正文

Wireshark & Packetdrill | TCP 基础快速重传续

admin
admin V管理员 /0 评论 /58 阅读
0601
此篇文章发布距今已超过22天，您需要注意文章的内容或图片是否可用！

要经验，不要经验主义

实验目的

基于 packetdrill TCP 三次握手脚本，通过构造模拟服务器端场景，继续测试基础快速重传相关。

基础脚本

# cat tcp_fast_retransmit_1_000.pkt 0   socket(..., SOCK_STREAM, IPPROTO_TCP) = 3+0  setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0+0  bind(3, ..., ...) = 0+0  listen(3, 1) = 0+0 < S 0:0(0) win 10000 <mss 1000>+0 > S. 0:0(0) ack 1 <...>+0.01 < . 1:1(0) ack 1 win 10000+0 accept(3, ..., ...) = 4#

快速重传

TCP 快速重传是一种机制，用于在检测到数据丢失时快速触发重传，以提高数据传输效率。当 TCP 接收端收到一个失序的报文段时，它会发送重复的确认（ACK）给发送端。如果发送端连续收到三个重复的 ACK ，就会认为某个报文段可能已经丢失，从而立即重传该报文段，而无需等待超时计时器到期。这种机制可以显著减少因丢包导致的传输延迟，同时与拥塞控制机制配合，避免因重传引发网络拥塞加剧。

考虑到目前真实环境更多都是基于 SACK 开启下的各类重传（包括但不限于 RACK、ER、TLP 等等），因此在无 SACK 的场景下的快速重传，我个人把它定义为基础快速重传。

基础测试

数据接收端

首先观察下丢包场景下数据接收端响应 ACK 的发送行为，计划注入 6 个 MSS 1000 字节大小的数据段，但模拟两个数据段丢失，假设是第 2 个和第 4 个数据段。

# cat tcp_fast_retransmit_1_007.pkt `ethtool -K tun0 tso off ethtool -K tun0 gso off`0   socket(..., SOCK_STREAM, IPPROTO_TCP) = 3+0  setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0+0  setsockopt(3, SOL_TCP, TCP_NODELAY, [1], 4) = 0+0  bind(3, ..., ...) = 0+0  listen(3, 1) = 0+0 < S 0:0(0) win 10000 <mss 1000>+0 > S. 0:0(0) ack 1 <...>+0.01 < . 1:1(0) ack 1 win 10000+0 accept(3, ..., ...) = 4+0.01 < P. 1:1001(1000) ack 1 win 10000+0 < P. 2001:3001(1000) ack 1 win 10000+0 < P. 4001:5001(1000) ack 1 win 10000+0 < P. 5001:6001(1000) ack 1 win 10000#

运行脚本观察发送端现象，会触发出 3 个 Dup ACK。

# packetdrill tcp_fast_retransmit_1_007.pkt ## tcpdump -i any -nn port 8080tcpdump: data link type LINUX_SLL2tcpdump: verbose output suppressed, use -v[v]... for full protocol decodelistening on any, link-type LINUX_SLL2 (Linux cooked v2), snapshot length 262144 bytes22:31:45.026968 tun0  In  IP 192.0.2.1.52979 > 192.168.77.65.8080: Flags [S], seq 0, win 10000, options [mss 1000], length 022:31:45.027001 tun0  Out IP 192.168.77.65.8080 > 192.0.2.1.52979: Flags [S.], seq 294502670, ack 1, win 64240, options [mss 1460], length 022:31:45.037074 ?     In  IP 192.0.2.1.52979 > 192.168.77.65.8080: Flags [.], ack 1, win 10000, length 022:31:45.047127 ?     In  IP 192.0.2.1.52979 > 192.168.77.65.8080: Flags [P.], seq 1:1001, ack 1, win 10000, length 1000: HTTP22:31:45.047145 ?     Out IP 192.168.77.65.8080 > 192.0.2.1.52979: Flags [.], ack 1001, win 64000, length 022:31:45.047154 ?     In  IP 192.0.2.1.52979 > 192.168.77.65.8080: Flags [P.], seq 2001:3001, ack 1, win 10000, length 1000: HTTP22:31:45.047159 ?     Out IP 192.168.77.65.8080 > 192.0.2.1.52979: Flags [.], ack 1001, win 64000, length 022:31:45.047164 ?     In  IP 192.0.2.1.52979 > 192.168.77.65.8080: Flags [P.], seq 4001:5001, ack 1, win 10000, length 1000: HTTP22:31:45.047166 ?     Out IP 192.168.77.65.8080 > 192.0.2.1.52979: Flags [.], ack 1001, win 64000, length 022:31:45.047170 ?     In  IP 192.0.2.1.52979 > 192.168.77.65.8080: Flags [P.], seq 5001:6001, ack 1, win 10000, length 1000: HTTP22:31:45.047172 ?     Out IP 192.168.77.65.8080 > 192.0.2.1.52979: Flags [.], ack 1001, win 64000, length 022:31:45.047241 ?     Out IP 192.168.77.65.8080 > 192.0.2.1.52979: Flags [R.], seq 1, ack 1001, win 64000, length 022:31:45.047262 ?     In  IP 192.0.2.1.52979 > 192.168.77.65.8080: Flags [R.], seq 6001, ack 1, win 10000, length 0#

