Wireshark mailing list archives

Re: libtshark + scripting language support

From: "Thierry Emmanuel" <Emmanuel.Thierry () technicolor com>
Date: Fri, 20 Aug 2010 10:51:48 +0200

Hello,

I can briefly, according to the humble experience I have had with Wireshark, give some (technical) pros and cons about 
linking directly with libwireshark :
- Advantages : Firstly, some features you can't access by the other way. You have the dissect tree under his native 
structure, so you can browse it or even dump it. I think it's easier to manipulate data under that format. Secondly, 
you can pass to the library text coming from any medium, and decode it as any protocol. I capture the data from sockets 
and I pass only the application layer data to Wireshark, and ask to it to decode it as the protocol I want.
- Drawbacks : Firstly, you are dependent from bugs present in Wireshark. Wireshark isn't, in my humble opinion, 
suitable for long time captures. I have had memory leaks on part of the code I didn't control (honestly, on a huge 
software like Wireshark, I don't think that such bugs could be avoided). Secondly, Wireshark isn't made to work as a 
simple library, there are lots of dependences between epan and some other parts of the code. I don't think the main 
consideration for the team is to make an independent build of the epan (But I wrote in my previous message, I think it 
would be a great enhancement for this project).

Best regards



-----Original Message-----
From: wireshark-dev-bounces () wireshark org [mailto:wireshark-dev-bounces () wireshark org] On Behalf Of Mark 
Landriscina
Sent: jeudi 19 août 2010 21:13
To: Developer support list for Wireshark
Subject: Re: [Wireshark-dev] libtshark + scripting language support


Eloy,

I've picked up on two separate threads in this discussion. First is, "does my approach work for me?" I think that it 
does and I appreciate your comments on this topic. Second is,"is there much utility is offering my code up as another 
wireshark related project?" I thought that the answer to this question might be 'yes'. However, I am hearing that there 
might be better/more useful approaches if the goal is to start a project that makes wireshark dissection capabilties 
more available to scripting languages such as Python, Ruby, etc. Also, the Wireshark listserve admin gods haven't 
exactly been beating down my door to get ahold of my code ...

Given the above, I was thinking about the following course of action.

1. Use what I have already written for my current project. The folks that I am working with are already familiar with 
tshark and having a tshark-like Python experience works for them. Also, I am still in the "this is so cool!" phase of 
my relationship with this code.

2. Future work, however, might be more interesting/usable by others if I try to develop something that more directly 
exposes Wireshark dissection functionality and makes it more easily available to scripting languages. Not sure how I 
would do this yet, however.

I do have a quick related question for you. I noticed that you you support specific versions of  libwireshark in your 
project. Has keeping up with the evolution of wireshark code been terribly onerous?

Regards,
Mark 

----- Original Message -----
From: Eloy Paris <peloy () chapus net>
Date: Thursday, August 19, 2010 1:54 pm
Subject: Re: [Wireshark-dev] libtshark + scripting language support
To: wireshark-dev () wireshark org

Hi Mark,

On 08/19/2010 09:02 AM, Mark Landriscina wrote:

Hmmm... I think that you, Guy, and Eloy have given me good cause to
go back and rethink my approach. I really appreciate the detailed
feedback.

I'll take a closer look at your suggestion to pull together
dissection capabilities as an independent lib. If you happen to have
any notes/thoughts/past work on this sitting around and wouldn't mind
sharing, I'd certainly love to see it.


I actually think there is nothing wrong with your approach, and in 
fact, 
someone suggested that I launch tshark in Network Expect instead of 
(actually, perhaps "in addition to") using libwireshark services.

The reason is that building Network Expect and linking against 
libwireshark may be challenging for some users that don't have a lot 
of 
experience and by just launching tshark building becomes a lot easier.

I'd say that if what you have works for you, keep working on it and 
making it better. There are many ways to skin a cat and one is not 
necessarily better than the other.

