Skip to content

device-oriole: add opt-in Bluetooth test overlay and bring-up plan - #872

Draft
Gr-linkk wants to merge 1 commit into
derekross:mainfrom
Gr-linkk:oriole-bluetooth-test-overlay
Draft

Gr-linkk wants to merge 1 commit into
derekross:mainfrom
Gr-linkk:oriole-bluetooth-test-overlay

Conversation

@Gr-linkk

Copy link
Copy Markdown
Contributor

Prepare the Pixel 6 trial requested in derekross/linux-omarchy-mobile#159. Its merged board description supplies the wiring. The opt-in overlay enables the USI controller, UART and BCM4389 child together. Default images keep them disabled.

The targeted W=1 device-tree build and relevant schema checks passed. Comparing the compiled trees found exactly three status changes. The historical BLE keyboard result on the private kernel used a different clock setup. Pairing, keyboard input and restart on the proposed public configuration are pending. The plan records the firmware/address prerequisites and stop conditions for that trial.

No firmware or private address is included. Repository policy and whitespace checks passed.

Enable the three existing nodes for the Pixel 6 Bluetooth trial. Record the clock and provisioning differences from the previously tested integration setup.

Assisted-by: LLM
@derekross

Copy link
Copy Markdown
Owner

Thank you. The overlay is as small as it should be, and it does exactly what you say: I compiled it and applied it to a gs101-oriole.dtb from one of my kernel builds, and the two trees differ in three status lines and nothing else. The plan is also careful to call the earlier result history from another clock setup rather than a pass, which I appreciate.

It is a draft, so only what it needs before the trial:

  1. A runbook with commands. Each step needs its commands, an Expect, a stop condition and a rollback, runnable from this repository:

    • build and write through the project's path: DTB_OVERLAYS="bluetooth-test" in your local descriptor, then bin/omamobile images and push, rather than "the qualified Oriole deployment";
    • the blacklist as a file and a line, and the step that removes it. The package's bring-up file pairs blacklist with install … /bin/false; for hci_uart use blacklist alone, or the load by hand in step 2 is refused;
    • the dead-man by name. "The existing watchdog" isn't in the public package. Arm a systemd-run --on-active=… systemctl reboot timer and start the log stream to the laptop, then load the driver, then confirm over a fresh login (tools/phone-ssh --fresh), then stop the timer, in that order;
    • the address the way the 6 Pro's package does it: read from the bootloader's device tree at run time, after the controller is powered on, never stored, and kept out of any log you paste here.
  2. What the phone holds today. Our device table records that this phone's slot B was marked and vbmeta_b verification switched off without the project's go, and your rollback relies on stock on slot A. Please make the read-only step record the running slot, both slots' successful, unbootable and retry values, what is on slot A, and the backup check. If slot A isn't stock, that comes first.

  3. The trial's kernel is the published one you built against (26867fc4e826), with uname -r in the result.

Smaller: the first boot with the overlay brings up a new UART before any Bluetooth driver loads, so give it its own Expect (the UART's probe line, USB still up). And the title's prefix should be the path (device/oriole BLUETOOTH-PLAN, overlays: …), since the device package isn't touched.

I did not re-run your schema validation, and I can't see the private plans the text mentions.

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