那么理论会触发出数据发送端的基础快速重传，修改脚本如下。

# cat tcp_fast_retransmit_1_008.pkt `ethtool -K tun0 tso off ethtool -K tun0 gso off`0   socket(..., SOCK_STREAM, IPPROTO_TCP) = 3+0  setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0+0  setsockopt(3, SOL_TCP, TCP_NODELAY, [1], 4) = 0+0  bind(3, ..., ...) = 0+0  listen(3, 1) = 0+0 < S 0:0(0) win 10000 <mss 1000>+0 > S. 0:0(0) ack 1 <...>+0.01 < . 1:1(0) ack 1 win 10000+0 accept(3, ..., ...) = 4+0.01 < P. 1:1001(1000) ack 1 win 10000+0 < P. 2001:3001(1000) ack 1 win 10000+0 < P. 4001:5001(1000) ack 1 win 10000+0 < P. 5001:6001(1000) ack 1 win 10000+0.01 < P. 1001:2001(1000) ack 1 win 10000#

接收端在收到 Seq 1001:2001 的重传数据包后，直接响应了 ACK 数据包 AckNum 3001，这意味着接收端之前收到的乱序数据包 Seq 2001:3001 并没有丢弃，而是缓存起来了。

对于接收端，发出 ACK 数据包 AckNum 3001 就是说下一个希望接收 Seq 为 3001 的数据段，而对于发送端，收到这个 ACK 数据包，也就明白之前 Seq 3001:4001 的数据段也发生了丢失。

# packetdrill tcp_fast_retransmit_1_008.pkt ## tcpdump -i any -nn port 8080tcpdump: data link type LINUX_SLL2tcpdump: verbose output suppressed, use -v[v]... for full protocol decodelistening on any, link-type LINUX_SLL2 (Linux cooked v2), snapshot length 262144 bytes22:33:47.198978 tun0  In  IP 192.0.2.1.55827 > 192.168.129.221.8080: Flags [S], seq 0, win 10000, options [mss 1000], length 022:33:47.199005 tun0  Out IP 192.168.129.221.8080 > 192.0.2.1.55827: Flags [S.], seq 206416240, ack 1, win 64240, options [mss 1460], length 022:33:47.209075 tun0  In  IP 192.0.2.1.55827 > 192.168.129.221.8080: Flags [.], ack 1, win 10000, length 022:33:47.219137 ?     In  IP 192.0.2.1.55827 > 192.168.129.221.8080: Flags [P.], seq 1:1001, ack 1, win 10000, length 1000: HTTP22:33:47.219153 ?     Out IP 192.168.129.221.8080 > 192.0.2.1.55827: Flags [.], ack 1001, win 64000, length 022:33:47.219162 ?     In  IP 192.0.2.1.55827 > 192.168.129.221.8080: Flags [P.], seq 2001:3001, ack 1, win 10000, length 1000: HTTP22:33:47.219166 ?     Out IP 192.168.129.221.8080 > 192.0.2.1.55827: Flags [.], ack 1001, win 64000, length 022:33:47.219172 ?     In  IP 192.0.2.1.55827 > 192.168.129.221.8080: Flags [P.], seq 4001:5001, ack 1, win 10000, length 1000: HTTP22:33:47.219176 ?     Out IP 192.168.129.221.8080 > 192.0.2.1.55827: Flags [.], ack 1001, win 64000, length 022:33:47.219182 ?     In  IP 192.0.2.1.55827 > 192.168.129.221.8080: Flags [P.], seq 5001:6001, ack 1, win 10000, length 1000: HTTP22:33:47.219185 ?     Out IP 192.168.129.221.8080 > 192.0.2.1.55827: Flags [.], ack 1001, win 64000, length 022:33:47.229140 ?     In  IP 192.0.2.1.55827 > 192.168.129.221.8080: Flags [P.], seq 1001:2001, ack 1, win 10000, length 1000: HTTP22:33:47.229157 ?     Out IP 192.168.129.221.8080 > 192.0.2.1.55827: Flags [.], ack 3001, win 62000, length 022:33:47.229235 ?     Out IP 192.168.129.221.8080 > 192.0.2.1.55827: Flags [R.], seq 1, ack 3001, win 63000, length 022:33:47.229280 ?     In  IP 192.0.2.1.55827 > 192.168.129.221.8080: Flags [R.], seq 2001, ack 1, win 10000, length 0#