I don't think this will be a problem in your case, but also keep in 
mind 
that if you use libwireshark services (link against it) then you only 

have one license choice for your project (GPL). If you run tshark that 

is considered "at arm's length" and then your license can be anything 

you choose.

Cheers,

Eloy Paris.-
netexpect.org


----- Original Message -----
From: Thierry Emmanuel<Emmanuel.Thierry () technicolor com>
Date: Thursday, August 19, 2010 4:29 am
Subject: Re: [Wireshark-dev] libtshark + scripting language support
To: Developer support list for Wireshark<wireshark-dev () wireshark org>

Hello,

I share the same point of view than Guy.
For my current project, I have integrated Wireshark in a monitoring
probe, which permit to have a dissection without additional work, and
having a fully detailed packets capture.

I have worked very differently than you, considering Wireshark as a
library itself. If you take a look at the epan directory, you'll see
that you have all the tools you need to decode any kind of packet.

You can :
- Init the library with "epan_init" and "init_dissection" functions
- Find a dissector with "dissector_table_foreach_handle" and
"dissector_handle_get_protocol_index" functions
- Request the library to process your data against the protocol you
want with "call_dissector_only" (from layer 2 to 7, for example I
decode http or icmp packets as well)
- Access to any part of the dissected packet with the ptree and finfo
structures

I succeeded to use this by getting my packets from simple SOCK_STREAM,
SOCK_DGRAM, SOCK_RAW sockets, or from a libpcap binding (which is also
accessible from Wireshark). So this library gives per self a fully
usable interface, you just have to compile with header files located
in epan/ directory and some other in the root directory, and to link
against libwireshark.so.
So I don't think that you have simplified your work by wrapping tshark
into a new library.

But the debate is very interesting. When I registered to this list

to

ask how I could use independently the dissection code, one said to

me

that nobody would be able to help me because it wasn't a common use

of

Wireshark. And I think that it would be a great plus to provide the
dissecting stuff as an independent library. The dissecting abilities
of Wireshark are really huge, and it would be enormous if it was
independent. It would multiply its possibilities. I have worked a bit
on this kind of use so I would be glad to give help and comments if
you opened such a project.

Best regards.


-----Original Message-----
From: wireshark-dev-bounces () wireshark org [ On Behalf Of Mark Landriscina
Sent: mercredi 18 août 2010 21:37
To: wireshark-dev () wireshark org
Subject: Re: [Wireshark-dev] libtshark + scripting language support

Guy,

Only need to link to libtshark.a. No need to link to libwireshark,
etc. Tshark.c does actually contain a fair amount of other useful code
that ties all the dissection stuff nicely together. My original
approach was to just draw on libwireshark and libwiretap code directly
but found that I was simply rewriting a basic version tshark.

Reason for the named-pipe was that I wanted to launch several
instances of tshark from within Python have them doing different
things and then collect their dissections via separate data streams.
Writing the dissection data over a named pipe seemed like a clean,
painless way to do this. Additionally, I wanted a flexible way to
export the dissection data in the event that I decided to do something
else with this code such as compile libtshark as an executable
(tshark) instead of a lib. I'd still be able to have the tshark
executable export its dissection data to other applications in binary
form (as opposed to printing it out in pdml format and parsing text).


I'm still playing around with the code and different ideas, so pls
feel free to share any ideas for better approaches.

----- Original Message -----
From: wireshark-dev-request () wireshark org
Date: Wednesday, August 18, 2010 3:00 pm
Subject: Wireshark-dev Digest, Vol 51, Issue 22
To: wireshark-dev () wireshark org

Send Wireshark-dev mailing list submissions to
  wireshark-dev () wireshark org

To subscribe or unsubscribe via the World Wide Web, visit
  
or, via email, send a message with subject or body 'help' to
  wireshark-dev-request () wireshark org

You can reach the person managing the list at
  wireshark-dev-owner () wireshark org

When replying, please edit your Subject line so it is more specific
than "Re: Contents of Wireshark-dev digest..."


