[3452] in linux-net channel archive
Speaking of MTU problems
daemon@ATHENA.MIT.EDU (Dennis)
Mon Jun 24 11:22:42 1996
Date: Mon, 24 Jun 1996 10:12:41 -0400
To: linux-net@vger.rutgers.edu
From: dennis@etinc.com (Dennis)
Shouldn't the (physical) mtu of a device take into account to header
length? I noticed in 1.2.13 that sometimes it does, and sometimes
it doesn't ie if the MTU is 1500 and the hard_header_len is defined as,
say 16, sometimes it will send 1484 bytes of data to the device, and
others (specifically when generated internally, as with a PING packet)
it will send the full MTU when fragmenting. It caused a nasty bug a
while back when our buffer size was only the MTU size...much crashing
occurred :(
There are those that will argue that the MTU is the data size, but it is
difficult
to coordinate with directly connected devices when they have different
definitions
of the terms. (Ciscos have 2 mtus...phyisical and data). Perhaps its semantics,
but it shouldnt work differently in different scenarios.
Dennis
----------------------------------------------------------------------------
Emerging Technologies, Inc. http://www.etinc.com
Synchronous Communications Cards and Routers For
Discriminating Tastes. 56k to T1 and beyond. Frame
Relay, PPP, HDLC, and X.25 for BSD/OS, FreeBSD
and LINUX
