Date: Thu, 18 Oct 2001 16:49:02 -0700 From: Lars Eggert <larse@ISI.EDU> To: net@freebsd.org, Yu-Shun Wang <yushunwa@ISI.EDU> Subject: ARP & IP fragments Message-ID: <3BCF6A6E.5000302@isi.edu>
next in thread | raw e-mail | index | archive | help
Hi, we're seeing a strange thing happening, related to ARP and IP fragments. In the tcpdump below, ifc is trying to send a large UDP message (~3x MTU) to dee. It does not have dee's MAC address in its ARP cache (happens both after an ARP timeout or an explicit ARP cache flush.) The ARP request/reply succeed, but then only the third IP fragment makes it out onto the wire: 16:36:27.550259 arp who-has dee.isi.edu tell ifc.isi.edu 16:36:27.550519 arp reply dee.isi.edu is-at 0:2:b3:2c:3:d9 16:36:27.550538 ifc.isi.edu > dee.isi.edu: (frag 11094:102@2960) (ttl 64, len 122) The IP protocol stats (netstat -s -p ip) go up correctly on ifc, "output datagrams fragmented" by 1 and "fragments created" by 3. The fragments never seem to make it into bpf or out on the wire though, and I see no other stats change or errors appear. This is with 4.4-RELEASE, but has happened (at least) since 4.2-RELEASE. Repeating this when an ARP entry for dee is present results in all three fragments to be sent: 16:36:54.255363 ifc.isi.edu.1309 > dee.isi.edu.1323: udp 3054 (frag 53580:1480@0+) (ttl 64, len 1500) 16:36:54.255369 ifc.isi.edu > dee.isi.edu: (frag 53580:1480@1480+) (ttl 64, len 1500) 16:36:54.255376 ifc.isi.edu > dee.isi.edu: (frag 53580:102@2960) (ttl 64, len 122) Any clues? Thanks, Lars -- Lars Eggert <larse@isi.edu> Information Sciences Institute http://www.isi.edu/larse/ University of Southern California To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3BCF6A6E.5000302>