Skip to content

device/hotdog: the OnePlus 7T Pro (community port) - #931

Open
johannSo wants to merge 5 commits into
derekross:mainfrom
johannSo:hotdog-port
Open

johannSo wants to merge 5 commits into
derekross:mainfrom
johannSo:hotdog-port

Conversation

@johannSo

@johannSo johannSo commented Nov 19, 2026 •

Copy link
Copy Markdown

What and why

A community port of the OnePlus 7T Pro (hotdog, SM8150): the descriptor, plans with runbooks and results, bring-up overlays and kernel config fragments, the device package, and what the port needs from the shared tools. Plans: device/hotdog/BRINGUP-M1.md and the *-PLAN.md next to it.

Commits:

  1. bin/omamobile, initramfs: BOOT_METHOD=qcom-abl-v2 packs the one header-v2 boot.img this ABL boots (Image behind ABL's UNCOMPRESSED_IMG tag, DTB appended and in the v2 DTB section). fastboot, flash and push refuse it. New opt-in INITRAMFS_USERDATA_WAIT (1–600 s) for a UFS that comes up late.
  2. build-rootfs, phone-install-package: with --local, the contributor's local key ships as its own keyring so the phone trusts its [omamobile]; no __pycache__ copied to the phone.
  3. hexagonrpc, modemmanager, mmsd-tng: packages only omamobile-device-hotdog depends on.
  4. device/hotdog, device-hotdog 0.1.18: the port itself, plus the docs/DEVICES.md rows.
  5. device-hotdog 0.1.19: ships the opt-in files for the shell PR, and a package README.

Nothing here changes another phone

  • qcom-abl-v2 code runs only for that BOOT_METHOD.
  • Without /etc/omamobile-userdata-wait the init looks for userdata once with no sleep, as before. tests/initramfs-root-policy.py covers the packing, the validation, and the init both ways.
  • The local-key keyring is only added for CHANNEL=local with a key.
  • modemmanager is installed by nothing but this device package. The Pixels' omamobile-mmd conflicts with it. One caveat: a phone that has Arch's modemmanager installed by hand would take this build (Arch's recipe plus two upstream fixes) from [omamobile].
  • The port's first draft had its own USB_HOST_PANIC=0 init change. It's dropped here in favour of upstream's existing INITRAMFS_USB_RECOVERY=no.

For the maintainer to decide: CONTRIBUTING.md §3.1

This phone has no stock slot. A slot switch once moved its UFS boot LUN onto mismatched bootloader firmware, and recovering took a full EDL restore. So here:

  • Omarchy boots from boot_a, and slot B is left untouched.
  • A human writes the image with fastboot flash boot_a after it has RAM-booted.
  • device/hotdog/slot-a-mark re-marks slot A successful in the GPT, by hand and with the owner's yes, after such a flash. It records the sectors before and after.

Both are outside §3.1 as it stands. Per §3.1, this PR doesn't touch CONTRIBUTING or AGENTS; the decision is yours.

Kernel

The kernel (the SM8150 board on 7.3, plus drivers: AW8697 haptics, s6sy761 ear detection, a fastrpc fix, the q6voice series) is on local branches and not published yet. It will follow as linux-omarchy-mobile PRs or a fork branch. Related PRs: #932 (shell, opt-in), #933 (GNSS, opt-in), #934 (libcamera sensors), #935 (Waydroid on Adreno), #936 (MMS, draft: changes every phone). This PR does not depend on them; device-hotdog 0.1.19 ships the opt-in files #932 reads, which do nothing without it.

Device and build

  • Phone: OnePlus 7T Pro (hotdog), the owner's own
  • Kernel release (uname -r): 7.3.0-rc6-omamobile-push1-g7539ca3e06f6 (the port's local kernel, on boot_a)
  • Packages changed, with versions: omamobile-device-hotdog 0.1.19 (new), hexagonrpc 0.5.0.r20.gda7a374-2 (new), modemmanager 1.24.2-2.2 (new), mmsd-tng 2.6.4.r2.g3411171-1 (new)

Checks (no phone needed)

  • tools/pr-check origin/main passes
  • tests/all passes
  • tests/packages: omamobile-device-hotdog test/all passes
  • Each changed package has its version bumped, and the new version is in the commit subject
  • Commits have area prefixes and bodies that say why; AI help is a Co-Authored-By: trailer, never a Signed-off-by:

