Linux commands

Contents

• Disk operations
• File operations
• Network operations
• Managing users and groups
• System information
  
• Debian packages
• Miscellaneous
  

Resources

• Use man pages or -help or --help for details
• man 5 setlocale (format for accessing a manual page in a specific section)
  
• Command line help (similar to this file)
• Unix help (similar to this file -- just more stuff)
• Alphabetical Directory of Linux Commands
  
• See Configuring and administering Linux
• O'Reilly's Alphabetical Directory of 379 Linux Commands
• O'Reilly's Directory of Linux commands
• Guide for compiling software from source
• APT Cheat Sheet and Howto.
• Debian Quick Reference and Debian Reference

Disk operations -- see also newbie guide

fdisk -l	list all disks and partitions on your system, mounted or not
cfdisk /dev/hda	partition the drive
hddtemp /dev/hda	harddrive temperature (above 40C shortens life)
du -ah | sort -n	display hard disk usage per directory (can take ages!)
kdirstat	display hard disk usage (use this)
parted	a utility for partitioning; allows shrinking (cfdisk doesn't seem to)
fips	utility to reduce partitions?
resize_reiserfs	increase or decrease reiserfs partition
mke2fs	make an ext2 file system (careful)
mkfs.xfs /dev/sdc1	make an xfs file system (xfs_admin -L tv6 /dev/sdc1)
fsck .ext2 /dev/hda8	check data integrity (not tried this one -- be careful)
e2fsck	check integrity of an ext2 partition
e2salvage	utility for salvaging ext2 partitions (homepage) See also the package testdisk -- it may be better
tune2fs -j /dev/hda3	add a journal to ext2 (turn ext2 into ext3)
tune2fs -O ^has_journal /dev/hda4	remove a journal from an ext3 file system (you can add / remove a journal without affecting data)
reiserfsck	check integrity of a reiserfs partition
reiserfsck --check --logfile chck.log /dev/sdb1	run this first -- if it exits with status 0, you're fine
reiserfsck --fix-fixable --logfile fixable.log /dev/sdb1	run this if the previous line exits with status 1
mkswap /dev/hda2	create a swap file on an existing partition
swapon /dev/hda2	turn on the swap file
mount	list all mounted drives with permissions
mount -a	mount drives according to /etc/fstab (root only)
mount nicco	mount nicco as listed in /etc/fstab at default mount point (see samba)
mount /dev/hda8 vm	mount the physical drive partition /dev/hda8 in the directory (mount point) vm
mount spello:/mnt/tv1 /mnt/tv1	mount the tv1 drive from spello, not from cyberspace (when you need to specify)
mount /dev/hdb cdrom	To run a CD, insert it in the CD ROM and type this string
mount -n -o remount/rw /	force a remount of a read-only drive
mount -o loop -t iso9660 filename.img /mountpoint	To mount an ISO image via loopback:
umount mantaray	unmount the drive mantaray
hdparm -v /dev/hda	list disk parameters, such as whether DMA is enabled
df -h	displays the used and free space on all mounted drives
di	better than df
du -s	list the combined size of files in a directory
du --max-depth=1	list the size of all immediate subdirectories
insmode rd	insert the ramdisk module

Find the largest files on your drive:

find / -mount -size +10000k

This found a dozen files on /dev/hdc8. Or just issue

large

You can modify the /usr/local/bin/large script to suit your purpose.

Sort files by date -- or remove duplicates

ls -c -lt

sort temp.log | uniq > podcast.log

Undelete a file

• e2undel works on ext2 file systems -- see man e2undel for details
  
• If possible, unmount the partition where the file was deleted
• e2undel -d /dev/hda7 -a -s /tmp
  

Global replace in multiple files

To replace all instances of

editor@cogweb.net

with

"webmaster at cogweb dot net"

use the K-File-Replace Plugin in Quanta -- press Ctrl-Alt-f  (not tested)

You could also use this command in each directory (didn't bother to make it recursive):

perl -p -i -e 's/editor\@cogweb\.net/\"webmaster\ at\ cogweb\ dot\ net\"/g' *.html
perl -p -i -e 's/editor\ at\ cogweb\.net/\"webmaster\ at\ cogweb\ dot\ net\"/g' *.html

Of course you could put it in a small bash script to make it recursive; for a start see /vm/scripts/replace.sh.

umount

For stubborn hard drives or CDs that won't unmount, try running "lazy unmount" first:

umount -l

This removes the drive in "lazy" mode -- the mount-point "vanishes" for all programs whose working directories are not within that mount point. Once all the files are closed the filesystem is fully unmounted.