Today's Topics:

    1. Wiki weirdness? (Jeff Morriss)
    2. Re: Wiki weirdness? (Bill Meier)
    3. Re: Wiki weirdness? (Gerald Combs)
    4. libtshark + scripting language support (Mark Landriscina)
    5. Re: libtshark + scripting language support (Guy Harris)
    6. Re: libtshark + scripting language support (Eloy Paris)


----------------------------------------------------------------------

Message: 1
Date: Wed, 18 Aug 2010 11:29:11 -0400
From: Jeff Morriss<jeff.morriss.ws () gmail com>
Subject: [Wireshark-dev] Wiki weirdness?
To: Developer support list for Wireshark<wireshark-dev () wireshark org>
Message-ID:<4C6BFC47.1060207 () gmail com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed


The top part of the Wiki (that has a kind of tool bar with links to

the
page's Info, etc.) has gotten "weird" for me: instead of lining up
nicely the links are in a vertical list.

It looks the same on Firefox and IE and doesn't change if I'm logged

in
or not.  Anyone else seeing this?

------------------------------

Message: 2
Date: Wed, 18 Aug 2010 11:58:06 -0400
From: Bill Meier<wmeier () newsguy com>
Subject: Re: [Wireshark-dev] Wiki weirdness?
To: Developer support list for Wireshark<wireshark-dev () wireshark org>
Message-ID:<4C6C030E.8000806 () newsguy com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Jeff Morriss wrote:

The top part of the Wiki (that has a kind of tool bar with links

to

the

page's Info, etc.) has gotten "weird" for me: instead of lining up

nicely the links are in a vertical list.

It looks the same on Firefox and IE and doesn't change if I'm

logged

in

or not.  Anyone else seeing this?

Yep ....

------------------------------

Message: 3
Date: Wed, 18 Aug 2010 09:01:21 -0700
From: Gerald Combs<gerald () wireshark org>
Subject: Re: [Wireshark-dev] Wiki weirdness?
To: Developer support list for Wireshark<wireshark-dev () wireshark org>
Message-ID:<4C6C03D1.3070406 () wireshark org>
Content-Type: text/plain; charset=UTF-8

Bill Meier wrote:

Jeff Morriss wrote:

The top part of the Wiki (that has a kind of tool bar with links

to

the

page's Info, etc.) has gotten "weird" for me: instead of lining

up

nicely the links are in a vertical list.

It looks the same on Firefox and IE and doesn't change if I'm

logged in

or not.  Anyone else seeing this?


Yep ....


It should be fixed now. I was experimenting with caching yesterday,

and

left a bad configuration in place.


------------------------------

Message: 4
Date: Wed, 18 Aug 2010 13:34:55 -0400
From: Mark Landriscina<mlandri1 () jhu edu>
Subject: [Wireshark-dev] libtshark + scripting language support
To: wireshark-dev () wireshark org
Message-ID:<7310dfcb3a2d.4c6be17f () johnshopkins edu>
Content-Type: text/plain; CHARSET=US-ASCII

Hi,

I'd like to contribute some work that I've done to the wireshark
community and need some advice on the best way to do this, assuming

there is interest. If not, that would be good to know as well. I
suspect that it might be best to fork this off as a separate project

vs. incorporating it directly into ongoing SVN builds.

My initial goal was to modify the tshark (command line wireshark)

and

wrap it as a Python module. I wanted to expose tshark dissections

as

Python objects during packet capture or capture file processing. In

addition this, I found that it was quite easy to extend this idea

bit more, so that other scripting languages (in additional to

Python)

could leverage the same code base. See below for details.

My motivation was that I wanted to do some work with Scapy and

needed

to access application layer protocol dissections within Python

without

re-writing all the dissection code already available in
tshark/wireshark.