On the phone

  • I ran this on my own phone. It runs the port's branch: device-hotdog 0.1.18 and the same tools, from the same files except the two changes below.
  • Runbook steps passed, each with its date and kernel in the plan's results:
    • BRINGUP-M1 H0–H4
    • FINGERPRINT-PLAN F2–F8
    • MIC-PLAN
    • CAMERA-PLAN C0–C8, P0–P5
    • MODEM-PLAN R0–R9, V1–V4
    • KEYS-PLAN
    • GNSS-PLAN G0–G5
    • SUSPEND-PLAN S1–S5, S8
    • HAPTICS-PLAN H1–H7
    • SENSORS-PLAN
  • What I did not test:
    • an image built with this branch's descriptor keys (INITRAMFS_USB_RECOVERY=no and INITRAMFS_USERDATA_WAIT=300 replace the port's USB_HOST_PANIC=0 and its unconditional 300 s wait);
    • device-hotdog 0.1.19's two opt-in files on the phone;
    • Bluetooth audio, charging and NFC, which aren't in this port's plans yet.
  • Nothing in this change writes outside slot B and userdata. Not true for this phone: the boot image goes to boot_a by fastboot (by hand), and slot-a-mark writes slot A's GPT attributes (by hand). Linux itself writes only userdata: UFS_RW=userdata, every other partition read-only, rmtfs read-only. Nothing weakens the storage guard or reads registers from userspace.
tools/phone-smoke report

phone-smoke: hotdog

Kernel 7.3.0-rc6-omamobile-push1-g7539ca3e06f6, tools at 4b31273a, 1970-01-19T09:11Z.

Check Result Detail
boot (required) not ok kernel 7.3.0-rc6-omamobile-push1-g7539ca3e06f6 on slot a: our builds boot from slot b
storage guard (required) ok 117 partitions; only userdata and slot a's boot images writable
root (required) ok ext4 on userdata, 180G free
kernel health (required) ok no oops or BUG; 0 WARN, 39 error lines in dmesg (1 known, 38 new)
new dmesg errors not ok N N: ASoC: mux RX N N INP has no paths x6; msm_dpu N: [drm:recover_worker] ERROR N: hangcheck recover! x4; adreno N: [drm:N] ERROR gpu fault ring N fence N status N rb N/N N N/N N N/N x4
systemd units ok none failed
USB gadget (required) ok bound to a600000.usb
display ok card1-DSI-1 connected, 1440x3120
GPU ok render node on msm_dpu
session (required) ok Hyprland and the Omarchy shell running
lock ok PAM file present, the shell can lock (passwordPam true, locked true)
touch ok s6sy761
buttons ok Volume keys
Wi-Fi ok wlan0 on ath10k_snoc, link down
Bluetooth ok hci0, powered false
battery ok 63% Discharging, 31 °C
sound cards ok Pro
sensors ok AmbientLight Proximity
cameras ok 4 listed by libcamera
thermal ok hottest zone 47 °C

Packages: hyprland 0.57.0-3.3, libcamera 0.7.2-4.23, mesa 1:26.2.7-1, omacam 0.3.3-1, omamessages 0.2.2-2, omamobile-apps 0.2.3-1, omamobile-device-hotdog 0.1.18-1, omamobile-firstboot 0.1.6-1, omamobile-gnss 0.2.11-1, omamobile-keyring 20261026-1, omamobile-mirror 0.1.1-1, omamobile-omarchy 0.1.48-1, omamobile-owl-widgets 0.1.2-1, omamobile-settings 0.1.2-1, omamobile-shell 0.2.135-1, omamobile-voice 0.1.0-1, omamobile-waydroid 0.1.4-1, omaphone 0.2.3-1, omarchy 4.5.0-3, omatel-agent 0.2.4-3, quickshell 0.3.2.r7.g2b4501a-1

Notes on the report:

  • "boot: not ok" is this port's slot A (see §3.1 above).
  • The phone's clock read 1970: the RTC is reset at boot until NTP syncs, a known open item in BRINGUP-M1.
  • The new dmesg lines include four Adreno GPU faults with hangcheck recovery, which I haven't looked into yet.

