Subject: TCP retrans - do they work?
To: None <current-users@NetBSD.ORG>
From: Simon J. Gerraty <firstname.lastname@example.org>
Date: 03/11/1995 23:43:13
I've seen some reports of problems via ethernet similar to what I'm
seeing over PPP. Basically lots of data sitting in tcp Send-Q's
(according to netstat) but nothing being transmitted.
Other TCP sessions over the same link work fine, but so far all
attempts to FTP data to the other end of the link get stuck.
netstat -ina shows:
ppp0 296 <Link> 5374 73 6058 0 0
ppp0 296 144.136 18.104.22.168 5374 73 6058 0 0
Ie. a few read errors but not excessive. My uarts do not have FIFO's,
the link is running at 9600 bps (over X.25 would you believe) but it
all works except for these TCP sessions which do not send their data.
Now a lost ACK followed by failure of the TCP re-transmission logic
would account for the symptoms.
I've also got one session:
tcp 0 15104 22.214.171.124.1606 126.96.36.199.20 FIN_WAIT_1
that session has been like that for 20 minutes - since I aborted the
transfer after it hung (with that same 15104 bytes in Send-Q).
I'd like to try and track this down, but would appreciate any useful
suggestions of where to start looking etc.
Are there some kernel structs I could easily query to see what TCP is
doing? The kernel had ddb in it, but I've not really played with it
yet so again any useful tips appreciated.