I2C write failed

[stc1988]

Build environment: macOS
Moddable SDK version: 8.0.0
Target device: M5StackCoreS3, M5atom matrix

Steps to Reproduce

1. Connect an I2C sensor, such as an SHT3x or PAJ7620U2, to the Grove port.
2. Build and install the example app using this build command: mcconfig -d -m ...
3. See error

/Users/satoshi/Projects/moddable/modules/io/i2c/esp32/_i2c.c (282) # Break: Error: write failed (in write)!

Other information

Here is the log with instrumented build.

E (417) i2c.common: I2C bus id(0) has already been acquired
E (417) i2c.common: acquire bus failed
E (417) i2c.master: i2c_new_master_bus(1050): I2C bus acquire failed
E (421) i2c.common: i2c_release_bus_handle(250): Bus not freed entirely
XS abort: unhandled exception activate failed

I haven’t tested this on every device, but it does not occur on M5Stack devices and does occur on other devices.

[phoddie]

Is it possible you are mixing the pins/i2c with the ECMA-419 I²C on the devices that fail? Maybe ESP-IDF is detecting conflicts there – since both implementations assume they own the bus.

[stc1988]

I deteled the lines that using pins/i2c, but the situation remained unchanged.

[phoddie]

Thank you for performing the test. I did a similar test using M5 Atom Matrix and see different results. After commenting out the three lines you suggest, there is no longer an activation failure.

Here's what I did:

• Used helloworld
• Included "$(MODDABLE)/modules/io/manifest.json" in the helloworld manifest.
• Added the following to main.js (I don't have a Grove sensor, so this just pretends).

const dummy = new device.io.I2C({
	...device.I2C.default,
	address: 12,
	hz: 100_000
});
dummy.read(1);

With the MPU6886 enabled, the exception is activate failed – you see the same; when disabled, the exception is read failed, after the activation, because I have no I²C peripheral attached, but activation succeeded.

[stc1988]

Thank you for checking. I will try it in my environment as well.

If mixing the pins/i2c with the ECMA-419 I²C on the devices is the problem, does that mean we need to migrate away from pins/i2c in order to use ECMA-419 I²C?

[stc1988]

With the MPU6886 enabled, the exception is activate failed – you see the same; when disabled, the exception is read failed, after the activation, because I have no I²C peripheral attached, but activation succeeded.

I got the same result in my environment as well.

To clarify, I did an instrumented build again with the MPU6886 disabled, and in that case it was not activate failed; the exception was write failed.

Therefore, there may be issues other than mixing the pins/i2c with the ECMA-419 I²C on the devices.

[phoddie]

There is a conflict. That's resolved by the attached patch.

i2c.diff

From 2a6c1245135f9ab92bae3c9dbbd3dc688f84f1fa Mon Sep 17 00:00:00 2001
From: Peter Hoddie <peter@moddable.com>
Date: Mon, 20 Apr 2026 11:42:56 -0700
Subject: [PATCH] =?UTF-8?q?avoid=20conflict=20between=20pins=20&=20419=20I?=
 =?UTF-8?q?=C2=B2C=20phoddie/node-red-mcu#134=20#1597?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 modules/io/i2c/esp32/_i2c.c     | 24 +++++++++++++++++-------
 modules/pins/i2c/esp32/modI2C.c |  8 ++++++--
 2 files changed, 23 insertions(+), 9 deletions(-)

diff --git a/modules/io/i2c/esp32/_i2c.c b/modules/io/i2c/esp32/_i2c.c
index 7508e5968..7f18ad58d 100644
--- a/modules/io/i2c/esp32/_i2c.c
+++ b/modules/io/i2c/esp32/_i2c.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019-2025 Moddable Tech, Inc.
+ * Copyright (c) 2019-2026 Moddable Tech, Inc.
  *
  *   This file is part of the Moddable SDK Runtime.
  *
@@ -56,7 +56,8 @@ static I2C gI2CActive;
 static i2c_master_bus_handle_t 	gBus;
 static i2c_master_dev_handle_t	gDevice;
 
-static uint8_t i2cActivate(I2C i2c);
+__attribute__((weak)) void modI2CUninit(void *);	// pins I2C
+uint8_t i2cActivate(I2C i2c);
 static uint8_t usingPins(uint32_t data, uint32_t clock);
 
 static void _xs_i2c_mark(xsMachine* the, void* it, xsMarkRoot markRoot);
@@ -81,7 +82,7 @@ void _xs_i2c_constructor(xsMachine *the)
 	I2C i2c;
 	int data, clock, hz, address;
 	int timeout = 32000;		// 400 ms (same as default I2C_SLAVE_TIMEOUT_DEFAULT)
-	uint8_t pullup = GPIO_PULLUP_ENABLE;
+	uint8_t pullup = 1;
 	int port = I2C_NUM_0;
 
 	if (NULL == gI2CMutex)
@@ -117,7 +118,10 @@ void _xs_i2c_constructor(xsMachine *the)
 			xsRangeError("duplicate address");
 	}
 
