fuser - Linux Command - Unix Command

fuser - Linux Command - Unix Command

fuser - identify processes using files or sockets
SYNOPSIS
fuser [-a|-s] [-4|-6] [-n space] [-signal] [-kimuv] name ...
fuser -l
fuser -V
DESCRIPTION
fuser displays the PIDs of processes using the specified files or file systems. In the default display mode, each file name is followed by a letter denoting the type of access:
c
current directory.
e
executable being run.
f
open file. f is omitted in default display mode.
r
root directory.
m
mmap'ed file or shared library.
fuser returns a non-zero return code if none of the specified files is accessed or in case of a fatal error. If at least one access has been found, fuser returns zero.

In order to look up processes using TCP and UDP sockets, the corresponding name space has to be selected with the -n option. By default fuser will look in both IPv6 and IPv4 sockets. To change the default, behavour, use the -4 and -6 options. The socket(s) can be specified by the local and remote port, and the remote address. All fields are optional, but commas in front of missing fields must be present:

[lcl_port][,[rmt_host][,[rmt_port]]]

Either symbolic or numeric values can be used for IP addresses and port numbers.

stty erase

RE: vi question on OSX: "Try this shell command:


stty erase


The Backspace key should now work on the command line and in vi. If you
want to make this fix permanent, you'll need to add it to your .cshrc
file (or appropriate file for whatever shell you are using; .cshrc is
for the default tcsh shell). The trick, of course, is that if you have
already fixed the problem in your shell, entering the same command while
editing your .cshrc file becomes more difficult--you don't get the ^?
anymore, it just backspaces. You can get around this by hitting
Control-v and then the Backspace key. So while editing a new line in
.cshrc, type the following:


stty erase ^v


And you should see:


stty erase ^?


The save, quit, and open a new shell to confirm that this worked.Try this shell command:


stty erase


The Backspace key should now work on the command line and in vi. If you
want to make this fix permanent, you'll need to add it to your .cshrc
file (or appropriate file for whatever shell you are using; .cshrc is
for the default tcsh shell). The trick, of course, is that if you have
already fixed the problem in your shell, entering the same command while
editing your .cshrc file becomes more difficult--you don't get the ^?
anymore, it just backspaces. You can get around this by hitting
Control-v and then the Backspace key. So while editing a new line in
.cshrc, type the following:


stty erase ^v


And you should see:


stty erase ^?


The save, quit, and open a new shell to confi"

Unix for Advanced Users

nohup

Sometimes you actually want processes to run after you logout. One way to do this is to invoke a command called nohup. nohup is short for no hang-up. Say you want to search a huge disk for a file but you want to logout and get a pizza while it is searching. You could type:

nohup find /bigdisk -name file

and the results will be stored in a file called nohup.out.

Unix commands reference

System status check
Listed here are a few system monitoring commands which should give you a rough idea of how the server is running.

# server information
uname -a

# server config information
prtconf
sysdef -i

# server up time
uptime

# disk free, listed in KB
df -kt

# mounted devices
mount

# network status
netstat -rn

# network configuration info
ifconfig -a

# processes currently running
ps -elf

# user processes
w
whodo
who am i
finger
ps

# virtual memory statistics
vmstat 5 5

# system activity reporter (Solaris/AIX)
sar 5 5

# report per processor statistics (Solaris)
mpstat 5 5
psrinfo

# swap disk status (Solaris)
swap -l

# shared memory
ipcs -b





Solaris note: SAR numbers can be misleading: as memory use by processes is freed, but not considered 'available' by the reporting tool. Solaris support has recommended using the SR (swap rate) column of vmstats to monitor the availability of memory. When this number reaches 150+, a kernel panic may ensue.
System startup
The kernel is loaded by the boot command, which is executed during startup in a machine-specific way. The kernel may exist on a local disk, CD-ROM, or network. After the kernel loads, the necessary file systems are mounted (located in /etc/vfstab), and /sbin/init is run, which brings the system up to the "initdefault" state set in /etc/inittab. Subsystems are started by scripts in the /etc/rc1.d,/etc/rc2.d, and /etc/rc3.d directories.
System shutdown