那么继续模拟重传 Seq 3001:4001 的数据段，修改脚本如下。

# cat tcp_fast_retransmit_1_009.pkt `ethtool -K tun0 tso off ethtool -K tun0 gso off`0   socket(..., SOCK_STREAM, IPPROTO_TCP) = 3+0  setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0+0  setsockopt(3, SOL_TCP, TCP_NODELAY, [1], 4) = 0+0  bind(3, ..., ...) = 0+0  listen(3, 1) = 0+0 < S 0:0(0) win 10000 <mss 1000>+0 > S. 0:0(0) ack 1 <...>+0.01 < . 1:1(0) ack 1 win 10000+0 accept(3, ..., ...) = 4+0.01 < P. 1:1001(1000) ack 1 win 10000+0 < P. 2001:3001(1000) ack 1 win 10000+0 < P. 4001:5001(1000) ack 1 win 10000+0 < P. 5001:6001(1000) ack 1 win 10000+0.01 < P. 1001:2001(1000) ack 1 win 10000+0.01 < P. 3001:4001(1000) ack 1 win 10000#

接收端在收到 Seq 3001:4001 的重传数据包后，直接响应了 ACK 数据包 AckNum 6001，这同样说明接收端之前收到的乱序数据包 Seq 4001:6001 也都没有丢弃，而是缓存起来了，至此所有的数据段都正常接收并进行了确认。

# packetdrill tcp_fast_retransmit_1_009.pkt ## tcpdump -i any -nn port 8080tcpdump: data link type LINUX_SLL2tcpdump: verbose output suppressed, use -v[v]... for full protocol decodelistening on any, link-type LINUX_SLL2 (Linux cooked v2), snapshot length 262144 bytes21:55:35.519055 tun0  In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [S], seq 0, win 10000, options [mss 1000], length 021:55:35.519131 tun0  Out IP 192.168.153.142.8080 > 192.0.2.1.42777: Flags [S.], seq 1066490332, ack 1, win 64240, options [mss 1460], length 021:55:35.529258 ?     In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [.], ack 1, win 10000, length 021:55:35.539330 ?     In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [P.], seq 1:1001, ack 1, win 10000, length 1000: HTTP21:55:35.539356 ?     Out IP 192.168.153.142.8080 > 192.0.2.1.42777: Flags [.], ack 1001, win 64000, length 021:55:35.539366 ?     In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [P.], seq 2001:3001, ack 1, win 10000, length 1000: HTTP21:55:35.539370 ?     Out IP 192.168.153.142.8080 > 192.0.2.1.42777: Flags [.], ack 1001, win 64000, length 021:55:35.539375 ?     In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [P.], seq 4001:5001, ack 1, win 10000, length 1000: HTTP21:55:35.539378 ?     Out IP 192.168.153.142.8080 > 192.0.2.1.42777: Flags [.], ack 1001, win 64000, length 021:55:35.539384 ?     In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [P.], seq 5001:6001, ack 1, win 10000, length 1000: HTTP21:55:35.539389 ?     Out IP 192.168.153.142.8080 > 192.0.2.1.42777: Flags [.], ack 1001, win 64000, length 021:55:35.549410 ?     In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [P.], seq 1001:2001, ack 1, win 10000, length 1000: HTTP21:55:35.549487 ?     Out IP 192.168.153.142.8080 > 192.0.2.1.42777: Flags [.], ack 3001, win 62000, length 021:55:35.559351 ?     In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [P.], seq 3001:4001, ack 1, win 10000, length 1000: HTTP21:55:35.559385 ?     Out IP 192.168.153.142.8080 > 192.0.2.1.42777: Flags [.], ack 6001, win 59000, length 021:55:35.559459 ?     Out IP 192.168.153.142.8080 > 192.0.2.1.42777: Flags [R.], seq 1, ack 6001, win 64000, length 021:55:35.559488 ?     In  IP 192.0.2.1.42777 > 192.168.153.142.8080: Flags [R.], seq 4001, ack 1, win 10000, length 0#

