The Linux Socket Filter: Sniffing Bytes over the Network

A feature added to the kernel with the 2.2 release, this LSF can be programmed to let the kernel decide to which packets access should be granted. Here's how.

Sniffing packets over your LAN is an invaluable tool for debugging network problems or collecting measurements. Sometimes the commonly available tools, such as tcpdump or ethereal, will not exactly fit your needs and writing your own sniffer can be of great help. Thanks to the LPF, you can do this in a simple and efficient way.


Gianluca Insolvibile has been a Linux enthusiast since kernel 0.99pl4. He currently deals with networking and digital video research and development. He can be reached at



Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

only sniff or discard too

xandi's picture


nice article. i like the simplicity of it. However, I am wondering whether this technique can be used to create firewalls ? can i discard packets based on the criteria that I choose. Libpcap won't help because it creates a copy of the packet so the packet does reach where it is intended.


In both listing1 and

pepox's picture

In both listing1 and listing2 when i run them i only see "reply" packets, the "request" packets are ignored..

for example:

i run in a console:
zxc@zxc-l:~$ ping -c1

while in another was running:
zxc@zxc-l:~$ sudo ./listing2 (or listing1 it's the same output)
98 bytes read
Source MAC address: 00:af:bc:xx:xx:xx
Destination MAC address: 00:1a:2b:xx:xx:xx
Source host
Dest host
Source,Dest ports 0,10327
Layer-4 protocol 1

However if i do the same with tcpdump, i get this:
zxc@zxc-l:~$ sudo tcpdump -i eth1 -ntUl
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth1, link-type EN10MB (Ethernet), capture size 65535 bytes
IP > ICMP echo request, id 44322, seq 1, length 64
IP > ICMP echo reply, id 44322, seq 1, length 64

In one word, with your examples i can only see reply packets, in fact always my output is:
Dest host (my computers ip).

So my question is, if there is a way to "sniff" both reply packets and also request packets over the network...

Hoping to receive answer...

Thank you very much....


Abelardo Rodríguez-Trilla's picture

Hi Gianluca

Excelent article!. I'm writting from Venezuela, and I wanted to know how do I sniff packets without using the PF_PACKET family. I ask you this because I need to do that without root permissions. Thanks

Great work !

John Kallmer's picture

Really great tutorial !

I also found video tutorials on sniffing at . started by a Vivek Ramachandran, they are quite elaborate in coverage and literally spoon feed topics like sniffing, packet injection etc.

Might help ...take a look !

Some changed in list of include files

andyS1976's picture

Without changes for last C programm compillator gives
warning: implicit declaration of function ‘htons’

If this code changed for C++ programm
(replace exit(X) on return X ) the compiller giver error:
‘htons’ was not declared in this scope

So to improve code include files list might be as follow:
// add this header
// changed #include on


some problem with formating

andyS1976's picture

#include < stdio.h >
#include < string.h >
#include < errno.h >
#include < unistd.h >
// add this header
#include < asm/types.h >
#include < sys/socket.h >
#include < sys/types.h >
#include < netinet/in.h >
// changed #include < linux/in.h > on
#include < netinet/in.h >
#include < linux/if_ether.h >
#include < net/if.h >
#include < linux/filter.h >
#include < sys/ioctl.h >

Setting promiscuous mode

Anonymous's picture

Great article!
I'd just like to point out that you should not use ioctl() for setting the promiscuous mode. If you do, you're responsible for disabling the promiscuous mode after you're done. Unfortunately, you have no way of knowing if another socket also requested the promiscuous mode while your code was running. Thus, resetting the Ethernet flags to the original value could mess things up.

Instead, you should use setsockopt() with SOL_PACKET, PACKET_ADD_MEMBERSHIP and have PACKET_MR_PROMISC as the argument. This way the kernel will track the promiscuous mode usage and turn it off automatically.

Thanks for the simple way of explanation

Rajesh George's picture

Your article made concept so simple. Thanks


Jagannath's picture


your document "sniffing bytes over network" is very nice, you have written very clearly so, that we can uderstand. i am a student of master in networking, doing project on filters.

thanking you for your good, nice,and easy to understand document


Hi Dude....Thu purush naji

Anonymous's picture

Hi Dude....

Thu purush nahi ho.... Maha purush ho..... Sniff sniff sniff and do the right thing for vx180.... Good luck

Re: Kernel Korner: The Linux Socket Filter: Sniffing Bytes over

Anonymous's picture

thanks much for this informative article on a poorly documented subject. this tied together a lot of the bits and pieces i've been sifting through. i'd advise anyone seeking to learn more about creating your own filters to keep this article + source in one hand and the Van Jacobsen/McCanne paper in the other. -

Sniffing Bytes over the Serial port

bahrian's picture

ACtually i m bit new with the socket programming stuff..Actually wat i want to read the bytes from the socket using read () subroutine (I am using Fedora Envionment)but when i exc the program; it stops at the same position where i defined the read sub routine and does not giving me anything..could u plz comment on this..thx in advance.

Thanx a lot for your

Pradeep Reddy's picture

Thanx a lot for your informative article.