes those two images anywhere, the PR is the place to find out.
    • DECON at 0x1c300030: agreed. Read it first, on a booted phone, and only then write it. Same deal as the TRIG_CON note on Pixel Fold (felix, gs201) port: first light booted, contributors welcome #3: let's make that write a per-device device.conf value rather than code in init.
    • Panfrost: our Mesa package builds Mesa's panfrost driver, which serves both the Panfrost kernel driver (the G78's) and Panthor (our G710's), so userspace is ready. The kernel side (the GPU node and supplies from the gs101-mainline fork) would come into omamobile as its own branch.
    • Still needing a phone: your list is right: the live DT, ro.boot.dtbo_idx, soc_id, getvar all, /proc/iomem. bin/omamobile capture collects all of these into a private folder; please post facts, not the files.
  8. bilacc commented on Oct 16, 2026

    @bilacc
    ContributorAuthor

    Stage 0, part 2: facts from a retail Pixel 6

    I queried my own stock Pixel 6 (bootloader still locked) with read-only commands: adb shell getprop, /dev/block/by-name and fastboot getvar all. Nothing was written or flashed, and no serial numbers are included below.

    Question Answer Source
    Board dtbo entry ro.boot.dtbo_idx = 0, which is board_id 0x20304, board_rev 0x10000 in the stock dtbo.img getprop + factory dtbo
    Hardware revision MP1.0 ro.boot.hardware.revision, getvar hw-revision
    init_boot / vendor_kernel_boot Neither exists. The A/B boot partitions are boot, vendor_boot, dtbo and pvmfw getvar all, /dev/block/by-name
    Build CP3A.261005.002.A1, patch 2026-10-05: the same build as the factory image I read earlier getprop
    Bootloader slider-17.0-15681480, secure-boot:PRODUCTION, current slot a, locked getvar all
    soc_id still unknown: /sys/firmware/devicetree/base/soc_id needs root —

    This confirms the values in PR #5's device/oriole/device.conf: DTBO_BOARD_ID=0x20304, DTBO_BOARD_REV=0x10000, BOOT_IMAGES="boot vendor_boot dtbo" and OS_PATCH_LEVEL=2026-10. I pushed a commit to the PR that marks them as checked [V] and records the partition list and bootloader state in NOTES.md.

    Stage 0 still has two open items: the soc_id (A0 or B0), and the live device tree and /proc/iomem. All of those need root, which in turn means unlocking the bootloader, and unlocking wipes the phone. If someone with an already-unlocked or rooted Pixel 6 can run bin/omamobile capture, please post the facts here (not the files).

  9. derekross commented on Oct 16, 2026

    @derekross
    Owner

    Thanks, @bilacc, that settles the board entry and the partition list from a real phone, and the read-only approach is exactly right. #5 is merged with these values. For the soc_id and the live DT, no rush: anyone with an already-unlocked Pixel 6 who can run bin/omamobile capture, please post the facts here (not the files). Until then, images can ship both silicon revisions the way cheetah does.

  10. 4 remaining items

  11. PeterXMR commented on Nov 2, 2026

    @PeterXMR
    Contributor

    Pixel 6a (bluejay): the first-light board DTS is merged, and it needs someone with a spare 6a to boot it.

    Status: the descriptor and notes landed in #265, and gs101-bluejay.dts is in omamobile since linux-omarchy-mobile#179 (merged
    today on the oriole precedent, built but never booted). It is oriole's file with bluejay's model, compatible, soc_id and board entry, all from Google's GPL bluejay tree
    (notes).

    I own a 6a but keep it locked, and unlocking wipes it, so I can't run the boot myself. If you have an unlocked 6a you can spare, the ask is a slot B boot only, nothing else is touched (kernel
    clone on omamobile):

    DEVICE=bluejay bin/omamobile kernel
    DEVICE=bluejay bin/omamobile initramfs
    DEVICE=bluejay bin/omamobile images
    DEVICE=bluejay bin/omamobile flash
    

    Read CONTRIBUTING §3 first. The screen stays on the boot logo (no display register is
    written on bluejay yet), so the sign of life is the USB network device at 172.16.42.1 and SSH. If ABL falls back to slot A, that is the expected failure and stock comes back on its own.

