tcp: mitigate ACK loops for connections as tcp_request_sock
authorNeal Cardwell <ncardwell@google.com>
Fri, 6 Feb 2015 21:04:39 +0000 (16:04 -0500)
committerDavid S. Miller <davem@davemloft.net>
Sun, 8 Feb 2015 09:03:12 +0000 (01:03 -0800)
commita9b2c06dbef48ed31cff1764c5ce824829106f4f
tree7e1dc6d03d3f2ba41a3a867fad3bf59c49665599
parent032ee4236954eb214651cb9bfc1b38ffa8fd7a01
tcp: mitigate ACK loops for connections as tcp_request_sock

In the SYN_RECV state, where the TCP connection is represented by
tcp_request_sock, we now rate-limit SYNACKs in response to a client's
retransmitted SYNs: we do not send a SYNACK in response to client SYN
if it has been less than sysctl_tcp_invalid_ratelimit (default 500ms)
since we last sent a SYNACK in response to a client's retransmitted
SYN.

This allows the vast majority of legitimate client connections to
proceed unimpeded, even for the most aggressive platforms, iOS and
MacOS, which actually retransmit SYNs 1-second intervals for several
times in a row. They use SYN RTO timeouts following the progression:
1,1,1,1,1,2,4,8,16,32.

Reported-by: Avery Fay <avery@mixpanel.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/tcp.h
include/net/tcp.h
net/ipv4/tcp_minisocks.c