# shutdown the server in 60 seconds, restart system in administrative state
# (Solaris)
/usr/sbin/shutdown -y -g60 -i1 "System is begin restarted"

# shutdown the server immediately, cold state
# (Solaris)
/usr/sbin/shutdown -y -g0 -i0

# shutdown AIX server, reboot .. also Ctrl-Ctrl/Alt
shutdown -Fr



# restart the server
/usr/sbin/reboot

User accounts
Adding a unix account involves creating the login and home directory, assigning a group, adding a description, and setting the password. The .profile script should then be placed in the home directory.

# add jsmith account .. the -m parm forces the home dir creation
useradd -c "Jim Smith" -d /home/jsmith -m -s "/usr/bin/ksh" jsmith

# change group for jsmith
chgrp staff jsmith

# change jsmith password
passwd jsmith

# change jsmith description
usermod -c "J.Smith" jsmith

# remove ksmith account
userdel ksmith

# display user accounts
cat /etc/passwd

/* here is a sample .profile script, for sh or ksh */
stty istrip
stty erase ^H
PATH=/usr/bin:/usr/ucb:/etc:/usr/lib/scripts:/usr/sbin:.
export PATH
PS1='BOXNAME:$PWD>'
export PS1



Displaying files

# display file contents
cat myfile

# determine file type
file myfile

# display file, a screen at a time (Solaris)
pg myfile

# display first 100 lines of a file
head -100 myfile

# display last 50 lines of a file
tail -50 myfile

# display file that is changing, dynamically
tail errlog.out -f


File permissions
Permission flags: r = read, w = write, x = execute Permissions are displayed for owner, group, and others.

# display files, with permissions
ls -l
# make file readable, writeable, and executable for group/others
chmod 777 myfile

# make file readable and executable for group/others
chmod 755 myfile

# make file inaccessible for all but the owner
chmod 700 myfile

# make file readable and executable for group/others,
# user assumes owner's group during execution
chmod 4755 myfile

# change permission flags for directory contents
chmod -R mydir

# change group to staff for this file
chgrp staff myfile

# change owner to jsmith for this file
chown jsmith myfile


Listing files
See scripting examples for more elaborate file listings.

# list all files, with directory indicator, long format
ls -lpa

# list all files, sorted by date, ascending
ls -lpatr

# list all text files
ls *.txt


Moving/copying files
See scripting examples for moving and renaming collections of files.

# rename file to backup copy
mv myfile myfile.bak

# copy file to backup copy
cp myfile myfile.bak

# move file to tmp directory
mv myfile /tmp

# copy file from tmp dir to current directory
cp /tmp/myfile .


Deleting files
See scripting examples for group dissection routines.

# delete the file
rm myfile

# delete directory
rd mydir

# delete directory, and all files in it
rm -r mydir


Disk usage

# display disk free, in KB
df -kt

# display disk usage, in KB for directory
du -k mydir

# display directory disk usage, sort by largest first
du -ak / | sort -nr | pg


Using tar

# display contents of a file
tar tvf myfile.tar

# display contents of a diskette (Solaris)
volcheck
tar tvf /vol/dev/rdiskette0/unnamed_floppy

# copy files to a tar file
tar cvf myfile.tar *.sql

# format floppy, and copy files to it (Solaris)
fdformat -U -b floppy99
tar cvf /vol/dev/rdiskette0/floppy99 *.sql

# append files to a tar file
tar rvfn myfile.tar *.txt

# extract files from a tar filem to current dir
tar xvf myfile.tar


Starting a process
This section briefly describes how to start a process from the command line.

