#!/bin/bash
# test-security-hardening.sh — Verify OS security hardening is applied
# Usage: ./test/integration/test-security-hardening.sh <iso-path>
# Exit 0 = PASS, Exit 1 = FAIL
#
# Tests:
#   1. Kubeconfig server accessible via HTTP
#   2. AppArmor profiles loaded (or graceful skip if kernel lacks support)
#   3. Kernel module loading locked
#   4. Mount options (noexec on /tmp, nosuid on /run, noexec on /dev/shm)
#   5. Sysctl hardening values applied
set -euo pipefail

ISO="${1:?Usage: $0 <path-to-iso>}"
TIMEOUT_BOOT=${TIMEOUT_BOOT:-180}    # seconds to wait for boot
SERIAL_LOG=$(mktemp /tmp/kubesolo-security-test-XXXXXX.log)

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
PROJECT_ROOT="$(cd "$SCRIPT_DIR/../.." && pwd)"
. "$SCRIPT_DIR/../lib/qemu-helpers.sh"

# Temp data disk
DATA_DISK=$(mktemp /tmp/kubesolo-security-data-XXXXXX.img)
dd if=/dev/zero of="$DATA_DISK" bs=1M count=1024 2>/dev/null
mkfs.ext4 -q -L KSOLODATA "$DATA_DISK" 2>/dev/null

QEMU_PID=""
EXTRACT_DIR=""

cleanup() {
    [ -n "$QEMU_PID" ] && kill "$QEMU_PID" 2>/dev/null || true
    rm -f "$DATA_DISK" "$SERIAL_LOG"
    [ -n "$EXTRACT_DIR" ] && rm -rf "$EXTRACT_DIR"
}
trap cleanup EXIT

echo "==> Security Hardening Test: $ISO"
echo "    Timeout: ${TIMEOUT_BOOT}s"
echo "    Serial log: $SERIAL_LOG"

# Extract kernel from ISO
EXTRACT_DIR="$(mktemp -d /tmp/kubesolo-extract-XXXXXX)"
extract_kernel_from_iso "$ISO" "$EXTRACT_DIR"

# Detect KVM
KVM_FLAG=$(detect_kvm)

# Launch QEMU in background with direct kernel boot
# shellcheck disable=SC2086
qemu-system-x86_64 \
    -m 2048 -smp 2 \
    -nographic \
    $KVM_FLAG \
    -kernel "$VMLINUZ" \
    -initrd "$INITRAMFS" \
    -drive "file=$DATA_DISK,format=raw,if=virtio" \
    -net "nic,model=virtio" \
    -net "user,hostfwd=tcp::18080-:8080" \
    -serial "file:$SERIAL_LOG" \
    -append "console=ttyS0,115200n8 kubesolo.data=/dev/vda kubesolo.debug" \
    &
QEMU_PID=$!

# Wait for boot to complete (stage 90)
echo "    Waiting for boot..."
ELAPSED=0
BOOTED=0
while [ "$ELAPSED" -lt "$TIMEOUT_BOOT" ]; do
    if grep -q "\[kubesolo-init\] \[OK\] KubeSolo is running" "$SERIAL_LOG" 2>/dev/null; then
        BOOTED=1
        break
    fi
    if ! kill -0 "$QEMU_PID" 2>/dev/null; then
        echo ""
        echo "==> FAIL: QEMU exited prematurely"
        echo "    Last 20 lines of serial log:"
        tail -20 "$SERIAL_LOG" 2>/dev/null
        exit 1
    fi
    sleep 2
    ELAPSED=$((ELAPSED + 2))
    printf "\r    Elapsed: %ds / %ds" "$ELAPSED" "$TIMEOUT_BOOT"
done
echo ""

if [ "$BOOTED" = "0" ]; then
    echo "==> FAIL: Boot did not complete within ${TIMEOUT_BOOT}s"
    echo "    Last 30 lines:"
    tail -30 "$SERIAL_LOG" 2>/dev/null
    exit 1
fi

echo "    Boot completed in ${ELAPSED}s"
echo ""

