Input/output error using Python module SMBus, a Raspberry Pi and an Arduino

Question

I have connected a Raspberry Pi and Rainbowduino together with a homemade I²C level shifter, and installed the Python module SMBus, the Raspberry Pi can communicate with the Rainbowduino, but every so often I get an input/output error message when trying the command bus.write_i2c_block_data(address, signal, data).

It says:

IOError: [Errno 5] Input/output error

Why does it happen and how do I fix or ignore these errors?

Answer 1

Long story short alot of people are plagued by this, I found a very simple work around is the following.

It will let you ignore the error and keep tx/rx-ing, calling i2cdetect seems to reinitialize the bus somehow instead of the arduino disappearing from it.

i posted an explanation of how i found this solution here (waiting mod approval right now) http://www.raspberrypi.org/phpBB3/viewtopic.php?f=41&t=52517

try:
    bus.write_i2c_block_data(address, signal, data)
except IOError:
    subprocess.call(['i2cdetect', '-y', '1'])
    flag = 1     #optional flag to signal your code to resend or something

Even though this allows the Pi to keep transmitting bad data is still being sent to the arduino. The simplest way I found to get around this was to add an extra check-sum byte to the end of my data blocks.

I added up each byte of the message inside a byte variable allowing the value to rollover, then assign the check-sum byte whatever value necessary to sum the whole message out to zero.

The arduino can then check each incoming transmission by summing all the bytes. If the message does not sum out to zero, it is ignored as an erroneous transmission.

I also assigned my messages a one byte message id which increments after each successful transmission, eliminating the possibility of accidental double sends. But that may not really be necessary.

Answer 2

I am creating a buzzing server with a Raspberry Pi and Arduino UNO with i2c and has encountered the same problem. My design is that when the Pi receives a connection request from a socket(by some external machines on the network), it writes '1' to the Arduino and Arduino will enable a loop in loop() by changing a global variable. after the write Pi will continuously read byte from Arduino to check for button state. when the Pi wants to stop reading it sends '0' to stop the loop and reset all counters and LED.

What happens is that Python will through IOError randomly when writing byte. On the Serial monitor with Arduino I noticed that the last byte received is 1 instead of 0, which is what the pi should have sent. Upon looking at i2cdetect -y 1 I noticed the address is wrong and I tried Jon's method, but as user3126397 mentioned, the bad data has been sent and the Arduino has halted. I attempted his modprobe and that only suppressed the error message and the Arduino is still in halt state.

I originally suspect that the data gone sour because of incomplete read/write and therefore added a Serial.println() to check the argument byteCount in onReceive(). Without altering any other code I observed that the no. of successful operations before IOError increased a lot. Therefore I tried to add more println() to test the correlation and noticed a dramatic increase in failure. Finally I commented all the Serial statements and I ended up able to use the server without faults for a considerable number of times(I tested something like 30 times and still no IOError).

I suspect, with regard to user3126397's solution on resetting the baudrate and my observation on the Serial.println() relationship, that error is indeed caused by synchronisation issues between the pi and Arduino(as Serial is relatively slow and causes more delay in the program, thus increases the chance of failure.

Answer 3

Depending on your RPi, you might use bus = SMBus(0) or bus = SMBus(1) to initialize the SMBus.

I hope this solves your problem.

Answer 4

I wrote this 19 hours ago when I thought I had a fix for the IOError problem: .............................................................................

I've been plagued with the same Input/Output error when using bus.write_byte talking to an Arduino. I tried Jon's i2cdetect fix but found that by this time in the code the damage had been done, bad numbers had already gotten to the Arduino and somehow disabled it.

What did work was resetting the i2c baudrate, using

    sudo modprobe -r i2c_bcm2708
    sudo modprobe i2c_bcm2708 baudrate=100010

After this no more I/O errors!! I'd be very interested in theories on what could cause this apparent baudrate mismatch. Hope this helps!

.................................................................................

Well after much more testing what I found was that the baudrate fix did make the rate of errors go down but did not eliminate them.

It now seems clear that Jon's i2cdetect call applied right after the error will initialize the RPi so that it can continue. But you also have to deal with the fact that bad data was probably sent to the Arduino and you need to detect this and fix it, and also reinitialize Wire (and in my case the servo driver), so that even though Pi got an error, that the data will continue to get through and be usable. Hope this helps.

Answer 5

I recently came across this same issue. When I disabled the teensy's serial interface, the errors disappeared entirely.

I'm using an RPi 2 with a teensy 3.2 communicating over i2c at 2.4mhz, sending 33-byte payloads at a rate of about 38 Kbps.