Glossary:

& - run in background
nohup (No Hang Up) - lets process continue, even if session is disconnected


# run a script, in the background
runbackup &

# run a script, allow it to continue after logging off
nohup runbackup &


# Here nohup.out will still be created, but any output will
# show up in test70.log. Errors will appear in nohup.out.

nohup /export/spare/hmc/scripts/test70 > test70.log &



# Here nohup.out will not be created; any output will
# show up in test70.log. Errors will appear test70.log also !

nohup /export/spare/hmc/scripts/test70 > test70.log 2>&1 &






Killing a process

1) In your own session; e.g. jobs were submitted, but you never logged out:

ps # list jobs
kill -9 < process id> # kill it



2) In a separate session

# process ID appears as column 4
ps -elf | grep -i

kill -9 < process id> # kill it



3) For device (or file)

# find out who is logged in from where

w

# select device, and add /dev ... then use the fuser command

fuser -k /dev/pts/3




Redirecting output
Output can be directed to another program or to a file.

# send output to a file
runbackup > /tmp/backup.log

# also redirect error output
runbackup > /tmp/backup.log 2> /tmp/errors.log

# send output to grep program
runbackup | grep "serious"


Date stamping, and other errata
Other errata is included in this section

# Date stamping files
# format is :
# touch -t yyyymmddhhmi.ss filename

touch -t 199810311530.00 hallowfile

# lowercase functions (ksh)
typeset -u newfile=$filename

# date formatting, yields 112098 for example
date '+%m%d%y'

# display a calendar (Solaris / AIX)
cal

# route output to both test.txt and std output
./runbackup | tee test.txt

# sleep for 5 seconds
sleep 5

# send a message to all users
wall "lunch is ready"

# edit file, which displays at login time (message of the day)
vi /etc/motd

# edit file, which displays before login time (Solaris)
vi /etc/issue

Chat log (interesting)

