[15806] in Athena Bugs
Re: No path MTU discovery in Solaris TCP?
daemon@ATHENA.MIT.EDU (John Hawkinson)
Mon Feb 23 17:42:55 1998
Date: Mon, 23 Feb 1998 17:42:52 -0500 (EST)
To: John Hawkinson <jhawk@MIT.EDU>
Cc: bugs@MIT.EDU
In-Reply-To: "[15801] in Athena Bugs"
From: John Hawkinson <jhawk@MIT.EDU>
Ah. So it looks like this situation only arises when a machine that does
not offer an MSS connects to a Solaris machine.
This is not the normal case, but is still a relatively frequent occurance
(for instance, anytime someone does an ftp from the a Solaris machine
to an older machine (e.g. SunOS) and does a put).
I'd recommend boosting tcp_mss_def to the max ifmtu. Three different approaches
come to mind:
1. Dynamically check the max ifmtu at boot time and twiddle
things.
2. Hardcode tcp_mss_def to 1500 on the assumption that everybody
is basically using ethernet. (Do this in /etc/system).
3. Hardcode tcp_mss_def to 9180 on the assumption that that's
the largest ifmtu we'll ever see under Solaris and it doesn't
do any harm to set tcp_mss_def higher than the ifmtu since
mtu discovery should pick the min(tcp_mss_def,ifmtu) when
sizing packets. This would be /etc/system.
Not sure which is preferred. Probably 3...
--jhawk