如上实验，模拟的是第 2 个和第 4 个数据段的丢失，同样，假设是第 3 个和第 5 个数据段丢失，又或者是别的丢失场景，实际处理过程也是一样。

实验现象也说明了一些东西，基础快速重传实际不像一些资料上所说的有问题，像是不知道重传是重传一个，还是重传所有的问题，确切来说发送端是清楚的。至于说没有 SACK 的支持下，仍是可以通过 ACK 清楚哪个丢失了从而进行重传，只不过效率比较低，需要依靠一个又一个 ACK 来告知再重传，而有了 SACK 的支持，实际上也是通过 ACK&SACK 信息告知，从而清楚哪几个数据段丢失了，之后才能更有效率的进行重传。

数据发送端

再观察下丢包场景下数据发送端的数据发送行为，如果仍是连续发送 6 个 MSS 1000 字节大小的数据段，模拟两个数据段丢失，其中一个是第 2 个数据段，如下脚本，会发现无论另一个丢失的分段是第 3-6 中的哪一个，返回的 ACK 实际都是一样的，即 Dup ACK，也就是这种情况下对于发送端是无法得知接收端到底是收到了第 3-6 中的哪 3 个数据段。

# cat tcp_fast_retransmit_1_010.pkt `ethtool -K tun0 tso off ethtool -K tun0 gso off`0   socket(..., SOCK_STREAM, IPPROTO_TCP) = 3+0  setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0+0  setsockopt(3, SOL_TCP, TCP_NODELAY, [1], 4) = 0+0  bind(3, ..., ...) = 0+0  listen(3, 1) = 0+0 < S 0:0(0) win 10000 <mss 1000>+0 > S. 0:0(0) ack 1 <...>+0.01 < . 1:1(0) ack 1 win 10000+0 accept(3, ..., ...) = 4+0.01 write(4, ..., 6000) = 6000+0.01 < . 1:1(0) ack 1001 win 10000+0 < . 1:1(0) ack 1001 win 10000+0 < . 1:1(0) ack 1001 win 10000+0 < . 1:1(0) ack 1001 win 10000#

