Bug#457638: confused by router that always responds as the destination host
On Wed, Jan 02, 2008 at 04:01:18PM +0100, Rogier Wolff wrote: > How traceroute decides that it reaches the destination I don't > remember. Traceroute however takes ages: It sends out one packet, > waits for the result or times out, and then tries the next hop. > (after trying the same hop 3 times). Dunno about all versions of traceroute, but at least a couple of them send out packets to the first hop to a high UDP port number, then the next hop to that port number + 1, then the third hop to port number + 2, etc. ICMP doesn't have port numbers, but you can tuck freeform data into the echo request packet, which in theory makes it back in the reply. For most routers. I think. -- Robert Woodcock - [EMAIL PROTECTED] "We would like to return to work with our writers. If we cannot, we would like to express our ambivalence, but without our writers we are unable to express something as nuanced as ambivalence." -- Jon Stewart and Stephen Colbert -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#457638: confused by router that always responds as the destination host
On Mon, Dec 24, 2007 at 03:46:02PM -0500, Joey Hess wrote: > Robert Woodcock wrote: > > On Mon, Dec 24, 2007 at 10:07:47AM -0500, Joey Hess wrote: > > > ICMP (-I) works, TCP SYN (-T) doesn't. First, This router is completely broken. It foges packets from the destination. As everybody else does it differently I'm pretty sure you can find an RFC that tells you this is wrong. We can't program against all possible router faults. How traceroute decides that it reaches the destination I don't remember. Traceroute however takes ages: It sends out one packet, waits for the result or times out, and then tries the next hop. (after trying the same hop 3 times). Mtr tries to do things a bit more parallel. This means that it sends out request for multiple hops at once. It determins it's reached the destination when at a certain hop, it gets back a reply from the destination. This is exactly what you've seen. In practise, we COULD probably detect this situation: mtr sends out requests for 5 hops further than it knows it is reasonable. So the first round, it sends out a request for hops 1 through 5. If it then gets a reply from hop 4, it will in the next round probe for 1 through 9. On the other hand, it concludes it reached the destination on the reply to the first probe. This probably happens before the probe for hop 2 goes out (at 200ms). Although you might want different results, mtr is behaving "as designed", and the fact that you get odd results is the fault of your odd network, and not mtr's. Roger. -- ** [EMAIL PROTECTED] ** http://www.BitWizard.nl/ ** +31-15-2600998 ** **Delftechpark 26 2628 XH Delft, The Netherlands. KVK: 27239233** *-- BitWizard writes Linux device drivers for any device you may have! --* Q: It doesn't work. A: Look buddy, doesn't work is an ambiguous statement. Does it sit on the couch all day? Is it unemployed? Please be specific! Define 'it' and what it isn't doing. - Adapted from lxrbot FAQ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#457638: confused by router that always responds as the destination host
Robert Woodcock wrote: > On Mon, Dec 24, 2007 at 10:07:47AM -0500, Joey Hess wrote: > > ICMP (-I) works, TCP SYN (-T) doesn't. > > > > (That's probably the last time I'll be able to use this network.) > > Well shoot. If it's not too late to run 'tcpdump -n -s 1500 -w dumpfile' > while you do the tests Sorry, too late. :-) -- see shy jo signature.asc Description: Digital signature
Bug#457638: confused by router that always responds as the destination host
On Mon, Dec 24, 2007 at 10:07:47AM -0500, Joey Hess wrote: > ICMP (-I) works, TCP SYN (-T) doesn't. > > (That's probably the last time I'll be able to use this network.) Well shoot. If it's not too late to run 'tcpdump -n -s 1500 -w dumpfile' while you do the tests, then do those (one for mtr and one for traceroute -I) and email them to me - otherwise, we'll probably have to wait for someone else to say "hey this happens to me too!" -- Robert Woodcock - [EMAIL PROTECTED] "We've worked with our purchasing team to bring in many companies to compete on our main cabin tidbit item (pretzels). To date, no one has been able to match our current cost, about 3 cents per package." -- US Airways management -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#457638: confused by router that always responds as the destination host
