zephyr-cp: enable BLE on the Raspberry Pi Pico W

[tyeth]

Stacked on #4 (Pico 2 W BLE). Rebased onto main 121489fe70 on 2026-09-08 — the RP2040 boot fix this once needed from #13 is on main (ffd62e1c59), so the image boots on its own.

Same CYW43439 as the Pico 2 W, and the shared-bus HCI driver needed no board-specific code. Only per-board devicetree/config changes were required:

• Overlay: hang the infineon,cyw43-bt-hci node off the stock infineon,airoc-wifi node and point zephyr,bt-hci at it.
• Overlay: the settings partition is too small for NVS. It was originally 2 KB at 0x180800, neither erase-sector aligned nor large enough for one of the RP2040's 4 KB sectors; main has since made it one aligned 4 KB sector at 0x17f000, which is still one short of the two nvs_mount() insists on (-EINVAL), so bt_enable() fails before opening the HCI driver. storage becomes 8 KB at 0x17e000, taking the extra sector from the code partition. nvm (0x180000) and circuitpy (0x181000) stay where ports/raspberrypi has them, as cptools/check_partitions.py requires, so CIRCUITPY is not reformatted.
• Conf: BT_EXT_ADV=n (this controller rejects the Bluetooth 5 extended commands — see zephyr-cp: BT_EXT_ADV defaults on for every board, asserting a controller capability that is not universal #8) plus a deeper system workqueue stack.

HW_STACK_PROTECTION is deliberately left off here. It earned its place on the Cortex-M33, where it turned several silent corruptions into clean named faults, but it costs RAM and this is the constrained board (and the RP2040 has no MPU anyway).

Verified on hardware

=== BLE SCAN TEST ===
adapter address: <Address 2c:cf:67:d8:d4:a5>
adapter name: CIRCUITPYd4a5
scanning 20s ...
     113ms  rssi= -83  <Address 54:32:04:0c:d7:56>  HUB-D754
  >>> FOUND TARGET: HUB-D754 ... rssi -80 / -83 / -88 / -91  (sustained)

First advertiser at 113 ms, then continuous tracking. The board's WiFi MAC is 2C:CF:67:D8:D4:A3 and the BLE adapter comes up as ...:D8:D4:A5 — MAC + 2, this part's OTP offset, derived through the CYW43439 shared-bus path (the Pico 2 W's is MAC + 1).

That run used the pre-rebase layout (settings at 0x181000). The 0x17e000 placement is build-tested only so far.

Size (post-rebase head)

BOOT_FLASH:       256 B /       256 B  100.00%
FLASH:      1,181,372 B / 1,564,416 B   75.52%
RAM:          228,044 B /      264 KB   84.36%   (~41 KB free)

BLE costs about 26 KB of RAM here. Tight but it fits with no stack trimming.

CI cannot link this until tyeth/zephyr#1 lands and the manifest points at a revision with the driver — see #16.

🤖 Generated with Claude Code

[tyeth]

⚠️ CI board builds fail to link on this branch — expected, tracked as #16.

undefined reference to __device_dts_ord_8…`` from hci_core.c: the DT node and `CONFIG_BT` are present, but the CYW43 HCI driver that defines that device is not in the zephyr revision the manifest pins (`adafruit/zephyr @ 62e7a3764`). Needs tyeth/zephyr#1 to land and the manifest bumped. Builds fine locally against the fork branch carrying the driver, and `ci/pico2w-ble-assets` is green only because of the CI-only override (#12).

Not the same failure as #11, which is the tests / zephyr native_sim job.

🤖 Generated with Claude Code

[tyeth]

Rebased onto main 121489fe70 via the rebased #4 (force-pushed with --force-with-lease). New head 8d2023c6b5, previously 8ac1521d25. One commit, as before.

What moved. main d0f27bfc97 relocated the files this PR edits:

was	now
boards/rpi_pico_rp2040_w.overlay	boards/raspberrypi/rpi_pico_w_zephyr/board.overlay
boards/rpi_pico_rp2040_w.conf	boards/raspberrypi/rpi_pico_w_zephyr/board.conf

Git followed the renames; board.conf and autogen_board_info.toml applied cleanly. board.overlay had the same two conflicts as #4 (#include "../../../app.overlay", and the partition layout) and got the same resolution: storage is 8K at 0x17e000, taken from the code partition (now 0x100..0x17e000); nvm stays at 0x180000 and circuitpy at 0x181000, which main's new check_partitions.py parity check requires for boards with a counterpart. So, unlike the original commit, CIRCUITPY is not reformatted. Commit message reworded accordingly.

ranges; was not re-added by this branch — main already has it (ffd62e1c59), which also means the "also needs #13" dependency in the description is gone; #13 is closed as redundant and this branch boots on its own. The BOOT_FLASH region below shows boot2 is linked.

Build on the new head (make BOARD=raspberrypi_rpi_pico_w_zephyr, local zephyr tree with the driver):

	value
BOOT_FLASH	256 B / 256 B
FLASH	1,181,372 B / 1,564,416 B = 75.52%
RAM	228,044 B / 264 KB = 84.36%
CONFIG_BT / CONFIG_BT_CYW43_SHARED_BUS	y / y
CONFIG_BT_EXT_ADV	not set
CONFIG_HW_STACK_PROTECTION	absent (deliberate, and RP2040 has no MPU)
DT_CHOSEN_zephyr_bt_hci	.../pio0_spi0/airoc_wifi_0/cyw43_bt_hci
check_partitions.py	ok (5 partitions), matches ports/raspberrypi/raspberry_pi_pico_w

RAM is essentially unchanged from the description (84.32% → 84.36%); flash is up from 68.67% because the code partition is 8K smaller and main grew. Still ~41 KB of RAM to spare.

Caveats: the hardware run in the description used the earlier 0x181000 settings placement; 0x17e000 is build-tested only. CI will still fail to link here (#16): the branch now inherits main's west.yml pin (adafruit/zephyr @ 52dc937c7c), which predates the HCI driver.

🤖 Generated with Claude Code