board::esp32_devkitc

Constants

ADC_EXTERNAL_HIGH_GPIO

const ADC_EXTERNAL_HIGH_GPIO: Option<u8>;

A GPIO this board holds at a hard, low-impedance high, or None.

The ADC self-test needs one, and the chip cannot supply it: a pad in analog mode has its digital buffers bypassed, and the internal pull-up is tens of kilohms into the SAR’s sampling capacitor – measured at 4% of full scale rather than 80. So the board has to provide the high, which makes it a manifest fact rather than something the test can assume.

A board without one is not broken; the test skips, and says so.

A bare DevKitC has nothing on GPIO39 (SENSOR_VN), and GPIO34-39 are input-only with no internal pull at all, so the pin floats. Jumper GPIO39 to 3V3 and change this to Some(39) to run the test here.

BOARD

const BOARD: crate::Board;

This board as one value; see [crate::Board].

A bare DevKitC has no onboard IMU or addressable LED, so both are None; it is the reference board for the on-chip loopback self-tests, so selftest carries every free pad.

References: Board

BOARD_NAME

const BOARD_NAME: &str;

DMA_POOL_BYTES

const DMA_POOL_BYTES: usize;

The DMA pool’s size is set by the build’s memory map (tools/build/src/map.rs, radio-keyed against the ROM bound), not by this constant, and the runtime reads the real size from the linker symbols the map emits. This value is a manifest fact only: it mirrors the Wi-Fi-only pool the map places (12 KiB; a radio-bt build gets 8 KiB, the ROM bound being tighter once the controller takes the bottom of DRAM). See #140.

HAS_BT

const HAS_BT: bool;

HAS_WIFI

const HAS_WIFI: bool;

Radios this board’s module physically carries.

A fact about the hardware, in the same way a pin number is — which is why it lives in the manifest rather than being assumed from the SoC family. Every ESP32 part FlintOS supports today has both, but the ESP32-S2 has no Bluetooth at all and would declare HAS_BT = false, so applications that ask for a radio the board has not got fail to build instead of failing to connect.

The kernel checks these against the radio-* features; see kernel/src/radio.rs.

LOOPBACK_AUX_GPIOS

const LOOPBACK_AUX_GPIOS: Option<(u8, u8)>;

Two more electrically-free pads for loopback self-tests that need more than the single [LOOPBACK_SCRATCH_GPIO], as (a, b).

A bus’s init insists on distinct pins, so a folded single-pad loopback still needs spare pads to get through init before the second signal is folded onto the scratch pad: SPI uses both (a clock pad and a placeholder MISO), UART uses one (a placeholder RX). None carries an external wire — each is only ever driven-and-ignored or only read. GPIO23 and GPIO19 are free on a bare DevKitC — the pair apps/tests/spidma has always looped on.

A board that declares None skips those loopback tests.

LOOPBACK_SCRATCH_GPIO

const LOOPBACK_SCRATCH_GPIO: Option<u8>;

A GPIO that is electrically free, for on-chip loopback self-tests (TWAI, I2S) that route a peripheral’s output and input to the same pad through the matrix. Nothing is connected — the loopback is internal — so the pin only has to be one no other peripheral or the board itself drives.

A board that declares None skips those tests. GPIO21 is unbonded to anything on a bare DevKitC (the I2C convention pins carry nothing by default), clear of the strapping and flash pins.

MCPWM_SELFTEST_GPIOS

const MCPWM_SELFTEST_GPIOS: Option<[u8; 3]>;

Three free pads for the MCPWM self-test, as [pwm_a, pwm_b, fault], or None to skip. No wire: the driver routes PWM0A/PWM0B out to the first two pads and taps them back through the matrix into PCNT (edge counting) and the MCPWM capture unit (dead-time timing); the fault pad is driven high in software and routed into the MCPWM fault input to prove the hardware shutdown. GPIO 21/23/19 are the free loopback pads, clear of the strapping and flash pins; reused here because the self-tests run one at a time.

PCNT_LOOPBACK_GPIO

const PCNT_LOOPBACK_GPIO: Option<u8>;

A single electrically-free pad for the PCNT self-test. The test drives it in software and routes it straight back to a PCNT unit’s signal input through the matrix, so — like the other loopback pads — nothing external is connected and the pin only has to be one no other peripheral drives. GPIO22 is free on a bare DevKitC (the I2C convention’s SCL pin, carrying nothing by default) and clear of the strapping and flash pins.

A board that declares None skips the PCNT self-test.

PHY_MAX_TX_POWER_DBM

const PHY_MAX_TX_POWER_DBM: i32;

Maximum radio transmit power, in dBm.

A regulatory and thermal choice, not a chip fact, which is why it is a board manifest entry rather than a constant in the radio crate. It feeds the six transmit-power entries of the PHY’s initialisation table (radio_esp32::phy_init), clamped per modulation exactly as esp-idf’s LIMIT macro does.

20 dBm is esp-idf’s own default, and saturates every band. Lower it for a board that runs hot, has a poor antenna, or ships somewhere with a tighter limit than the part can reach.

PSRAM_FREE_GPIOS

const PSRAM_FREE_GPIOS: [u8; 2];

GPIO16/17 are free on WROOM-32 (no onboard PSRAM to reserve them for), unlike WROVER. Not wired to anything in this manifest; listed for boards/drivers that want to make use of them.

SPI_SLAVE_LOOPBACK_GPIOS

const SPI_SLAVE_LOOPBACK_GPIOS: Option<[u8; 4]>;

Four electrically-free pads for the on-chip SPI master↔slave loopback, as [sck, mosi, miso, cs], or None.

The self-test drives SPI2 as master and SPI3 as slave and joins their signals through the GPIO matrix with no external wire: master SCK-out → slave SCK-in on the first pad, master MOSI-out → slave MOSI-in on the second, slave MISO-out → master MISO-in on the third, and a master-driven CS → slave CS-in on the fourth. The real CS edge frames each transaction — the slave commits its received data when CS deasserts.

Separate from [LOOPBACK_SCRATCH_GPIO]/[LOOPBACK_AUX_GPIOS] so the single-pad folded loopbacks and this two-controller one cannot fight over what a pad means. The pads are the free GPIOs 21/23/19 plus a PSRAM-free 16 — nothing is wired to them and the tests never run at once.

A board that declares None skips the test.

TICK_PERIOD_US

const TICK_PERIOD_US: u32;

TOUCH_SELFTEST_GPIO

const TOUCH_SELFTEST_GPIO: Option<u8>;

A free touch-capable pad for the touch-sensor self-test, or None to skip.

Only ten pins can sense touch (T0–T9 = GPIO 4/0/2/15/13/12/14/27/32/33), so unlike the other loopback pads this cannot reuse the general free GPIOs. The test does not need a wire — it reads the pad’s own parasitic capacitance to prove the controller measures — only a touch-capable pin no other peripheral drives. GPIO27 is T7, is not a strapping or flash pin, and carries nothing on a bare DevKitC.