This is what I have done to date (all Linux for now, but am porting

to

Windows):

a. Modified tshark code base and compiled it as a library,
libtshark.a. This is the original tshark executable, more or less,
with some notable additions. In particular, after packet dissection,

the epan dissection tree data is copied off into another tree
structure that I've defined. This t_dissect_node tree is then
serialized and written out over a named-pipe. The name of the
named-pipe is defined by the user at run-time. The code to

unserialize

the t_dissect_node tree is also part of libtshark.a. Also, I have
incorporated some additional helper code that makes tree navigation

easier. A function named 'run' is called to start tshark and accepts

as parameters tshark command line args.

b. A compiled Python shared library, _tsharkPY.so. I used SWIG to
generate the Python bindings. Hence one could take the SWIG

interface

file that I wrote for Python (tsharkPY.i) and modify for use with
other SWIG supported languages: Ruby, Java, etc.

c. tsharkPY.py is the Python module file created by SWIG, leverages

my

tsharkPY.i SWIG interface file.

All the above is based off of the most recent SVN builds and
generation of the two lib files above has been incorporated into the

existing Wireshark build process. Hence, all you have to do is run
'make' and you get libtshark.a and _tsharkPY.so. 'make install' puts

these files into your Python lib path as defined by libtool. I do

need

some help tweaking this, however. Right now, libtool wants to put
these in /usr/local/lib/python2.6/site-packages/. However, they need

to be placed in /usr/lib/python2.6/site-packages/. Any thoughts

(other

than hard coding the correct path)?

Some basic Python code to use the Python module is as follows.

import tsharkPY

#fork tshark. tshark will publish its dissections to 'tshark_pipe'
FIFO. Will read and dissect 3 packets from mycapfile.
tsharkPY.run(["python","-W", "tshark_pipe","-c","3","-r","mycapfile"])

#subscribe to 'tshark_pipe'FIFO
tsharkPY.subscribe("tshark_pipe")

packets = []

#grab packets one at a time from tshark and save them in 'packets'

array

while(1):

     #get packet from "tshark_pipe" FIFO
     p = tsharkPY.get_next_packet("tshark_pipe")

     #check for closed pipe/EOF. break out of loop when detected.
     if(p is None):
         #unsubscribe from tshark_pipe FIFO. cleans up FIFO file and

does some other house keeping.
         tsharkPY.unsubscribe("tshark_pipe")
         break

     #create protocol set, array, and dictionary objects and make

them

part of t_dissect_node object
     p.create_protocol_containers()

     #create dictionary containing field names of all the nodes in

the

packet tree that has 'p' as its root.
     p.create_node_dict()

     #append t_dissect_node object to 'packets' array
     packets.append(p)


print "Protocol sets: unordered list of protocols found in packet."
for packet in packets:                          #iterate over array

of

t_dissect_node trees. Each tree is one packet's worth of data.
     for proto in packet.protocol_set:           #iterate over each
protocol name (string) in t_dissect_tree's protocol set object.
         print proto,                            #print protocol name
     print                                       #print extra line


print "\nProtocol array: ordered array of protocol-level
t_dissect_node references."
for packet in packets:                          #iterate over array

of

t_dissect_node trees. Each tree is one packet's worth of data.
     for node in packet.protocol_array:          #iterate over
t_dissect_node object references in packet's protocol array.
         if node.field_name is not None:         #if node.field_name

exists (is not NULL), print value
             print node.field_name,
     print

print "\nProtocol dictionary: hash table indexed by protocol name.
provides access to t_dissect_node references for protocol level

nodes

in dissection tree."
for packet in packets:
#iterate over array of t_dissect_node trees. Each tree is one

packet's

worth of data.
     d_keys = packet.protocol_dict.keys()

