Problems with WLAN connectivity - solved

rkorell

Dear @schlomm ,
I initially had no clue at all regarding root cause :-)
And the finding “undervoltage” was never expected but came out off my logfiles.

After a LOT of tinkering and playing with syptomatic “solutions” system kept to be unstable so I decided to dig in and do some logging to identify root cause.

For this I wrote a shellscript and installed a system service which collects this data every five minutes.

shellscript:

sudo nano /usr/local/bin/wlan-diagnose.sh

content:

#!/bin/bash
LOGFILE="/var/log/wlan-diagnose.log"
DATE=$(date '+%a %d %b %H:%M:%S %Z %Y')
WLAN_IF="wlan0"

echo "===== $DATE =====" >> $LOGFILE

# IP-Adresse
echo "--- IP-Adresse ---" >> $LOGFILE
ip addr show $WLAN_IF >> $LOGFILE 2>&1

# Link-Status
echo "--- Link Status ---" >> $LOGFILE
iw dev $WLAN_IF link >> $LOGFILE 2>&1

# Default Route
echo "--- Routing ---" >> $LOGFILE
ip route >> $LOGFILE 2>&1

# Wpa_supplicant Status
echo "--- wpa_supplicant ---" >> $LOGFILE
systemctl status wpa_supplicant --no-pager >> $LOGFILE 2>&1

# Letzte wpa_supplicant Logs
echo "--- wpa_supplicant journal (letzte 20 Zeilen) ---" >> $LOGFILE
journalctl -u wpa_supplicant -n 20 --no-pager >> $LOGFILE 2>&1

# Kernel/Treiber Logs
echo "--- dmesg wlan0 ---" >> $LOGFILE
dmesg | tail -n 20 >> $LOGFILE 2>&1

# Ping-Test
PING_TARGET="8.8.8.8"
ping -I $WLAN_IF -c3 -W3 $PING_TARGET >> $LOGFILE 2>&1

echo "" >> $LOGFILE

set as executable:

sudo chmod +x /usr/local/bin/wlan-diagnose.sh

systemd-timer for this diagnosis script:

sudo nano /etc/systemd/system/wlan-diagnose.timer

content:

[Unit]
Description=WLAN Diagnose alle 5 Minuten

[Timer]
OnBootSec=1min
OnUnitActiveSec=5min
Persistent=true

[Install]
WantedBy=timers.target

service file:

sudo nano /etc/systemd/system/wlan-diagnose.service

content:

[Unit]
Description=WLAN Diagnose Service

[Service]
Type=oneshot
ExecStart=/usr/local/bin/wlan-diagnose.sh

activate the service:

sudo systemctl daemon-reload
sudo systemctl enable --now wlan-diagnose.timer

Created logfile: /var/log/wlan-diagnose.log

possible command for filtering for errors:

grep -i "fail\|error\|disconnect" /var/log/wlan-diagnose.log

in my personal case directly after starting the service the undervoltage warnings appeared in the logfile:

Sep 24 19:23:02 MagicMirrorPi5 wpa_supplicant[702]: wlan0: CTRL-EVENT-CONNECTED - Connection to f8:bc:0e:51:50:48 completed [id=0 id_str=] Sep 24 19:23:02 MagicMirrorPi5 wpa_supplicant[702]: bgscan simple: Failed to enable signal strength monitoring --- dmesg wlan0 --- [ 385.672898] hwmon hwmon4: Voltage normalised [ 399.780700] hwmon hwmon4: Undervoltage detected! [ 401.796721] hwmon hwmon4: Voltage normalised [ 403.812728] hwmon hwmon4: Undervoltage detected! [ 405.831888] hwmon hwmon4: Voltage normalised [ 425.988994] hwmon hwmon4: Undervoltage detected! [ 428.008109] hwmon hwmon4: Voltage normalised [ 434.052979] hwmon hwmon4: Undervoltage detected! [ 438.087587] hwmon hwmon4: Voltage normalised [ 442.117090] hwmon hwmon4: Undervoltage detected! [ 444.133104] hwmon hwmon4: Voltage normalised [ 452.198182] hwmon hwmon4: Undervoltage detected! [ 454.213171] hwmon hwmon4: Voltage normalised [ 470.341318] hwmon hwmon4: Undervoltage detected! [ 478.405369] hwmon hwmon4: Voltage normalised [ 488.485467] hwmon hwmon4: Undervoltage detected! [ 490.505469] hwmon hwmon4: Voltage normalised [ 514.693689] hwmon hwmon4: Undervoltage detected! [ 516.709733] hwmon hwmon4: Voltage normalised [ 520.744884] hwmon hwmon4: Undervoltage detected! PING 8.8.8.8 (8.8.8.8) from 172.23.56.157 wlan0: 56(84) bytes of data. 64 bytes from 8.8.8.8: icmp_seq=1 ttl=116 time=13.2 ms 64 bytes from 8.8.8.8: icmp_seq=2 ttl=116 time=33.6 ms 64 bytes from 8.8.8.8: icmp_seq=3 ttl=116 time=27.5 ms --- 8.8.8.8 ping statistics --- 3 packets transmitted, 3 received, 0% packet loss, time 2003ms rtt min/avg/max/mdev = 13.220/24.756/33.576/8.529 ms

So I had identified my root cause with first strike.

