Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
83 changes: 83 additions & 0 deletions device/tegu/NOTES.md
Original file line number Diff line number Diff line change
Expand Up @@ -125,6 +125,89 @@ with nothing but Bluetooth starting resets again at that time, start here.
(434-445 mA cable out, as-is) and tonight; treat about 300 mA cable out
as the baseline.

## 2026-10-28: the second 9a (the phone behind #74)

These results are from a second retail Pixel 9a, the one behind
<https://github.com/derekross/omarchy-mobile/pull/74>. They don't replace
the palmtop's results in the entries around this one.

**Slot A, as it stands:** stock `CP3A.261005.005` (Android 17, October
2026). All five live slot A boot images match the official October images
by hash. Selecting A and asking for recovery reached Android's `No command`
screen, recovery ADB and its shell answered (slot A, `CP3A.261005.005`),
and recovery went back to fastboot and B back to Omarchy with no image
written and no userdata reset. That verifies stock's recovery boot, not a
full Android boot: Arch holds `userdata`, so going back to Android means
resetting it, as with the installer. An earlier draft of this entry said A
"is not currently a qualified rollback"; that was before the recovery check
and no longer holds. Recovery for these trials also relied on retained,
checksum-verified slot B image sets.

**Fastboot is by the buttons on this phone:** with the phone off, hold
Volume Down, add Power, and keep both held about 20 s. Software
`reboot-to bootloader` came back to a normal boot of B instead.

### Prototype kernel, simple framebuffer

Kernel `7.3.0-rc6-omamobile-push1-g9ea0e19f8dc2`, from an unpublished
branch on public `omamobile` `33267e524` plus the clock and binding changes
then in review and prototype board, USB, UFS and touch changes. It is not
the tree in linux-omarchy-mobile #80 and #81.

`DEVICE=tegu bin/omamobile flash` wrote only `boot_b`, `init_boot_b`,
`vendor_boot_b`, `vendor_kernel_boot_b` and `dtbo_b`, selected B and
rebooted; a second trial used an initramfs without builder SSH keys. Both
reached Arch, Hyprland and the Omarchy shell on the existing `userdata`.
The display was ABL's simple framebuffer with a DECON trigger write, on
October 2026 headers; none of that is in the shared descriptor.

`tools/phone-smoke`: boot, storage guard (62 guarded partitions), ext4
root, kernel health, USB and session passed; display 1080x2424 and
Synaptics touch found. No GPU render node; Wi-Fi, Bluetooth, battery and
charging, audio, sensors and cameras absent or skipped.

The touch start marker: a normal reboot registered touch and cleared the
marker; a reboot with the marker kept on purpose left touch off and logged
the skip; `--retry`, with kmsg streamed to the laptop and the watchdog
armed, brought touch back and cleared it. A fresh SSH login worked
afterwards.

### Public base, oldest headers

The public board and touch base `9386dfaa1d5d` with this directory's
`kernel.config` boots as `7.3.0-rc6-omamobile-push1-g9386dfaa1d5d`. With
Android 16 / September 2025 headers and no AVB footers, B went back to
fastboot (`bootfailure`, `enter-reason` "boot failure"), with no log that
names the cause. The same payloads, byte for byte, with hash footers
(the image tool's AOSP test key, rollback index 0, OS version 16, security
patch 2025-09-05) booted Arch with USB and SSH. ACPM reported 16 usable
channels at 0.22 s; touch registered at 36.1 s. Smoke at 97 s: boot, all 62
storage guards, root, kernel health, USB, touch and buttons passed.

So this newer phone boots the oldest shared header values with
index-zero footers. That says nothing about footers on the palmtop, and the
shared descriptor is unchanged. This base had no display driver (Plymouth
failed, the Google logo stayed).

### Native display

Display tree `1175a53bf9b2` booted as
`7.3.0-rc6-omamobile-push1-gde2e0689c517`, same headers and footers, no
trigger write. The first boot found that Arch's `kernel-modules-hook`
cleanup, during the previous kernel's boot, had deleted the unowned modules
of the next kernel; the package's `linux-modules-cleanup` drop-in (carried
into 0.2.0) skips that cleanup. With the modules restored: `card0-DSI-1`,
Omarchy on camera, touch confirmed by hand. A later reboot kept all 1,782
modules, brought up the panel at 3.66 s at 1080x2424, 60 Hz (the cap this
branch's package set; 0.2.0 drops it), and the lock screen came back on
DPMS wake. Smoke at 223 s: boot, storage guard, root, kernel health, USB,
the session, touch and buttons passed.

Still open on this phone: Plymouth's splash crashes in `strcmp` with no
console; no GPU render node (this kernel has no GPU node); Wi-Fi,
Bluetooth, battery and charging, audio, sensors and cameras unqualified; no
clean install; B not marked successful. Stock A was never written.

## 2026-10-28, evening: USB-C host mode through TCPM

Kernel `7.3.0-rc6-omamobile-tegu-s2mpg15-m1-g991042e2eff3-dirty` (the
Expand Down
13 changes: 13 additions & 0 deletions device/tegu/kernel.config
Original file line number Diff line number Diff line change
@@ -0,0 +1,13 @@
# SPDX-License-Identifier: GPL-2.0-only
# The USB gadget must come up in the initramfs before rootfs modules exist.
CONFIG_PHY_SNPS_EUSB2=y
# The public tegu board supplies that PHY through S2MPG14 over ACPM.
# Keep the whole supply chain available before mounting userdata.
CONFIG_EXYNOS_MBOX=y
CONFIG_EXYNOS_ACPM_PROTOCOL=y
CONFIG_MFD_SEC_ACPM=y
CONFIG_REGULATOR_S2MPS11=y
# Native DRM takes over the TG4C panel; no initramfs trigger write is needed.
CONFIG_DRM_PANEL_GOOGLE_TG4C=m
# Touch as a module, as tested on the second 9a (NOTES 2026-10-28).
CONFIG_TOUCHSCREEN_SYNAPTICS_TCM_SPI=m