#!/bin/bash
# Wake-on-LAN and shutdown script for f3s hosts (f0, f1, f2, f3)
# and optional shutdown for Raspberry Pi nodes (pi0–pi3)
#
# Before any shutdown of f3s Beelink hosts (f0/f1/f2/f3), this script will
# attempt to umount all NFS filesystems currently mounted on this machine
# (earth). NFS mounts here are tunnelled via stunnel (127.0.0.1:2323 →
# 192.168.1.138:2323 CARP VIP on f0/f1). If any NFS mount is active and
# cannot be umounted, the shutdown is aborted to prevent data loss or a
# hung filesystem.
#
# Usage:
#   wol-f3s              # Wake f0, f1, and f2
#   wol-f3s f0           # Wake only f0
#   wol-f3s f1           # Wake only f1
#   wol-f3s f2           # Wake only f2
#   wol-f3s f3           # Wake only f3
#   wol-f3s shutdown     # Shutdown f0, f1, and f2
#   wol-f3s shutdown-f3  # Shutdown only f3
#   wol-f3s shutdown-pis # Shutdown all four Raspberry Pis
#   wol-f3s shutdown-all # Shutdown f0, f1, f2, and Pis

# MAC addresses
F0_MAC="e8:ff:1e:d7:1c:ac"  # f0 (192.168.1.130)
F1_MAC="e8:ff:1e:d7:1e:44"  # f1 (192.168.1.131)
F2_MAC="e8:ff:1e:d7:1c:a0"  # f2 (192.168.1.132)
F3_MAC="e8:ff:1e:d7:f3:d7"  # f3 (192.168.1.133)

# IP addresses
F0_IP="192.168.1.130"
F1_IP="192.168.1.131"
F2_IP="192.168.1.132"
F3_IP="192.168.1.133"

# Raspberry Pi IP addresses (no WoL support, shutdown only)
PI0_IP="192.168.1.125"
PI1_IP="192.168.1.126"
PI2_IP="192.168.1.127"
PI3_IP="192.168.1.128"

# Shelly Plug M Gen 3 controlling the rack fans (HTTP RPC API).
# It is powered on when waking all hosts and powered off when shutting all
# hosts down. The plug uses digest auth (user "admin"); the password is read
# from ~/.shelly_plug if that file exists.
#
# The shutdown paths need root (to umount NFS), so this script is typically
# run elevated. When elevated, $HOME becomes root's home, where ~/.shelly_plug
# does not exist — so resolve the file relative to the invoking user's home
# ($SUDO_USER under sudo on Linux, $DOAS_USER under doas on NetBSD Pis),
# falling back to $HOME otherwise. Uses ~-expansion (portable) rather than
# getent, which does not exist on NetBSD.
SHELLY_IP="192.168.1.28"
_shelly_user="${SUDO_USER:-${DOAS_USER:-}}"
if [[ -n "$_shelly_user" ]]; then
    _shelly_home="$(eval echo "~${_shelly_user}")"
    SHELLY_PASS_FILE="${_shelly_home:-$HOME}/.shelly_plug"
else
    SHELLY_PASS_FILE="${HOME}/.shelly_plug"
fi

# SSH user
SSH_USER="paul"

# Broadcast address for your LAN
BROADCAST="192.168.1.255"

wake() {
    local name=$1
    local mac=$2

    if [[ "$mac" == "XX:XX:XX:XX:XX:XX" ]]; then
        echo "⚠️  $name MAC address not configured yet"
        return 1
    fi

    echo "Sending WoL packet to $name ($mac)..."
    wol -i "$BROADCAST" "$mac"
}

# shelly_set turns the rack-fan Shelly plug on or off.
#   $1 — "true" to power on, "false" to power off
shelly_set() {
    local on=$1
    local label="on"
    [[ "$on" == "false" ]] && label="off"

    local auth=()
    if [[ -r "$SHELLY_PASS_FILE" ]]; then
        local pass
        pass="$(head -n 1 "$SHELLY_PASS_FILE" | tr -d '\r\n')"
        [[ -n "$pass" ]] && auth=(--digest -u "admin:$pass")
    fi

    if [[ ${#auth[@]} -eq 0 ]]; then
        echo "  ⚠️  No Shelly password found at $SHELLY_PASS_FILE"
        echo "     — cannot switch plug $label (digest auth required)."
        return 1
    fi

    echo "Turning rack-fan Shelly plug $label ($SHELLY_IP)..."
    # --fail is essential: without it curl returns 0 even on an HTTP 401
    # (auth failure), which would falsely report success while the plug is
    # left untouched.
    if curl -s --fail --max-time 5 "${auth[@]}" \
        "http://${SHELLY_IP}/rpc/Switch.Set?id=0&on=${on}" >/dev/null; then
        echo "  ✓ Shelly plug switched $label"
    else
        echo "  ✗ Failed to switch Shelly plug at $SHELLY_IP (auth or network error)"
        return 1
    fi
}