-	xsmcGet(xsVar(0), xsArg(0), xsID_hz);
+	if (!GPIO_IS_VALID_OUTPUT_GPIO(clock) || !GPIO_IS_VALID_OUTPUT_GPIO(data))
+		xsRangeError("unusable pins");
+
+xsmcGet(xsVar(0), xsArg(0), xsID_hz);
 	hz = xsmcToInteger(xsVar(0));
 	if ((hz <= 0) || (hz > 20000000))
 		xsRangeError("invalid hz");
@@ -129,7 +133,7 @@ void _xs_i2c_constructor(xsMachine *the)
 
 	if (xsmcHas(xsArg(0), xsID_pullup)) {
 		xsmcGet(xsVar(0), xsArg(0), xsID_pullup);
-		pullup = xsmcToBoolean(xsVar(0)) ? GPIO_PULLUP_ENABLE : GPIO_PULLUP_DISABLE;
+		pullup = xsmcToBoolean(xsVar(0));
 	}
 
 	if (xsmcHas(xsArg(0), xsID_port)) {
@@ -318,13 +322,16 @@ void _xs_i2c_writeRead(xsMachine *the)
 
 uint8_t i2cActivate(I2C i2c)
 {
+	if (C_NULL == gI2CMutex)		//@@ this check only needed for "make pins release bus" case below.
+		return 1;
+
 	xSemaphoreTake(gI2CMutex, portMAX_DELAY);
 
 	if (gI2CActive) {
 		if (i2c == gI2CActive)
 			return 1;
 
-			i2c_master_bus_rm_device(gDevice);
+		i2c_master_bus_rm_device(gDevice);
 		i2c_del_master_bus(gBus);
 		gBus = NULL;
 		gDevice = NULL;
@@ -332,12 +339,15 @@ uint8_t i2cActivate(I2C i2c)
 	}
 
 	if (i2c) {
+		if (modI2CUninit)
+			modI2CUninit(C_NULL);		// make pins release bus
+
 		i2c_master_bus_config_t busC = {
 			.i2c_port = i2c->port,
 			.sda_io_num = i2c->data,
 			.scl_io_num = i2c->clock,
 			.clk_source = I2C_CLK_SRC_DEFAULT,
-	      .glitch_ignore_cnt = 7,
+			.glitch_ignore_cnt = 7,
 			.flags.enable_internal_pullup = i2c->pullup,
 		};
 		if (ESP_OK != i2c_new_master_bus(&busC, &gBus)) {
diff --git a/modules/pins/i2c/esp32/modI2C.c b/modules/pins/i2c/esp32/modI2C.c
index 99c18d4a1..a7607134f 100644
--- a/modules/pins/i2c/esp32/modI2C.c
+++ b/modules/pins/i2c/esp32/modI2C.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2025 Moddable Tech, Inc.
+ * Copyright (c) 2016-2026 Moddable Tech, Inc.
  *
  *   This file is part of the Moddable SDK Runtime.
  * 
@@ -30,6 +30,7 @@
 
 // N.B. Cannot save pointer to modI2CConfiguration as it is allowed to move (stored in relocatable block)
 
+__attribute__((weak)) uint8_t i2cActivate(void *);	// ECMA-419 I2C
 static uint8_t modI2CActivate(modI2CConfiguration config);
 
 static uint32_t gHz;
@@ -80,6 +81,9 @@ uint8_t modI2CActivate(modI2CConfiguration config)
 {
 	int data, clock, hz;
 
+	if (i2cActivate)
+		i2cActivate(C_NULL);		// make ECMA-419 release bus
+
 #if defined(MODDEF_I2C_SDA_PIN) && defined(MODDEF_I2C_SCL_PIN)
 	data = (-1 == config->sda) ? MODDEF_I2C_SDA_PIN : config->sda;
 	clock = (-1 == config->scl) ? MODDEF_I2C_SCL_PIN : config->scl;
@@ -116,7 +120,7 @@ uint8_t modI2CActivate(modI2CConfiguration config)
 #else
 		busC.flags.enable_internal_pullup = 0;
 #endif
-		if (ESP_OK != i2c_new_master_bus(&busC, &gBus))
+	if (ESP_OK != i2c_new_master_bus(&busC, &gBus))
 			return 1;
 		
 		gClock = clock;
-- 
2.46.0

I don't have a way to reproduce the write failure. I did check for an inexpensive Grove module that uses I²C on Amazon, but no luck. Suggestions?

[stc1988]

This patch works fine!