    Please post here:

    • uname -r;
    • the tools/phone-smoke report;
    • ro.boot.dtbo_idx from stock, if you still have it, to confirm the MP 1.0 entry 0x20506/0x10000, and which soc_id the phone has.

    No serials, MACs or captures, facts only. I'll turn the result into a dated entry in device/bluejay/NOTES.md and open the device package PR on top of it.

  12. derekross commented on Nov 2, 2026

    @derekross
    Owner

    Thank you, @PeterXMR, for pushing the 6a along: the descriptor and the board file are carefully
    sourced, and a tester with a spare 6a is exactly what it needs next.

    Before anyone flashes, though, one thing from bluejay's own notes needs settling first: the AVB
    question. device/bluejay/NOTES.md lists AVB_FOOTERS as open and says "no flash until this is
    settled", because we don't know whether a 6a's stock vbmeta_b chains boot and vendor_boot.
    If it does, footer-less images won't boot, and on the 7 Pro that cost us a few rounds of
    fallbacks. The good news is that it needs no phone: the current bluejay factory zip's
    vbmeta.img (avbtool info_image, the chain partition descriptors) and the boot and
    vendor_boot footers answer it, and the dtbo.img table in the same zip tells us which entry is
    0x20506. Could you gather those facts (facts only, no files) and add them to the notes in a PR?
    Then the call for a tester can include the right image settings.

    When it goes out, could you post it as its own issue ("Device port: Pixel 6a (bluejay)") and
    link it here? #2 is the Pixel 6's thread. And please put the read-only stock capture from
    CONTRIBUTING §5.1 first in the steps, before the flash, as the notes describe.

  13. PeterXMR commented on Nov 2, 2026

    @PeterXMR
    Contributor

    The Pixel 6a now has its own thread, #348, with the tester call reordered as asked: the read-only stock capture (CONTRIBUTING §5.1) first, and no flash until the AVB facts from the factory zip land in the notes. This thread stays the Pixel 6's.

  14. Svffer commented on Nov 8, 2026

    @Svffer

    Hi, everyone! I have a retail Pixel 6 with an unlocked bootloader (oriole, MP1.0, gs101 B0), on which I'm running Omarchy Mobile from slot B. So I can answer the Stage 0 questions open since October and test oriole changes. My English isn't very good, so I'm using a translator.

    Stage 0 facts (from my phone, facts only):

    • soc_id = 0x9845001 (B0); dtbo entry 0 = board_id 0x20304 / rev 0x10000;
    • live DT and /proc/iomem captured: the stock reserved-memory regions match iomem;
    • the stock vbmeta chains boot, so AVB_FOOTERS="boot:0x4000000" is required, and it works.

    What runs on it today (all local, nothing pushed yet):

    • boot from slot B with the root on userdata; USB networking and SSH;
    • display under gs-decon at 60 and 90 Hz, brightnld init);
    • the stock vbmeta chains boot, so AVB_FOOTERS="boot:0x4000000" is required, and it works.

    **What runs on it today (all local, nothing pushe

    • boot from slot B with the root on userdata; U
    • display under gs-decon at 60 and 90 Hz, brightnld init);
    • touchscreen (ST FTS "FTM5"), buttons;
    • GPU (panfrost);
    • battery gauge, charging, thermal zones;
    • Wi-Fi (BCM4389 on PCIe GEN4A_1) and Bluetooth, including A2DP audio;
    • s2idle suspend, woken by the power button (about 275 mA asleep; real standby would need SYS_SLEEP).

