I2C fail and crashe ATMEGA4809 when i2cdetect linux command is launched

[goodchip]

I've a Raspberry 3 board in I2C master configuration, wired at 3 arduino boards :

• 2 arduino pro mini (as control sample device) at 0x03 and 0x04 slave i2c address
• 1 arduino Uno WiFi rev 2 (as 4809 test device) at 0x11 slave i2c address

NB: the crypto chip on the Uno WiFi Rev 2 has been desolded to remove any potential influence and does not appear in the addresses

The 3 boards have the same demo i2c sketch in his EEPROM :

#include <Wire.h>

void setup() {
  Wire.begin(0x~thisSlaveAddr~);
  Wire.onRequest(reqIRQ);
}

void loop() {
  delay(1000);
}

void reqIRQ() {
  Wire.write(0x01);
}

IMPORTANT : I use the Wire.h library I submitted in pull request to make the sketch working : #22

I use the following command on Pi3 to list the slaves presents on the bus:

i2cdetect -y 0

When I launch this command for the 1st time, it's happening: (normal condition)

     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          03 04 -- -- -- -- -- -- -- -- -- -- -- 
10: -- 11 -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 

But, at the 2nd launch, the 4809 chip disappear:

     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          03 04 -- -- -- -- -- -- -- -- -- -- -- 
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 

And, at the 3th launch, the i2c BUS freeze totally for undetermined time:

     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          -- -- -- -- -- -- -- -- -- -- -- -- -- 
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 

A simple reset of the Uno WiFi rev 2 board, restore the freeze.

Interestingly, the following command does NOT cause the problem:

i2cdetect -r -y 0

The problem only concerns the i2cdetect command, the other master / slave communications seem to work properly.

[facchinm]

@goodchip I'll try to reproduce the issue later this week; are you able to reproduce the issue without a Rpi3 , only using arduinos? (maybe running an i2cscan sketch)

[goodchip]

hi @facchinm, yes, I'm trying to do this this afternoon.

PS: I also have problems of reliability with the I2C in slave mode: after a few hours of operation, the BUS falls, whereas I do not have this problem with the ATMEGA328p, I have the feeling that it could there have a connection to the i2cdetect problem, but I remain cautious for the moment.

[goodchip]

hi @facchinm, I've results today:

I've reproduce the issue only with 1 arduino Nano and my Arduino Uno WiFi Rev2:

15:28:41.889 -> Scanning...
15:28:41.889 -> I2C device found at address 0x08  !
15:28:41.922 -> done
15:28:41.922 -> 
15:28:46.878 -> Scanning...

After the 1st scan, the BUS crashes: SDA is continuous at the level 0V and SCL too.

NB: the ATMEGA328P emulation is off at the compilation.

---

Sketch on Arduino Nano : (MASTER, I2C scanner)

// --------------------------------------
// i2c_scanner
//
// Version 1
//    This program (or code that looks like it)
//    can be found in many places.
//    For example on the Arduino.cc forum.
//    The original author is not know.
// Version 2, Juni 2012, Using Arduino 1.0.1
//     Adapted to be as simple as possible by Arduino.cc user Krodal
// Version 3, Feb 26  2013
//    V3 by louarnold
// Version 4, March 3, 2013, Using Arduino 1.0.3
//    by Arduino.cc user Krodal.
//    Changes by louarnold removed.
//    Scanning addresses changed from 0...127 to 1...119,
//    according to the i2c scanner by Nick Gammon
//    http://www.gammon.com.au/forum/?id=10896
// Version 5, March 28, 2013
//    As version 4, but address scans now to 127.
//    A sensor seems to use address 120.
// Version 6, November 27, 2015.
//    Added waiting for the Leonardo serial communication.
//
//
// This sketch tests the standard 7-bit addresses
// Devices with higher bit address might not be seen properly.
//
 
#include <Wire.h>
 
 
void setup()
{
  Wire.begin();
 
  Serial.begin(9600);
  while (!Serial);             // Leonardo: wait for serial monitor
  Serial.println("\nI2C Scanner");
}
 
 
void loop()
{
  byte error, address;
  int nDevices;
 
  Serial.println("Scanning...");
 
  nDevices = 0;
  for(address = 1; address < 127; address++ )
  {
    // The i2c_scanner uses the return value of
    // the Write.endTransmisstion to see if
    // a device did acknowledge to the address.
    Wire.beginTransmission(address);
    error = Wire.endTransmission();
 
    if (error == 0)
    {
      Serial.print("I2C device found at address 0x");
      if (address<16)
        Serial.print("0");
      Serial.print(address,HEX);
      Serial.println("  !");
 
      nDevices++;
    }
    else if (error==4)
    {
      Serial.print("Unknown error at address 0x");
      if (address<16)
        Serial.print("0");
      Serial.println(address,HEX);
    }    
  }
  if (nDevices == 0)
    Serial.println("No I2C devices found\n");
  else
    Serial.println("done\n");
 
  delay(5000);           // wait 5 seconds for next scan
}

Sketch on Arduino Uno WiFi Rev2: (SLAVE, I2C RECEIVER)

// Wire Slave Receiver
// by Nicholas Zambetti <http://www.zambetti.com>

// Demonstrates use of the Wire library
// Receives data as an I2C/TWI slave device
// Refer to the "Wire Master Writer" example for use with this

// Created 29 March 2006

// This example code is in the public domain.


#include <Wire.h>

void setup() {
  Wire.begin(8);             // join i2c bus with address #8
  Wire.onReceive(receiveEvent); // register event
  Serial.begin(9600);           // start serial for output
}

void loop() {
  delay(100);
}

// function that executes whenever data is received from master
// this function is registered as an event, see setup()
void receiveEvent(int howMany) {
  while (1 < Wire.available()) { // loop through all but the last
    char c = Wire.read(); // receive byte as a character
    Serial.print(c);         // print the character
  }
  int x = Wire.read();    // receive byte as an integer
  Serial.println(x);         // print the integer
}

[facchinm]

Great, thanks., I'll take a look immediately.

[facchinm]

@goodchip I found the problem but I'd love some help to make sure the fix doesn't hurt other stuff 😄

The offending line is

ArduinoCore-megaavr/libraries/Wire/src/utility/twi.c

Line 582 in 0b4b24f

TWI0.SCTRLA &= ~(TWI_APIEN_bm | TWI_PIEN_bm);

Commenting it the scan doesn't stop (it makes sense since the interrupts are not being disabled 🙂 )
I think that in most situations simply removing that line should be ok, but some tests on the field would be great.

[goodchip]

Yes, we'll have to look at the edge effects.
I'm seriously looking at it on Monday.
Have a nice week end @facchinm!

[goodchip]

@facchinm, I tell you about my progress in my investigations:

• the i2cdetect now works perfectly at home, RPi included.
• the deletion of the line has been extremely beneficial for another problem: before, I could not interrogate a 4809 in Slave i2c for more than 30 hours (one interrogation every 200ms): it ended with crash the BUS in the same way as for the i2cdetect. I have now exceeded 50 hours. Wait & See, but it's very promising.

I will continue my tests with several slaves on the BUS to give you a stronger feedback. This week.

Many thanks for your reactivity! ;)

[goodchip]

Test always running fine this morning, totalize now 80 hours ( 1 master RPi, 1 slave 328p not interrogated for the moment, 1 slave 4809 interrogated every 200ms ).

[facchinm]

I think it can be safely merged then, doing it now