我正在尝试通过 LoRa 隧道传输 IP 数据包。为此，我使用两个运行Meshtastic。使用 ICMP 我能够确定大约有 30％ 的数据包丢失并且延迟大约为 18 秒（不是毫秒！）。
$ time ping 10.115.253.197 -c 20 -W 600
PING 10.115.253.197 (10.115.253.197) 56(84) bytes of data.
64 bytes from 10.115.253.197: icmp_seq=1 ttl=64 time=12612 ms
64 bytes from 10.115.253.197: icmp_seq=2 ttl=64 time=15457 ms
64 bytes from 10.115.253.197: icmp_seq=4 ttl=64 time=14536 ms
64 bytes from 10.115.253.197: icmp_seq=5 ttl=64 time=14669 ms
64 bytes from 10.115.253.197: icmp_seq=6 ttl=64 time=14803 ms
64 bytes from 10.115.253.197: icmp_seq=8 ttl=64 time=17756 ms
64 bytes from 10.115.253.197: icmp_seq=9 ttl=64 time=19245 ms
64 bytes from 10.115.253.197: icmp_seq=10 ttl=64 time=19234 ms
64 bytes from 10.115.253.197: icmp_seq=11 ttl=64 time=19280 ms
64 bytes from 10.115.253.197: icmp_seq=12 ttl=64 time=19564 ms
64 bytes from 10.115.253.197: icmp_seq=14 ttl=64 time=21571 ms
64 bytes from 10.115.253.197: icmp_seq=16 ttl=64 time=21286 ms
64 bytes from 10.115.253.197: icmp_seq=17 ttl=64 time=21525 ms
64 bytes from 10.115.253.197: icmp_seq=20 ttl=64 time=29039 ms
--- 10.115.253.197 ping statistics ---
20 packets transmitted, 14 received, 30% packet loss, time 19209ms
rtt min/avg/max/mdev = 12611.611/18612.706/29038.582/4040.933 ms, pipe 16
real 0m50.196s
user 0m0.007s
sys 0m0.068s
然后我尝试通过隧道运行 UDP，并且成功了。
$ ncat --verbose --wait 10m --udp 10.115.180.153 8886
Ncat: Version 7.94 ( https://nmap.org/ncat )
Ncat: Connected to 10.115.180.153:8886.
Hello World!
^C
$ ncat --listen --verbose --udp --source-port 8886
Ncat: Version 7.94 ( https://nmap.org/ncat )
Ncat: Listening on [::]:8886
Ncat: Listening on 0.0.0.0:8886
Ncat: Connection from 10.115.253.197:47920.
Hello World!
^C
然而，当我尝试建立 TCP 隧道时，却出现了超时的情况。
$ ncat --verbose --wait 10m 10.115.180.153 8886
Ncat: Version 7.94 ( https://nmap.org/ncat )
Ncat: Connection timed out.
然后我做了一个流量转储（在服务器端）并且可以看到大量的 TCP 重传。此外，TCP 三次握手似乎不起作用。
我现在想知道我需要调整哪些 TCP 参数才能使隧道正常工作？我在两台 Linux 计算机上运行了测试。IP 隧道的 MTU 为 200。