Files
wadamesh/variants/thinknode_m9/M9Keyboard.cpp
T
Christopher Van HooseandClaude Fable 5 68e3b45932 M9 + V4-R8: 2026-08-19 audit passes
ThinkNode M9 (see M9_PORT.md audit section): restore nav-mode HW Back,
Lua-app key forwarding + d-pad support, popup-registry gates widened,
SD-backed data store + chat history, ADC2 battery sample filter,
keyboard poll throttle + backlight, map tile cache to SD, spectrum
sweep LR1110 fix, power-menu cleanup for buttonless hardware.

Heltec V4 R8 (see R8_AUDIT.md): chat history/battery log follow SD
adoption, reinsert watch no longer unmounts live store, tile cache
prefers SD, CAP_ROTATABLE reboot trap defused, SD CS parked at boot.

Shared: atomic threads-index write (tmp+rename), recursive
scroll-into-view during keypad nav, textarea focus highlight.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-20 07:25:18 -04:00

116 lines
4.3 KiB
C++

// ThinkNode M9 keyboard driver — see M9Keyboard.h.
#if defined(HAS_M9_KEYBOARD) && defined(ESP32)
#include "M9Keyboard.h"
#include <Arduino.h>
#include <Wire.h>
#ifndef PIN_KB_ADDR
#define PIN_KB_ADDR 0x6c
#endif
// Controller register map — from Elecrow's keyboard-controller firmware
// (ThinkNode-M9-KB-platformio, src/main.cpp). The keyboard is a separate
// ESP32-S2 running a matrix scanner as an I2C SLAVE with addressed registers:
// reads MUST be preceded by a 1-byte register-address write, otherwise the
// slave answers with whatever register was addressed last (0x00 = HW version
// after its reset — a constant 0x03, which is exactly why the bare-read
// protocol produced "no keys do anything").
#define M9_KB_REG_HW_VER 0x00 // constant 0x03 on this controller FW
#define M9_KB_REG_KEY 0x01 // pending key byte, 0x00 = none (single slot)
#define M9_KB_REG_BACKLIGHT 0x02 // PWM duty 0..255 (controller lights on keypress)
#define M9_KB_REG_FW_VER 0xFE // constant 0x10 on this controller FW
static volatile uint8_t s_ring[16];
static volatile uint8_t s_head = 0;
static volatile uint8_t s_tail = 0;
static bool s_inited = false;
static TwoWire* s_bus = nullptr; // set once the controller answers
static uint32_t s_last_probe_ms = 0;
static bool kbReadReg(TwoWire& w, uint8_t reg, uint8_t* out) {
w.beginTransmission((uint8_t)PIN_KB_ADDR);
w.write(reg);
if (w.endTransmission() != 0) return false; // NACK / bus error
if (w.requestFrom((int)PIN_KB_ADDR, 1) != 1) return false;
*out = (uint8_t)w.read();
return true;
}
static bool kbProbe(TwoWire& w, const char* name) {
uint8_t hw = 0, fw = 0;
if (!kbReadReg(w, M9_KB_REG_HW_VER, &hw)) return false;
kbReadReg(w, M9_KB_REG_FW_VER, &fw); // best-effort, for the log only
Serial.printf("M9 keyboard: controller on %s (hw=0x%02X fw=0x%02X)\n", name, hw, fw);
return true;
}
// The schematic reading puts the keyboard on its own bus (host 20/21); the
// shared sensor bus (Wire, 7/6) is probed as a fallback so a wrong pin
// reading shows up in the boot log instead of as a dead keyboard.
static void kbTryFindBus() {
if (kbProbe(Wire1, "Wire1 (20/21)")) s_bus = &Wire1;
else if (kbProbe(Wire, "Wire (7/6)")) s_bus = &Wire;
}
void m9KeyboardBegin() {
Wire1.begin(PIN_KB_SDA, PIN_KB_SCL);
// The controller is its own MCU on the switched peripheral rail and may
// still be booting here — poll re-probes once a second until it answers.
kbTryFindBus();
if (!s_bus) Serial.println("M9 keyboard: no controller yet (will keep probing)");
s_inited = true;
}
void m9KeyboardPoll() {
if (!s_inited) return;
// Rate-limit: the UI loop free-runs, and unthrottled this was a blocking
// ~0.5 ms write+read on the 100 kHz bus EVERY iteration (hundreds/s) — pure
// waste on the UI thread. The controller latches one key, so the poll only
// has to outpace keystrokes: 15 ms (~66/s) is ample for burst typing and
// removes ~97% of the bus traffic. Wraparound-safe; first call passes.
{
static uint32_t s_next_poll_ms = 0;
const uint32_t now_ms = millis();
if ((int32_t)(now_ms - s_next_poll_ms) < 0) return;
s_next_poll_ms = now_ms + 15;
}
if (!s_bus) {
uint32_t now = millis();
if (now - s_last_probe_ms < 1000) return;
s_last_probe_ms = now;
kbTryFindBus();
if (!s_bus) return;
}
// Read the latched key; when one IS pending, drain a couple more reads in
// the same poll (bounded) — the controller holds a single slot, so a second
// key struck inside the 15 ms window would otherwise be lost.
for (int i = 0; i < 3; i++) {
uint8_t key = 0;
if (!kbReadReg(*s_bus, M9_KB_REG_KEY, &key)) return;
if (key == 0x00 || key == 0xFF) return; // no key / undefined-register reply
const uint8_t nh = (uint8_t)((s_head + 1) & 15);
if (nh != s_tail) { // drop if the ring is full
s_ring[s_head] = key;
s_head = nh;
}
}
}
int m9KeyboardReadKey() {
if (s_tail == s_head) return 0; // empty
const uint8_t key = s_ring[s_tail];
s_tail = (uint8_t)((s_tail + 1) & 15);
return key;
}
void m9KeyboardSetBacklight(uint8_t duty) {
if (!s_inited || !s_bus) return;
s_bus->beginTransmission((uint8_t)PIN_KB_ADDR);
s_bus->write(M9_KB_REG_BACKLIGHT);
s_bus->write(duty);
s_bus->endTransmission();
}
#endif