Posts Tagged ‘recursion’

How to count how many files are in a directory with find on Linux

Tuesday, February 21st, 2012

how to count how many directories are on your linux server

Did you ever needed to count, how many files in a directory are there?
Having the concrete number of files in a directory is not a seldom task but still very useful especially for scripts or simply for the sake of learning

The quickest and maybe the easiest way to count all files in a directory in Linux is with a combination of find and wc commands:

Here is how;

linux:~# cd ascii
linux:~/ascii# find . -type f -iname '*' -print |wc -l

This will find and list all matched files in any directory and subdirectories, print them out and count them with wc command.
The -type f argument instructs find to look only for files.

Other helpful variance of finding and listing all files in a directory and subdirectories is to list and count all the files with a certain file extension under a directory. For example, lets list all text files (.txt) contained in a directory and all level sub-directories:

linux:~/ascii# find . -type f -iname '*.txt' -print |wc -l

If you need to check the number of files in a directory for multiple directories on a server and you're aiming at doing it efficienly, issung above find .. | wc code will definitely be not a good choice. If used it will generate heavy load for the system and along with that will complete the execution in ages if issued on a large number of files containing dirs.

Thanksfully if efficiency is targetted, there is a command written in C called tree which is more efficient than find.
To count the number of files in dir but using tree :

linux:~# cd ascii
linux:/ascii# tree | tail -n 1
32 directories, 407 files

By default tree prints info for both the number of found files and directories.
To print out only the files matched, awk comes handy, e.g.:

linux:/ascii# tree |tail -n 1| awk '{ print $3 }'407

To list only the number of files in a directory without its existing sub-directories ls + wc use is also possible:

linux:~/ascii# ls -l | grep ^- | wc -l68

This result the above command would produce is +1 more than the real number of files, as it counts the directory ".." as one file (in UNIX / LINUX everything is file).

A short one liner script that can calculate all files correctly by substracting 1 is and hence present correct result on number of files is like so:

linux:~/ascii# var=$(ls -l | grep ^- | wc -l); var=$(($var - 1)); echo $var

ls can be used to calculate the number of 1-st level sub-directories under certain directory for instance:

linux:~/ascii# ls -l |grep ^d|wc -l

You see the ascii directory has 25 subdirectories in its 1st level.

To check symlinks under a directory with ls the command would be:

linux:~/ascii# ls -l | grep ^l | wc -l

Note above 3 ls | grep … examples, will not work properly if the directory contains files with SUID or some special properties set.
Hence to get the same 3 results for active files, directories and symbolic links, a one liner similar to the one below can be used instead:

linux:~/ascii# for t in files links directories; do echo `find . -type ${t:0:1} | wc -l` $t; done 2> /dev/null
407 files
0 links
33 directories

This will show statistics about all files, links and directories for all directory sub-levels.
Just in case if there is need to only count files, links and directories without directory recursion enabled, use:

linux:~/ascii# for t in files links directories; do echo `find . -maxdepth 1 -type ${t:0:1} | wc -l` $t; done 2> /dev/null
68 files
0 links
26 directories

Anyways the above bash loop will be slow, for directories containing thousands of files. For better performance the equivallent of above bash loop rewritten in perl would be:

linux:~/ascii# ls -l |perl -e 'while(<>){$h{substr($_,0,1)}+=1;} END {foreach(keys %h) {print "$_ $h{$_}\n";}}'
- 68
d 25
t 1
In any case the most preferrable and efficient way to count files en directories is by using tree command.
In my view using always tree command instead of code "hacks" is smart idea.

In Slackware tree command is part of the base install, on Debian and CentOS Linux, tree cmd is not part of the base system and requires install via apt / yum e.g.:

debian:~# apt-get --yes install tree

[root@centos:~ ]# yum --yes install tree

Happy counting 😉

Disable DNS recursion and AXFR requests in BIND on Debian Linux and FreeBSD / How to test a nameserver if AXFR requests are allowed with dig command

Monday, March 15th, 2010

I am playing with bind on a newly configured server and therefore doing my best to configure the nameserver in a good manner. In that manner of thoughts I remembered about the good old “recursion” which could pose a security hole in your DNS systems. I won’t buffle on how bad it is for a BIND domain resolver to have Domain recursion switched on, there is plenty of information you can read further online. Anyways here is a brief overview on recursion:
Recursive DNS is essentially the opposite of Custom DNS. Custom DNS is an authoritative DNS service that allows others to find your domain, and Recursive DNS allows you to resolve other people’s domains.

So considering the above definition if you decide to leave the default behaviour of the Bind nameserver (which by the way is also default behaviour of many other DNS servers including Microsoft DNS), this would mean that your DNS will be left open for the whole world to be able to serve resolve requests for any domain name requested by end users. In other words somebody out there might decide to use your nameserver to resolve all internet domains, like:, etc.

It is wise to enable recursion only for localhost on your bind name server, So to achieve that on Debian:
Open /etc/bind/named.conf.options and insert into it
Right before the options {

acl recurseallow {;; };

Also in the options {} include the following lines:

allow-recursion { recurseallow; };recursion yes;

On FreeBSD you need to include the same in /var/named/etc/namedb/named.conf by default or any other location if you have some specific named.conf file location.

Another truly Vital things to include in /etc/bind/named.conf.options on Debian Lenny among options {} is:

auth-nxdomain no;

Including this in the options {} configuration block would completely disable AXFR transfer requests on your nameserver on FreeBSD the procedure is absolutely analogous, just open /var/named/etc/namedb/named.conf and include the auth-nxdomain no; in the options configuration block.

To stress out the importance of disable AXFR it’s important to know that if you don’t disable the AXFR which is enabled by default in many nameservers out there you’re risking that a malicious person could list the whole zone files for each and every of the configured domains in the DNS server and consequently the attacker can learn a lot about the DNS topology of your network etc.
So to complete the article I’m gonna give an example on how the dig command can be used in order to check a certain DNS server if it has enabled the AXFR requests (e.g. if it’s vulnerable to this type of DNS information leak).

dig axfr

In the above example = is a random name server hosting a specific DNS domain = is the DNS domain name / (a.k.a. zone file) hosted on

If everything is configured properly in your the namesever you’re running the axfr test against you should see something like:

; <<>> DiG 9.6.1-P1 <<>> axfr
; (1 server found)
;; global options: +cmd
; Transfer failed.