Repository navigation
Conversation
Ignore local development playbooks and captures so tegu work can retain private context without exposing it in public changes. Co-Authored-By: Codex <codex@openai.com>
Provide tegu-only userspace policy for the first native root: keep the storage guard in force, configure USB networking, and bound a hung shutdown. Keep slot-success marking and peripheral startup absent until the board and recovery gates pass. The package stages cleanly on the Linux build host; tests/all passes there. This does not make the staged descriptor installable.
The package needs its reboot-to helper at build time, but packages/.gitignore ignores src directories. Track this tegu copy explicitly so a clean checkout can build the package.
The currently running stock Pixel 9a and the verified CP3A.261005.005 factory init_boot header agree on Android 17 and the October 2026 patch level. Keep the image selector disabled until the board and AVB gates pass.
The tegu USB gadget runs in the guarded initramfs before root filesystem modules are available. Build the Synopsys eUSB2 PHY into the kernel so the high-speed link can probe at that stage.
The shared SOC_COMPATIBLE packer is now on upstream main. Keep tegu gated until its public DTB and AVB values are qualified, while removing stale review wording from both descriptor copies.
Tegu first light has no usable wall clock before network time. On the first full Omarchy boot, the clock predated the bundled package keys, so pacman-key skipped them and firstboot finished with a trust warning. Raise only a stale clock to the installed keyring file timestamp before omamobile-firstboot. This lets the signed local repository initialise without depending on a particular build host or network connection. Verified with the tegu package test, an arm64 package build, and on-device unit verification and start. The existing installed keyring was repaired after setting current time.
ABL leaves the tegu command-mode display trigger masked after its splash. The live register read 0x3070; 0x3061 sent the simpledrm framebuffer to the panel while Hyprland and Omarchy Shell were running. Pack the board-checked trigger pair into the tegu initramfs so the desktop appears after boot.
The package installs a copy of the device descriptor. The booted tegu descriptor gained the board-checked DECON trigger, but the package copy did not, so package tests failed and an installed package would leave the screen at the boot splash. Keep the installed descriptor identical to the boot-tested source. The tegu package test passes in the Linux build environment.
The S3908 child waits for S2MPG14 regulators. Keep ACPM and PMIC modules from autoloading during early boot, then load them through a tegu timer after the USB rescue path is available. Fail the service if touch input does not appear. On the Pixel 9a, a native B cold boot reached systemd before the timer, then registered Synaptics TouchComm input at 35 seconds without manual commands. The package staging test passed on Linux and the phone stayed responsive over USB.
Select both observed tegu IPOP revisions, apply the staged watchdog, USB, UFS and touch overlays, and use the October Android 17 partition sizes and AVB metadata. Keep package descriptor synchronized with the image builder. The public recipe rebuilt all five images and the overlaid DTB byte-for-byte from the working touch-child bundle with the same frozen kernel and initramfs. The device cold booted Omarchy and registered S3908 touch after the delayed ACPM start; physical taps and swipes were confirmed. The public tegu kernel and signed root image remain separate release dependencies.
The boot recipe and device package now reach native Omarchy with working touch on slot B. Keep the first-light summary aligned with the qualified source while naming the public kernel, signed root image and clean install still needed for a release.
|
Welcome, and thank you for the package. It's a careful piece of work: the eight files you took from the Pixel 7 Pro's package are byte-identical to main's, the test refuses boot-success, bootctl and every hardware start, and I ran The board DTS first. The four overlays name 28 nodes ( One descriptor, two phones. The notes. Rather than rewriting the header of the existing notes, please add a dated entry below the existing sections with the kernel release ( The touch timer. Smaller things, for when you get to them:
Thanks again for taking this on. Having two 9a owners is a real asset, as long as you're both building the same tree, so getting the DTS into linux-omarchy-mobile is the next step that unblocks everything else here. |
|
On the descriptor values, the rule I'll apply so neither of you has to wait for the other: the descriptor carries the values of the oldest stock build any owner runs on slot A, which today is W4HO-ham's 2025-09 (what main has), and a pull request never raises them. A newer rollback index or patch level in slot B's images can only ever lock an older stock out of slot A, and slot A is the way back for everyone. If your phone's bootloader refuses the older values, that is a finding for the notes (with the exact fastboot or ABL message), and we'd look at deriving the values from the owner's own slot A at install time rather than pinning newer ones in the repository. The AVB footer and the trigger line follow the same rule: in only once both of you have seen them boot. |
The current mainline kernel builds exynos_mailbox as a module. Starting ACPM before it leaves all 16 channels unusable, so S2MPG14 cannot supply the Synaptics controller. Load the mailbox before ACPM and keep the startup scoped to tegu. The new B kernel booted Omarchy and registered touch after loading the mailbox first. The tegu package staging test passes on Linux; pkgrel 2 upgrades the on-device package. Use coreutils for the descriptor test so a minimal build root does not need cmp.
