"Linux Gazette...making Linux just a little more fun!"

Here are a couple of scripts, one shell script, and one awk script, that will find all file names that have identical inode numbers in your system. Why would you want to know this? Well, there are three reasons I can think of.

First, it's generally useful when you're administering a system to know where all of your file links are. Since an inode in Linux isn't freed until the last link to it is removed, it can be helpful to know how many filenames have to be removed in order to _really_ kill a particular file.

Second, knowing all of the hard links in your system can be useful when you have an intrusion. Suppose someone breaks into your system and installs a packet sniffer, keystroke sniffer, or some other nasty piece of work. If he wants to guarantee that he'll be able to carry on his business even if you find his program, he might just create some hard links to it in some out of the way places in the system. Then, even if you find his original program and kill it, he can simply start it running again under one of the alternate file names he created, assuming he can still access your system. However, if you look at the inode number of the original program, then run these scripts, any links to the file will be immediately evident. Granted, you can get the same effect by running "find", if you know the inode number you're looking for, but this won't get you information about _all_ the links in your system, which leads me to the final reason for doing this ...

If you're going to take a proactive approach to system security, you'll be running a lot of file testing and scanning programs, to guard against tampering. If you run these scripts on your system when you're sure it's "pristine", and keep this list of hard links secure, you can run the scripts on your system periodically, perhaps once a day, and compare the new list of links to the baseline. Any differences seen, especially in the "system" directories, such as /bin, /lib, /etc, and so on, should be investigated without delay.

The scripts should run on any system using bash or zsh, and every Linux distrbution provides some version of awk. The file listing mechanism does, however, rely on the use of Mark Baranowski and James Gleason's Enhanced ls program, els. Unlike ls, els gives you complete control over the output format of the file listing. This is an amazing program, and I recommend getting it whether you use my scripts or not. Its only shortcoming is that it doesn't display different file types in different colors, but, this is a minor point in light of the program's power. It's available by ftp at ftp://perseus.elen.utah.edu/pub/markb/els.tar.Z. The current version is 1.44.

Below are the scripts. Adjust the list of directories in "findhardlinks.sh" to suit your preferences, and adjust the location of the awk interpreter in "showsame.awk" to match what you have.

To use them, first run "findhardlinks.sh" from the command line, as root. A file called "allfiles.lst.srt" will be created in /temp(or whatever other directory you'd like) . Be patient; if you have a lot of files and directories, this could take a few minutes. When this is done, copy "showsame.awk" to the same directory as "allfiles.lst.srt" and run

        showsame.awk allfiles.lst.srt

This will create a file called "outfile.txt", which contains the list of identical links. Run this through "uniq" to eliminate duplicated lines, like so

        uniq outfile.txt > outfile.nodup

and you're done. The file "outfile.nodup" is the list of links you want. And yes, I could have put all this together into one big script, but I wanted to show how the pieces work. Feel free to combine and streamline these scripts to your heart's content.

One last point: hard links only make sense between filenames on the _same_ filesystem, so when you put your list of directories into the shell script, be sure they're all part of the same filesystem, such as "/".

findhardlinks.sh

#These directories are all under "/" on my system; you may have to change
#the list to suit your configuration. 

for i in bin boot etc home lib opt root sbin tmp usr var
        do

#This line generates a file listing for all the specified directories and
#their subdirectories, showing each files i-node number, and its complete
#path name, and puts all of it into "allfiles.lst"

        els -a -i +R +NF $i >> /temp/allfiles.lst
        done
cd /temp
sort allfiles.lst > allfiles.lst.srt 
rm allfiles.lst

showsame.awk

#! /usr/bin/awk -f
BEGIN {

        getline  #get the first line of the file, then,
        n1=split($0, test1) #put its fields into an array, then,

#increment the line pointer by one

}
{
#this is the i-node comparison loop

        n2=split($0, test2) #put the next line into an array
        if (test1[1] == test2[1]) { #see if the i-node numbers of the two
                                    #lines are the same(field 1 of both
                                    #strings)

#if they're the same, print the contents of both arrays

                for (i=1; i <= n1; i++) {
                        printf ("%s ", test1[i]) >> "outfile.txt"
                                        } #note (space) after %s; this puts
                                          #back spaces between fields lost
                                          #when the strings were split into
                                          #the arrays

                        printf ("\n") >> "outfile.txt"
                for (i=1; i <= n2; i++)  {
                        printf ("%s ", test2[i]) >> "outfile.txt"
                                        }
                        printf ("\n") >> "outfile.txt"    
                                   }

#Now put the most recent string into the array that holds the previous
#string; this allows us to do comparison between each line and the one
#before it.

        for (i=1; i <= n2; i++) { 
                test1[i] = test2[i]
                                }
        n1 = n2 #set the count of fields in the array that is holding what
                #is now the previous string to the number of fields in that
                #string

        next    #go get the next string and do it again
}