#dump

key list for packet's protocol_dict object
     for k in d_keys:
#iterate over key valus
         node = packet.protocol_dict[k]                          #get

reference to each protocol level node in series
         if node is not None and node.field_name is not None:    #if

successful in retrieving node using current key, print node's field_name
             print node.field_name,
     print

print "\nPacket debug print"
for packet in packets:                          #iterate over array

of

t_dissect_node trees. Each tree is one packet's worth of data.
     packet.print_tree()                         #print

t_dissect_node

tree info for current packet

print "\nPacket data as Python char list."
for packet in packets:
#iterate over array of t_dissect_node trees. Each tree is one

packet's

worth of data.
     try:
         p = packet.first_child.next.last_child

#find

a node in tree that probably has data
         data_list = p.binary_blob                               #get

node data as a list of chars
         print data_list

#print

list

     except:
         pass
#ignore any exceptions thrown from above code

print "\nNode dictionary: dictionary that hashes all nodes in node
tree by their field names (if defined). If duplicate field_names
exist, only the first one encountered is added."
for packet in packets:                          #iterate over array

of

t_dissect_node trees. Each tree is one packet's worth of data.
     d_keys = packet.node_dict.keys()            #dump key list
     for k in d_keys:                            #iterate over key

list

         print k,                                #print each key
     print "\n"

print "\nFind node by its field name. Looking for 'ip.dst_host' in
second packet"
node = packets[1].node_dict['ip.dst_host']

  #find node in second packet that has its field_name param set to

'ip.dst_host'.

if (node is not None):
     print node.field_name+" found! Showname is '"+node.showname+"'"

  #if found, print some stuff from t_dissect_node structure
print



------------------------------

Message: 5
Date: Wed, 18 Aug 2010 11:05:37 -0700
From: Guy Harris<guy () alum mit edu>
Subject: Re: [Wireshark-dev] libtshark + scripting language support
To: Developer support list for Wireshark<wireshark-dev () wireshark org>
Message-ID:<B766DD58-4AA7-42FE-8CF9-5B36656FFAF9 () alum mit edu>
Content-Type: text/plain; charset=us-ascii


On Aug 18, 2010, at 10:34 AM, Mark Landriscina wrote:

I'd like to contribute some work that I've done to the wireshark

community and need some advice on the best way to do this, assuming

there is interest. If not, that would be good to know as well. I
suspect that it might be best to fork this off as a separate project

vs. incorporating it directly into ongoing SVN builds.


My initial goal was to modify the tshark (command line wireshark)

and wrap it as a Python module. I wanted to expose tshark

dissections

as Python objects during packet capture or capture file processing.

In

addition this, I found that it was quite easy to extend this idea

bit more, so that other scripting languages (in additional to

Python)

could leverage the same code base. See below for details.


My motivation was that I wanted to do some work with Scapy and

needed to access application layer protocol dissections within

Python

without re-writing all the dissection code already available in
tshark/wireshark.


This is what I have done to date (all Linux for now,


...which hopefully really means "all UN*X for now", so that it

largely

Just Works on Solaris, *BSD, Mac OS X, HP-UX, etc.

but am porting to Windows):

a. Modified tshark code base and compiled it as a library,

libtshark.a. This is the original tshark executable, more or less,
with some notable additions. In particular, after packet dissection,

the epan dissection tree data is copied off into another tree
structure that I've defined.

The tshark executable image, by default, actually contains no code

to

parse packets or to read capture files; it's linked with two
dynamically linked libraries, libwireshark (which contains all the
dissection code) and libwiretap (which contains all the capture-file

reading code).

What code other than that code is in your libtshark.a?  Or does
anything linked with libtshark.a also have to be linked with
libwireshark and libwiretap?

This t_dissect_node tree is then serialized and written out over

named-pipe. The name of the named-pipe is defined by the user at
run-time. The code to unserialize the t_dissect_node tree is also

part

of libtshark.a.

So what's the reason for the named pipe?