# Conflicts: # tests/all
|
This is amazing feedback ty @derekross. My Clanker is still clanking and opened this PR last night while I was asleep - but will read through and address all the feedback. We should open a comms channel @W4HO-ham so we don't overlap - what suits best? I'm on TG at this expiring link handle or twitter @ harkl_ . Ping me and we can connect. |
Restore the shared descriptor to the oldest owner firmware values and leave image selection gated. Remove prototype-only overlays in favor of the public board and touch series. Preserve the first owner notes and append the second phone results with their limits. Persist a touch-start marker before loading drivers, retain it on failure, and require an explicit retry after an incomplete start. Handle the public and prototype DT paths, remove the ineffective regulator blacklist, and move the clock floor to a separate firstboot change. Linux tests/all passes with environment-dependent skips. Image layout and overlay checks also pass with the AOSP and kernel build tools. Mocked package tests cover module order, marker persistence, retry, timeout and disabled nodes. Hardware smoke evidence predates this review change and is recorded separately. Co-Authored-By: Codex <codex@openai.com>
Phones without an RTC can start before the installed signing keys were created. Use the newest public keyring timestamp in the verified root image as a clock floor before GnuPG creates or imports keys, without moving an accurate clock backwards. The Linux firstboot tests cover an old clock, an accurate clock, absent keyrings, and ordering before pacman-key. Move this shared policy out of the tegu package. Co-Authored-By: Codex <codex@openai.com>
The package check uses cmp, which is absent from the minimal phone root. Declare diffutils as a check dependency so makepkg can run the full tests in a fresh build environment. Built with checks enabled on aarch64 tegu, installed 0.1.3-2, and ran the installed keyring setup against an isolated keyring root. Clock seeding and real pacman-key initialization and population passed. Co-Authored-By: Codex <codex@openai.com>
Keep the tegu review focused on device support. The shared omamobile-firstboot clock floor and native check dependency are carried by the firstboot-clock-floor branch instead. Co-Authored-By: Codex <codex@openai.com>
Co-Authored-By: Codex <codex@openai.com>
The public board needs ACPM for early USB power, so gate the modular touchscreen itself behind the synced startup marker. Pin the public board, touch and native display source tree and provide a worktree preparation command instead of depending on private patches. Native aarch64 package checks, installation integrity and retry pass. The existing private kernel also passed automatic startup, cross-boot marker skip, guarded retry and stock A recovery boot. Public-source reproduction and display schema checks pass; public-kernel hardware qualification remains pending, so the descriptor stays gated with the oldest shared metadata. Co-Authored-By: Codex <codex@openai.com>
The footer-free images returned to fastboot. Adding index-zero AVB footers to byte-identical payloads booted the public board and touch kernel with Android 16 and September 2025 metadata on the newer phone. Record actual smoke results and limits without changing the shared descriptor or claiming cross-owner display qualification. Co-Authored-By: Codex <codex@openai.com>
…policy Arch kernel-modules-hook removes unowned future-kernel modules, leaving a flashed kernel without display, touch or firewall modules. Skip that cleanup while the tegu package is installed. Enable native DSI brightness commands as on the other Tensor packages and cap refresh at the tested 60 Hz. Native aarch64 package checks and integrity checks passed. A subsequent public-kernel boot retained all 1782 modules, automatically initialized DRM and touch, and passed every required phone-smoke check. Camera and owner testing confirmed the visible Omarchy lock screen and touch; backlight restore and firewall passed. Record the remaining Plymouth failure, missing GPU/peripherals and cross-owner footer qualification limits. Co-Authored-By: Codex <codex@openai.com>
|
Thanks for working through everything. I went through my earlier points one by one against the new head:
Minor: This stays a draft until the kernel PRs land and W4HO-ham has seen the footers boot, but from my side the package itself is close. |
|
@h4rkl, I missed one thing in my last review, sorry. Google's tegu device tree lists three panels (TG4A, TG4B and TG4C), and gs-dsim-cmd can't read the panel's ID yet, so it can't tell which one it's driving. TG4B's rate switch sends more than TG4C's, so a TG4B phone given TG4C's commands would get half a rate switch. On derekross/linux-omarchy-mobile#90 we agreed the guard goes in the tegu device package: turn on A modprobe.d # omamobile-tegu-display.conf