运行脚本观察发送端的重传现象，在收到 3 个 Dup ACK 的情况下，会触发出基础快速重传。

# packetdrill tcp_fast_retransmit_1_010.pkt ## tcpdump -i any -nn port 8080tcpdump: data link type LINUX_SLL2tcpdump: verbose output suppressed, use -v[v]... for full protocol decodelistening on any, link-type LINUX_SLL2 (Linux cooked v2), snapshot length 262144 bytes22:12:04.726952 tun0  In  IP 192.0.2.1.53799 > 192.168.29.171.8080: Flags [S], seq 0, win 10000, options [mss 1000], length 022:12:04.726978 tun0  Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [S.], seq 4229715043, ack 1, win 64240, options [mss 1460], length 022:12:04.737098 ?     In  IP 192.0.2.1.53799 > 192.168.29.171.8080: Flags [.], ack 1, win 10000, length 022:12:04.747314 ?     Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [.], seq 1:1001, ack 1, win 64240, length 1000: HTTP22:12:04.747318 ?     Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [P.], seq 1001:2001, ack 1, win 64240, length 1000: HTTP22:12:04.747324 ?     Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [.], seq 2001:3001, ack 1, win 64240, length 1000: HTTP22:12:04.747325 ?     Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [P.], seq 3001:4001, ack 1, win 64240, length 1000: HTTP22:12:04.747328 ?     Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [.], seq 4001:5001, ack 1, win 64240, length 1000: HTTP22:12:04.747329 ?     Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [P.], seq 5001:6001, ack 1, win 64240, length 1000: HTTP22:12:04.757438 ?     In  IP 192.0.2.1.53799 > 192.168.29.171.8080: Flags [.], ack 1001, win 10000, length 022:12:04.757494 ?     In  IP 192.0.2.1.53799 > 192.168.29.171.8080: Flags [.], ack 1001, win 10000, length 022:12:04.757505 ?     In  IP 192.0.2.1.53799 > 192.168.29.171.8080: Flags [.], ack 1001, win 10000, length 022:12:04.757512 ?     In  IP 192.0.2.1.53799 > 192.168.29.171.8080: Flags [.], ack 1001, win 10000, length 022:12:04.757535 ?     Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [.], seq 1001:2001, ack 1, win 64240, length 1000: HTTP22:12:04.757639 ?     Out IP 192.168.29.171.8080 > 192.0.2.1.53799: Flags [F.], seq 6001, ack 1, win 64240, length 022:12:04.757665 ?     In  IP 192.0.2.1.53799 > 192.168.29.171.8080: Flags [R.], seq 1, ack 1001, win 10000, length 0#

修改脚本，模拟收到基础快速重传的响应 ACK，譬如 ACK Num 3001 ，继续观察发送端的重传现象。

# cat tcp_fast_retransmit_1_011.pkt `ethtool -K tun0 tso off ethtool -K tun0 gso off`0   socket(..., SOCK_STREAM, IPPROTO_TCP) = 3+0  setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0+0  setsockopt(3, SOL_TCP, TCP_NODELAY, [1], 4) = 0+0  bind(3, ..., ...) = 0+0  listen(3, 1) = 0+0 < S 0:0(0) win 10000 <mss 1000>+0 > S. 0:0(0) ack 1 <...>+0.01 < . 1:1(0) ack 1 win 10000+0 accept(3, ..., ...) = 4+0.01 write(4, ..., 6000) = 6000+0.01 < . 1:1(0) ack 1001 win 10000+0 < . 1:1(0) ack 1001 win 10000+0 < . 1:1(0) ack 1001 win 10000+0 < . 1:1(0) ack 1001 win 10000+0.01 < . 1:1(0) ack 3001 win 10000#

可以看到发送端在收到 ACK 数据包 AckNum 3001 之后，紧接着就重传了 Seq 3001:4001 的数据包。

# cat tcp_fast_retransmit_1_007.pkt `ethtool -K tun0 tso off ethtool -K tun0 gso off`0   socket(..., SOCK_STREAM, IPPROTO_TCP) = 3+0  setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0+0  setsockopt(3, SOL_TCP, TCP_NODELAY, [1], 4) = 0+0  bind(3, ..., ...) = 0+0  listen(3, 1) = 0+0 < S 0:0(0) win 10000 <mss 1000>+0 > S. 0:0(0) ack 1 <...>+0.01 < . 1:1(0) ack 1 win 10000+0 accept(3, ..., ...) = 4+0.01 < P. 1:1001(1000) ack 1 win 10000+0 < P. 2001:3001(1000) ack 1 win 10000+0 < P. 4001:5001(1000) ack 1 win 10000+0 < P. 5001:6001(1000) ack 1 win 10000#0

最后当然还可以追加，模拟注入确认所有数据段的 ACK 数据包 AckNum 6001 ，这个就不再继续实验了。

实验总结

通过数据接收端和数据发送端的结合实验验证，进一步了解了无 SACK 场景下的基础快速重传行为。

往期推荐

后台回复「TT」获取 Wireshark 提示和技巧系列合集
后台回复「TS」获取 Wireshark Troubleshooting 系列合集
如需交流或加技术群，可后台直接留言，我会在第一时间回复，谢谢！


推荐站内搜索：最好用的开发软件、免费开源系统、渗透测试工具云盘下载、最新渗透测试资料、最新黑客工具下载……
ZhouSa.com