amd64/123603: tcp_do_segment and Received duplicate SYN

[John Baldwin]

On Monday 12 May 2008 03:45:16 am John wrote:
> >Number:         123603
> >Category:       amd64
> >Synopsis:       tcp_do_segment and Received duplicate SYN
> >Confidential:   no
> >Severity:       critical
> >Priority:       high
> >Responsible:    freebsd-amd64
> >State:          open
> >Quarter:
> >Keywords:
> >Date-Required:
> >Class:          sw-bug
> >Submitter-Id:   current-users
> >Arrival-Date:   Mon May 12 07:50:01 UTC 2008
> >Closed-Date:
> >Last-Modified:
> >Originator:     John
> >Release:        FB7.0 (x64)
> >Organization:
>
> NULL
>
> >Environment:
>
> FreeBSD mail.mydomain.com 7.0-RELEASE FreeBSD 7.0-RELEASE #0: Thu Mar  6
> 12:04:57 HKT 2008     root at mydomain.com:/usr/src/sys/amd64/compile/FB7NEW 
> amd64
>
> >Description:
>
> A FreeBSD 7.0 (x64) Lighttpd Web Server with most-updated ports and patchs.
> when a client connect and view a static HTML file, At the first time
> (before web server idle time), it needs to wait a long time to establish a
> connection, OR when this server try to download file from Internet, there
> are lots of logs messages just like below:
>
> May 12 11:57:54 mail kernel: TCP: [55.66.77.88]:41792 to [11.22.33.44]:80
> tcpflags 0x2<SYN>; syncache_add: Received duplicate SYN, resetting timer
> and retransmitting SYN|ACK
> May 12 15:17:53 mail kernel: TCP: [193.166.3.2]:45979 to
> [11.22.33.44]:63372 tcpflags 0x10<ACK>; tcp_do_segment: FIN_WAIT_1:
> Received 1448 bytes of data after socket was closed, sending RST and
> removing tcpcb May 12 15:17:53 mail kernel: TCP: [193.166.3.2]:21 to
> [11.22.33.44]:55007 tcpflags 0x18<PUSH,ACK>; tcp_do_segment: FIN_WAIT_2:
> Received 13 bytes of data after socket was closed, sending RST and removing
> tcpcb
>
> >How-To-Repeat:
>
> any type of  connection will generate above log messages.

You can either comment out all the log(LOG_DEBUG, ...) calls 
in /sys/netinet/tcp*.c or change your /etc/syslog.conf to not send kern.debug 
messages to the console.

I think these messages should probably be conditional on a kernel option FWIW.

-- 
John Baldwin