feat: add usb watchdog to rpi running zwave stick

ansible/provision_iot.yaml

@@ -9,5 +9,6 @@
     - role: apt
     - role: udev
     - role: ser2net
+    - role: usb-watchdog
     - role: wmbusmeters
     - role: readonly

ansible/roles/udev/tasks/main.yaml

@@ -6,10 +6,10 @@
       # Creates consistent device symlinks regardless of USB port
 
       # IMST iU891A-XL USB stick (Cypress chipset)
-      SUBSYSTEM=="tty", ATTRS{idVendor}=="04b4", ATTRS{idProduct}=="0003", ATTRS{manufacturer}=="IMST", SYMLINK+="wmbus_stick", GROUP="dialout", MODE="0664"
+      SUBSYSTEM=="tty", ATTRS{idVendor}=="04b4", ATTRS{idProduct}=="0003", ATTRS{manufacturer}=="IMST", SYMLINK+="wmbus_stick", GROUP="dialout", MODE="0664", TAG+="systemd", ENV{SYSTEMD_WANTS}+="wmbusmeters-restart.service"
 
       # Aeotec Z-Wave USB Stick Gen7 (Silicon Labs CP2102N chipset)
-      SUBSYSTEM=="tty", ATTRS{idVendor}=="10c4", ATTRS{idProduct}=="ea60", ATTRS{manufacturer}=="Silicon Labs", SYMLINK+="zwave_stick", GROUP="dialout", MODE="0664"
+      SUBSYSTEM=="tty", ATTRS{idVendor}=="10c4", ATTRS{idProduct}=="ea60", ATTRS{manufacturer}=="Silicon Labs", SYMLINK+="zwave_stick", GROUP="dialout", MODE="0664", TAG+="systemd", ENV{SYSTEMD_WANTS}+="ser2net-restart.service"
     dest: /etc/udev/rules.d/99-usbdevices.rules
     owner: root
     group: root
@@ -18,3 +18,37 @@
   notify:
     - reload udev
     - trigger udev
+
+- name: Create ser2net restart service for udev trigger
+  copy:
+    content: |
+      [Unit]
+      Description=Restart ser2net after Z-Wave stick reconnect
+
+      [Service]
+      Type=oneshot
+      ExecStartPre=/bin/sleep 2
+      ExecStart=/bin/systemctl restart ser2net
+    dest: /etc/systemd/system/ser2net-restart.service
+    owner: root
+    group: root
+    mode: "0644"
+  become: yes
+  notify: reload udev
+
+- name: Create wmbusmeters restart service for udev trigger
+  copy:
+    content: |
+      [Unit]
+      Description=Restart wmbusmeters after W-MBus stick reconnect
+
+      [Service]
+      Type=oneshot
+      ExecStartPre=/bin/sleep 2
+      ExecStart=/bin/systemctl restart wmbusmeters
+    dest: /etc/systemd/system/wmbusmeters-restart.service
+    owner: root
+    group: root
+    mode: "0644"
+  become: yes
+  notify: reload udev

ansible/roles/usb-watchdog/handlers/main.yaml

@@ -0,0 +1,4 @@
+---
+- name: reload systemd
+  ansible.builtin.systemd:
+    daemon_reload: true

ansible/roles/usb-watchdog/tasks/main.yaml

@@ -0,0 +1,49 @@
+---
+- name: Install usb-watchdog script
+  ansible.builtin.template:
+    src: templates/usb-watchdog.sh.j2
+    dest: /usr/local/bin/usb-watchdog
+    owner: root
+    group: root
+    mode: "0755"
+
+- name: Install usb-watchdog service
+  ansible.builtin.copy:
+    content: |
+      [Unit]
+      Description=USB stick watchdog (Z-Wave and W-MBus)
+      After=local-fs.target
+
+      [Service]
+      Type=oneshot
+      ExecStart=/usr/local/bin/usb-watchdog
+    dest: /etc/systemd/system/usb-watchdog.service
+    owner: root
+    group: root
+    mode: "0644"
+  notify: reload systemd
+
+- name: Install usb-watchdog timer
+  ansible.builtin.copy:
+    content: |
+      [Unit]
+      Description=Run USB stick watchdog every 30s
+
+      [Timer]
+      OnBootSec=30
+      OnUnitActiveSec=30
+
+      [Install]
+      WantedBy=timers.target
+    dest: /etc/systemd/system/usb-watchdog.timer
+    owner: root
+    group: root
+    mode: "0644"
+  notify: reload systemd
+
+- name: Enable usb-watchdog timer
+  ansible.builtin.systemd:
+    name: usb-watchdog.timer
+    enabled: true
+    state: started
+    daemon_reload: true

ansible/roles/usb-watchdog/templates/usb-watchdog.sh.j2

@@ -0,0 +1,59 @@
+#!/bin/bash
+# USB stick watchdog for Z-Wave and Wireless M-Bus dongles
+# Recovery: USB reset, then reboot if device was previously working this boot.
+
+ALIVE_DIR="/tmp/usb-watchdog"
+mkdir -p "$ALIVE_DIR"
+
+check_device() {
+    local name="$1"
+    local device="$2"
+    local vendor="$3"
+    local product="$4"
+    local service="$5"
+    local alive_flag="$ALIVE_DIR/$name"
+
+    if [ -e "$device" ]; then
+        touch "$alive_flag"
+        return 0
+    fi
+
+    logger -t usb-watchdog "$name: $device not found, attempting USB reset"
+
+    # Try USB unbind/rebind
+    for usbdev in /sys/bus/usb/devices/*/idVendor; do
+        dir=$(dirname "$usbdev")
+        if [ "$(cat "$dir/idVendor" 2>/dev/null)" = "$vendor" ] && \
+           [ "$(cat "$dir/idProduct" 2>/dev/null)" = "$product" ]; then
+            devname=$(basename "$dir")
+            echo "$devname" > /sys/bus/usb/drivers/usb/unbind 2>/dev/null
+            sleep 2
+            echo "$devname" > /sys/bus/usb/drivers/usb/bind 2>/dev/null
+            sleep 3
+            break
+        fi
+    done
+
+    if [ -e "$device" ]; then
+        logger -t usb-watchdog "$name: recovered after USB reset"
+        touch "$alive_flag"
+        systemctl restart "$service"
+        return 0
+    fi
+
+    # Only reboot if device was working before
+    if [ ! -f "$alive_flag" ]; then
+        logger -t usb-watchdog "$name: never seen since boot, not rebooting (manual intervention required)"
+        return 1
+    fi
+
+    logger -t usb-watchdog "$name: was working but not recovered, rebooting"
+    reboot
+}
+
+needs_reboot=0
+
+check_device "zwave" "/dev/zwave_stick" "10c4" "ea60" "ser2net" || needs_reboot=1
+check_device "wmbus" "/dev/wmbus_stick" "04b4" "0003" "wmbusmeters" || needs_reboot=1
+
+exit $needs_reboot

ansible/roles/wmbusmeters/tasks/main.yaml

@@ -67,10 +67,16 @@
     recurse: yes
   become: yes
 
+- name: Configure wmbusmeters
+  ansible.builtin.command:
+    cmd: ./configure
+    chdir: "{{ wmbusmeters_home }}/wmbusmeters-src"
+  become: yes
+
 - name: Build wmbusmeters
   make:
     chdir: "{{ wmbusmeters_home }}/wmbusmeters-src"
-    jobs: "{{ ansible_processor_cores }}"
+    jobs: "{{ ansible_facts['processor_cores'] }}"
   become: yes
   environment:
     HOME: "{{ wmbusmeters_home }}"