    What it's built on. Almost everything is other people's work, adapted for oriole:

    • upstream gs101-oriole.dts and bilacc's stage 0 work (device-oriole: descriptor and notes from the factory image; images for devices without init_boot #5);
    • robotosaith's Pixel 6 Pro (raven) work, my main reference: the gs101 display series (linux-omarchy-mobile#116), the G78 under
      panfrost, the battery node, the BCM4389 Bluetooth, gs101's UFS suspend, and raven's plans forWi-Fi, Bluetooth and suspend, which I followed step by step;
    • Gr-linkk's oriole first-light package (device-oriole: add minimal first-light policy package #222) and the oriole Bluetooth node (linux-omarchy-mobile#159);
    • Derek's Pixel 7 Pro (cheetah) drivers in `omamolds on too: gs-decon, gs-dsim-cmd, the PCIe
      controller and PHY, brcmfmac's BCM4389 support, ale fix;
    • Alejandro Garcia's S6E3FC3 driver for the Pixel 7's P10 panel, the same display chip as oriole's;
    • Google's GPL kernel sources for the oriole-specnds and timings, gs101's PCIe PHY tuning, and ST's
      FTS touchscreen driver.

    My own changes are mostly small, and I made them with the help of an AI coding assistant: gs101 match data for the PCIe drivers, an oriole variant of the S6E3FC3 driver with a 60/90 Hz switch, oriole's DSIM timers, oriole DTS blocks, and device notes. Some
    of it is bring-up only (oriole's mode table in gseen driver) and isn't ready for merging. I tested every step on the phone and will sign off whatever I submit.

    What's the best way to proceed? I'm new to prnt to get in the way of anyone's work.Specifically:

    • @Gr-linkk, my oriole package changes sit on topnd them to your branch, or wait until it's merged?
    • Should my display changes wait for tools/phone-mirror: the panel size from the phone #116, or can it?
    • One open question on my side: slot B is never marked successful on oriole, so every boot uses up one of ABL's boot attempts. How should slot B's success mark work on oriole?

    I'm happy to post tools/phone-smoke reports or e who's interested.

  15. Gr-linkk commented on Nov 8, 2026

    @Gr-linkk
    Contributor

    Hey @Svffer, thanks for reaching out and crediting our work! It would be great to coordinate.

    Our published first-light branch is here:
    https://github.com/Gr-linkk/omarchy-mobile/tree/oriole-first-light-package

    It’s submitted as #222. That branch doesn’t contain our complete local Pixel 6 setup—we’re still extracting and qualifying further contributions.

    Locally, we’ve tested the native desktop, GPU acceleration, Wi-Fi, Bluetooth data/keyboard input, battery telemetry, and guarded Linux recovery. We also have a tested slot-B boot-success approach that may help with your question. Battery charge accuracy remains under investigation, and cellular services aren’t working yet.

    Could you publish your current source branches or patches and share the kernel commit you tested? In particular, I’d like to compare your additions to our first-light package and your display, charging and suspend work.

    You don’t need to wait for #222 to merge to share your branch. Once the changes are visible, we can compare dependencies and agree on the best PR arrangement. Please keep contributions on your own branch for now so we can review and coordinate them.

    I’m still organizing our local work and its test evidence, but I can gather the relevant source for boot success and recovery next. Glad to have another Pixel 6 contributor involved!

  16. Svffer commented on Nov 8, 2026

    @Svffer

    Thanks, @Gr-linkk! My branches are up:

    The kernel I tested is 50a7df7b9 on oriole-touch (7.3.0-rc6-omamobile-push1-g50a7df7b911e), with omamobile-device-oriole 0.2.14 and omamobile-shell 0.2.102 on the phone. My package started from your #222 at 0.2.7.

    Where to look:

    • display: device/oriole/DISPLAY-PLAN.md; kernel commits for the S6E3FC3 oriole variant, the DSIM timers and gs-decon (the gs-decon mode table and pitch commits are marked NOT FOR MERGE);
    • charging: the battery node 3913dd770 and the charger options in the package;
    • suspend: device/oriole/SUSPEND-PLAN.md, the sleep policy and hook in the package (0.2.12-0.2.14);
    • Wi-Fi and Bluetooth: device/oriole/WIFI-PLAN.md, device/oriole/BLUETOOTH-PLAN.md.

    I'd be glad to see your slot-B boot-success approach. I'll keep everything on my branches until we agree on the PR arrangement.