If this doesn't work, you can also kill processes that hold up the unmounting:

fuser -m /mnt/vm        # whatever the mount point is

to find out which PIDs are using it. Once you kill those processes you should be able to umount it -- typeically, however, they are simply konsole instances that you should log out of, or change directory out of the mounted drive.

To accomplish both in one stroke, issue

fuser -k /mnt/vm

the -k option should identify and kill the processes that are running on that filesystem (file or directory).

Note that if you have exported a file system through NFS, you won't be able to unmount it afterwards unless you first unexport it:

exportfs -u clitunno:/mnt/tv1

To reexport the volume, just issue "exportfs clitunno:/mnt/tv1". You can of course also unexport all volumes, but this is generally not what you want:

just stop nfs-kernel-server

Then unmount the drive and reexport the other volumes (this is clutzy):

just start nfs-kernel-server

If you don't leave the nfs-kernel-server down for more than a few seconds or maybe minutes, the connections will resume cleanly without human intervention, but it's still better to use exportfs to do a volume at a time.

DOS commands: note that PartitionMagic 6 will corrupt reiserfs partitions. See also Lilo.

Environmental variables

To add a new directory to the default path, issue

export PATH=$PATH:/usr/local/bin

See bash programming examples in /vm/video/scripts

Set default editor (/usr/bin/mcedit-debian is nice for cron, or you could use nano):

update-alternatives --config editor

Characters turned to odd symbols

cat'ing a binary file can cause all the characters you type to show up as odd symbols. Cause: The binary file contained a 016 (so, Shift Out) character. Solution: Print a 017 (Shift In) character. You can use

echo -ne "\\017"

to do this. Even better, make it a shell script called 'fix' for easy access. Reason: Shift Out is canceled by a Shift In.

You can also try to issue

reset

That should be enough. If not, write

setterm -reset

if you just lose a cursor; this is aliased to ungarble on my user.

Terminating processes (see also Freeze and crash recovery strategies)

There are several ways to terminate a program in Linux, either because it is hanging, or because you're maintaining the system remotely, or for some other reason.

Method 1: On the command line

In a console, issue "ps aux" to find the PID (process ID), and issue

        kill <PID>

If that doesn't do the trick, try

        kill -9 <PID>

You can also do this through the program top, which displays all running processes. Press k and enter the PID to terminate the process.

Finally, you can kill all processes related to a particular application by issuing

killall <application name>

The application name has to be the one listed in top or ps aux.

Method 2: Same thing with a GUI

Alternatively, in KDE you can press Ctrl+Esc and get a window with all running processes, select the offender and press Kill.

Method 3: Forget the PID, I just want to get rid of the thing

Finally, in KDE you can also press Ctrl+Alt+Esc and get a "skull and bones" icon -- move it over the offending application and click to terminate it.

kill all processes for some user:

su - johndoe -c 'kill -9 -1'

Terminating a hanging application does not cause problems in Linux -- there's no point in rebooting.

Frozen terminals

If everything freezes, check out instructions in  /usr/src/linux/Documentation/sysrq.txt (assuming you have another computer handy).

Sysrq can also be used to set logging levels:

ALT+SYSRQ+0 should set the log level so that you don't see anything.  Depending on what priority those messages are logged at, ALT+SYSRQ+2 or 3 might be enough, and still let critical messages show on the console.  See also setterm(1).

If a reboot or shutdown command isn't working, try "reboot -nf" (doesn't sync) or else

echo s > /proc/sysrq-trigger
echo u > /proc/sysre-trigger
echo s > /proc/sysrq-trigger
echo b > /proc/sysrq-trigger

This will "sync", "umount/remount read-only", "sync", "immediate hardware reboot".  Should always work, as long as sysrq is enabled in a 2.6 kernel. On the keyboard, this should work by hitting Alt+PrtScr+S, +U, +S, +B

Submit a bug report

Launch "gdb <application that crashed>" as root, and then type:
 (gdb) set args --gst-debug-level=0 >/dev/null
 (gdb) run

 You should lose the gdb prompt for a second, then the program should
 segfault, and you should get the gdb prompt back.  Then run in gdb:
 (gdb) thread apply all bt full

 and paste the output in the bug report.

