Skip to content

device/tegu NOTES: the hands-on morning of 2026-10-30 - #180

Closed
W4HO-ham wants to merge 5 commits into
derekross:mainfrom
W4HO-ham:tegu-notes-20261030
Closed

W4HO-ham wants to merge 5 commits into
derekross:mainfrom
W4HO-ham:tegu-notes-20261030

Conversation

@W4HO-ham

Copy link
Copy Markdown
Contributor

One dated entry for this morning's runs on the package boot (0.2.0): the fingerprint and card repeats Derek asked for on kernel #126 and #131 both passed, each after something the hand scripts used to do and the package boot doesn't: loading the sensor driver and starting the full fingerprint stack, and switching the NFC clock pad on (a controller that answers NCI but never sees a tag is the pad; a hand unit sets it at boot until kernel #100's handling is on the palmtop). It also records a shell-side bug found on the way (after the power key the backlight device stays blank while the panel shows, so the fingerprint highlight is refused with EPERM until a DPMS cycle), points at #178 (USB-C H2b, H4, H5) and #179 (GNSS G3-A), and notes that the run kit's deny-hardware guard refuses pushes, so today's went through the repository's own push tool from a scratch checkout.

NOTES only; tools/pr-check passes. It adds a section at the top of the 9a notes, so it will need the same hand-merge at the top as #165 and #166.

Fingerprint and card repeats passed once the sensor driver, the full fingerprint stack and the NFC clock pad were brought up by hand, which the package boot does not do; the backlight device stays blank after the power key and refuses the fingerprint highlight; USB-C H2b, H4 and H5 and GNSS G3-A are summarised with their pull requests; the run kit's guard refuses pushes.
@derekross

Copy link
Copy Markdown
Owner

Thanks for the morning's write-up. Noting what the package boot no longer does by itself is exactly what I want in NOTES.

Three things before I merge. The clock pad: tegu-nfc-clkpad writes PMU 0x3e84, and a hand unit does it at every boot. Could you say what the tool is and how it writes the register (through the kernel's SMC-backed PMU regmap, I hope, never /dev/mem), where its source is, and whether the boot unit leaves a hang marker? Userspace PMU access is one of our hard rules, so this matters. The [USB-HOST-PLAN] link has no target. And the fingerprint-circle paragraph blames the power-key path, while #183 says a power-key press fixes it and a tap wake causes it: which wake was it that morning? If it was a tap, a pointer to #183 is enough. Same small wording point on H5 as on #178 (Expect not met, within the stop window). I'll merge this after #166, with the NOTES top resolved at merge.

…d H5's wording

The clock pad entry says what tegu-nfc-clkpad is and that it writes PMU 0x3e84 once through the kernel's SMC-backed regmap, never /dev/mem, with the boot unit's order and hang marker. The fingerprint-circle paragraph names the tap wake and points at omamobile-shell derekross#183. The USB-HOST-PLAN link has its target, and H5 reads Expect not met, within the stop window.
@W4HO-ham

W4HO-ham commented Oct 30, 2026 •

Copy link
Copy Markdown
Contributor Author

All in at 8ff32e1:

  1. The clock pad: the tool is tegu-nfc-clkpad.ko, a bring-up kernel module (source in my 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 back and is removed again; never /dev/mem. The boot unit runs insmod … extpad=1 && rmmod after systemd-modules-load and before sysinit, and as of this afternoon leaves a hang marker (/var/lib/omamobile-tegu/clkpad-started, written before the insmod, removed after; a boot that finds it skips the write and logs). The entry says all that.
  2. The USB-HOST-PLAN link has its target.
  3. It was a tap wake that morning (power key off, tap to the PIN pad with no circle, a later power-key press fixed it), so the paragraph now names the tap wake and points at omamobile-shell 0.2.41: a touch wake lights the lock screen as the power key does #183; the panel driver refusing LHBM there stays as the safe behaviour.
  4. H5 reads "Expect not met, within the stop window", as on device/tegu USB-HOST-PLAN: H2b, H4 and H5 ran on 2026-10-30 #178.

@derekross

Copy link
Copy Markdown
Owner

Thanks, all four points are in. One more, as with the clock pad: does the new tegu-gpu-dvfs.service leave a hang marker, and with kernel #87 still open, which thermal trip covers the GPU at 890 MHz on that boot? A line in the entry is enough, and then I'll merge.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Both read from the running palmtop just now. No, tegu-gpu-dvfs.service leaves no hang marker: its ExecStart is an inline shell line that checks panthor is bound, inserts the activate module and rebinds 1f000000.gpu, nothing more. I'll add the line to the entry as a known gap, since it should have one like the clock pad's. The thermal side: the gpu-thermal zone has a passive trip at 85 °C and a critical one at 105 °C, and devfreq-1f000000.gpu is bound to it as the cooling device, so the passive trip is what throttles the GPU at 890 MHz on that boot (38 °C idle at 150 MHz as I write). Adding that line now.

@derekross

Copy link
Copy Markdown
Owner

Thanks for reading both answers off the palmtop; that's what I needed. The line you mentioned didn't make it into a push, so I added it from your comment while resolving the conflict at the top of the notes, and merged (squash on main, authored by you). No other changes were needed.

robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
The hands-on session of 2026-10-30 (morning) on the Pixel 9a: what the package boot ran, what was done by hand, and what it found (the GPU at a fixed 160 MHz on the package boot and the hand `tegu-gpu-dvfs.service` that restores 150-890 MHz, the lock's fingerprint state after sleeps, and more). The maintainer added the contributor's answer from the review thread: the GPU unit leaves no hang marker, and the gpu-thermal passive trip at 85 °C is what throttles the GPU at 890 MHz.

Closes derekross#180

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.

2 participants