Skip to content

device-alioth: add experimental POCO F3 source build and guarded installer - #25

Closed
bc-abe wants to merge 3 commits into
derekross:mainfrom
bc-abe:device/alioth-initial-support
Closed

bc-abe wants to merge 3 commits into
derekross:mainfrom
bc-abe:device/alioth-initial-support

Conversation

@bc-abe

@bc-abe bc-abe commented Oct 22, 2026 •

Copy link
Copy Markdown

What and why

Add an experimental private source-build and guarded installer for an unlocked POCO F3 (alioth). Its Android-v0 boot image uses an appended DTB, so it needs a dedicated install path. The runbook is device/alioth/FLASH-RUNBOOK.md; provenance and gates are in device/alioth/BRINGUP-M1.md.

Review changes:

  • Read the root label with LC_ALL=C dumpe2fs, with a real ARM BusyBox 1.37/ext4 regression check. Existing installations now need the omamobile-root label; other userdata stays in rescue without repair or mounting.
  • Gate clock, interactive-firstboot and coredump workarounds to alioth. Its drop-in orders after time-set.target without pulling that target in. Preserve main's touch/autologin changes.
  • Read the fastboot serial once and pin every later check and write to it. Probe help and dry runs disclose the dtbo_b erase and persistent slot-B selection.
  • Keep private keys in keys/alioth/, signed packages in out/alioth/repo, and generic root images in out/alioth/. Maintainer paths retain their defaults.
  • Keep the F3 matrix/Wanted entries at testing pending a clean install. Move its instructions outside the Pixel quick start. Use one shared reboot helper source, a userdata-only descriptor allowlist, and remove unused device dependencies/Waydroid chains.

The kernel is a prebuilt 6.15.1-sm8250 kernel from mainlining/linux's nikroks/alioth tree at 022f7f866a9377c34301e60ba131121c624d2e56, extracted from the checksum-pinned 2026-09-28 reference release with matching modules/runtime. No binary or firmware is redistributed. The notes record the path toward building it here. Firmware packaging and replacing shared-init string patches with descriptor hooks remain follow-ups.

Device and build

  • Phone: POCO F3 (alioth), existing native installation.
  • Kernel release: 6.15.1-sm8250.
  • Packages changed: omamobile-device-alioth 0.1.1, omamobile-firstboot 0.1.3, omamobile-shell 0.2.24.
  • Fresh build: Passed: ten locally signed packages, 15 Omarchy patches, rootfs smoke, personalized ext4/sparse verification, normal/rescue boot image checks, ARM userspace and actual SSH configuration checks, and a dedicated-key login into the packed ARM Dropbear. No phone accessed by this build; clean physical installation remains pending.

Checks (no phone needed)

  • tools/pr-check origin/main passes.
  • tests/all passes, including Pixel installer/image/storage tests.
  • Touched package suites pass: device-alioth, firstboot and shell (189 shell checks).
  • Package versions and version-bearing area-prefixed commit subjects updated.
  • Commits describe the evidence and credit AI assistance without AI sign-off.

Additional checks: 19 installer gates; real ext4 provisioning with spaces and target-account validation; actual owner-key generation preserving maintainer sentinels; and real ARM BusyBox 1.37 label behaviour. A foreign-key cached neard package was quarantined; signature verification stayed enabled.

On the phone

  • A clean install of this exact rebuilt root assembly remains pending.
  • Existing phone: USB SSH, slot B, storage guard, ext4 root, display/GPU/touch, Wi-Fi and kernel health checked. Its installed userspace differs from this PR assembly.
  • The dedicated installer dry run executes no phone command; write targets remain literal userdata, boot_b, and dtbo_b erase.
  • Nothing widens the storage allowlist or writes radio, bootloader, vbmeta, NV or slot A.

Camera/haptic runtime experiments on the existing phone are a separate follow-up and do not change this PR's initial hardware claims. Calls, messaging/data, GPS, cameras/audio, haptics, fingerprint and suspend are not claimed supported by this initial profile. No keys, images, device identifiers or raw captures are included.