install gs_dsim_cmd /usr/lib/omamobile/omamobile-tegu-dsim-load#!/bin/sh
# Native DSI commands only for the panel they were written for (TG4C).
case " $(cat /proc/cmdline) " in
*" exynos_drm.panel_name=google-tg4c"*)
exec /usr/bin/modprobe --ignore-install gs_dsim_cmd allow_tx=1 ;;
esac
exec /usr/bin/modprobe --ignore-install gs_dsim_cmdPlease add a test case for each branch, with the command line passed in so it doesn't read the host's Two small things while you're there:
This is still a draft and waits on the kernel PRs either way, so there's no hurry. |
1 similar comment
|
@h4rkl, I missed one thing in my last review, sorry. Google's tegu device tree lists three panels (TG4A, TG4B and TG4C), and gs-dsim-cmd can't read the panel's ID yet, so it can't tell which one it's driving. TG4B's rate switch sends more than TG4C's, so a TG4B phone given TG4C's commands would get half a rate switch. On derekross/linux-omarchy-mobile#90 we agreed the guard goes in the tegu device package: turn on A modprobe.d # omamobile-tegu-display.conf
install gs_dsim_cmd /usr/lib/omamobile/omamobile-tegu-dsim-load#!/bin/sh
# Native DSI commands only for the panel they were written for (TG4C).
case " $(cat /proc/cmdline) " in
*" exynos_drm.panel_name=google-tg4c"*)
exec /usr/bin/modprobe --ignore-install gs_dsim_cmd allow_tx=1 ;;
esac
exec /usr/bin/modprobe --ignore-install gs_dsim_cmdPlease add a test case for each branch, with the command line passed in so it doesn't read the host's Two small things while you're there:
This is still a draft and waits on the kernel PRs either way, so there's no hurry. |
…wenty-sixth rounds (#127) * docs/review: 19 lessons and 3 corrections from the twenty-fifth round The confirmed candidates from round 25 (L-260 to L-278): the 9a slot B marking as a devinfo question, cross-repository links, squash-merged stacks and version files, stale hand-built prebuilts, and before-merge Pixel 7 Pro runs on omamobile-int. C-16 to C-18 correct our own earlier comments on kernel #98, kernel #80 and #74; L-241 is corrected in place. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * docs/review: 15 lessons from the twenty-sixth round L-279 to L-293: the 9a's slot B hold and what made devinfo writable (#125), per-device flags checked by building both command lines (#101), env defaults in the shared build library and CI waiting for approval (#79), GLib retry paths and radio stops checked at the controller (#99), silent re-pushes and stacks with several tops (the tegu stack), and two of our own: merge only after every check has finished (#126), and an empty diff is checked for empty inputs (kernel #123). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
|
Hi h4rkl, W4HO here, the other 9a. Thanks for building on the public kernel PRs; that's the right base. So we don't duplicate, here is where my phone stands on the same kernel line: display at 120 Hz on the panel driver, GPU (Panthor, DVFS), touch, battery and charging, Wi-Fi, Bluetooth, haptics and torch, speakers and sensors through the AoC, three cameras, NFC (reads a payment card), fingerprint (unlocks the lock screen), modem ONLINE, s2idle with wake. GNSS is the one link still open (an SPI timing problem, being worked on #138). The bring-up for all of that runs from scripts and units on the phone that I never packaged; I'm packaging them now as I'd suggest we treat that one as the device package and rebase the pieces of this PR onto it. Two of yours I'll carry over as they are: the On the AVB footer question: I'll run the footered-image test on my phone (the 2025-09 bootloader) and post the result here, so the shared descriptor can move. |
|
The AVB footer result from the 2025-09 phone, and it turns out every boot here has been the test: my push kit's descriptor carries So for the shared descriptor, per Derek's rule (the oldest owner's values, never raised), these three lines are what I'd propose; they're exactly what main's |
|
The package I mentioned is up: #152, omamobile-device-tegu 0.2.0. It carries your |
|
@h4rkl, W4HO-ham has opened #152, a fuller The package goes through #152. It covers what yours does and more, so once it merges the package part of this PR is superseded, and I'd close this one then (or you can turn the branch into the follow-ups below, whichever you prefer). This PR also conflicts with main in What I'd still like from you, as separate pull requests:
Before you install 0.2.0 on your phone, please wait until #152's current round of changes is in. As it stands it would set The questions I asked earlier about this PR's own files ( |