--- Day changed

--- Log opened Thu Feb 09 23:59:01 2006
00:13 -!- womble [~mpalmer@203.219.190.106] has quit [Ping timeout: 480 seconds]
01:26 < dave_> hmm, maybe not.
01:27 < dave_> I can get my switch to please a config eth0=daemon
01:27 < dave_> but anything beyond that just prink's some transmit failures :/
01:29 < dave_> uml_net_start_xmit: failed(-111)
01:39 -!- shehjar_ [~shehjar@60-240-80-162-nsw-pppoe.tpgi.com.au] has joined #uml
02:40 -!- Newsome [~sorenson@216-190-206-130.customer.csolutions.net] has quit [Quit: Linux: Now with employee pricing!]
02:41 -!- ams [1000@pv108234.reshsg.uci.edu] has joined #uml
02:56 -!- ram [~ram@pool-71-117-233-57.ptldor.fios.verizon.net] has joined #uml
02:57 -!- Shaun2 [Shaun@204.15.132.253] has quit [Ping timeout: 480 seconds]
03:10 -!- ams [1000@pv108234.reshsg.uci.edu] has left #uml [Leaving]
03:43 -!- ram [~ram@pool-71-117-233-57.ptldor.fios.verizon.net] has quit [Remote host closed the connection]
04:22 -!- Shoragan [~shoragan@d072.apm.etc.tu-bs.de] has joined #uml
05:56 -!- shehjar_ [~shehjar@60-240-80-162-nsw-pppoe.tpgi.com.au] has quit [Quit: shehjar_]
08:20 -!- da303king [~heiko@p4523e3d4.np.schlund.de] has joined #uml
08:47 -!- Shoragan [~shoragan@d072.apm.etc.tu-bs.de] has quit [Quit: Leaving]
08:57 < dgraves> dave_: perhaps uml_net with a tuntap would be easier?
08:59 < dave_> dgraves: the point there was that I'm writing my own switch daemon; so bypassing it isn't very interesting :-)
09:02 < dgraves> oh. yeah, i see that. i doubt there's a spec. if you talk to jdike later today, he wrote it, i think.
09:16 < dave_> it's okay for the time being. I figured out how to get a communication going between the VM and my switch.
09:17 < dgraves> oh, good!
09:17 < dave_> the code of the kernel driver is much clearer than the uml_router code; uml_router having to handle 4 different revs of the communication protocol makes it fairly obfuscate.
09:17 < dgraves> now you just need to make it take over the world.
09:17 < dave_> working on that too.
09:44 -!- Newsome [~sorenson@obelix.cs.byu.edu] has joined #uml
09:58 -!- silug [~steve@206.80.72.37] has quit [Ping timeout: 480 seconds]
10:04 -!- da303king [~heiko@p4523e3d4.np.schlund.de] has left #uml [? DIVISION BY ZERO]
10:50 -!- hfb [~hfb@pool-71-116-252-188.lsanca.dsl-w.verizon.net] has joined #uml
11:18 -!- ram [~ram@pool-71-117-233-57.ptldor.fios.verizon.net] has joined #uml
11:18 -!- hfb [~hfb@pool-71-116-252-188.lsanca.dsl-w.verizon.net] has quit [Quit: Leaving]
11:42 -!- jdike [~jdike@pool-71-245-228-225.bstnma.fios.verizon.net] has joined #uml
11:42 < jdike> Hi guys
11:46 < dgraves> mourning jdike.
11:46 < jdike> Who are we mourning?
11:47 < jdike> your spelling abilities?
11:47 < dgraves> the trout i'm about to nuke you with.
11:47 * dgraves nukes jdike with a trout bomb.
11:47 < jdike> Uh oh
11:47 * dgraves nukes jdike with a trout mirv too.
11:47 * jdike investigates trout shelters
11:48 < dgraves> doggone it. don't make me nuke you more! will you just *stop moving*???!!!
11:48 < jdike> I'm gonna have a shelter
11:48 < jdike> I ain't gonna care what sorts of trouts you rain down upon me
11:49 * dgraves investigates sturgeon bombs then.
11:52 < jdike> man
11:52 < jdike> sturgeon shelters are *expensive*
11:53 < dgraves> yeah. so are the bombs.
11:53 < dgraves> don't make me.
11:56 -!- hfb [~hfb@pool-71-116-252-188.lsanca.dsl-w.verizon.net] has joined #uml
12:11 -!- ram [~ram@pool-71-117-233-57.ptldor.fios.verizon.net] has quit [Ping timeout: 480 seconds]
12:37 -!- hfb [~hfb@pool-71-116-252-188.lsanca.dsl-w.verizon.net] has quit [Quit: Leaving]
13:02 -!- the_hydra [~a_mulyadi@202.59.168.5] has joined #uml
13:02 < the_hydra> hi
13:04 < dgraves> hiya the_hydra
13:05 < the_hydra> hello allan
13:11 -!- Newsome [~sorenson@obelix.cs.byu.edu] has left #uml [Linux: Now with special employee pricing!]
13:13 < the_hydra> jdike: today I re-read about what you had said about userspace() in process.c
13:13 < the_hydra> jdike: for me, seems like this is similar to entry.S
13:13 -!- acklen [~kno@64.62.148.249] has quit [Read error: Operation timed out]
13:17 -!- tomimo [~kurre@a80-186-64-246.elisa-laajakaista.fi] has quit [Ping timeout: 480 seconds]
13:34 -!- acklen [~kno@64.62.148.249] has joined #uml