Read-only phone-smoke report — existing runtime, not the new root assembly

Kernel: 6.15.1-sm8250. The Camera app was open during this inventory; the camera-list result below records that limitation. Separate runtime tests captured main-camera frames, without claiming a clean install of this PR.

ok - boot: kernel 6.15.1-sm8250, slot b, up 593s
ok - storage guard: 119 partitions; only userdata and slot B's boot images writable
ok - root: ext4 on userdata, 202G free
ok - kernel health: no oops or BUG; 0 WARN, 108 error lines in dmesg
ok - systemd units: none failed
ok - USB gadget: bound to a600000.usb
ok - display: card0-DSI-1 connected, 1080x2400
ok - GPU: render node on msm_dpu
ok - session: Hyprland and the Omarchy shell running
ok - lock: PAM file present, the shell can lock (passwordPam true, locked false)
ok - touch: Focaltech FT8756 Touchscreen
ok - buttons: gpio-keys
ok - Wi-Fi: wlp1s0 on ath11k_pci, link up
ok - Bluetooth: hci0, powered true
ok - battery: 50% Charging, 37 °C
ok - sound cards: xiaomialioth
skip - sensors: no sensor proxy
not ok - cameras: libcamera lists none (or a camera app holds them)
ok - thermal: hottest zone 77 °C

@bc-abe
bc-abe marked this pull request as ready for review October 22, 2026 15:29

@derekross derekross left a comment

Copy link
Copy Markdown
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for this. It's a careful port. The alioth installer's fixed write targets, the phone-bound backup receipts, the explicit wipe flag and the dry-run default are exactly the kind of guards we want, and keeping the userdata-only allowlist in the F3 initramfs is a nice touch.

One blocker before we can merge, because it affects the reference device:

initramfs/init label check breaks Pixel boot. The Pixel initramfs has no util-linux; /sbin/blkid is busybox, which ignores -s LABEL -o value and prints the full dev: LABEL="…" UUID="…" TYPE="ext4" line. I checked this with our current cheetah initramfs busybox (1.37) under qemu, so with this change every Pixel would stay in the initramfs. The F3 only works because your initramfs adds util-linux, and test-storageguard.py stubs blkid. Could you read the label with dumpe2fs -h instead, the way grow_root already does, and add a test against the real busybox behaviour? Our dev phone's userdata does carry the omamobile-root label, so the check itself is fine once it reads the label correctly. It is still a new requirement for existing installs, so please mention it in the commit message.

Also requested:

  • Shared rootfs/firstboot changes (masking systemd-firstboot, the coredump drop-in, clock-epoch, date -s, Wants=time-set.target) change cheetah images too. Please move them into their own commit with the Pixel rationale, or gate them per device. Per systemd.special(7), time-set.target should only be ordered after, not pulled in.
  • Pin the serial in tools/alioth/install.py. Read serialno once and pass -s to every fastboot call, so the checks and the writes can't reach different devices when more than one phone is in fastboot.
  • Keep F3 outputs and keys out of the maintainer paths. init-keys.py reuses keys/gnupg/keys/fingerprint (our release signing key), and build.py rebuilds omamobile-keyring into the shared out/repo. Please use keys/alioth/, out/alioth/repo and out/alioth/ for the root image. Today an F3 root becomes the default pick for tests/all and install-phone.
  • Status table. Please mark the DEVICES.md columns as "testing" until the clean install of this branch passes, and keep the F3 out of the Pixel quick start for now.
  • Probe wording. Please make --probe say in its help and dry-run text that it erases dtbo_b and leaves slot B active.

Smaller items (fine as follow-ups): package the extracted firmware instead of overlaying unowned files; replace the string-patching of the shared init with descriptor hooks; make UFS_RW in device.conf match the userdata-only allowlist you actually ship; one reboot-to.c; drop the Waydroid chains and unused ufw/bluez depends; bump firstboot to 0.1.1; squash to area-prefixed commits.

