mailing list archives
[patch] traceroute release
From: "Eddie Bell" <ejlbell () gmail com>
Date: Mon, 28 Aug 2006 20:26:44 +0100
Here is the last patch from my summer of code. It is a parallel
traceroute patch, activated with --traceroute. It was mainly developed
on linux so any windows testing would be appreciated. Bugs reports
or feature ideas are welcome
o The patch for 4.20ALPHA4 is attached
o Source distribution can be downloaded here
o Windows binary distribution can be downloaded here
For anyone interested in how it works, ill past the information
included in Traceroute.cc below
* A multi-protocol parallel traceroute implementation for nmap.
* For more information on how traceroutes work:
* Nmaps traceroute takes in a list of scanned targets and determines a valid
* responsive port to trace to based on the scan results, scan protocol and
* various pieces of protocol data.
* It first sends a probe to the target port. From the reply traceroute is able
* to infer how many hops away the target is. Nmap starts the trace by sending
* a packet with a TTL equal to that of the hop distance guess. If it gets an
* ICMP_TTL_EXCCEDED message back it know the hop distance guess was under so
* nmap will continue sending probes with incremental TTLs until it receives a
* reply from the target host.
* Once a reply from the host is received nmap sets the TTL to once below the
* hop guess and continues to send probes with decremental TTLs until it reaches
* TTL 0. Then we have a complete trace to the target. If nmap does not get a
* hop distance probe reply, the trace TTL starts at one and is incremented
* until it hits the target host.
* Forwards/Backwards tracing example
* hop guess:20
* send:20 --> ICMP_TTL_EXCCEDED
* send:21 --> ICMP_TTL_EXCCEDED
* send:22 --> Reply from host
* send:19 --> ICMP_TTL_EXCCEDED
* send:1 --> ICMP_TTL_EXECCEDED
* The forward/backwards tracing method seems a little convoluted at first but
* there is a reason for it. The first host traced in a Target group is
* designated as the reference trace. All other traces
* (once they have reached their destination host) are compared against the
* reference trace. If a match is found the trace is ended prematurely and the
* remaining hops are assumed to be the same as the reference trace. This
* normally only happens in the lower TTls, which rarely change. On average nmap
* sends 5 less packets per host. If nmap is tracing related hosts
* (EG. 126.96.36.199/24) it will send a lot less packets. Depending on the network
* topology it may only have to send a single packet to each host.
* Forwards/Backwards tracing saves bandwidth, speeds up the scan and stops local
* routers from rate limiting ICMP_TTL_EXCCEDED messages
* Traceroute employs a dynamic timing model similar to nmaps main
* but a little more light weight. It keeps track of sent, received and dropped
* packet, then adjusts timing parameters accordingly. The parameters
are; number of
* retransmissions, delay between each sent packet and the amount of time to wait
* for a reply. This are initialised based on the timing level (-T0 to -T5)
* The output from each trace is consolidated to save space, XML
logging and debug
* mode ignore consolidation. There are two types of consolidation timeout and
* reference trace.
* Timed out
* 1 ... Hop 1 timed out
* 1 ... 6 Hop 1 to 6 timed out
* Reference trace
* 1 -- Hop 1 has been taken from the reference trace
* 1 --> 6 Hop 1 to 6 have been taken from the reference trace
* Traceroute also optimises performance by using DNS caching (nmap_dns.cc).
* Each host name resolved is stored in a cache and each time it is used a
* cache hit variable is incremented. When the cache is full, entries with
* the lowest cache hit counter are evicted. This algorithm is called
* "least frequently used with ageing". Dns name resolution is slow
* (not because of nmap) so caching improves traceroute times especially
* because many of the host names will reoccur frequently.
* Traceroute does not work with connect scans or idle scans and has trouble
* with ICMP_TIMESTAMP and ICMP_ADDRESSMASK scans because so many host filter
* them out. The quickest seems to be SYN scan.
* Bugs and Future Features:
* o The code, currently, only works with ipv4.
* o Send both UDP and TCP hop distance probes no matter what the
* scan protocol
Sent through the nmap-dev mailing list
Archived at http://SecLists.Org
- [patch] traceroute release Eddie Bell (Aug 28)