# umount_nfs_mounts tries to umount all currently mounted NFS filesystems on
# this local machine (earth). It reads /proc/mounts to find active nfs/nfs4
# mounts, attempts to umount each one, and returns 1 if any mount could not
# be umounted. This prevents shutting down f3s servers while NFS is still in
# use, which would leave the filesystem in a hung/stale state.
umount_nfs_mounts() {
    # Collect all active NFS mount points from /proc/mounts (covers nfs and nfs4)
    local nfs_mounts=()
    while IFS= read -r mountpoint; do
        nfs_mounts+=("$mountpoint")
    done < <(awk '$3 ~ /^nfs/ { print $2 }' /proc/mounts)

    if [[ ${#nfs_mounts[@]} -eq 0 ]]; then
        echo "  No NFS filesystems currently mounted — nothing to umount."
        return 0
    fi

    local failed=0
    for mp in "${nfs_mounts[@]}"; do
        echo "  Umounting NFS filesystem: $mp ..."
        if umount "$mp" 2>/dev/null; then
            echo "  ✓ Umounted $mp"
        else
            # If it's no longer listed in /proc/mounts, it's already gone.
            if ! grep -q "^[^ ]* $mp nfs" /proc/mounts; then
                echo "  ✓ $mp was already unmounted"
            else
                echo "  ✗ Failed to umount $mp (in use or already unmounted?)"
                failed=1
            fi
        fi
    done

    if [[ $failed -ne 0 ]]; then
        echo ""
        echo "✗ One or more NFS filesystems could not be umounted."
        echo "  Aborting shutdown to prevent data loss or a hung filesystem."
        echo "  Please check running processes that may have open files on NFS,"
        echo "  then retry."
        return 1
    fi

    return 0
}

shutdown_host() {
    local name=$1
    local ip=$2

    echo "Shutting down $name ($ip)..."
    ssh -o ConnectTimeout=5 "$SSH_USER@$ip" "doas poweroff" 2>/dev/null && \
        echo "  ✓ Shutdown command sent to $name" || \
        echo "  ✗ Failed to reach $name (already down?)"
}

ACTION="${1:-all}"

case "$ACTION" in
    f0)
        wake "f0" "$F0_MAC"
        ;;
    f1)
        wake "f1" "$F1_MAC"
        ;;
    f2)
        wake "f2" "$F2_MAC"
        ;;
    f3)
        wake "f3" "$F3_MAC"
        ;;
    all|"")
        # Power on the rack fans before booting the hosts.
        shelly_set "true"
        wake "f0" "$F0_MAC"
        wake "f1" "$F1_MAC"
        wake "f2" "$F2_MAC"
        ;;
    shutdown|poweroff|down)
        # Umount local NFS filesystems first; abort if any are stuck.
        # The NFS mounts on earth tunnel through stunnel to the f3s CARP VIP,
        # so they will hang or corrupt if the servers go down while mounted.
        echo "Checking for locally mounted NFS filesystems..."
        umount_nfs_mounts || exit 1
        # Mute Gogios monitoring alerts for a day on both OpenBSD gateways
        ssh rex@blowfish.buetow.org touch /tmp/f3s_taken_down
        ssh rex@fishfinger.buetow.org touch /tmp/f3s_taken_down
        shutdown_host "f0" "$F0_IP"
        shutdown_host "f1" "$F1_IP"
        shutdown_host "f2" "$F2_IP"
        # Power off the rack fans once all hosts are shutting down.
        shelly_set "false"
        echo ""
        echo "✓ Shutdown commands sent to all Beelinks."
        exit 0
        ;;
    shutdown-f3|poweroff-f3|down-f3)
        # Umount local NFS filesystems first; abort if any are stuck.
        # f3 does not host the NFS CARP VIP (that is f0/f1), but umounting
        # first is still the safe default to avoid stale state.
        echo "Checking for locally mounted NFS filesystems..."
        umount_nfs_mounts || exit 1
        shutdown_host "f3" "$F3_IP"
        echo ""
        echo "✓ Shutdown command sent to f3."
        exit 0
        ;;
    shutdown-pis)
        # Raspberry Pis do not serve NFS, so no umount check is needed here.
        shutdown_host "pi0" "$PI0_IP"
        shutdown_host "pi1" "$PI1_IP"
        shutdown_host "pi2" "$PI2_IP"
        shutdown_host "pi3" "$PI3_IP"
        echo ""
        echo "✓ Shutdown commands sent to all Pis."
        exit 0
        ;;
    shutdown-all)
        # Umount local NFS filesystems first; abort if any are stuck.
        # The NFS mounts on earth tunnel through stunnel to the f3s CARP VIP,
        # so they will hang or corrupt if the servers go down while mounted.
        echo "Checking for locally mounted NFS filesystems..."
        umount_nfs_mounts || exit 1
        # Mute Gogios monitoring alerts for a day on both OpenBSD gateways
        ssh rex@blowfish.buetow.org touch /tmp/f3s_taken_down
        ssh rex@fishfinger.buetow.org touch /tmp/f3s_taken_down
        shutdown_host "f0" "$F0_IP"
        shutdown_host "f1" "$F1_IP"
        shutdown_host "f2" "$F2_IP"
        shutdown_host "pi0" "$PI0_IP"
        shutdown_host "pi1" "$PI1_IP"
        shutdown_host "pi2" "$PI2_IP"
        shutdown_host "pi3" "$PI3_IP"
        # Power off the rack fans once all hosts are shutting down.
        shelly_set "false"
        echo ""
        echo "✓ Shutdown commands sent to all machines."
        exit 0
        ;;
    *)
        echo "Usage: $0 [f0|f1|f2|f3|all|shutdown|shutdown-f3|shutdown-pis|shutdown-all]"
        exit 1
        ;;
esac

echo ""
echo "✓ WoL packets sent. Machines should boot in a few seconds."
echo "  Wait ~30 seconds, then try: ssh paul@192.168.1.130"