Date: Sun, 24 Jan 1999 16:56:05 -0800 From: John-Mark Gurney <gurney_j@efn.org> To: current@FreeBSD.ORG Subject: bug with TCP over lossy lines? Message-ID: <19990124165605.05029@hydrogen.nike.efn.org>
next in thread | raw e-mail | index | archive | help
well, I've seen a bug where the machine will actually recieve a packet from the remote machine that ack's data, but for some reason it will not get to the stack, and the local machine will continue to send old data... the local machine is metriclient-1.uoregon.edu, the remote machine is jupiter.uoregon.edu... I'm connected via a Ricochet wireless modem... it opperates in the 900Mhz range and causes (and I believe is susceptible to) interference to/from 900Mhz phones... here is a tcpdump run on metriclient-1: 16:37:16.547109 metriclient-1.uoregon.edu.3016 > jupiter.uoregon.edu.ssh: . 10880:12340(1460) ack 41 win 17520 (DF) [tos 0x8] 16:37:21.707361 jupiter.uoregon.edu.ssh > metriclient-1.uoregon.edu.3016: . ack 13800 win 8760 (DF) [tos 0x8] 16:37:21.709420 metriclient-1.uoregon.edu.3016 > jupiter.uoregon.edu.ssh: . 13800:15260(1460) ack 41 win 17520 (DF) [tos 0x8] 16:37:21.711135 metriclient-1.uoregon.edu.3016 > jupiter.uoregon.edu.ssh: . 15260:16720(1460) ack 41 win 17520 (DF) [tos 0x8] 16:37:29.666465 jupiter.uoregon.edu.ssh > metriclient-1.uoregon.edu.3016: . ack 15260 win 8760 (DF) [tos 0x8] 16:37:29.666911 metriclient-1.uoregon.edu.3016 > jupiter.uoregon.edu.ssh: . 16720:18180(1460) ack 41 win 17520 (DF) [tos 0x8] 16:37:29.668611 metriclient-1.uoregon.edu.3016 > jupiter.uoregon.edu.ssh: . 18180:19640(1460) ack 41 win 17520 (DF) [tos 0x8] 16:37:40.866012 jupiter.uoregon.edu.ssh > metriclient-1.uoregon.edu.3016: . ack 16720 win 8760 (DF) [tos 0x8] 16:37:46.545981 metriclient-1.uoregon.edu.3016 > jupiter.uoregon.edu.ssh: . 15260:16720(1460) ack 41 win 17520 (DF) [tos 0x8] 16:38:20.544650 metriclient-1.uoregon.edu.3016 > jupiter.uoregon.edu.ssh: . 15260:16720(1460) ack 41 win 17520 (DF) [tos 0x8] why are we sending data that the remote has already ack'd?? it does finally recover as jupiter will send another ack packet for even more data, but it doesn't explain why we send data that the remote host ack'd... the line that I'm going over has a VERY high latency right now.. ping is returning rtt's in the 4-8second range... I am running natd on the local host, but I have seen something VERY similar happen on my notebook running an older 3.0-current... FreeBSD hydrogen.fircrest.net 3.0-CURRENT FreeBSD 3.0-CURRENT #9: Sun Dec 6 11:52:59 PST 1998 jmg@hydrogen.fircrest.net:/a/home/johng/FreeBSD-checkout/30r/sys/compile/hydrogen i386 and the kernel has been built with these files: tcp.h: $Id: tcp.h,v 1.10 1998/07/13 11:09:51 bde Exp $ tcp_debug.c: $Id: tcp_debug.c,v 1.13 1998/08/17 01:05:25 bde Exp $ tcp_debug.h: $Id: tcp_debug.h,v 1.7 1997/02/22 09:41:37 peter Exp $ tcp_fsm.h: $Id: tcp_fsm.h,v 1.10 1997/08/16 19:15:38 wollman Exp $ tcp_input.c: $Id: tcp_input.c,v 1.81 1998/09/11 16:04:03 wollman Exp $ tcp_output.c: $Id: tcp_output.c,v 1.31 1998/07/13 11:53:59 bde Exp $ tcp_seq.h: $Id: tcp_seq.h,v 1.8 1997/02/22 09:41:41 peter Exp $ tcp_subr.c: $Id: tcp_subr.c,v 1.47 1998/09/06 08:17:35 phk Exp $ tcp_timer.c: $Id: tcp_timer.c,v 1.28 1998/04/24 09:25:35 dg Exp $ tcp_timer.h: $Id: tcp_timer.h,v 1.13 1997/09/07 05:26:48 bde Exp $ tcp_usrreq.c: $Id: tcp_usrreq.c,v 1.38 1998/08/23 03:07:15 wollman Exp $ tcp_var.h: $Id: tcp_var.h,v 1.48 1998/08/24 07:47:39 dfr Exp $ tcpip.h: $Id: tcpip.h,v 1.7 1998/09/26 14:26:59 dfr Exp $ -- John-Mark Gurney Voice: +1 541 684 8449 Cu Networking P.O. Box 5693, 97405 Live in Peace, destroy Micro$oft, support free software, run FreeBSD Don't trust anyone you don't have the source for To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990124165605.05029>