On the prebuilt reference kernel: since nothing is redistributed it's within our rules, but we'd like DEVICES.md to say plainly that the F3 kernel is a third-party pinned binary, with a path towards building it here.

derekross added a commit that referenced this pull request Oct 26, 2026
* DEVICES: status as of 2026-10-26, from today's merged pull requests

The matrix and the wanted list had fallen behind what merged since
2026-10-22:

- Pixel Fold buttons, no -> works (volume and power): the volume keys
  (linux-omarchy-mobile#11, felix NOTES 2026-10-22) and the power key
  (#41: event0 reports KEY_POWER, logind ignores it, the phone stays up).
- Pixel Fold GPU, no -> testing: Panthor bound and glmark2 ran on the
  cover (#5), but #41's smoke test found no render node, so not on by
  default. A new felix NOTES entry records both, since the 2026-10-22
  entry still says the power key is not described.
- Pixel Fold touch: still WIP; links the cover touch reports (#48, in
  review).
- Pixel 9a boot: still WIP; "private kernel" is now "SoC patches merging,
  board DTS not public" (#18-#31, #34, #46 merged; issue #28).
- Wanted: the Pixel 7a reached first light (#43, issue #36; descriptor
  in omarchy-mobile#37/#38); the Pixel 6 Pro boots our kernel to
  Hyprland on the G78 (#2, #3; package in #41); the Pixel 8 Pro's UFS
  plan and the zuma DT are in review (omarchy-mobile#42, kernel #13);
  the Pixel 11 Pro Fold has a tracking issue (#40).

CAMERA-PLAN's status line still said the telephoto is "out of focus";
it is fixed-focus, sharp at a distance (2026-10-26, as DEVICES says).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* README, ROADMAP, CONTRIBUTING: status as of 2026-10-26

- README: a 2026-10-26 entry for omamobile-shell 0.2.17 and 0.2.18
  (#43, #46: home pill, recents on swipe and hold, Power wakes a dark
  screen, the touch audit's quick wins), the reserved-memory overlap fix
  (linux-omarchy-mobile#58) and the community ports' merged steps. It
  now says Omarchy Mobile's own packages aren't hosted yet, since the
  2026-10-22 entry read as if `omarchy update` fetched them over Wi-Fi:
  pkgs.omamobile.org is set up (30052e8) but nothing is published. The
  first non-Pixel ports (#25, #47) are still in review, not running.
- ROADMAP: the updates row and OTA track name pkgs.omamobile.org as set
  up and empty; the telephoto's lens holds its I2C bus when driven
  (not "doesn't answer" or "NACKs"); the shell row gains #43 and #46;
  §3.7 no longer calls the installer "not started" while §1 has it in
  testing; §3.8 step 9 lists the ports under way.
- CONTRIBUTING §5.2: zuma and zumapro are no longer "not started"
  (zuma: kernel #15-#17, #49, #50; zumapro: #18-#28, #34, #46).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* DEVICES, CONTRIBUTING: the Pixel 8 Pro in the matrix, and the naming rule

husky's notes merged in #23 and its package and UFS plan in #42, so by
DEVICES.md's own rule it moves from Wanted into the matrix: boot WIP
(our kernel on slot B, the zuma DT in review as kernel #13), storage
testing (UFS at HS-G4 behind the guard, Arch on userdata, UFS-PLAN
U1-U3), a simple-framebuffer console, USB and SSH. It also makes the
website's "under way on N more phones" count it.

CAMERA-PLAN's older "telephoto lens works" entry is marked superseded
by the fixed-focus finding, and kept as the record of that test.