File operations

File permission changes that involve changing users and groups will generally not take effect until the user has logged completely out of his or her session, from x-windows all the way out of the console.

cat	display the contents of a file
chmod 664 <filename>	change permissions
chmod 4755 <filename>	set suid root for executable
chown <user.group> <filename>	change who owns a directory or file
chmod u+s /usr/sbin/<filename>	set uid root (allow non-root users to run the program with root privileges)
cp	copy <original> <target> -- see also scp
cp -R	copy directories and files recursively -- for large jobs, use tar
cp -s <file> <file>	make a symbolic link between two files (or directories)
dd if=/dev/sda  of=/dev/sdb	clone an OS drive
diff <file1> <file2>	show the difference between two files (cf. kompare, xdiff, et al)
diff -q -r <dir1> <dir2>	list which files are different in two directories
diff  -ruNp	create a diff patch for lkml etc
dmesg	see the boot messages (dmesg | lpr -- print the file)
du -s	list the combined size of files in a directory
du | grep keyword	find file names that contain the keyword string
egrep '^flags.*(vmx|svm)' /proc/cpuinfo	determine if CPU supports virtualization (elegant regular expression)
egrep '(txt|avi)' filename	use egrep to locate the string 'txt' and/or 'avi' in a file -- see bash
file <file name>	determine file type and list detailed characteristics
find -type l	find symbolic links
find / -name -mount string* -print | less	find file names (works, but slow -- see locate)
find -type f	find file names
find -iname string*	find library files? Not sure about the iname switch.
find ~/ -name \*finger-eat\* -print	This will find files with the string finger-eat in my directory recursively
find / -iname \* 2> /dev/null	list all files, pipe errors to /dev/null
fmt -u myessay.txt	remove double spaces in text files
fmt one -w 50000 >one2.txt	remove newlines in paragraphs -- better to use kword's old style import filter, then cut & paste txt2html --extract one >one.html and cut & paste
ftff string	find files in the current directory tree -- a case insensitive and fault tolerant way of 'find . -name xxxx -print'.
ftwhich	finds files in current PATH, fault tolerant search algorithm
gawk	Find and Replace text within file(s)
grep -C 3 <string>	Find a string and the 3 surrounding lines (context)
grep -n <string> <file>	Search the contents of a file or set of files
grep <string> * -r	Search all files in all subdirectories for the string
rgrep <string> * -r	Search for literal string, no regex (eg a$*b?)
grep ^$ *	Search for blank lines
grep  "[0-9]" memo	Search for numbers -- see guide
gs -dNOPAUSE -sDEVICE=pdfwrite -sOutputFile=merged.pdf *.pdf -c quit	Merges all .pdf files in the directory; you can also use pdftk (does lots) or pdfmerge
gs -dNOPAUSE -sDEVICE=pswrite -sOutputFile=foo%2d.ps x*.ps -c quit	Splits a ps (or pdf) file into individual pages
gunzip <filename.wad.gz>	untar command -- it makes its own directory; cf. tar (note that lynx can read .gz files)
finger john@foo.bar.com | grep -i -A 2 smith | less	Ignore case and output 2 lines for each match
ldd <program name>	list dependent libraries
locate <file>	fast find, uses a database -- details below
locate /etc/*session	find all files in all subdirectories under /etc that contain "session"
look	display lines beginning with a given string
ls -c	with -lt: sort by, and show, ctime (time of last modification of file status information) with -l: show ctime and sort by name otherwise: sort by ctime
ls -lh	human-readable
ls -x	list file names only, multiple columns
ls -1	list file names only, in a single colum
ls -R	list all subdirectories with all files in multiple colums
ls -lR	list all subdirectories with all files in single colums, with times
ls -ltc	list files by modification date!
ls -l -S -r	list files by file size, reversed
lsof	list open files
ldd <program>	list library dependencies .
ln -sf <file1> <file2>	creates a symbolic link to file1 called file2 (FAT32 file systems don't handle this) -- cf. realpath
lsmod	list loaded kernel modules
md, rd	make and remove directories
modinfo <module>	show module information
rmmod -r video1394	unload the module video1394
insmod <module>	load any module
modprobe <module>	load a module that is defined in /etc/modules.conf
/etc/modprobe.preload	list modules preloaded in the kernel (not tested)
ldconfig	update library links
pico -w +644 <file>	open file at line 644
ps ax | grep xfs	finds processes and searches for xfs, listing only that
realpath <file>	show the real path of a file, not symlinks
rpm -Uvh <filename>	install an rpm file
rpm -q <package>	find package version installed
rm -r	remove files and directories recursively
rm */<string>	remove all files matching string in all immediate subdirectories
scp <original> <paco:/target> scp file cogweb@cogweb:	copy a file to another computer using ssh (use sftp to copy scattered files in different directories) -- you may need the colon to indicate this is remote
lsmod | sort -nrk 2	sort numeric reverse by second column
sort temp.log | uniq > podcast.log	sort and remove duplicates
strings `which xfs` | grep "^/"	lists files that contain xfs, including the configuration file
tar -C /home/httpd_real -c	the best way to copy whole directory trees?
tar -cvvf foo.tar foo/	put all files in the directory foo/ into the tarball foo.tar (watch your space!)
tar zxvf <filename.tar.gz>	untar command -- it makes its own directory
touch <file> -d "2005-06-10 04:50"	change the last modification date of a file
uniq -cd phone_list.txt	remove duplicates in a list (source)
wc <filename>	word count (lines, words, and characters)
lynx -dump http://some.web.page | wc	word count in online document through lynx