In the meantime (today) I had severe additional problems (also “identified” by this mentioned log) - but this was a kernel/device driver problem which I cannot solve today.
But this leads to a modified recovery script because the version from yesterday only tried to restart the WPA_Supplicant which was not sufficient for my problem today.

[EDIT - Sep, 8th, 2025: deleted old recovery script because usage of ping without qualified path produced an error by the script itself. For this reason the script is not as useful as I thought. Sorry for confusion! ]

Hope this helps you.
Do not hesitate to ask for further information …

Warmest regards,
Ralf

rkorell

addition:
the recovery script is: /usr/local/bin/wlan-recovery.sh

set executable:

sudo chmod +x /usr/local/bin/wlan-recovery.sh

Systemd-Service

sudo nano /etc/systemd/system/wlan-recovery.service

content:

[Unit]
Description=WLAN Recovery Script
After=network.target

[Service]
Type=oneshot
ExecStart=/usr/local/bin/wlan-recovery.sh

timer:

sudo nano /etc/systemd/system/wlan-recovery.timer

content:

[Unit]
Description=Run WLAN Recovery every 5 minutes

[Timer]
OnBootSec=1min
OnUnitActiveSec=5min
Persistent=true

[Install]
WantedBy=timers.target

activate this service:

sudo systemctl daemon-reload
sudo systemctl enable --now wlan-recovery.timer

logfile: /var/log/wlan-recovery.log

schlomm

@rkorell Thanks for all these detailed Information! I’ll setup your scripts on my MagicMirror Instances - I have 6 different ones to manage and at least one has some weird problems - maybe also relating to power issues.

Thanks und Vielen Dank :)

rkorell

@schlomm Sehr gern und VIEL ERFOLG!

melde Dich, wenn Du noch was benötigst.
Ralf

rkorell

Dear @schlomm , team,

as I learned today sometimes system limitations are hard and leads to unwanted results.
I even got more serioous trouble with my pi and his WIFI so I had to dig in deeper.
The aproach until now - because of “growing” up - is a “recovery” and a “diagnose” part.
This leads to - surprise, surprise :-) - inconsistent data and so dignosis is merely impossible.
For this reason I converged both approches into one script and implemented a 4-stages error-handling and consecutive escalation (until reboot).

At this stage i had to recognize: almost EVERY test results immediatly in an error at stage 1 (ping) and was resolved at stage 2 (L2/L3 - ICMP problem – checking status of wlan interface).
To identify root cause for this I - again - dig down deeply (ChatGPT was NOT that helpful!) and found: At systemd level (on my system!?) ping is not in PATH !!!
So a fully qualified call solved this problem - and most of my “problems” are solved !

If you are using “ping”, too and stuck in problems in scripts - keep this in mind: “usr/bin/ping” might be really helpful for you.

If you are interested in, here my current recovery-script - including some useful logging information:

/usr/local/bin/wlan-recovery.sh
#!/bin/bash
# ============================================================================
# WLAN Recovery Script (Monolithische Version)
# Autor: Dr.  Ralf Korell, MD 
# Datum: 2025-10-07
#
# Dieses Script wird per systemd-Timer regelmäßig aufgerufen.
# Es prüft die WLAN-Verbindung in mehreren Stufen und führt nur dann
# Recovery-Aktionen aus, wenn wirklich eine Unterbrechung vorliegt.
#
# Features:
#   - Mehrstufige Diagnose (Ping, iw, IP, Route)
#   - Schutz vor Fehlalarmen und Selbstabschüssen
#   - SSH/VNC-Safe-Mode (keine Unterbrechung aktiver Sessions)
#   - Logrotation + Statistikdatei
# ============================================================================

# === Konfiguration ==========================================================
LOGFILE="/var/log/wlan-recovery.log"
STATSFILE="/var/log/wlan-recovery.stats"
MAX_LOG_SIZE=50000              # ~50 KB, dann Logrotation
PING_TARGET="172.23.56.1"
MAX_CONSECUTIVE_FAILS=2         # bevor Recovery startet
COOLDOWN_FILE="/tmp/wlan-recovery.cooldown"
COOLDOWN_MINUTES=5