The rule that public copy describes a base by its parts and doesn't
name other mobile Linux projects was only written in the website's
repository. It is in CONTRIBUTING §6.3 and AGENTS.md now, with the
exception for licensed sources (raised by #47).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
@derekross

Copy link
Copy Markdown
Owner

Hi @bc-abe, a short follow-up for when you pick this up again. Main has moved since my review, so here is what the rebase needs on top of the points there:

  • Conflicts with main in docs/DEVICES.md, docs/RELEASES.md, packages/omamobile-shell/version and tests/all.
  • Versions: main's omamobile-shell is now 0.2.23, and its omamobile-firstboot is already 0.1.2 with different contents, so this branch's changes need new versions (shell 0.2.24 and firstboot 0.1.3, or later) and matching >= depends. Both packages changed on main too (the touch work in Touch plan phase 0: a lock that engages, menu rows that can't break the phone, a keyboard that comes back #48), so please re-run their tests after the rebase.
  • DEVICES.md: the Wanted list now has a POCO F3 line pointing at this PR; please keep it in step with the matrix row (and the row at "testing" until a clean install of this branch passes, as before).
  • Wording: CONTRIBUTING §6.3 now asks us to describe a system by its parts, so please call the kernel "a prebuilt 6.15.1-sm8250 kernel from <tree, tag>" in the matrix and README. The links themselves are fine.

The busybox blkid fix in initramfs/init is still the one thing that would break the Pixels, so it's the one I'd most like to see first. Thanks!

bc-abe and others added 2 commits October 26, 2026 17:33
Require omamobile-root before any repair, resize or mount of userdata.
This is a new requirement for existing installations: an unlabelled root
stays in rescue rather than being modified. Read the label with dumpe2fs,
since Pixel BusyBox blkid ignores util-linux label/output selectors.

Exercise the boot function with real ext4 images and ARM BusyBox 1.37,
including its actual full-line blkid output and rejected Android labels.

Co-Authored-By: Codex <noreply@openai.com>
Allow private signing/repository/image paths while keeping maintainer
defaults. Support a board package profile and locally extracted runtime.
Use optional device firewall and clock helpers; gate the RTC lower bound,
interactive-firstboot mask and coredump workaround to alioth. Do not pull
in time-set.target on Pixels or alter their firstboot ordering.

Preserve upstream touch and autologin changes with shell 0.2.24 and
firstboot 0.1.3. Rebased shell and firstboot suites pass.

Co-Authored-By: Codex <noreply@openai.com>
derekross added a commit that referenced this pull request Oct 26, 2026
…lags as a named exception

Two decisions the Qualcomm ports (#17/#55-#59 flame, #52 coral, #25
alioth) were waiting on:

- AGENTS.md said never to run an unsuffixed fastboot flash, but our own
  tools/install-phone already flashes userdata that way behind its guards.
  The rule now says so: userdata has no slots, and it is flashed only
  through install-phone.
- On the Pixel 4 family the A/B slot flags live in GPT attributes, so
  marking slot B bootable is a partition-table write the storage guard
  can't block. It is allowed only as a narrow exception named per device
  in CONTRIBUTING §3.1 (LUN, sectors, bit, CRCs), done by hand once per
  boot-image install, with the flags recorded before and after, a
  rollback for a corrupted GPT and a way back. Never automatic. No device
  has an entry yet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
derekross added a commit that referenced this pull request Oct 26, 2026
…lags as a named exception (#60)

* AGENTS, CONTRIBUTING: userdata through install-phone; Qualcomm slot flags as a named exception

Two decisions the Qualcomm ports (#17/#55-#59 flame, #52 coral, #25
alioth) were waiting on:

- AGENTS.md said never to run an unsuffixed fastboot flash, but our own
  tools/install-phone already flashes userdata that way behind its guards.
  The rule now says so: userdata has no slots, and it is flashed only
  through install-phone.
- On the Pixel 4 family the A/B slot flags live in GPT attributes, so
  marking slot B bootable is a partition-table write the storage guard
  can't block. It is allowed only as a narrow exception named per device
  in CONTRIBUTING §3.1 (LUN, sectors, bit, CRCs), done by hand once per
  boot-image install, with the flags recorded before and after, a
  rollback for a corrupted GPT and a way back. Never automatic. No device
  has an entry yet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* DEVICES: the PinePhone Pro's codename and starting point

The Wanted row had "-" as its codename, which omamobile.org prints next
to the SoC ("-, Rockchip RK3399S"). It takes mainline's board name,
pinephone-pro, and names that device tree, already in our kernel, as the
starting point (promised on #47).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Build the pinned prebuilt 6.15.1-sm8250 kernel/runtime into private owner
images. Keep F3 keys under keys/alioth and outputs under out/alioth,
including its signed repository. Do not reuse release signing keys.