The Pixel 9a's (tegu) bring-up moves from hand-copied files into a device package, on in-tree drivers only. Built on h4rkl's #74. At boot (vendor wants links), two units: - omamobile-tegu-graphics: loads the GPU (panthor), display (gs_decon), haptics, torch, battery gauge and Bluetooth UART, each blacklisted against autoload, after a synced hang marker. Each step keys on its own node; a load or bind that returns and fails is logged and skipped (a GPU without mali_csffw.bin keeps the display), and only a step that never returns keeps the marker, so the next boot leaves them off until --retry. gs_dsim_cmd.allow_tx (DSI commands, 120 Hz) is turned on before gs_decon only when the bootloader names the google-tg4c panel (with or without its id suffix); any other panel stays read-only at 60 Hz. The session's side ships as /usr/share/uwsm/env.d (KMS card, hardware GL only when the GPU came up). - omamobile-tegu-usb-rescue: closes the MAX77779's USB data switch after a replug while no Type-C driver owns it, through a C helper that checks the board, the chip's identity and VBUS (twice), never forces an I2C address, and reads back its one write; it reads VBUS first, so the cable out costs a register read and no log lines. Also: the Bluetooth address from the bootloader's DT, a charge limit from /etc/omamobile/charge-limit, s2idle only, the reboot watchdog (no runtime watchdog), the storage rule, usb0 as a fallback route, Arch's module cleanup held off, and sleep hooks: Wi-Fi kept or reloaded as flagged; NFC's poll loop stopped before a sleep and restarted after (powering the ST54L off only behind /etc/omamobile/nfc-power-off-across-sleep, off by default). The AoC start ships without a wants link until a kernel enables its nodes. Nothing marks a slot or touches devinfo, persist, the RPMB or Trusty's storage. Tested on the owner's 9a (private kernel 7.3.0-rc6, five package boots on 10-29, phone-smoke 19/19 on ffbf958); test/all runs every script in a scratch root. Co-authored-by: W4HO-ham <87742609+W4HO-ham@users.noreply.github.com> Co-authored-by: h4rkl <89045610+h4rkl@users.noreply.github.com>
|
Closing this now that #152 has merged: omamobile-device-tegu 0.2.0 is built on this PR (eight of its files are yours byte for byte, and the squash credits you as co-author). Thank you for laying the groundwork. Two things from here would still be very welcome as a new |
derekross#152 landed omamobile-device-tegu 0.2.0, which supersedes this branch's package, so the package and tests/all come from main unchanged. Following the review on derekross#74, kernel.sources and prepare-kernel are dropped; the branch keeps the second phone's notes and device/tegu/kernel.config. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Following the review on derekross#74, now that derekross#152 carries the package: the second phone's results go into NOTES as one dated entry in newest-first order, and device/tegu/kernel.config joins the repository so the published kernel branch plus this fragment is the recipe. The entry now states slot A's current state once, at the top: stock's five boot images match the official October images and its recovery boots, which supersedes the earlier "not a qualified rollback" line. It also notes that reboot-to bootloader comes back to a normal boot on this phone, next to the button combination. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…its boots (#240) The second Pixel 9a's record (the phone behind #74): a dated NOTES entry with slot A's state stated once and plainly (stock CP3A.261005.005, live images matching Google's by hash, a recovery boot verified, a full Android boot not), the corrected "not a qualified rollback" line and the `reboot-to bootloader` note, plus `device/tegu/kernel.config`, the fragment those boots ran. The maintainer marked the fragment bring-up only (its shared lines go once #227 lands; touch as a module is a test choice) and gave it the MIT licence line the other fragments carry. Closes #240 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…rekross#152) The Pixel 9a's (tegu) bring-up moves from hand-copied files into a device package, on in-tree drivers only. Built on h4rkl's derekross#74. At boot (vendor wants links), two units: - omamobile-tegu-graphics: loads the GPU (panthor), display (gs_decon), haptics, torch, battery gauge and Bluetooth UART, each blacklisted against autoload, after a synced hang marker. Each step keys on its own node; a load or bind that returns and fails is logged and skipped (a GPU without mali_csffw.bin keeps the display), and only a step that never returns keeps the marker, so the next boot leaves them off until --retry. gs_dsim_cmd.allow_tx (DSI commands, 120 Hz) is turned on before gs_decon only when the bootloader names the google-tg4c panel (with or without its id suffix); any other panel stays read-only at 60 Hz. The session's side ships as /usr/share/uwsm/env.d (KMS card, hardware GL only when the GPU came up). - omamobile-tegu-usb-rescue: closes the MAX77779's USB data switch after a replug while no Type-C driver owns it, through a C helper that checks the board, the chip's identity and VBUS (twice), never forces an I2C address, and reads back its one write; it reads VBUS first, so the cable out costs a register read and no log lines. Also: the Bluetooth address from the bootloader's DT, a charge limit from /etc/omamobile/charge-limit, s2idle only, the reboot watchdog (no runtime watchdog), the storage rule, usb0 as a fallback route, Arch's module cleanup held off, and sleep hooks: Wi-Fi kept or reloaded as flagged; NFC's poll loop stopped before a sleep and restarted after (powering the ST54L off only behind /etc/omamobile/nfc-power-off-across-sleep, off by default). The AoC start ships without a wants link until a kernel enables its nodes. Nothing marks a slot or touches devinfo, persist, the RPMB or Trusty's storage. Tested on the owner's 9a (private kernel 7.3.0-rc6, five package boots on 10-29, phone-smoke 19/19 on ffbf958); test/all runs every script in a scratch root. Co-authored-by: W4HO-ham <87742609+W4HO-ham@users.noreply.github.com> Co-authored-by: h4rkl <89045610+h4rkl@users.noreply.github.com>
…its boots (derekross#240) The second Pixel 9a's record (the phone behind derekross#74): a dated NOTES entry with slot A's state stated once and plainly (stock CP3A.261005.005, live images matching Google's by hash, a recovery boot verified, a full Android boot not), the corrected "not a qualified rollback" line and the `reboot-to bootloader` note, plus `device/tegu/kernel.config`, the fragment those boots ran. The maintainer marked the fragment bring-up only (its shared lines go once derekross#227 lands; touch as a module is a test choice) and gave it the MIT licence line the other fragments carry. Closes derekross#240 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add Pixel 9a (
tegu) v0.1.1 boot, native-display and guarded-touch policy, using a pinned public kernel source recipe instead of private board patches. Keep the shared descriptor on Android 16 / September 2025 metadata, with its selector and AVB/display-register overrides inactive pending cross-owner qualification.The package gates the touchscreen module behind a synced persistent marker and explicit retry. It preserves tarball-installed kernel modules against Arch's old-kernel cleanup, enables native DSI brightness commands following the other Tensor packages, and limits refresh to the tested 60 Hz. The shared clock change is separate in #97, entirely within omamobile-firstboot. New commits include the co-author trailer. Upstream main f5da1df is merged.
On-device validation on the second contributor's phone used the pinned public board/touch/native-display tree with Android 16 / September 2025 headers and private AVB footers at rollback index 0. Footer-free images returned to fastboot; adding footers to byte-identical payloads booted. Native DRM appeared on camera, the owner confirmed touch, and a subsequent boot automatically initialized display and touch while retaining all 1,782 modules. Required smoke checks passed. Earlier guarded tests verified persistent-marker skip across reboot and explicit retry. Stock A recovery boot and return through fastboot were verified without writing A or resetting userdata.
Offline/native build checks: source-tree reproduction, kernel builds, targeted DRM W=1 without diagnostics, full-schema display validation, native aarch64 package build/check and integrity checks, Linux tests, and PR policy checks. The base ACPM W=1 check has inherited kernel-documentation warnings. The user is coordinating the older-stock owner's footer test; no reply or contact has been sent by the agent.
This remains draft: shared footer activation awaits that owner's result, clean installation and signed release artifacts are unqualified, and automatic B-success marking is disabled. Plymouth's early splash daemon still crashes, there is no GPU render node, and the remaining peripherals in the report are absent or untested. These results establish a working native display/touch development build on existing Arch userdata, not an installable release for every 9a.
phone-smoke: tegu
Kernel
7.3.0-rc6-omamobile-push1-gde2e0689c517, tools snapshot53a106f, 2026-10-27T22:51Z.Packages: hyprland 0.56.2-7, libcamera 0.7.2-4, mesa 1:26.2.5-1, omacam 0.3.3-1, omamobile-apps 0.2.0-1, omamobile-device-tegu 0.1.1-6, omamobile-firstboot 0.1.3-2, omamobile-keyring 20261026-1, omamobile-mirror 0.1.1-1, omamobile-omarchy 0.1.12-1, omamobile-owl-widgets 0.1.1-1, omamobile-shell 0.2.28-1, omarchy 4.0.6-1, quickshell 0.3.2.r7.g2b4501a-1