13:45 < jdike> yeah, pretty much
13:45 < jdike> very different in form, though
13:46 < the_hydra> jdike: yes, different form
14:06 -!- tomimo [~kurre@a80-186-64-246.elisa-laajakaista.fi] has joined #uml
14:06 < the_hydra> jdike: besides the different header files they use, is there any other difference between _user.c and _kern.c file?
14:06 < jdike> well, one is userspace code and the other is kernel code
14:07 < jdike> that's kind of a fundamental difference, even if it just amounts to the headers they use
14:12 < the_hydra> jdike: hm, for example, userspace() is put on process.c and the trap handler is on process_kern.c?
14:13 < jdike> right, userspace is full of ptrace and wait, so needs to be userspace
14:14 < jdike> the trap handler needs to call the IRQ system and page fault handler, so needs to be kernelspace
14:20 < the_hydra> those two codes (_user.c and _kern.c) are actually executed by two different threads?
14:22 < jdike> no
14:22 < jdike> there are direct calls between them
14:24 < the_hydra> hm i see
14:24 < jdike> it's just what headers they can use
14:25 < jdike> and they communicate (data passed across that boundary) using generic C types
14:25 < the_hydra> hm ok
14:34 -!- kos_tom [~thomas@humanoidz.org] has joined #uml
14:43 -!- the_hydra [~a_mulyadi@202.59.168.5] has quit [Quit: ]
14:46 -!- ams [1000@pv108234.reshsg.uci.edu] has joined #uml
15:09 -!- silug [~steve@206.80.72.37] has joined #uml
15:10 < dgraves> When Bob Kahn, Vint Cerf, and God were inventing TCP/IP, they didn't put much thought into how their technology might handle HDTV streams.
15:12 < jdike> No?
15:12 < jdike> shocking
15:13 < dgraves> ::L::
15:14 < jdike> Your porn is coming in a bit choppy?
15:15 < dgraves> i think you've got me confused with the_hydra, jdike. :)
15:15 < dgraves> its my martial arts flicks. :-P
15:15 < jdike> you're the one complaining
15:15 * dgraves checks.
15:15 < dgraves> wow! i *was* the one complaining!
15:15 < jdike> Ah, yes
15:15 < dgraves> dang. now how'd that happen?
15:15 < jdike> "martial arts"
15:16 -!- Newsome [~sorenson@216-190-206-130.customer.csolutions.net] has joined #uml
15:16 < dgraves> that's why i want more storage space!
15:16 < dgraves> actually, hdtv might just *kill* the porn industry...
15:17 < dgraves> have you ever thought about what those people might look like in HD?
15:17 * dgraves gags and wretches.
15:18 < jdike> hehe
15:18 < drrdarr> i have a question about disk writes in UML. Aren't write requests added to the linux elevator queue, and then serviced?
15:19 < drrdarr> A block write getting flushed to the disk (say /dev/hda6, a physical disk partition) traverses the following functions in the kernel: id
15:19 < drrdarr> e_do_request() -> start_request() -> do_request() -> do_rw_disk().
15:19 < dgraves> drrdarr: is this hd uml or sd uml?
15:19 * dgraves runs away.
15:19 < drrdarr> Does something change in the way block writes happen to virtual devic
15:19 < drrdarr> es in UML? The virtual devices should also be considered like a physical device right?
15:19 < jdike> yup, it's the same
15:19 < drrdarr> I am sorry, i do not understand the difference to sd and hd.. i am quite a newbie
15:20 < jdike> ignore him
15:20 < jdike> he's just a troublemaker
15:20 < drrdarr> :)
15:20 < jdike> except that UML uses the noop elevator
15:20 < drrdarr> o okay
15:20 * dgraves burns jdike's cookies then.
15:21 < drrdarr> i am working on an application that blocks disk writes and makes a copy of every block that is written to disk.
15:21 < drrdarr> i wrote it in the original kernel, but it hangs the system for some reason.
15:21 < drrdarr> when i implement the same thing in UML, I am not even able to see the disk block writes..
15:21 * dgraves is intrigued and waits for jdike's answer.
15:22 < drrdarr> I call my copy function from inside start_request() in drivers/ide/ide.c
15:22 < jdike> how does it hook into the block later?
15:22 < drrdarr> it taps all block writes in the original system..
15:22 < drrdarr> i get the data from the buffer head structure
15:22 < drrdarr> that is currently about to be removed from the elevator
15:23 < drrdarr> i just make a copy, and then leave the control back to the point in start_request() .
15:23 < drrdarr> and ddd for some reason always gives an error... errno = 3