Require phone-bound original backups, validated image payloads, explicit
userdata-loss approval, and guarded rescue/readback gates. Read serialno
once and pin subsequent fastboot checks and writes to that device. Fixed
write targets are userdata and boot_b, with dtbo_b erase; probe help and
dry runs disclose that erase and persistent slot B selection.

Keep positive hardware status at testing pending a clean install of this
assembly. Document the exact prebuilt kernel provenance and source-build
path outside the Pixel quick start. No keys, images, firmware, captures
or per-phone identifiers are included.

Validation: tests/all, shell (189 checks), firstboot, 19 installer gates,
real ext4 provisioning, owner-key isolation, ARM BusyBox 1.37 label checks.
Fresh signed package/root build, personalized image/boot verification, ARM
userspace checks and packed Dropbear key login pass. Clean physical
reinstall remains pending.

Co-Authored-By: Codex <noreply@openai.com>
@bc-abe
bc-abe force-pushed the device/alioth-initial-support branch from f336beb to 2a9b1cd Compare October 27, 2026 00:02
@derekross

Copy link
Copy Markdown
Owner

Thank you, @bc-abe. I went through the new head against both my review and the follow-up, and you've done everything I asked: the label check reads dumpe2fs -h the way grow_root does and says it's a new requirement; the firstboot mask, coredump drop-in and clock epoch are gated to alioth and the drop-in only orders after time-set.target; the serial is read once and every fastboot call is pinned to it; keys and outputs are under keys/alioth and out/alioth; the matrix is at "testing" with the kernel's provenance named; the probe wording, the single reboot-to.c, the trimmed depends, the userdata-only allowlist and the three area-prefixed commits are all there. I ran tests/all and the three package suites on your head; all pass. That's a lot of careful work and I appreciate it.

What's left is new, and two of the three items are mine to decide, so I'll be explicit:

  1. A rule changed after your push, and I owe you that. On the evening of the 26th, after my comment, AGENTS.md gained an explicit exception for userdata: it may be flashed, but only through tools/install-phone (serial pinned, codename typed, allowlist, dry run first). Your install.py has the same four guards and only ever writes userdata, boot_b and the dtbo_b erase, so the question is whether the F3's boot method becomes a branch inside install-phone, or whether CONTRIBUTING §3.1 names your installer as a per-device exception with those guards. I'll decide that this week and say which here; please don't rework the installer until I do.
  2. The keyring package. build-packages now rebuilds omamobile-keyring with your build-keyring files in it, so out/alioth/repo and the image carry a package with our name and version but a different trust list, and a phone with it trusts your key for [omamobile]. I'd rather a development key could never be mistaken for ours. device-panther: Pixel 7 first port, and contributor builds with a local key #79 (the Pixel 7 port) adds the path I want for this: tools/local-key-init <codename>, build-packages --local <codename> and build-rootfs --local <codename>, where the key is trusted in the build image only and the keyring package stays as published. Once device-panther: Pixel 7 first port, and contributor builds with a local key #79 is in (soon), please rebase onto it, drop the OMAMOBILE_*_DIR knobs and the keyring copy, and use --local alioth. Your ROOTFS_PROFILE, ROOTFS_DIST_PACKAGES and --overlay additions can stay as they are.
  3. Versions. Main's omamobile-shell is at 0.2.28 and 0.2.29 is taken by an open PR, so your 0.2.24 needs a new number (0.2.30 or later); omamobile-firstboot 0.1.3 is also claimed by flame 0.1.0: hooks for shell 0.2.24 and firstboot 0.1.3 #55, so take 0.1.4 unless that one lands first. The >= depends follow. The branch currently conflicts with main in docs/DEVICES.md and packages/omamobile-shell/version.

Small: the alioth-specific cp --remove-destination reboot-to.c in tools/build-packages can move into the package's own build once the symlink is handled there.