------------------------------

Message: 6
Date: Wed, 18 Aug 2010 14:22:22 -0400
From: Eloy Paris<peloy () chapus net>
Subject: Re: [Wireshark-dev] libtshark + scripting language support
To: Developer support list for Wireshark<wireshark-dev () wireshark org>
Message-ID:<4C6C24DE.3090309 () chapus net>
Content-Type: text/plain; charset=UTF-8; format=flowed

Hi Mark,

On 08/18/2010 01:34 PM, Mark Landriscina wrote:

[...]

My motivation was that I wanted to do some work with Scapy and needed
to access application layer protocol dissections within Python
without re-writing all the dissection code already available in
tshark/wireshark.


I am not a Python guy but my understanding is that there is Python
support in Wireshark trunk (perhaps in 1.4.x). Did you look into

that


and determined that it wasn't good enough for what you need? Just

curious.

a. Modified tshark code base and compiled it as a library,
libtshark.a. This is the original tshark executable, more or less,
with some notable additions. In particular, after packet dissection,
the epan dissection tree data is copied off into another tree
structure that I've defined. This t_dissect_node tree is then
serialized and written out over a named-pipe. The name of the
named-pipe is defined by the user at run-time. The code to
unserialize the t_dissect_node tree is also part of libtshark.a.
Also, I have incorporated some additional helper code that makes

tree

navigation easier. A function named 'run' is called to start tshark
and accepts as parameters tshark command line args.


Any reason you chose to integrate tshark instead of libwireshark,
which
is what does all the dissection work, as Guy mentioned? I would

guess


that it is because it is easier to execute tshark than to fully
integrate libwireshark, but then I don't understand why you need to

make
tshark a library instead of just executing it from within Python.

I actually had a similar need and my approach was to interface with

libwireshark. You can check out my work at

Cheers,

Eloy Paris.-
netexpect.org


------------------------------

_______________________________________________
Wireshark-dev mailing list
Wireshark-dev () wireshark org



End of Wireshark-dev Digest, Vol 51, Issue 22
*********************************************

___________________________________________________________________________
Sent via:    Wireshark-dev mailing list<wireshark-dev () wireshark org>
Archives:
Unsubscribe:

___________________________________________________________________________
Sent via:    Wireshark-dev mailing list<wireshark-dev () wireshark org>
Archives:
Unsubscribe:

___________________________________________________________________________
Sent via:    Wireshark-dev mailing list<wireshark-dev () wireshark org>
Archives:    
Unsubscribe:


___________________________________________________________________________
Sent via:    Wireshark-dev mailing list <wireshark-dev () wireshark org>
Archives:    
Unsubscribe:

___________________________________________________________________________
Sent via:    Wireshark-dev mailing list <wireshark-dev () wireshark org>
Archives:    http://www.wireshark.org/lists/wireshark-dev
Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
             mailto:wireshark-dev-request () wireshark org?subject=unsubscribe
___________________________________________________________________________
Sent via:    Wireshark-dev mailing list <wireshark-dev () wireshark org>
Archives:    http://www.wireshark.org/lists/wireshark-dev
Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
             mailto:wireshark-dev-request () wireshark org?subject=unsubscribe

Current thread:

libtshark + scripting language support Mark Landriscina (Aug 18)
- Re: libtshark + scripting language support Guy Harris (Aug 18)
- Re: libtshark + scripting language support Eloy Paris (Aug 18)
- <Possible follow-ups>
- Re: libtshark + scripting language support Mark Landriscina (Aug 18)
  - Re: libtshark + scripting language support Thierry Emmanuel (Aug 19)
    - Re: libtshark + scripting language support Mark Landriscina (Aug 19)
    - Re: libtshark + scripting language support Eloy Paris (Aug 19)
    - Re: libtshark + scripting language support Mark Landriscina (Aug 19)
    - Re: libtshark + scripting language support Thierry Emmanuel (Aug 20)
    - Re: libtshark + scripting language support Eloy Paris (Aug 20)
    - Re: libtshark + scripting language support Mark (Aug 23)
    - Re: libtshark + scripting language support Guy Harris (Aug 20)