From owner-freebsd-questions Mon May 31 11:52: 8 1999 Delivered-To: freebsd-questions@freebsd.org Received: from stennis.ca.sandia.gov (stennis.ca.sandia.gov [146.246.243.44]) by hub.freebsd.org (Postfix) with ESMTP id C27B814EA9 for ; Mon, 31 May 1999 11:52:06 -0700 (PDT) (envelope-from bmah@stennis.ca.sandia.gov) Received: (from bmah@localhost) by stennis.ca.sandia.gov (8.9.3/8.9.3) id LAA24784; Mon, 31 May 1999 11:51:51 -0700 (PDT) Message-Id: <199905311851.LAA24784@stennis.ca.sandia.gov> X-Mailer: exmh version 2.1.0 04/14/1999 To: slava Cc: freebsd-questions@FreeBSD.ORG Subject: Re: [off-topic] DF bit and IP In-reply-to: Your message of "Mon, 31 May 1999 10:35:24 -0000." From: bmah@CA.Sandia.GOV (Bruce A. Mah) Reply-To: bmah@CA.Sandia.GOV X-Face: g~c`.{#4q0"(V*b#g[i~rXgm*w;:nMfz%_RZLma)UgGN&=j`5vXoU^@n5v4:OO)c["!w)nD/!!~e4Sj7LiT'6*wZ83454H""lb{CC%T37O!!'S$S&D}sem7I[A 2V%N&+ X-Url: http://www.ca.sandia.gov/~bmah/ Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Mon, 31 May 1999 11:51:51 -0700 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG If memory serves me right, slava wrote: > Does TCP always gets encapsulated in an IP header with DF bit set? > I know this is needed for path MTU discovery to make tcp more efficient > but is this implemented on all OSes? Not all OSs implement PMTU discovery, but I admit that I don't recall off the top of my head who does and who doesn't. > What if something in the way is blocking the icmp packet-too-big type > to the initiator of a TCP connection and it never finds out about a small > MTU in the path? As far as the sending TCP is concerned, the TCP segments that would have generated those ICMP packets just vanished. It then tries again (very likely with the same packet sizes), and the packets vanish again, until it finally concludes the receiving host is dead and gives up. This is Very Bad (TM). > Will it retry with a smaller MTU itself? Not in most cases, no, because TCP can't tell the difference between "packet lost due to congestion" and "packet apparently lost because the ICMP packet-too-big message was blocked". Some ISPs think they're being smart by blocking all ICMP packets, but doing so plays heck with PMTU discovery. At one point, someone sent me the pointer below, which has some more explanation: http://www.freelabs.com/~whitis/isp_mistakes.html (I dimly recall, however, a survey of PMTU sizes along various paths which showed that a high fraction of paths supported sending of Ethernet-MTU-sized packets.) Bruce. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message