  17. Gr-linkk commented on Nov 8, 2026

    @Gr-linkk
    Contributor

    Thanks @Svffer — I suggest:

    1. Keep #222 minimal. Your independent display contributions can proceed separately, preserving the NOT FOR MERGE boundaries.

    2. Our historical Oriole boot-success evidence addendum is posted: DEVI 3.15, full-record verification showing only B’s successful bit changed, A unchanged, and success retained across a repeat boot. The safety limitation matters: the procedure briefly made the whole boot LUN writable. It is evidence, not a procedure to copy.

    3. Finish a short Oriole evidence addendum with Derek, coordinated with #252 and #483. The guard change and tool review come first; Oriole is not covered by the current tool. If you already have sanitized layout/version, slot-status or retry-transition evidence recorded, that would help fill the gaps. No new metadata commands or tests are requested; full dumps and private backups stay private.

  18. Svffer commented on Nov 8, 2026

    @Svffer

    Thanks @Gr-linkk, agreed on all three points. I've added what I had recorded to #252.

  19. Ruraam commented on Nov 18, 2026

    @Ruraam

    Hi everyone,

    I have a retail Pixel 6 (oriole, MP1.0, gs101) with an unlocked bootloader and would like to run Omarchy Mobile natively from slot B, keeping slot A as a fallback.

    Read-only diagnostics (fastboot getvar all, no writes):

    • product: oriole
    • unlocked: yes
    • current-slot: a
    • version-bootloader: slider-16.4-14097577
    • version-baseband: g5123b-145971-251030-B-14356419
    • secure-boot: PRODUCTION
    • 128 GB storage, 8 GB RAM, both slots marked successful and bootable
    • Currently running LineageOS on slot A

    I have read device/oriole/NOTES.md and docs/DEVICES.md. I understand that the public omamobile-device-oriole package is not a supported install on its own, that the kernel and packages used for the successful native boot were never published as a complete tree, and that the vbmeta_b / devinfo changes were made without project approval. I do not want to reproduce any of that blindly.

    Four questions before I touch the phone:

    1. Is there a branch, bundle, or set of patches I can actually build for oriole today? If not, what is the realistic path to a reproducible build?
    2. What is the supported way to boot a custom boot_b without writing vbmeta_b or devinfo? My constraint is to leave vbmeta untouched.
    3. My bootloader is slider-16.4, older than the slider-17.0-15681480 referenced in the notes for factory build CP3A.261005.002.A1. Which factory build / firmware should I use, and is there an anti-rollback risk here?
    4. Which partitions exactly need to be modified to boot Linux on oriole? My rule is to leave vbmeta, persist, modem, radio and bootloader untouched, and to keep slot A as the recovery reference.

    I am happy to provide read-only captures from my device (getvar, partition facts, boot logs) and to test changes once the maintainers consider the port safe enough to try. I will not post serial numbers, IMEI or any per-phone calibration data.

    Thanks for the work on this port.

  20. Ruraam commented on Nov 18, 2026

    @Ruraam

    Additional technical findings regarding the Oriole port:

    After reviewing the A/B boot flow and the project's AVB implementation, we identified a potentially safer approach to booting Linux without flashing modified vbmeta partitions.

    The project already supports AVB_FOOTERS in device.conf, using AVB hash footers signed with a key accepted by an unlocked bootloader.

    We would like to clarify the following points for Pixel 6 (oriole):

    1. Can the existing AVB_FOOTERS mechanism be used on Oriole while keeping both stock vbmeta partitions untouched?
    2. Does the current Oriole bootloader (slider-16.4-14097577) support this approach?
    3. Can the Linux slot be marked successful through the standard boot control mechanism, without manually modifying devinfo?
    4. Is there a supported way to complete the first boot and slot validation without modifying protected partitions?

    Our goal is to develop an independent Oriole build while preserving the original Android slot as a recovery path.

    We will not modify vbmeta, devinfo, persist, modem or bootloader partitions without an explicitly documented and validated procedure.

    Any guidance on the correct Oriole implementation would be greatly appreciated.