🤖 Generated with Claude Code

johannSo and others added 5 commits November 19, 2026 10:37
…ata wait

The OnePlus 7T Pro's ABL boots one header-v2 boot.img: the uncompressed
Image behind ABL's "UNCOMPRESSED_IMG" tag with the DTB appended and in
the v2 DTB section, the command line in the header. BOOT_METHOD=qcom-abl-v2
packs it. BOOT_IMAGES must be boot and UFS_RW userdata alone, and
fastboot, flash and push refuse the method: there is no slot B of ours
on that phone, and a human writes the image by fastboot after it has
RAM-booted (device/hotdog/BRINGUP-M1.md).

UFS can come up late there: after a RAM boot from the fastboot screen,
userdata appeared 50 s after SSH, and the init's single look found no
partition. INITRAMFS_USERDATA_WAIT (1-600 s) packs a limit, and the init
then looks once a second with a line every 10 s, and lets the guard see
the new partition before it checks it is writable.

Every other phone is unchanged: the new code runs only for
BOOT_METHOD=qcom-abl-v2, and without /etc/omamobile-userdata-wait the
init looks once with no sleep, as before (tests/initramfs-root-policy.py
checks both, and that no other descriptor packs the file).

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

A contributor's port builds its own packages with a local key
(tools/local-key-init). build-inside removes the build's keyring from the
image, so the phone didn't trust that key, and every refresh failed at
[omamobile] with "key ... is unknown", omarchy update with it (hotdog,
2026-11-04). With --local, the key now ships as a keyring of its own
(omamobile-local), which firstboot's pacman-key --populate trusts. Images
built for any other channel are unchanged.

phone-install-package no longer copies the laptop's Python caches: a
bin/__pycache__ from a test run failed the shell's check() on the phone.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- hexagonrpc 0.5.0.r20.gda7a374-2: the FastRPC file server the SLPI's
  sensor stack reads its registry and configuration through, with two
  fixes (the listener keeps serving after an unsupported method; rpcd
  serves /sys/project_info). device/hotdog/SENSORS-PLAN.md.
- modemmanager 1.24.2-2.2: Arch's recipe with two fixes from upstream
  ModemManager, the netlink use-after-free (9317cbc4: it crashed on every
  second data connection) and the multiplexed-bearer cleanup assertion
  (0edcb916: it aborted at every shutdown with data on).
  device/hotdog/MODEM-PLAN.md.
- mmsd-tng 2.6.4.r2.g3411171-1: the MMS daemon (a user service), built
  from a pinned upstream commit two past the 2.6.4 tag.

Only omamobile-device-hotdog depends on them. The Pixels' modem stack,
omamobile-mmd, conflicts with modemmanager, so no Pixel installs it; a
phone that has Arch's modemmanager installed by hand would take this
build from [omamobile] at its next update.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A community port of the OnePlus 7T Pro (hotdog, SM8150): the descriptor,
the bring-up overlays and kernel config fragments, the plans with their
runbooks and results, the device package and the camera table tools.

On the owner's phone, from boot_a: display, GPU, touch, USB networking,
Wi-Fi, buttons and the alert slider, haptics, the torch, all four
cameras with autofocus and the pop-up, speakers, earpiece and
microphones, calls, SMS, MMS and mobile data, GNSS through the modem's
QMI LOC, ambient light through the SLPI, the screen off at the ear in
calls, s2idle suspend and the optical fingerprint sensor. Each plan's
status line and results give the kernel release and date.

This phone has no stock slot: a slot switch once moved its UFS boot LUN
onto mismatched bootloader firmware and took a full EDL restore, so
Omarchy boots from boot_a, slot B is left as it is, and the human writes
the boot image by fastboot after a RAM boot (BRINGUP-M1). slot-a-mark
re-marks slot A successful in the GPT, by hand and with the owner's yes,
after such a flash. Both are outside CONTRIBUTING.md §3.1 as it stands,
and are the maintainer's decision; nothing here changes that section.

The kernel (the SM8150 board on 7.3) is on local branches and not
published yet. Nothing in this commit changes another phone.

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

omamobile-shell 0.2.136 takes the Mobile network page and the data tile
through ModemManager, and turns the screen off at the ear in calls, only
where a device package ships /usr/share/omamobile/mobile-network-modemmanager
and /usr/share/omamobile/call-proximity. This package ships both, so the
OnePlus 7T Pro keeps what it ran with shell 0.2.135 of the port's branch,
and no other phone gets either.

Also a README for the package, as the package layout asks (CONTRIBUTING
§6.3): what each part is and the plan it came from.

Co-Authored-By: Claude Opus 5.5 <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.

1 participant