Lightsleep works, but crashes randomly when wifi is connected.

[frazzmatazz]

I'm experiencing a weird problem where lightsleep on a ESP32 works, but it crashes/reboots randomly. I've distilled down to example code below.

I'm making an alarmclock (with alarms by weekday) and I'd love a working lightsleep for battery operation.

from machine import Pin, lightsleep, SLEEP
#import time
#import ntptime
import esp32
from utime import sleep

wake_pin = Pin(13, Pin.IN, Pin.PULL_UP)
esp32.wake_on_ext0(pin = wake_pin, level = esp32.WAKEUP_ALL_LOW)

count = 0
while True:

   count = count+1
   print(count)
   sleep(0.2)
   lightsleep(1000)

and it crashes/reboots and produces this output usually within 20 to 100 loops, this run went till nearly 1000. The sleep(0.2) duration just seems to change how long i have to wait, not the probability of a crash. The dump on failure is reproducible:

990
991
992
993
994
995
ets Jul 29 2019 12:21:46

rst:0x10 (RTCWDT_RTC_RESET),boot:0x13 (SPI_FAST_FLASH_BOOT)
configsip: 0, SPIWP:0xee
clk_drv:0x00,q_drv:0x00,d_drv:0x00,cs0_drv:0x00,hd_drv:0x00,wp_drv:0x00
mode:DIO, clock div:1
load:0x3fff0030,len:5128
load:0x40078000,len:15884
ho 0 tail 12 room 4
load:0x40080400,len:4
load:0x40080404,len:3696
entry 0x400805c4
Started webrepl in normal mode

I'm using a "WEMOS Lite V1.0.0". It reports this info with esptool:

Chip is ESP32-D0WDQ6-V3 (revision v3.1)
Features: WiFi, BT, Dual Core, 240MHz, VRef calibration in efuse, Coding Scheme None
Crystal is 40MHz

I was getting crashes with both these firmware: ESP32_GENERIC-20241129-v1.24.1.bin and firmware-LOLIN_D32-v1.24.1-1-g50c8864e7.bin. The 2nd one I downloaded while looking for something specific for this chip.

[frazzmatazz]

PS: My boot.py looks like this, and connects to wifi and starts a webrepl.

# This file is executed on every boot (including wake-boot from deepsleep)
#import esp
#esp.osdebug(None)
import webrepl
webrepl.start()

#import os
import network
from utime import sleep
import secrets

# intitialise WIFI
network.country("AU")
sta_if = network.WLAN(network.STA_IF)

sta_if.active(False)
sta_if.active(True)

#sta_if.config(pm=sta_if.PM_NONE) # disable power management, seems to need active(TRUE) first
#sta_if.config(pm=sta_if.PM_PERFORMANCE) # disable power management, seems to need active(TRUE) first
sta_if.config(pm=sta_if.PM_POWERSAVE) # disable power management, seems to need active(TRUE) first

if not sta_if.isconnected():
     loopnum = 0
     try:
         sta_if.active(True)
         sta_if.config(txpower=10)
         sta_if.connect(secrets.SSID, secrets.wifi_password)
         print('Maybe connected now: {}...'.format(sta_if.status()))
         sleep(1)
     except:
         print("well that attempt failed")
         sta_if.disconnect()
         sta_if.active(False)
     loopnum = 0

     while loopnum < 5:
         print('Got status {}...'.format(sta_if.status()))
         if sta_if.status() == 1010: #CONNECTED
             break
         loopnum = loopnum + 1
         sleep(1)

I've just realised my boot.py does 'from utime import sleep' for me! I've disabled it with a exit() at the start and added that import line above.

It seems like having a wifi connection is making lightsleep crash.... I haven't reproduced it yet while wifi is off. Up to 3000 loops so far....

There's another thread here suggesting a lightsleep wifi problem: [(https://github.com/orgs/micropython/discussions/10899)]

[frazzmatazz]

One more update:
It's not pin related at all (though that took me a long time to get working). It crashes while peacefully lightsleeping without a pin interrupt, the short lightsleeps just make it easier to reproduce.

I think the crashes happen (only occasionally) as it exits lightsleep if there's a wifi connection. So when I was doing 60 second lightsleeps the crashes were infrequent enough (like once an hour) that I wasn't sure what was happening.

19
20
21
22
ets Jul 29 2019 12:21:46

rst:0x10 (RTCWDT_RTC_RESET),boot:0x13 (SPI_FAST_FLASH_BOOT)
configsip: 0, SPIWP:0xee
clk_drv:0x00,q_drv:0x00,d_drv:0x00,cs0_drv:0x00,hd_drv:0x00,wp_drv:0x00
mode:DIO, clock div:1
load:0x3fff0030,len:5128
load:0x40078000,len:15884
ho 0 tail 12 room 4
load:0x40080400,len:4
load:0x40080404,len:3696
entry 0x400805c4
Started webrepl in normal mode
Maybe connected now: 1001...

Maybe the obvious answer is get rid of the webrepl and boot.py and only use wifi when I need an NTP update.
The classic, "Doc it hurts when I do this". Dr: "Then don't do that".

[frazzmatazz]

PPS: I have now found references to the hardware docs that say you should turn off the wifi before going into lightsleep.
Not a bug, just a limitation. Good to know!