Masks

022 => drwxr-xr-x
033 => drwxr--r--
044 => drwx-wx-wx
055 => drwx-w--w-
066 => drwx--x--x
077 => drwx------

Find text in files

grep -i -I passion * -r

will find all non-binary ( -I ) files with the string passion, ignoring case ( -i ) in any file recursively. It searches whatever is mounted, including the novell file system.

Note that all packages that get installed get copied to the installer's ./kpackage directory. This might be convenient -- no need to keep the originals anywhere else.

Pressing ctrl-C in pico gives you the line number of the cursor

For helpful hints, see http://home.c2i.net/dark/linux.html

vi -- see http://ist.uwaterloo.ca/ec/unix/viprimer.txt.

• esc ZZ   -- save
  
• gg          -- go to start of file
• G            -- go to end of file
• ?             -- search backwards
• :%s/old/new/g -- replace old with new globally (\n is line feed)
  
• : set nowrapscan
• :q!          -- quit without saving
• :wq        -- write and quit
  

How do I find files on my system?

The simple answer is one of these:

locate string
find / -name *string* -mount -print | less

There are two methods to search for files on your Linux machine, one method being the locate command and the other being the find command. If you wish to use the locate command, you must first update the locate database by running the following command in a terminal as root:

/usr/bin/updatedb

This will create an index of all the files and their locations on your hard drive. Once updatedb has completed running, you may now search your drive for a specific file by running the following command in a terminal:

/usr/bin/locate filename

It may be useful to customize the locate and updatedb commands for various users.  Set permissions in /var/cache/local so that another user can write to it and add this sort of thing to the user's .bashrc

alias locate='locate -d /var/cache/locate/locatedb-steen'
alias updatedb='updatedb --output=/var/cache/locate/locatedb-steen --localpaths=/vx/fuma'

That lets the user maintain his own file database that contains his files rather than irrelevant system files. Meanwhile, user root can use the default setup and search all files.

Another way to search your linux system is with the find command. To find a file on your sytem with the find command, in a terminal run the follwoing:

/usr/bin/find / -name filename

Be sure to replace "filename" with the actual name of the file.

find / -name "httpd.conf" -ls

http://www.linuxhelp.net/

How to split files

  $ split -b 200k aa.tar.gz aa.tar.gz.pieces.
  $ for i in aa.tar.gz.pieces.* ; do echo $i | mutt <EMAIL: PROTECTED> -= a  $i  ; done
  $ rm aa.tar.gz.pieces.*
  If you don't have mutt you can use
  $ for i in aa.tar.gz.pieces.* ; do uuencode $i $i | mail <EMAIL: PROTECTED>= ns.it   ; done

Network

Talking to another user

• write <user name>
  
• talk <user name>     (requires the talkd daemon)
• talk root pts/3   (if you have several with the same user name)
  
• mesg y        (allow people to write to you)
• echo "say something" | wall      (writes to everyone on the machine)
  

Managing users and groups

System debugging

Compile in CONFIG_MAGIC_SYSRQ and then when the load goes up do

alt + sysrq + t

Debian admin (cf. popular admin packages)

See also

• local file on managing the local debian mirror using apt-move (or creating one from the cache)
• apt-pinning in Debian

• devscripts
• dh_make
• dpkg-dev
  
• fakeroot

• ./configure --help
• dh_make -s
• pico debian/rules for ./configuration options
• fakeroot dpkg-buildpackage

Gerd Knorr:  Just put the patch into dist/<some-name>.diff and rebuild.

