#!/bin/bash
# 关键假设验证：是不是 MTU / 分片问题？
#
# 线索：
#   * 新建 TLS 连接约 30% 失败，但连接一旦建好就 100% 成功
#   * 到 1.1.1.1 丢包 40%
#   -> 很像**大包被丢、小包能过**：TLS 握手要发大证书包才失败，
#      建好之后数据包小就没事。
#
# 如果真是 MTU 问题，把 MSS 夹到合适值就能大幅改善 —— 这是能落地的修复。
set -u

EDGE=198.41.192.107

echo "===== 1. 本机网卡 MTU ====="
ip link show 2>/dev/null | grep -E '^[0-9]+:' -A1 | grep -E 'mtu' \
    | sed 's/^/  /'
echo "  默认路由："
ip route get 1.1.1.1 2>/dev/null | head -2 | sed 's/^/    /'

echo
echo "===== 2. 逐步加大 ping 包，找不丢包的最大值（到 1.1.1.1）====="
for size in 1400 1420 1440 1452 1460 1472 1473 1500; do
    out=$(ping -c 3 -W 3 -M do -s "$size" 1.1.1.1 2>&1)
    loss=$(echo "$out" | grep -oE '[0-9]+% packet loss' | head -1)
    err=$(echo "$out" | grep -oiE 'frag needed|Message too long|Invalid argument' | head -1)
    printf "  payload %-5s (包大小 %-5s) 丢包 %-12s %s\n" \
        "$size" "$((size + 28))" "${loss:-?}" "${err:-}"
done

echo
echo "===== 3. 同样测到 Cloudflare 边缘 $EDGE ====="
for size in 1200 1300 1400 1452 1460 1472; do
    out=$(ping -c 3 -W 3 -M do -s "$size" "$EDGE" 2>&1)
    loss=$(echo "$out" | grep -oE '[0-9]+% packet loss' | head -1)
    err=$(echo "$out" | grep -oiE 'frag needed|Message too long' | head -1)
    printf "  payload %-5s (包大小 %-5s) 丢包 %-12s %s\n" \
        "$size" "$((size + 28))" "${loss:-?}" "${err:-}"
done

echo
echo "===== 4. 当前 MSS / 分片相关设置 ====="
echo "  iptables MSS 规则："
iptables -t mangle -L -n 2>/dev/null | grep -i mss | sed 's/^/    /' \
    || echo "    （没有）"
echo "  /proc/sys/net/ipv4/tcp_mtu_probing = $(cat /proc/sys/net/ipv4/tcp_mtu_probing 2>/dev/null)"
echo "  /proc/sys/net/ipv4/tcp_base_mss    = $(cat /proc/sys/net/ipv4/tcp_base_mss 2>/dev/null)"
echo "  /proc/sys/net/ipv4/tcp_window_scaling = $(cat /proc/sys/net/ipv4/tcp_window_scaling 2>/dev/null)"

echo
echo "===== 5. 到边缘的实际 TCP 握手成功率（20 次）====="
ok=0
for i in $(seq 1 20); do
    timeout 4 bash -c "exec 3<>/dev/tcp/$EDGE/7844" 2>/dev/null && ok=$((ok+1))
done
echo "  TCP 7844 握手 $ok/20"

echo
echo "===== 6. 到大站的 HTTPS 握手成功率（20 次，看证书大包）====="
ok=0
for i in $(seq 1 20); do
    timeout 10 curl -4 -s -o /dev/null https://www.cloudflare.com/cdn-cgi/trace \
        2>/dev/null && ok=$((ok+1))
done
echo "  HTTPS 握手 $ok/20"

echo
echo "===== 结论 ====="
echo "  如果小包通、大包不通（或丢包率随包大小骤增）-> 就是 MTU 问题，"
echo "  夹 MSS 就能修。如果各尺寸都差不多 -> 是链路整体丢包，夹 MSS 没用。"