Robert Woodcock wrote: > On Sun, Dec 23, 2007 at 09:36:32PM -0500, Joey Hess wrote: > > I think mtr must optimise away the rest, which seems like a sort of bug in > > this situation. > > It's important to note that mtr, by design, sends out ICMP packets, and > traceroute, by default, sends out UDP packets. > > I suspect if you use the traceroute packages' traceroute command with the > -I switch on a host behind such a router, you'll get identical results. ICMP (-I) works, TCP SYN (-T) doesn't. (That's probably the last time I'll be able to use this network.) -- see shy jo signature.asc Description: Digital signature
Bug#457638: confused by router that always responds as the destination host
On Sun, Dec 23, 2007 at 09:36:32PM -0500, Joey Hess wrote: > I think mtr must optimise away the rest, which seems like a sort of bug in > this situation. It's important to note that mtr, by design, sends out ICMP packets, and traceroute, by default, sends out UDP packets. I suspect if you use the traceroute packages' traceroute command with the -I switch on a host behind such a router, you'll get identical results. If packets sent out with a TTL of, say, 30, come back with a TTL exceeded message from the same host as packets with a TTL of 1, there is no software method of determining the existence of the other hops. So I think this is a case of garbage in, garbage out. Let me know what you get. -- Robert Woodcock - [EMAIL PROTECTED] "We would like to return to work with our writers. If we cannot, we would like to express our ambivalence, but without our writers we are unable to express something as nuanced as ambivalence." -- Jon Stewart and Stephen Colbert -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
Bug#457638: confused by router that always responds as the destination host
Package: mtr-tiny Version: 0.72-2 Severity: normal At the moment, I'm on a network where the router seems to always respond to traceroutes as if it's the destination host: traceroute to kitenet.net (80.68.85.49), 30 hops max, 40 byte packets 1 wren.kitenet.net (80.68.85.49) 3.672 ms 3.915 ms 4.176 ms 2 69-77-144-1.skybest.com (69.77.144.1) 27.488 ms 31.056 ms 32.551 ms 3 host73.embarqnow.net (207.14.228.73) 34.936 ms 37.630 ms 39.313 ms 4 165.166.133.57 (165.166.133.57) 54.685 ms 56.792 ms 61.700 ms 5 gi4-3.ccr01.clt01.atlas.cogentco.com (38.104.168.9) 62.331 ms 64.015 ms 66.463 ms 6 te2-2.ccr02.dca01.atlas.cogentco.com (154.54.7.141) 162.628 ms 161.886 ms 163.362 ms 7 te3-3.ccr04.jfk02.atlas.cogentco.com (154.54.6.14) 165.515 ms 42.146 ms 118.178 ms 8 gi13-0-0.core02.jfk02.atlas.cogentco.com (154.54.5.233) 101.473 ms 101.673 ms 101.843 ms 9 po2-0.core01.lon01.atlas.cogentco.com (66.28.4.190) 127.005 ms 130.158 ms 131.162 ms 10 te7-1.mpd01.lon01.atlas.cogentco.com (130.117.1.78) 134.613 ms 138.251 ms 141.777 ms 11 linx-226-gw.thdo.bogons.net (195.66.226.228) 145.628 ms 169.674 ms 169.834 ms 12 cr0-G1-1.rbsov.bogons.net (193.178.223.218) 170.023 ms 170.327 ms 170.434 ms 13 43-242.rbsov.bogons.net (85.158.43.242) 203.535 ms 203.637 ms 203.747 ms 14 gi1-0.core1.man.bytemark.co.uk (89.16.160.41) 203.976 ms 204.178 ms 204.317 ms 15 gi0-23.agg1.man.bytemark.co.uk (89.16.160.58) 204.541 ms 204.729 ms 124.986 ms 16 ppp.dh.bytemark.co.uk (89.16.161.184) 127.403 ms 130.764 ms 133.188 ms 17 wren.kitenet.net (80.68.85.49) 136.618 ms 140.763 ms 143.048 ms At first I tried mtr and it was very confusing, seemed to be one hop to anywhere: Host Loss% Snt Last Avg Best Wrst StDev 1. wren.kitenet.net 0.0%29 16.1 13.4 1.3 24.3 6.1 I think mtr must optimise away the rest, which seems like a sort of bug in this situation. -- System Information: Debian Release: lenny/sid APT prefers unstable APT policy: (500, 'unstable'), (500, 'testing'), (1, 'experimental') Architecture: i386 (i686) Kernel: Linux 2.6.22-3-686 (SMP w/1 CPU core) Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash Versions of packages mtr-tiny depends on: ii libc6 2.7-5 GNU C Library: Shared libraries ii libncurses5 5.6+20071215-1 Shared libraries for terminal hand mtr-tiny recommends no packages. -- no debconf information -- see shy jo signature.asc Description: Digital signature