task 8.2 · chapter 8 of 8: a broken path · 25 min
The mystery fault
objective
The lab breaks the path to web one of three ways, at random. Capture at three points along the path and find out where the packets stop and why, from the packets alone.
answers/8.2.ymlopen a file in VS Code: Ctrl+P (Cmd+P on a Mac), type its name, Enter
| The fault | make break X=mystery (don’t read what it prints: scroll it away) |
|---|---|
| Capture | make capture AT=laptop:eth1,r1:eth2,r2:eth2 NAME=8.2 X=webip |
| Afterwards | make fix |
| Your answers | answers/8.2.yml: broken_router (r1 or r2) and cause (acl or route) |
new to this? step by stepbeginner
make break X=mystery, thenmake capture AT=laptop:eth1,r1:eth2,r2:eth2 NAME=8.2 X=webip.- Filter
tcp && frame.interface_name == "..."for each point in turn. Where does the SYN stop appearing? Does a SYN-ACK appear anywhere? - Answer the router and the cause, then
make fixandmake check T=8.2.
more in meet the lab
Divide the path
Silence is a clue, if you know where it starts
When a connection to a server just hangs, the client’s capture shows one thing: SYNs going out, again and again, and nothing coming back (task 6.2 told you why that’s different from a refusal). The client can’t tell you any more than that. The packets went somewhere, and stopped.
The method is the one you’d use for a dead string of Christmas lights: check halfway along. Capture at several points on the path at the same time, then follow one packet from point to point. The fault is between the last place you see it and the first place you don’t.
in one sentence Capture at several points at once and follow one packet: the fault is between the last point that saw it and the first that didn’t.
Two kinds of silence
On this path, two very different faults look the same from the laptop:
| Cause | What happens | What the captures show |
|---|---|---|
| acl: an access list on a router drops the traffic | the SYN reaches the router and goes no further | the SYN on the points before the router, on none after it |
| route: a router has no route back to the laptop | the SYN gets all the way to web; web answers; the answer dies on the way back | the SYN everywhere, a SYN-ACK leaving web that never reaches the laptop, and an ICMP Network unreachable from that router back to web |
An access list (ACL) is a router’s list of permit and deny rules for packets. A router that denies a
packet can tell the sender, with an ICMP administratively prohibited message, or drop it silently. The lab’s
fault is set up with no ip unreachables, so it drops silently: the most common setting on real networks, and
the hardest to debug.
A warning about one file, three cables
When one packet is captured at three points, it’s in your file three times. Wireshark’s TCP analysis doesn’t
know they’re the same packet seen in three places: it marks the second and third copies as
Retransmission or Dup ACK. Filtering the display to one interface doesn’t help: the flags were worked
out over the whole file, before the filter. So use frame.interface_name to see where a packet got to, and
don’t trust the tcp.analysis flags in a file with several capture points. If you need them, save one
interface’s frames to a file of their own (File › Export Specified Packets, with that interface’s filter
applied) and open that.
Three points on the path
laptop ──[laptop:eth1]── sw1 ── r1 ──[r1:eth2]── r2 ──[r2:eth2]── web
The capture points, from the laptop outwards:
| Point | What it tells you |
|---|---|
laptop:eth1 | what the laptop sent and received |
r1:eth2 | what got through r1, heading to r2 (and what came back from r2) |
r2:eth2 | what got through r2 and reached web’s cable (and what web sent) |
The traffic is X=webip: the laptop opens http://192.168.2.10/ by its address, so a DNS problem can’t get in
the way. A good capture plan always takes away the things you aren’t testing.
Here’s how one of the three faults begins, all three points in one listing:
2 laptop:eth1 39608 → 80 [SYN] Seq=0 Win=64240 Len=0 MSS=1460 ...
3 r1:eth2 [TCP Retransmission] 39608 → 80 [SYN] Seq=0 Win=64240 Len=0 MSS=1460 ...
4 r2:eth2 [TCP Retransmission] 39608 → 80 [SYN] Seq=0 Win=64240 Len=0 MSS=1460 ...
5 r2:eth2 80 → 39608 [SYN, ACK] Seq=0 Ack=1 Win=65160 Len=0 MSS=1460 ...
7 laptop:eth1 [TCP Retransmission] 39608 → 80 [SYN] Seq=0 Win=64240 Len=0 MSS=1460 ...
Frames 3 and 4 aren’t retransmissions: they’re frame 2, seen further along.
Find it
There’s no follow-along this time: this is the whole chapter in one task.
Break it, and don’t read the commands it prints:
run in the terminalmake break X=mystery clearCapture at all three points while the laptop tries to open the page:
run in the terminalmake capture AT=laptop:eth1,r1:eth2,r2:eth2 NAME=8.2 X=webipOpen
/captures/8.2.pcapng, and for each point in turn filter:tcp && frame.interface_name == "laptop:eth1"(then
"r1:eth2", then"r2:eth2"). Where does the SYN stop? Is there a SYN-ACK anywhere?Write your verdict, then undo the fault:
editanswers/8.2.ymlCtrl+P → 8.2.ymlbroken_router: r? # r1 or r2 cause: ? # acl or routerun in the terminalmake fixmake check T=8.2still knows which fault it was after the fix. Want another go?make break X=mysteryagain picks a new one at random (there are three).
hint
SYN on laptop:eth1 only: it didn’t get through r1. On laptop:eth1 and r1:eth2 but not r2:eth2: it
didn’t get through r2.
hint
A SYN-ACK on r2:eth2 means web got the SYN and answered. Then the problem is on the way back. Filter icmp:
did a router say anything about it?
hint
Prove it on the router: click r2 on the Topology tab, then in the r2 tab show ip route and show access-lists.
show the solutionanswers/8.2.yml
or pull it into your lab: make solve T=8.2, then git diff --staged shows what changed
It depends on the fault you got:
| What you saw | broken_router | cause |
|---|---|---|
SYN on laptop:eth1 only | r1 | acl |
SYN on laptop:eth1 and r1:eth2, nowhere on r2:eth2 | r2 | acl |
SYN on all three, SYN-ACK and an ICMP Network unreachable on r2:eth2 only | r2 | route |
In the last case, r2 has lost its route back to 192.168.1.0/24, so web’s answer never gets back to r1. r2
says so, but to web, not to you: the ICMP Network unreachable goes to the sender of the packet it couldn’t
route, which is web. The laptop never hears about it. Notice
what doesn’t help: from the laptop, all three look identical, and ping from web to r2 works fine. Only the
captures at the right points tell them apart.
5 check: run it, then prove it
$ make check T=8.2when it's right, you'll see
╭─ task 8.2 · the mystery fault
│ ✓ captures/8.2.pcapng: taken on laptop:eth1, r1:eth2, r2:eth2, ... frames
│ ✓ the router where the packets stop (challenge): ...
│ ✓ the cause: acl (a filter drops them) or route (no way back) (challenge): ...
╰─ 3/3 checks passed verified