Skip to content
Closed
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
103 changes: 103 additions & 0 deletions device/tegu/NOTES.md
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,109 @@ hash/build ID; this dirty release is not a published kernel package.
`[V]` marks verified observations; `[I]` marks an inference still to test.
Raw captures, recordings, backups and vendor files remain off repository.

## 2026-10-30: the hands-on morning (fingerprint, card, USB-C host, GNSS)

Release `7.3.0-rc6-omamobile-tegu-s2mpg15-m1-g991042e2eff3-dirty`, device
package 0.2.0, the first morning of hands-on tests on the package boot. Each
run had one thing in its way that the package boot no longer does by itself.

- **Fingerprint (kernel #126's panel module, one-minute repeat): pass.** The
enrolled finger lit the circle and unlocked on the first press; three
presses of a wrong finger were refused; the right one matched again. Two
things had to be done by hand first: the sensor driver is not under
`/lib/modules` for this release (`goodix-fp.ko` lives in
`/var/lib/omamobile-tegu/`, loaded by `insmod`), and the stack has to be
started as `omamobile-fp-chroot start bridge`, not `start storage`. A
hand unit, `tegu-fp-stage.service` (insmod the driver, then the bridge
stage, after the graphics start), does both at boot since the afternoon,
so the lock screen takes a finger without a hand step; it is not in the
repository, like the clock-pad unit below.
- **No circle after a tap wake.** The power key turned the screen off, a
tap woke it to the PIN pad, and there the panel's backlight device still
read `bl_power` 4; setting it back to 0 was not enough, because the
backlight core's blank bit stayed set, `backlight_get_brightness()`
returned 0, and `local_hbm_mode` refused with EPERM (the driver's
allow_lhbm-and-brightness check). A DPMS off and on through Hyprland
cleared it, and so did a later power-key press, which is the wake the
shell handles fully. Cause and fix are in omamobile-shell
<https://github.com/derekross/omarchy-mobile/pull/183>: a touch wake from
s2idle got no `omamobile-keys wake`, so Hyprland never re-enabled the
display; it now runs the same wake as the power key. The panel driver
refusing LHBM in that state is the safe behaviour and stays.
- **NFC card read (kernel #131's modules): pass**, after two misses. The
controller answered every NCI command and neard reported Powered and
Polling (Mode stayed "Idle"), but no card was ever seen, because the NFC
clock pad (PMU 0x3e84) was off on this boot: the hand script that set it
at boot no longer runs on the package boot. `tegu-nfc-clkpad extpad=1`
(0x00000000 to 0x00010000) and a rebind of the driver, and the very next
60 s `nfctool -d nfc0 -p Initiator` poll found the card. The tool is
`tegu-nfc-clkpad.ko`, a bring-up kernel module (source in the owner's
logs, `logs/tegu-nfc-20261026/clkpad/tegu-nfc-clkpad.c`, not in the
repository): it takes the kernel's SMC-backed PMU regmap
(`exynos_get_pmu_regmap()`), does one `regmap_update_bits()` on 0x3e84
(EXTPAD_EN | PAD_EN), reads the register back, and is removed again;
never `/dev/mem`. A hand unit, `tegu-nfc-clkpad.service`, runs
`insmod … extpad=1 && rmmod` after `systemd-modules-load.service` and
before `sysinit.target`, and leaves a hang marker
(`/var/lib/omamobile-tegu/clkpad-started`, written before the insmod and
removed after; a boot that finds it skips the write and logs it). It
stays until kernel #100's in-driver handling is on the palmtop. Rule of
thumb: a controller that talks NCI but never sees a tag is the clock pad.
- **USB-C host (H2b, H4, H5): recorded in [USB-HOST-PLAN](USB-HOST-PLAN.md)**
(pull request #178). H2b passed (the PMIC block probes with its interrupt, the boost's
watchdog is serviced from it); H4 showed that unbinding the charger takes
the Type-C port down with it as a device-link consumer (a WARN in
`tcpm_port_unregister_pd`, VBUS to 0, the laptop link gone until the
dead-man reboot); H5 ran with its Expect not met, within the stop window
(the boost stays on across a reset, read at 3.13 s of the next boot, and
drops by itself when the RT9470's watchdog expires; the reworked driver
then takes it over).
- **GNSS G3-A (automatic chip select, feedback delay 0): same CHPP failure,
clean checksums**, recorded in [MODEM-PLAN](MODEM-PLAN.md) (pull request
#179). G3-A2 (stock's 32-bit words) is the one boot left on that trace.
- **Pushes from the run kit are refused by its own guard.** The kit's
`omamobile.local.conf` puts a `deny-hardware` directory of stubs (ssh,
fastboot, dd, ...) first on PATH, from the hands-free packing days; every
push since reads "phone not reachable". Today's pushes went through the
repository's own `bin/omamobile push` from a scratch checkout of main
carrying the kit's `device.conf`. After H5's boot the push wrote and
verified slot B but its reboot command never reached the palmtop: the USB
link had dropped again and Wi-Fi was unreachable too while touch still
worked; the owner rebooted from the menu and the new slot booted.
- **Sleep:** `tegu-mif-sleep.ko` loads at boot from a hand unit since last
night; a cable-out `power-ab` run with it is in progress (result in
[SLEEP-PLAN](SLEEP-PLAN.md) S1).
- **UFS did not come back from an s2idle, and the root went read-only
(afternoon).** A forced 10-minute `systemctl suspend` woke on its alarm
with `link startup failed 1` / `Host init failed 1` five times and
`ufshcd_wl_resume failed: 1` in the console; every disk read then failed,
ext4 went `emergency_ro`, and for an hour the palmtop answered SSH from
memory while `systemctl suspend`/`reboot` did nothing and the power key
could not light the screen. A sysrq `b` reset recovered it with a clean
journal replay. A palmtop that seems never to wake may be exactly this.
Record and rule in [SLEEP-PLAN](SLEEP-PLAN.md) S1 (#166).
- **The GPU ran at a fixed 160 MHz on the package boot.** `omamobile-tegu-graphics`
gives panthor the one inherited CMU point; the 10-27 table's 150-890 MHz
came from the hand `tegu-gpu-dvfs-activate` overlay module (adds stock's
G3D levels to `/opp-table-gpu`), which the package boot no longer applied.
Since this evening a hand unit, `tegu-gpu-dvfs.service`, applies it after
the graphics start with `activate=1` and re-binds `1f000000.gpu` before
the session starts (the module refuses to apply before panthor is bound;
the first two tries at boot failed on that), and devfreq shows 14 steps,
150 to 890 MHz, `simple_ondemand`. Kernel #85/#87 carry the proper form.
- **The lock's fingerprint state got stuck after the afternoon's sleeps.**
With the owner on the desktop, `omarchy-shell lock status` said
`locked:true, sessionLocked:true, fingerprint:true`, the lock's PAM
fingerprint verify ran in a loop every few seconds (fprintd always
claimed, `fprintd-verify` refused with AlreadyInUse), and
`omarchy-restart-shell` refused "while the session is locked". A fresh
`omarchy-shell lock lock` and a PIN unlock put the state back. Separately,
the lock shows no fingerprint marker over the sensor (only the PIN pad's
small icon), so the owner aims blind; the circle lights only on a touch
inside the sensor's 101 px radius at (540, 1756), which is why the
morning's unlock worked once the circle had shown the spot and the
afternoon's attempts saw no finger down at all.

## 2026-10-29: two resets at about 34 s, right after Bluetooth started

[V] During the night's GNSS G3 tries (kernel
Expand Down