# interne Speicherorte (nicht verändern)
STATEFILE="/tmp/wlan-recovery.state"
DATE_NOW=$(date "+%Y-%m-%d %H:%M:%S")

# === Hilfsfunktionen ========================================================

log() {
    echo "$DATE_NOW: $1" | tee -a "$LOGFILE"
}

rotate_log() {
    if [ -f "$LOGFILE" ] && [ $(wc -c <"$LOGFILE") -gt $MAX_LOG_SIZE ]; then
        mv "$LOGFILE" "$LOGFILE.old"
        echo "$DATE_NOW: Log rotated." > "$LOGFILE"
    fi
}

increment_stat() {
    local key="$1"
    local value
    value=$(grep "^$key=" "$STATSFILE" 2>/dev/null | cut -d= -f2)
    value=$((value + 1))
    grep -v "^$key=" "$STATSFILE" 2>/dev/null > "${STATSFILE}.tmp"
    echo "$key=$value" >> "${STATSFILE}.tmp"
    mv "${STATSFILE}.tmp" "$STATSFILE"
}

cooldown_active() {
    if [ -f "$COOLDOWN_FILE" ]; then
        local last=$(date -r "$COOLDOWN_FILE" +%s)
        local now=$(date +%s)
        local diff=$(( (now - last) / 60 ))
        [ $diff -lt $COOLDOWN_MINUTES ]
    else
        return 1
    fi
}

start_cooldown() {
    touch "$COOLDOWN_FILE"
}

ssh_or_vnc_active() {
    ss -tn state established | grep -Eq '(:22|:5900)'
}

# === Diagnosefunktionen =====================================================

is_connected_l2() {
    iw dev wlan0 link 2>/dev/null | grep -q "Connected to"
}

has_ip_l3() {
    ip -4 addr show wlan0 2>/dev/null | grep -q "inet "
}

has_route() {
    ip route get "$PING_TARGET" 2>/dev/null | grep -q "dev wlan0"
}

ping_ok() {
    /usr/bin/ping -I wlan0 -c 3 -W 2 "$PING_TARGET" >/dev/null 2>&1
}

# === Hauptlogik =============================================================

rotate_log

# Init Statsfile falls nicht vorhanden
[ -f "$STATSFILE" ] || echo -e "success=0\nrecoveries=0\nfailures=0" > "$STATSFILE"

# Lese bisherigen Fehlerzähler
fails=0
[ -f "$STATEFILE" ] && fails=$(cat "$STATEFILE")

# Diagnose
if ping_ok; then
    log "Ping erfolgreich. WLAN funktioniert."
    echo 0 > "$STATEFILE"
    increment_stat "success"
    exit 0
fi

# Wenn Ping fehlschlägt → weitere Prüfungen
log "Ping fehlgeschlagen → erweiterte Diagnose..."

if is_connected_l2 && has_ip_l3 && has_route; then
    log "L2/L3 ok → ICMP-Problem (kein Recovery)."
    increment_stat "failures"
    echo 0 > "$STATEFILE"
    exit 0
fi

# Hier gilt: echte Verbindung gestört
fails=$((fails + 1))
echo "$fails" > "$STATEFILE"

if [ $fails -lt $MAX_CONSECUTIVE_FAILS ]; then
    log "Erster Fehler ($fails/$MAX_CONSECUTIVE_FAILS) → Beobachten..."
    increment_stat "failures"
    exit 0
fi

# Wenn Cooldown läuft → überspringen
if cooldown_active; then
    log "Cooldown aktiv → Recovery übersprungen."
    exit 0
fi

# === Recovery-Stufen ========================================================

if ssh_or_vnc_active; then
    log "SSH/VNC aktiv → keine Recovery ausgeführt, nur geloggt."
    increment_stat "failures"
    exit 0
fi

log "Verbindung tatsächlich gestört → Recovery-Prozess gestartet."
increment_stat "recoveries"

# Stufe 1: sanfte Reassoziation
log "→ Stufe 1: wpa_supplicant Reassoziation..."
wpa_cli -i wlan0 reassociate >/dev/null 2>&1
sleep 5
if ping_ok; then
    log "Reassoziation erfolgreich."
    echo 0 > "$STATEFILE"
    start_cooldown
    exit 0
fi

# Stufe 2: Interface Toggle
log "→ Stufe 2: Interface Toggle..."
ip link set wlan0 down
sleep 2
ip link set wlan0 up
sleep 8
if ping_ok; then
    log "Interface Toggle erfolgreich."
    echo 0 > "$STATEFILE"
    start_cooldown
    exit 0
fi

# Stufe 3: Treiber-Reload
log "→ Stufe 3: Treiber-Reload..."
modprobe -r brcmfmac && modprobe brcmfmac
sleep 10
if ping_ok; then
    log "Treiber-Reload erfolgreich."
    echo 0 > "$STATEFILE"
    start_cooldown
    exit 0
fi

# Wenn alles fehlschlägt
log "Alle Recovery-Stufen fehlgeschlagen → Fehler bleibt bestehen."
increment_stat "failures"
start_cooldown
exit 1

[EDIT: in script above: changed ping count from -c 1 to -c 3 in:
/usr/bin/ping -I wlan0 -c 1 -W 2 “$PING_TARGET” >/dev/null 2>&1
]
Warmest regards,
Ralf

rkorell

In addition: It’s possible that the old recovery script was part of my problems - due to the above mentioned ping problem.
For this reason I’ve edited my earlier post and deleted the content of the script.
I’ve added an “edit note” instead.
Sorry for confusion and any inconvenience!

Regards,
Ralf