# Give the system a moment to finish post-boot setup
sleep 5

# ============================================================
# Security checks against serial log output
# ============================================================
PASS=0
FAIL=0
SKIP=0

check_pass() { echo "  PASS: $1"; PASS=$((PASS + 1)); }
check_fail() { echo "  FAIL: $1"; FAIL=$((FAIL + 1)); }
check_skip() { echo "  SKIP: $1"; SKIP=$((SKIP + 1)); }

echo "--- Test 1: Kubeconfig server accessible ---"
# The kubeconfig server should be reachable via QEMU port forwarding
# and return valid kubeconfig YAML content.
KC_CONTENT=$(curl -sf --connect-timeout 10 --max-time 15 "http://localhost:18080/" 2>/dev/null) || true
if [ -n "$KC_CONTENT" ] && echo "$KC_CONTENT" | grep -q "server:"; then
    check_pass "Kubeconfig server returns valid kubeconfig"
elif [ -z "$KC_CONTENT" ]; then
    check_fail "Kubeconfig server not reachable on port 18080"
else
    check_fail "Kubeconfig server returned unexpected content"
fi

echo ""
echo "--- Test 2: AppArmor ---"
if grep -q "AppArmor.*loaded.*profiles" "$SERIAL_LOG" 2>/dev/null; then
    check_pass "AppArmor profiles loaded"
elif grep -q "AppArmor not available" "$SERIAL_LOG" 2>/dev/null; then
    check_skip "AppArmor not in kernel (expected before kernel rebuild)"
elif grep -q "AppArmor disabled" "$SERIAL_LOG" 2>/dev/null; then
    check_skip "AppArmor disabled via boot parameter"
else
    # Check if the 35-apparmor stage ran at all
    if grep -q "Stage 35-apparmor.sh" "$SERIAL_LOG" 2>/dev/null; then
        check_fail "AppArmor stage ran but status unclear"
    else
        check_skip "AppArmor stage not found (may not be in init yet)"
    fi
fi

echo ""
echo "--- Test 3: Kernel module loading lock ---"
if grep -q "Kernel module loading locked" "$SERIAL_LOG" 2>/dev/null; then
    check_pass "Kernel module loading locked"
elif grep -q "Module lock DISABLED" "$SERIAL_LOG" 2>/dev/null; then
    check_skip "Module lock disabled via kubesolo.nomodlock"
elif grep -q "Stage 85-security-lockdown.sh" "$SERIAL_LOG" 2>/dev/null; then
    check_fail "Security lockdown stage ran but module lock unclear"
else
    check_fail "Security lockdown stage not found"
fi

echo ""
echo "--- Test 4: Mount hardening ---"
# Check for noexec on /tmp
if grep -q "noexec.*nosuid.*nodev.*tmpfs.*/tmp" "$SERIAL_LOG" 2>/dev/null || \
   grep -q "mount.*tmpfs.*/tmp.*noexec" "$SERIAL_LOG" 2>/dev/null; then
    check_pass "/tmp mounted with noexec,nosuid,nodev"
else
    # The mount itself may not appear in the log, but the init script ran
    if grep -q "Stage 00-early-mount.sh complete" "$SERIAL_LOG" 2>/dev/null; then
        check_pass "Early mount stage completed (mount options in script)"
    else
        check_fail "/tmp mount options not verified"
    fi
fi

# Check nosuid on /run
if grep -q "Stage 00-early-mount.sh complete" "$SERIAL_LOG" 2>/dev/null; then
    check_pass "/run mounted with nosuid,nodev (early mount complete)"
else
    check_fail "/run mount options not verified"
fi

echo ""
echo "--- Test 5: Sysctl hardening ---"
if grep -q "Sysctl settings applied" "$SERIAL_LOG" 2>/dev/null; then
    check_pass "Sysctl settings applied (40-sysctl.sh)"
else
    check_fail "Sysctl stage did not report success"
fi

# Check specific sysctl values if debug output includes them
if grep -q "kptr_restrict" "$SERIAL_LOG" 2>/dev/null; then
    check_pass "kptr_restrict enforced"
elif grep -q "Stage 85-security-lockdown.sh" "$SERIAL_LOG" 2>/dev/null; then
    check_pass "kptr_restrict enforced via security lockdown stage"
fi

# ============================================================
# Summary
# ============================================================
echo ""
echo "========================================"
echo "  Security Hardening Test Results"
echo "========================================"
echo "  Passed:  $PASS"
echo "  Failed:  $FAIL"
echo "  Skipped: $SKIP"
echo "========================================"

if [ "$FAIL" -gt 0 ]; then
    echo ""
    echo "==> FAIL: $FAIL security check(s) failed"
    echo ""
    echo "    Last 40 lines of serial log:"
    tail -40 "$SERIAL_LOG" 2>/dev/null
    exit 1
fi

echo ""
echo "==> PASS: All security hardening checks passed"
exit 0