PicoW [Errno 113] ECONNABORTED <class 'OSError'>

[gsexton]

CircuitPython version and board name

Adafruit Circuit Python 9.2.8 dated 07/02/2025 on PicoW

Code/REPL

I have a wireshark dump I can upload to a private location if you would like. I'm pretty sure the issue is in the network stack, and not the code.

Behavior

[07/02/2025 21:18:22] ERROR InfluxWriter._send_request(body=tempmon,sensor=otext temp=68.30) Exception: [Errno 113] ECONNABORTED <class 'OSError'>

Description

So I have a device that I'm using. It has a LCD display, and a TMP102 thermometer.

It continuously polls the thermometer and once per minute makes an HTTPS POST request to place the data on a server.

I frequently get:

[Errno 113] ECONNABORTED <class 'OSError'>
[Errno 113] ECONNABORTED <class 'OSError'> socket_resolve_host() returned -2
[Errno 104] ECONNRESET <class 'OSError'>

For example, 20:12:39 failed, 20:13,and 20:14 succeeded, and 20:15:40 failed.

This was happening with 9.2.6 as well. Also, the issue happens in two different environments with quite different AP's, networks, etc.

I don't have the error message, but at one point I had a crash message where the kernel itself crashed, and the pico was locked up.

I have another one running an ancient micropython version, and it just works. Is there any hope for this ever working?

Additional information

When it starts failing, I see a lot of SYN, SYN/ACK retransmits between the two hosts. I'm also seeing ARP requests for the Pico's address from the router, but there aren't any corresponding replies.

I can provide you with the tcpdump capture file if you would like. It has both successful, and failing connection cycles.

[dhalbert]

Could you try 10.0.0-alpha.7? Also make sure all the CircuitPython libraries are up to date.

A wireshark trace may be helpful, but a small test program that we can run to reproduce the may be the most helpful. Are you talking to the same server in both network environments? Is it a local server?

We use the same library to talk to the CYW43 co-processor that MicroPython uses. In fact, it was developed by them: https://github.com/georgerobotics/cyw43-driver. We may be ahead or behind of the library version compared with your MicroPython instance. Maybe check the open and closed issues there to see if anything sounds similar to your problem.

[gsexton]

You can grab the sanitized pcap file from:

https://www.mhsoftware.com/~gsexton/sanitized.pcap

[gsexton]

My apologies. I went through and updated all of the libraries, try 10 alpha 7, and it seems to be working as expected now.

[gsexton]

I'm re-opening this because it's still present. I'm using 10.0.0 beta 2. I've updated all python code that's running from the source repos, and re-compiled using the current mpy-cross command line utility. I routinely get:

Running in safe mode! Not running saved code.

You are in safe mode because:
CircuitPython core code crashed hard. Whoops!
Hard fault: memory access or instruction error.
Please file an issue with your program at github.com/adafruit/circuitpython/issues.
Press reset to exit safe mode.

Press any key to enter the REPL. Use CTRL-D to reload.

My program is reasonably large, and this error happens after 8-10 hours of operation. If you can give me an idea on how to localize the location more, I'll try that.

Addtionally, the connection issues persist:

[07/31/2025 04:20:58] ERROR InfluxWriter._send_request(body=tempmon,sensor=otext temp=69.83) Exception: [Errno 113] ECONNABORTED <class 'OSError'>
[07/31/2025 04:23:00] ERROR InfluxWriter._send_request(body=tempmon,sensor=otext temp=69.89) Exception: [Errno 113] ECONNABORTED <class 'OSError'>
[07/31/2025 04:25:00] ERROR InfluxWriter._send_request(body=tempmon,sensor=otext temp=69.80) Exception: [Errno 113] ECONNABORTED <class 'OSError'>

What's really sort of interesting is that every other attempt fails. Looking above, you can see 4:21:00, and 4:24 aren't represented. This is consistently seen.

It would be really nice if this could be made to work.

[dhalbert]

Is the server local or out on the Internet somewhere? I think the first thing to do would be to try to reproduce this with the simplest possible program, at a higher speed so it fails faster. That could be done POSTing to a local server so you don't annoy your IoT server.

[gsexton]

It's a local server. I wrote a test case that makes 10 HTTPS POST requests per second, and right now it's ticking past 9,000 with no errors.

The application has two other components. It has a ST7789 display that it writes data to, and a TMP102 I2C sensor that it reads once per second.

@dhalbert Is there any way that I can debug where in my main code set the hard fault is happening? I have a Pico debugger if that's useful.

CircuitPython core code crashed hard. Whoops!
Hard fault: memory access or instruction error.

[dhalbert]

You could set a breakpoint on reset_into_safe_mode(), and then look at the stack backtrace. It would be good to compile in debug mode so you can get more info. But sometimes that eliminates a bug because the code or timing is different.

I would say use your original program, but just speed up the POSTing to be every time you read the temperature sensor. If you can then get it to fail, then comment out the display code and see what happens, and if that doesn't cause the problem to go away, then uncomment the display code and comment out the TMP102 reading. The idea is to eliminate causes if possible.

[gsexton]

@dhalbert

I did get a hard-crash using a simplified program:

Request counter=15110
Exception: Repeated socket failures
Request counter=15120
...
Request counter=20870
Exception: [Errno 116] ETIMEDOUT
Request counter=20880
...
Request counter=21260
Request counter=21270
                                 Auto-reload is off.
Running in safe mode! Not running saved code.

You are in safe mode because:
CircuitPython core code crashed hard. Whoops!
Hard fault: memory access or instruction error.
Please file an issue with your program at github.com/adafruit/circuitpython/issues.
Press reset to exit safe mode.

Press any key to enter the REPL. Use CTRL-D to reload.

with this code:

import adafruit_requests
import json
import socketpool
import ssl
import time
import wifi
import gc

ssid="myysssi"
psk="mypasss"
while True:
    try:
        res = wifi.radio.connect(ssid, psk)
        print("Connect result=", res)
        if wifi.radio.connected:
            print("Connected to wifi!")
            break
        print("Trying connection again.")
        time.sleep(1)
    except Exception as e:
        print(e)
        time.sleep(1)

pool = socketpool.SocketPool(wifi.radio)
ctx = ssl.create_default_context()
ctx.check_hostname=False
requests = adafruit_requests.Session(pool, ctx)

baseline = gc.mem_alloc()
print(f"gc.mem_alloc()={gc.mem_alloc()} gc.mem_free() = {gc.mem_free()}")
with open('/godaddy.pem') as f:
    certs=f.read()
ctx.load_verify_locations(cadata=certs)

gc.collect()
print(f"Initial Post Cert Load: gc.mem_alloc()={gc.mem_alloc()} gc.mem_free() = {gc.mem_free()}")
data={"SomeValue": "SomeString", "FloatValue": 0.2, "IntValue": 42, "Slice": ["a","b","c"]}
counter = 0
while True:
    try:
        response = requests.post("https://somehost.somedomain.com/post_test.php", json=data, timeout=10)
        rdata = response.json()
        response.close()
        del response
        time.sleep(0.1)
        counter+=1
        if counter%10==0:
            print(f"Request {counter=}")
    except Exception as e:
        print(f"Exception: {str(e)}")
        time.sleep(1)

[gsexton]

The content of post_test.php is:

<?php

ob_implicit_flush(1);
function TestPost() {
    $wrValue = array("SomeField" => "SomeValue","OtherField" => 42);
    echo json_encode($wrValue);
    flush();
    
}
TestPost();
?>