Date: Sun, 24 Mar 1996 00:11:46 -0500 From: "Louis A. Mamakos" <louie@TransSys.COM> To: Brian Tao <taob@io.org> Cc: FREEBSD-HACKERS-L <freebsd-hackers@FreeBSD.org> Subject: Re: Changing Ethernet frame size to 576 bytes? Message-ID: <199603240511.AAA20790@wa3ymh.transsys.com> In-Reply-To: Your message of "Sat, 23 Mar 1996 22:27:38 EST." <Pine.BSF.3.91.960323222600.8944G-100000@cabal.io.org>
next in thread | previous in thread | raw e-mail | index | archive | help
There is much confusion evident in the email message. First, the 576 byte packet size mentioned is the *minimum* MTU which is "required" to be supported. Or something like that. In practice, most every link level encapsulation in use supports a 1500 byte MTU. The big exception to this these days is SLIP/PPP where a much smaller MTU is used to prevent interactive traffic from being "stuck" behind really large packets and screwing the echo response. Second, any fragmentation which would occur would happen in a IP router, which has an IP datagram it needs to transmit. If the interface which the packet needs to be forwarded over has an MTU smaller than the packet, then the fragmentation function occurs at that point. That is, unless the "don't fragment" bit is set. It's somewhat likely that it may be set, as the TCP in FreeBSD does MTU path discovery to "feel" the MTU along the path from the sender to the recpient's TCP. This is done by setting "Don't Fragment", and looking for ICMP messages coming back. When the path's MTU is discovered, the TCP can arrange to not send TCP segments larger than the MTU. Third, most long-haul links on the Internet today likely have 4K MTUs, which is usually used on FDDI and HSSI (to connect to DS3 links) interfaces. So, it's pretty unlikely that ethernet MTU sizes packets would be a problem. Fourth, you can verify that there's actually dropped fragments happening by looking on the receiving machine and doing a 'netstat -s' to look for the number of fragments received, as well as the number dropped due to reassembly timeouts. I suspect that the number will be 0. I don't think that you actually have a problem, or anything that you need to change.. louie > Anyone know what this guy is saying? I figured fragmentation and > reassembly would happen between the FTP server's Ethernet interface > and that of the router to the Internet. Is there any validity to this > guy's suggestion?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199603240511.AAA20790>