15:24 < drrdarr> and it locks my slackware.img file, so i can't spawn the UML again!
15:24 < jdike> that's just because there are still processes from the UML still there
15:24 < drrdarr> okay
15:25 < jdike> what version of UML?
15:26 < drrdarr> i downloaded 2.4.20 source
15:26 < dgraves> ::blink::
15:26 < drrdarr> I am sorry, i should have mentioned that earlier!
15:26 < jdike> can you use a 2.6 kernel?
15:27 < jdike> much easier to debug
15:27 < drrdarr> i would like to stick to 2.4.. dunno much about 2.6 to start all over again..
15:27 < jdike> and much easier to get help with, since everyone has forgotten how 2.4 works
15:27 < drrdarr> o ok
15:27 < dgraves> jdike: what's 2.4?
15:27 < jdike> see?
15:27 < dgraves> i don't know this mythical 2.4 thing you speak of...
15:27 < drrdarr> :)
15:28 * dgraves isn't much help with 2.6 though... but at least people don't fall off their seat when they say 2.4...
15:29 -!- ram [~ram@bi01p1.co.us.ibm.com] has joined #uml
15:29 < dgraves> oh good. i need more ram.
15:32 < drrdarr> jdike, i shall study the noop elevator and see if I am missing the control path (for a disk block write).. it would be great if i can see the values of the local variables in gdb.
15:32 < drrdarr> You said yesterday that it depends on gcc
15:32 < jdike> it has nothing to do with the elevator
15:33 < drrdarr> do i have to enable some flag before compilation?
15:33 < jdike> gcc can eliminate variables if it can prove they are the same as other variables it does preserve
15:33 < drrdarr> okay
15:33 < jdike> or if their values aren't needed yet or needed any more
15:33 < drrdarr> ok
15:35 < drrdarr> since i am new to UML, i thought that disk writes may be taking a different path in the code for virtual devices than physical devices.
15:35 < drrdarr> but since you said its not, i shall have to study the code again..
15:36 < jdike> exactly the same
15:36 < drrdarr> hmm
15:36 < jdike> things only change when they reach the disk driver
15:37 < drrdarr> okay.. i want to tap the write before the disk driver get the block (from the buffer cache, i think)
15:39 < drrdarr> also, another small question: is is true that the test (stub) program that calls the new system call i had written has to be compiled outside of UML, and then executed by mounting the host fs and accessing the executable there?
15:40 < jdike> no
15:40 < drrdarr> i tried writing the stub inside UML, but it could not find the system call entry point
15:40 < jdike> what was the exact error
15:41 < drrdarr> it used to complain that the system call was not implemented.
15:41 < drrdarr> you all helped me the other day with this stuff..
15:42 < drrdarr> i wrote the stub inside UML, made it include a file say somename.h which I put in /usr/include/sys
15:42 < jdike> what was the exact error
15:42 < drrdarr> just a moment
15:44 < drrdarr> perror result: Function not implemented
15:45 < jdike> that means UML didn't implement the system call
15:46 < drrdarr> I changed the /usr/include/asm/unistd.h to include the new system call.. this is the only place where i diverted from the instructions by Alexandros ..
15:47 < drrdarr> I had implemented all the other steps and built that kernel, and executed that..
15:47 < jdike> -ENOSYS means the system call wasn't implemented
15:47 < jdike> not present in the system call table
15:48 < drrdarr> but it seems to find the system call if i mount hostfs, and then execute the stub I had written and compiled outside UML
15:48 < drrdarr> okay
15:48 < jdike> in that case, the system call number was different inside UML than on the host or something
15:49 < drrdarr> okay, i shall check that once again.
15:49 < jdike> strace the two versions
15:49 < drrdarr> ok
15:49 < jdike> there should be some difference in that one system call
15:49 < drrdarr> ok
15:50 < drrdarr> will try it
15:53 < drrdarr> jdike, thx a lot.. going back to my block write implementation..

Test syntax of shell script

The shell has a number of flags that make debugging easier:

sh -n command

Read the shell script but don't execute the commands, i.e., check syntax.

sh -x command

Display commands and arguments as they're executed. In a lot of my shell scripts you'll see

# Uncomment the next line for testing
# set -x