Date: Sun, 9 Feb 1997 10:40:12 +0300 (MSK) From: bvt@mp.aha.ru (Boris Tyshkiewitch) To: jas@flyingfox.COM (Jim Shankland) Cc: bugs@freebsd.org, freebsd-isp@freebsd.org, freebsd-questions@freebsd.org, robert@nanguo.chalmers.com.au Subject: Re: I give up! no ideas left. Message-ID: <199702090740.KAA19046@mp.aha.ru> In-Reply-To: <199702090652.WAA07862@saguaro.flyingfox.com> from "Jim Shankland" at Feb 8, 97 10:52:01 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > Here's what I find when I connect to port 80 of your machine > (nanguo.chalmers.com.au) and do a "GET / HTTP/1.0": > > >From SunOS 4.1.4: everything works. > > >From FreeBSD 2.2-BETA_A: Connection hangs. My end is ESTABLISHED. > A packet trace reveals that after the initial, 3-way TCP handshake, > I receive the *second* data packet from you (covering bytes > 1440:2049, or something like that), but I never get the first > (bytes 1:1440). Of course, my end immediately does an ACK 1 > to signal that it got an out-of-sequence packet; but to no > avail. That packet simply never arrives. > > Same thing happens whether RFC1323 and RFC1644 are enabled or > not; so my tentative conclusion is that they are not a factor. > I have souch problem. squid cache works fine from FreeBSD 2.1.0 but connections hungs from 2.2-BETA. host in question was www.kenwoodusa.com After tricks with tcpdump I see picture like discribed above. I found differense. 2.1.0 begin session with mss 512, 2.2-BETA - with 1460. After setting MTU 1006 or less on 2.2 box they work fine. When I do traceroute to www.kenwoodusa.com, they failed (ping still ok). I think that some router in that place discard ICMP replayes, so MTU discovery not working .... Why 2.1.0 begin session from smaller MSS values that 2.2-BETA? Why 2.2-BETA do that? What is right way? Boris.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199702090740.KAA19046>