I'll post the decision on item 1 as soon as I have it. Thanks for your patience with a moving target.

@derekross

Copy link
Copy Markdown
Owner

The decision on item 1: the F3's boot method becomes a branch inside tools/install-phone, not a second installer with its own exception. One guarded path for userdata is the rule for every device, so that the serial pinning, the typed codename, the allowlist and the dry run are the same code everywhere and a future change to the guard reaches every phone. Concretely: install-phone reads the device's descriptor, and alioth's descriptor says how its boot image is written (the direct ABL path your install.py implements); the alioth-specific steps move into a function the descriptor selects, and install.py goes away. Your checks and dry-run output carry over as they are. If some part of the F3's flow truly can't fit that shape, say which part here and we'll look at it before you write code.

Items 2 and 3 stand as above: --local alioth once #79 is in, and the next free version numbers after a rebase. Thanks again.

@derekross

Copy link
Copy Markdown
Owner

A note to unblock this: #79 has merged, so contributor builds with a local key (--local alioth, RELEASES §10) are available now, which was the piece my 10-27 comment was waiting on. When you rebase onto main (it conflicts in five files after two days of merges), please also take the next free version numbers: omamobile-shell 0.2.40 and omamobile-firstboot 0.1.4, or whatever is free when you push. The other points from 10-27 (the install-phone branch) still stand.

@derekross

Copy link
Copy Markdown
Owner

A friendly check-in: this has been quiet for over a week and now conflicts with main. If you'd like to continue, a rebase onto main will get it reviewed again. If I don't hear back within a week I'll close it to keep the queue tidy; it can be reopened any time.

@bc-abe

bc-abe commented Nov 11, 2026

Copy link
Copy Markdown
Author

Yeah just leave me behind. I've been trying to get this update date but I am hitting different walls.

When I have more time, I'll jump back in.

Thanks for your hard work on this.

@derekross

Copy link
Copy Markdown
Owner

Thank you, @bc-abe, and no hard feelings at all: this was the first port to a phone that isn't a Pixel, and the tree has moved a lot under it since October. The work that's here was careful: the serial pinned on every fastboot call, the label check through dumpe2fs, and the keys and outputs kept under keys/alioth and out/alioth.

I'll close this to keep the queue tidy, as I said last week. It can be reopened any time, and your branch stays as it is.

If you come back to it, a smaller first step would be easier to land than a rebase of all 68 files: device/alioth/ with the descriptor and your notes, and the device package, with no installer and no changes to shared code. The installer as a branch of tools/install-phone and the root image can follow as their own PRs, and I'm glad to help with the shape of each before you write code.

Thanks again for all the work on this.

@derekross derekross closed this Nov 12, 2026
derekross added a commit that referenced this pull request Nov 12, 2026
…i 8 lines #505 left (#680)

The Pixel 8a's row still described first light. Since then Omarchy's
image runs from userdata (BRINGUP-M2 gate e, #456 and kernel #206
merged), and the buttons and the touchscreen passed gates f and g
(kernel #238 and #241 merged): Boot, Storage, Touch and Buttons say so.

For the Mi 8, three lines promised on #505: BRINGUP-M1's [NOTES.md] link
had no definition, the board DTS (kernel #220) is merged, not in review,
and the status line names the 2026-11-11 boot NOTES records.

The POCO F3 row said #25 was in review; it was closed as paused on
2026-11-11.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…i 8 lines derekross#505 left (derekross#680)

The Pixel 8a's row still described first light. Since then Omarchy's
image runs from userdata (BRINGUP-M2 gate e, derekross#456 and kernel derekross#206
merged), and the buttons and the touchscreen passed gates f and g
(kernel derekross#238 and derekross#241 merged): Boot, Storage, Touch and Buttons say so.

For the Mi 8, three lines promised on derekross#505: BRINGUP-M1's [NOTES.md] link
had no definition, the board DTS (kernel derekross#220) is merged, not in review,
and the status line names the 2026-11-11 boot NOTES records.

The POCO F3 row said derekross#25 was in review; it was closed as paused on
2026-11-11.

Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
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