I tried that and issued fakeroot dpkg-buildpackage from /share/software/tarballs/xawtv/xawtv-3.94 -- the source remained tarred and gzipped. The build process asked for the file to patch (it wasn't correct in the patch), and I entered "vbistuff/ntsc-cc.c", which it accepted and then built the Debian package. The patched file ended up in /share/software/tarballs/xawtv/xawtv-3.94/work/xawtv-3.94/vbistuff

apt-get source schroot
cd schroot-0.99.2
patch -p0 < /tmp/schroot-dchroot-compat-command.patch
dpkg-buildpackage -rfakeroot -us -uc
sudo dpkg -i ../dchroot_0.99.2-2_amd64.deb

Determine order of module loading:   snd_via82xx.index=0 saa7134.index=1

Miscellaneous

• Recurring jobs on a machine that is always on: cron
• Recurring jobs on a machine that is turned off: anacron
• Occasional jobs to be done at specific future times:
  
• at -- run the job at the time set, for example:
  
• at 5:00 -f /usr/local/bin/dv2xvid
• at now + 10 minutes -f record.sh
  
• atq -- see the at queue
• atq
• atrm -- terminate an at job
• atrm 2
• batch -- run a job when system resources free up
• batch -f /usr/local/bin/dv2xvid

CVS

cvs -d:pserver:cvs@cvs.xxx.com:/xxx login
cvs -z 3 -d:pserver:cvs@cvs.xxx.com:/xxx co module

The -z 3 sets compression level.

Examples of other useful options (see the CVS documentation for details):

• cvs update -A (update from the head of the tree, and ignore tags)
• cvs -n update (test only, don't make changes)
• cvs update -r Wine-YYMMDD (get the revision with the tag for a Wine snapshot)
• cvs update -D date (get the revision associated with an arbitrary date)

Job management

test $DISPLAY && dcop kdesktop default logout -- shut down x-windows remotely

cron jobs -- see crontab and /etc/cron

Lots of good info on command line management: http://amath.colorado.edu/computing/unix/jobs.html

To run a program in the background, add a & at the end of the command.

myprog > output.dat & -- run myprog in the background, sending any stdout to the file `output.dat'

math -noinit -batchinput -batchoutput < cmmds > rslts &

-- run Mathematica in the background, taking commands from the file `cmmds' and writing all output to the file `rslts'

Run two or more instances of x-windows at the same time on one machine:

first user # startx
second user # startx -- :1

The first user's session will be in Ctrl-Alt-F7
The second user's session will be in Ctrl-Alt-F8

These are useful for starting jobs at some future time:

at now + 3 hours -f /usr/local/bin/dv2xvid
at now + 2 minutes -f /usr/local/bin/grab

Normal procedure would be to create a shell script that calls the functions you want.
You can also use this to run something in the background:

at now -f /usr/local/bin/dv2xvid

while true; do saytime; sleep 600; done &
while true; do saytime; sleep 20; done &
while true; do saytime; sleep 1800; done &

jobs
fg 2
Ctrl-Z
jobs
bg 2
jobs

Interrupt signals

This is useful for instance for terminating dvgrab remotely.

• typing ctrl-c sends the interrupt signal SIGINT to your current foreground process
• The complete list of signals can be found on the signal(7) manual page
• Issue man 7 signal to see it or enter kill -1 for a short version of this list
• If no signal is specified on the kill command line, SIGTERM is sent
• If SIGTERM was caught or ignored, send SIGKILL, which will always work

It looks like you can be explicit about this:

kill -SIGINT 20371
kill -SIGTERM 20371 (default, doesn't need to be specified)
kill -SIGKILL 20371

Or you can use the numbers (details below):

kill -2 <dvgrab PID> (this is the SIGINT number)
kill -3 <less PID> (this is the SIGQUIT number -- useful for less)

• A related command is killall, which takes the name of the process as an argument
  
• This is a convenient way to terminate all processes with the same name
  
• You can specify path if you want to kill only a subset of processes by that name

Here's the list of the first few signals, from kill -l:

1) SIGHUP 2) SIGINT 3) SIGQUIT 4) SIGILL
5) SIGTRAP 6) SIGABRT 7) SIGBUS 8) SIGFPE
9) SIGKILL 10) SIGUSR1 11) SIGSEGV 12) SIGUSR2
13) SIGPIPE 14) SIGALRM 15) SIGTERM 17) SIGCHLD

Note that SIGINT is 2. Issuing killall -2 dvgrab will create clean files.