  21. Gr-linkk commented on Nov 19, 2026

    @Gr-linkk
    Contributor

    Thanks for digging into this, @Ruraam. We’ve reached the Omarchy desktop on our Pixel 6 with a private integration build. Getting someone else to build it from public sources is the next milestone I’d like to work on with you.

    If you’re up for it, could you try building the public Oriole kernel and generating the images on your computer? Share the two repository commit IDs and the first error you hit, or the names of the images it produced if it succeeds. I can try the same build on my side so we can turn the first reproducible blocker into a small PR.

    Let’s keep this step offline. We haven’t verified that AVB footers work on your bootloader with both vbmeta partitions untouched, and slot-success marking is still under review. A successful image build gives us something concrete to review before anyone flashes it.

  22. Ruraam commented on Nov 19, 2026

    @Ruraam

    Hi Gr-linkk,

    Following your request, I reproduced the public Oriole build fully offline on a separate Debian machine — no phone connected, no flashing, no AVB or slot-success changes.

    Repos / commits used (both verified against git ls-remote):

    • derekross/omarchy-mobile — main @ 17f9a3cd979c280431b2253e5af377e1b2377d55
    • derekross/linux-omarchy-mobile — omamobile @ 26867fc4e826a03d1095341af50b6165d4ea5164

    My own forks (Ruraam/…) were checked too: they are 0 commits ahead / 533 and 502 behind upstream — plain older snapshots, so the upstream commits above were used for the build.

    Environment: Debian forky/sid, Docker 28.5.2 (docker.io), image omamobile-build:24.04 built from docker/Dockerfile (FROM Ubuntu 24.04 @sha256:fe8156099…), toolchain from tools/fetch-toolchain --verify (clang-r596125, Android clang 22.0.2).

    Official procedure, all rc=0: fetch-toolchain --verify → docker build … → omamobile kernel → modules → initramfs → images.

    Images produced (out-public/images/):

    Image Size SHA-256
    boot.img 18,472,960 9c393a15f4f22f0ef2589a56801070352b0fb6441e43c51499ce114402bb5649
    vendor_boot.img 6,877,184 ebdb1c7f74ff2d25380c29f8ce09f885bafe65f9a5125e97e12d2d024924a8f4
    dtbo.img 4,096 f89588777e61c16b564feb0d04319805fa7d5d5922622e52a173e57e21376fb6

    Header v4 confirmed for boot.img (kernel only, empty generic ramdisk) and vendor_boot.img (ramdisk 6,678,371 + DTB 191,220). No init_boot / vendor_kernel_boot, as documented for Oriole.

    Comparison with my earlier (non-public, partly instrumented) oriole build: dtbo.img is byte-identical. The three images all differ for expected reasons (local AVB footer, an extra omamobile.rescue cmdline token, a stripped-down driver set). A full .config diff gave only 44 differences out of ~10,180 options (0.4 %) — almost all auto-detected toolchain probes (clang 22 vs Debian clang 21) plus Broadcom WiFi/BT drivers disabled locally. No debug/instrumentation option differs at all. So the ~16 kB kernel size delta is explained by the compiler, not by configuration. This makes porting our local traces onto the public base straightforward, but of course nothing here has been flashed or boot-tested.

    Small things worth feeding back:

    1. bin/omamobile calls docker run without sudo; a build machine where the user is not in the docker group hits permission denied at in_container() (line 307). Not a code bug — just an environment prerequisite worth documenting.
    2. KERNEL_BASE must point at the kernel clone's parent when using a custom KERNEL_DIR; otherwise the kernel tree is not bind-mounted into the container and make defconfig fails with No rule to make target 'defconfig'. This cost me two attempts and is easy to miss.
    3. omamobile modules prints depmod missing (kmod not in the container image) but still packs 1,827 modules successfully — non-blocking.
    4. No docker build command is documented; I derived docker build -f docker/Dockerfile --build-arg UID=1000 --build-arg GID=1000 -t omamobile-build:24.04 . from the Dockerfile.

    No source files were modified, no patch/commit/PR was made, and nothing was flashed. Happy to share full logs or run further read-only checks if useful.

    Thanks,

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions