Unix工具箱命令备忘录

Running kernel and system information
# uname -a                           
# lsb_release -a                     
# cat /etc/SuSE-release              
# cat /etc/debian_version            
Use /etc/ DISTR-release with DISTR= lsb (Ubuntu), redhat, gentoo, mandrake, sun (Solaris), and so on. See also /etc/issue.
# uptime                             
# hostname                           
# hostname -i                        
# man hier                           
# last reboot                        

Hardware Informations

Kernel detected hardware
# dmesg                              
# lsdev                              
# dd if=/dev/mem bs=1k skip=768 count=256 2>/dev/null | strings -n 8 

Linux

# cat /proc/cpuinfo                  
# cat /proc/meminfo                  
# grep MemTotal /proc/meminfo        
# watch -n1 'cat /proc/interrupts'   
# free -m                            
# cat /proc/devices                  
# lspci -tv                          
# lsusb -tv                          
# lshal                              
# dmidecode                          

FreeBSD

# sysctl hw.model                    
# sysctl hw                          
# sysctl hw.ncpu                     
# sysctl vm                          
# sysctl hw.realmem                  
# sysctl -a | grep mem               
# sysctl dev                         
# pciconf -l -cv                     
# usbdevs -v                         
# atacontrol list                    
# camcontrol devlist -v              

Load, statistics and messages

The following commands are useful to find out what is going on on the system.
# top                                
# mpstat 1                           
# vmstat 2                           
# iostat 2                           
# systat -vmstat 1                   
# systat -tcp 1                      
# systat -netstat 1                  
# systat -ifstat 1                   
# systat -iostat 1                   
# tail -n 500 /var/log/messages      
# tail /var/log/warn                 

Users

# id                                 
# last                               
# who                                
# groupadd admin                     
# useradd -c "Colin Barschel" -g admin -m colin
# usermod -a -G         
# groupmod -A           
# userdel colin                      
# adduser joe                        
# rmuser joe                         
# pw groupadd admin                  
# pw groupmod admin -m newmember     
# pw useradd colin -c "Colin Barschel" -g admin -m -s /bin/tcsh 
# pw userdel colin; pw groupdel admin
Encrypted passwords are stored in /etc/shadow for Linux and Solaris and /etc/master.passwd on FreeBSD. If the master.passwd is modified manually (say to delete a password), run # pwd_mkdb -p master.passwd to rebuild the database.

To temporarily prevent logins system wide (for all users but root) use nologin. The message in nologin will be displayed (might not work with ssh pre-shared keys).
# echo "Sorry no login now" > /etc/nologin       
# echo "Sorry no login now" > /var/run/nologin   

Limits

Some application require higher limits on open files and sockets (like a proxy web server, database). The default limits are usually too low.

Linux

Per shell/script

The shell limits are governed by ulimit. The status is checked with ulimit -a. For example to change the open files limit from 1024 to 10240 do:
# ulimit -n 10240                    
The ulimit command can be used in a script to change the limits for the script only.

Per user/process

Login users and applications can be configured in /etc/security/limits.conf. For example:
# cat /etc/security/limits.conf
*   hard    nproc   250              
asterisk hard nofile 409600          

System wide

Kernel limits are set with sysctl. Permanent limits are set in /etc/sysctl.conf.
# sysctl -a                          
# sysctl fs.file-max                 
# sysctl fs.file-max=102400          
# echo "1024 50000" > /proc/sys/net/ipv4/ip_local_port_range  
# cat /etc/sysctl.conf
fs.file-max=102400                   
# cat /proc/sys/fs/file-nr           

FreeBSD

Per shell/script

Use the command limits in csh or tcsh or as in Linux, use ulimit in an sh or bash shell.

Per user/process

The default limits on login are set in /etc/login.conf. An unlimited value is still limited by the system maximal value.

System wide

Kernel limits are also set with sysctl. Permanent limits are set in /etc/sysctl.conf or /boot/loader.conf. The syntax is the same as Linux but the keys are different.
# sysctl -a                          
# sysctl kern.maxfiles=XXXX          
kern.ipc.nmbclusters=32768           
kern.maxfiles=65536                  
kern.maxfilesperproc=32768
kern.ipc.somaxconn=8192              
# sysctl kern.openfiles              
# sysctl kern.ipc.numopensockets     
# sysctl net.inet.ip.portrange.last=50000 
# netstat -m                         
See The FreeBSD handbook Chapter 11 http://www.freebsd.org/handbook/configtuning-kernel-limits.html for details.

Solaris

The following values in /etc/system will increase the maximum file descriptors per proc:
set rlim_fd_max = 4096               
set rlim_fd_cur = 1024               

Runlevels

Linux

Once booted, the kernel starts init which then starts rc which starts all scripts belonging to a runlevel. The scripts are stored in /etc/init.d and are linked into /etc/rc.d/rcN.d with N the runlevel number.
The default runlevel is configured in /etc/inittab. It is usually 3 or 5:
# grep default: /etc/inittab                                         
id:3:initdefault:
The actual runlevel can be changed with init. For example to go from 3 to 5:
# init 5                             
  • 0       Shutdown and halt
  • 1       Single-User mode (also S)
  • 2       Multi-user without network
  • 3       Multi-user with network
  • 5       Multi-user with X
  • 6       Reboot
Use chkconfig to configure the programs that will be started at boot in a runlevel.
# chkconfig --list                   
# chkconfig --list sshd              
# chkconfig sshd --level 35 on       
# chkconfig sshd off                 
Debian and Debian based distributions like Ubuntu or Knoppix use the command update-rc.d to manage the runlevels scripts. Default is to start in 2,3,4 and 5 and shutdown in 0,1 and 6.
# update-rc.d sshd defaults          
# update-rc.d sshd start 20 2 3 4 5 . stop 20 0 1 6 .  
# update-rc.d -f sshd remove         
# shutdown -h now (or # poweroff)    

FreeBSD

The BSD boot approach is different from the SysV, there are no runlevels. The final boot state (single user, with or without X) is configured in /etc/ttys. All OS scripts are located in /etc/rc.d/ and in /usr/local/etc/rc.d/ for third-party applications. The activation of the service is configured in /etc/rc.conf and /etc/rc.conf.local. The default behavior is configured in /etc/defaults/rc.conf. The scripts responds at least to start|stop|status.
# /etc/rc.d/sshd status
sshd is running as pid 552.
# shutdown now                       
# exit                               
# shutdown -p now                    
# shutdown -r now                    
The process init can also be used to reach one of the following states level. For example # init 6 for reboot.
  • 0       Halt and turn the power off (signal USR2)
  • 1       Go to single-user mode (signal TERM)
  • 6       Reboot the machine (signal INT)
  • c       Block further logins (signal TSTP)
  • q       Rescan the ttys(5) file (signal HUP)

Windows

Start and stop a service with either the service name or "service description" (shown in the Services Control Panel) as follows:
net stop WSearch
net start WSearch                    
net stop "Windows Search"
net start "Windows Search"           

Reset root password

Linux method 1

At the boot loader (lilo or grub), enter the following boot option:
init=/bin/sh
The kernel will mount the root partition and init will start the bourne shell instead of rc and then a runlevel. Use the command passwd at the prompt to change the password and then reboot. Forget the single user mode as you need the password for that.
If, after booting, the root partition is mounted read only, remount it rw:
# mount -o remount,rw /
# passwd                             
# sync; mount -o remount,ro /        
# reboot

FreeBSD method 1

On FreeBSD, boot in single user mode, remount / rw and use passwd. You can select the single user mode on the boot menu (option 4) which is displayed for 10 seconds at startup. The single user mode will give you a root shell on the / partition.
# mount -u /; mount -a               
# passwd
# reboot

Unixes and FreeBSD and Linux method 2

Other Unixes might not let you go away with the simple init trick. The solution is to mount the root partition from an other OS (like a rescue CD) and change the password on the disk.
  • Boot a live CD or installation CD into a rescue mode which will give you a shell.
  • Find the root partition with fdisk e.g. fdisk /dev/sda
  • Mount it and use chroot:
# mount -o rw /dev/ad4s3a /mnt
# chroot /mnt                        
# passwd
# reboot

Kernel modules

Linux

# lsmod                              
# modprobe isdn                      

FreeBSD

# kldstat                            
# kldload crypto                     

Compile Kernel

Linux

# cd /usr/src/linux
# make mrproper                      
# make oldconfig                     
# make menuconfig                    
# make                               
# make modules                       
# make modules_install               
# make install                       
# reboot

FreeBSD

Optionally update the source tree (in /usr/src) with csup (as of FreeBSD 6.2 or later):
# csup 
I use the following supfile:
*default host=cvsup5.FreeBSD.org  # www.freebsd.org/handbook/cvsup.html#CVSUP-MIRRORS
*default prefix=/usr 
*default base=/var/db
*default release=cvs delete tag=RELENG_7
src-all
To modify and rebuild the kernel, copy the generic configuration file to a new name and edit it as needed (you can also edit the file GENERIC directly). To restart the build after an interruption, add the option NO_CLEAN=YES to the make command to avoid cleaning the objects already build.
# cd /usr/src/sys/i386/conf/
# cp GENERIC MYKERNEL
# cd /usr/src
# make buildkernel KERNCONF=MYKERNEL
# make installkernel KERNCONF=MYKERNEL
To rebuild the full OS:
# make buildworld                    
# make buildkernel                   
# make installkernel
# reboot
# mergemaster -p                     
# make installworld
# mergemaster -i -U                  
# reboot
For small changes in the source you can use NO_CLEAN=yes to avoid rebuilding the whole tree.
# make buildworld NO_CLEAN=yes       
# make buildkernel KERNCONF=MYKERNEL NO_CLEAN=yes

Repair grub

So you broke grub? Boot from a live cd, [find your linux partition under /dev and use fdisk to find the linux partion] mount the linux partition, add /proc and /dev and use grub-install /dev/xyz. Suppose linux lies on /dev/sda6:
# mount /dev/sda6 /mnt               
# mount --bind /proc /mnt/proc       
# mount --bind /dev /mnt/dev         
# chroot /mnt                        
# grub-install /dev/sda              

Processes

Listing | Priority | Background/Foreground | Top | Kill

Listing and PIDs

Each process has a unique number, the PID. A list of all running process is retrieved with ps.
# ps -auxefw                         
However more typical usage is with a pipe or with pgrep:
# ps axww | grep cron
  586  ??  Is     0:01.48 /usr/sbin/cron -s
# ps axjf                            
# ps aux | grep 'ss[h]'              
# pgrep -l sshd                      
# echo $$                            
# fuser -va 22/tcp                   
# pmap PID                           
# fuser -va /home                    
# strace df                          
# truss df                           

Priority

Change the priority of a running process with renice. Negative numbers have a higher priority, the lowest is -20 and "nice" have a positive value.
# renice -5 586                      
586: old priority 0, new priority -5
Start the process with a defined priority with nice. Positive is "nice" or weak, negative is strong scheduling priority. Make sure you know if /usr/bin/nice or the shell built-in is used (check with # which nice).
# nice -n -5 top                     
# nice -n 5 top                      
# nice +5 top                        
While nice changes the CPU scheduler, an other useful command ionice will schedule the disk IO. This is very useful for intensive IO application (e.g. compiling). You can select a class (idle - best effort - real time), the man page is short and well explained.
# ionice c3 -p123                    
# ionice -c2 -n0 firefox             
# ionice -c3 -p$$                    
The last command is very useful to compile (or debug) a large project. Every command launched from this shell will have a lover priority. $$ is your shell pid (try echo $$).
FreeBSD uses idprio/rtprio (0 = max priority, 31 = most idle):
# idprio 31 make                     
# idprio 31 -1234                    
# idprio -t -1234                    

Background/Foreground

When started from a shell, processes can be brought in the background and back to the foreground with [Ctrl]-[Z] (^Z), bg and fg. List the processes with jobs.
# ping cb.vu > ping.log
^Z                                    
# bg                                 
# jobs -l                            
[1]  - 36232 Running                       ping cb.vu > ping.log
[2]  + 36233 Suspended (tty output)        top
# fg %2                              
Use nohup to start a process which has to keep running when the shell is closed (immune to hangups).
# nohup ping -i 60 > ping.log &

Top

The program top displays running information of processes. See also the program htop from htop.sourceforge.net (a more powerful version of top) which runs on Linux and FreeBSD ( ports/sysutils/htop/). While top is running press the key h for a help overview. Useful keys are:
  • u [user name] To display only the processes belonging to the user. Use + or blank to see all users
  • k [pid] Kill the process with pid.
  • 1 To display all processors statistics (Linux only)
  • R Toggle normal/reverse sort.

Signals/Kill

Terminate or send a signal with kill or killall.
# ping -i 60 cb.vu > ping.log &
[1] 4712
# kill -s TERM 4712                  
# killall -1 httpd                   
# pkill -9 http                      
# pkill -TERM -u www                 
# fuser -k -TERM -m /home            
Important signals are:
  • 1       HUP (hang up)
  • 2       INT (interrupt)
  • 3       QUIT (quit)
  • 9       KILL (non-catchable, non-ignorable kill)
  • 15     TERM (software termination signal)

File System

Disk info | Boot | Disk usage | Opened files | Mount/remount | Mount SMB | Mount image | Burn ISO | Create image | Memory disk | Disk performance

Permissions

Change permission and ownership with chmod and chown. The default umask can be changed for all users in /etc/profile for Linux or /etc/login.conf for FreeBSD. The default umask is usually 022. The umask is subtracted from 777, thus umask 022 results in a permission 0f 755.
1 --x execute                        
2 -w- write                          
4 r-- read
  ugo=a                              
# chmod [OPTION] MODE[,MODE] FILE    
# chmod 640 /var/log/maillog         
# chmod u=rw,g=r,o= /var/log/maillog 
# chmod -R o-r /home/*               
# chmod u+s /path/to/prog            
# find / -perm -u+s -print           
# chown user:group /path/to/file     
# chgrp group /path/to/file          
# chmod 640 `find ./ -type f -print` 
# chmod 751 `find ./ -type d -print` 

Disk information

# diskinfo -v /dev/ad2               
# hdparm -I /dev/sda                 
# fdisk /dev/ad2                     
# smartctl -a /dev/ad2               

Boot

FreeBSD

To boot an old kernel if the new kernel doesn't boot, stop the boot at during the count down.
# unload
# load kernel.old
# boot

System mount points/Disk usage

# mount | column -t                  
# df                                 
# cat /proc/partitions               

Disk usage

# du -sh *                           
# du -csh                            
# du -ks * | sort -n -r              
# ls -lSr                            

Who has which files opened

This is useful to find out which file is blocking a partition which has to be unmounted and gives a typical error of:
# umount /home/
umount: unmount of /home             
   failed: Device busy

FreeBSD and most Unixes

# fstat -f /home                     
# fstat -p PID                       
# fstat -u user                      
Find opened log file (or other opened files), say for Xorg:
# ps ax | grep Xorg | awk '{print $1}'
1252
# fstat -p 1252
USER     CMD          PID   FD MOUNT      INUM MODE         SZ|DV R/W
root     Xorg        1252 root /             2 drwxr-xr-x     512  r
root     Xorg        1252 text /usr     216016 -rws--x--x  1679848 r
root     Xorg        1252    0 /var     212042 -rw-r--r--   56987  w
The file with inum 212042 is the only file in /var:
# find -x /var -inum 212042
/var/log/Xorg.0.log

Linux

Find opened files on a mount point with fuser or lsof:
# fuser -m /home                     
# lsof /home
COMMAND   PID    USER   FD   TYPE DEVICE    SIZE     NODE NAME
tcsh    29029 eedcoba  cwd    DIR   0,18   12288  1048587 /home/eedcoba (guam:/home)
lsof    29140 eedcoba  cwd    DIR   0,18   12288  1048587 /home/eedcoba (guam:/home)
About an application:
ps ax | grep Xorg | awk '{print $1}'
3324
# lsof -p 3324
COMMAND   PID    USER   FD   TYPE DEVICE    SIZE    NODE NAME
Xorg    3324 root    0w   REG        8,6   56296      12492 /var/log/Xorg.0.log
About a single file:
# lsof /var/log/Xorg.0.log
COMMAND  PID USER   FD   TYPE DEVICE  SIZE  NODE NAME
Xorg    3324 root    0w   REG    8,6 56296 12492 /var/log/Xorg.0.log

Mount/remount a file system

For example the cdrom. If listed in /etc/fstab:
# mount /cdrom
Or find the device in /dev/ or with dmesg

FreeBSD

# mount -v -t cd9660 /dev/cd0c /mnt  
# mount_cd9660 /dev/wcd0c /cdrom     
# mount -v -t msdos /dev/fd0c /mnt   
Entry in /etc/fstab:
# Device                Mountpoint      FStype  Options         Dump    Pass#
/dev/acd0               /cdrom          cd9660  ro,noauto       0       0
To let users do it:
# sysctl vfs.usermount=1  

Linux

# mount -t auto /dev/cdrom /mnt/cdrom   
# mount /dev/hdc -t iso9660 -r /cdrom   
# mount /dev/scd0 -t iso9660 -r /cdrom  
# mount /dev/sdc0 -t ntfs-3g /windows   
Entry in /etc/fstab:
/dev/cdrom   /media/cdrom  subfs noauto,fs=cdfss,ro,procuid,nosuid,nodev,exec 0 0

Mount a FreeBSD partition with Linux

Find the partition number containing with fdisk, this is usually the root partition, but it could be an other BSD slice too. If the FreeBSD has many slices, they are the one not listed in the fdisk table, but visible in /dev/sda* or /dev/hda*.
# fdisk /dev/sda                     
/dev/sda3   *        5357        7905    20474842+  a5  FreeBSD
# mount -t ufs -o ufstype=ufs2,ro /dev/sda3 /mnt
/dev/sda10 = /tmp; /dev/sda11 /usr   

Remount

Remount a device without unmounting it. Necessary for fsck for example
# mount -o remount,ro /              
# mount -o ro /                      
Copy the raw data from a cdrom into an iso image:
# dd if=/dev/cd0c of=file.iso

Add swap on-the-fly

Suppose you need more swap (right now), say a 2GB file /swap2gb (Linux only).
# dd if=/dev/zero of=/swap2gb bs=1024k count=2000
# mkswap /swap2gb                    
# swapon /swap2gb                    
# swapoff /swap2gb                   
# rm /swap2gb

Mount an SMB share

Suppose we want to access the SMB share myshare on the computer smbserver, the address as typed on a Windows PC is \\smbserver\myshare\. We mount on /mnt/smbshare. Warning> cifs wants an IP or DNS name, not a Windows name.

Linux

# smbclient -U user -I 192.168.16.229 -L //smbshare/    
# mount -t smbfs -o username=winuser //smbserver/myshare /mnt/smbshare
# mount -t cifs -o username=winuser,password=winpwd //192.168.16.229/myshare /mnt/share
Additionally with the package mount.cifs it is possible to store the credentials in a file, for example /home/user/.smb:
username=winuser
password=winpwd
And mount as follow:
# mount -t cifs -o credentials=/home/user/.smb //192.168.16.229/myshare /mnt/smbshare

FreeBSD

Use -I to give the IP (or DNS name); smbserver is the Windows name.
# smbutil view -I 192.168.16.229 //winuser@smbserver    
# mount_smbfs -I 192.168.16.229 //winuser@smbserver/myshare /mnt/smbshare

Mount an image

Linux loop-back

# mount -t iso9660 -o loop file.iso /mnt                
# mount -t ext3 -o loop file.img /mnt                   

FreeBSD

With memory device (do # kldload md.ko if necessary):
# mdconfig -a -t vnode -f file.iso -u 0
# mount -t cd9660 /dev/md0 /mnt
# umount /mnt; mdconfig -d -u 0                         
Or with virtual node:
# vnconfig /dev/vn0c file.iso; mount -t cd9660 /dev/vn0c /mnt
# umount /mnt; vnconfig -u /dev/vn0c                    

Solaris and FreeBSD

with loop-back file interface or lofi:
# lofiadm -a file.iso
# mount -F hsfs -o ro /dev/lofi/1 /mnt
# umount /mnt; lofiadm -d /dev/lofi/1                   

Create and burn an ISO image

This will copy the cd or DVD sector for sector. Without conv=notrunc, the image will be smaller if there is less content on the cd. See below and the dd examples.
# dd if=/dev/hdc of=/tmp/mycd.iso bs=2048 conv=notrunc
Use mkisofs to create a CD/DVD image from files in a directory. To overcome the file names restrictions: -r enables the Rock Ridge extensions common to UNIX systems, -J enables Joliet extensions used by Microsoft systems. -L allows ISO9660 filenames to begin with a period.
# mkisofs -J -L -r -V TITLE -o imagefile.iso /path/to/dir
On FreeBSD, mkisofs is found in the ports in sysutils/cdrtools.

Burn a CD/DVD ISO image

FreeBSD

FreeBSD does not enable DMA on ATAPI drives by default. DMA is enabled with the sysctl command and the arguments below, or with /boot/loader.conf with the following entries:
hw.ata.ata_dma="1"
hw.ata.atapi_dma="1"
Use burncd with an ATAPI device ( burncd is part of the base system) and cdrecord (in sysutils/cdrtools) with a SCSI drive.
# burncd -f /dev/acd0 data imagefile.iso fixate      
# cdrecord -scanbus                  
# cdrecord dev=1,0,0 imagefile.iso

Linux

Also use cdrecord with Linux as described above. Additionally it is possible to use the native ATAPI interface which is found with:
# cdrecord dev=ATAPI -scanbus
And burn the CD/DVD as above.

dvd+rw-tools

The dvd+rw-tools package (FreeBSD: ports/sysutils/dvd+rw-tools) can do it all and includes growisofs to burn CDs or DVDs. The examples refer to the dvd device as /dev/dvd which could be a symlink to /dev/scd0 (typical scsi on Linux) or /dev/cd0 (typical FreeBSD) or /dev/rcd0c (typical NetBSD/OpenBSD character SCSI) or /dev/rdsk/c0t1d0s2 (Solaris example of a character SCSI/ATAPI CD-ROM device). There is a nice documentation with examples on the FreeBSD handbook chapter 18.7 http://www.freebsd.org/handbook/creating-dvds.html.
                       
# growisofs -dvd-compat -Z /dev/dvd=imagefile.iso     
# growisofs -dvd-compat -Z /dev/dvd -J -R /p/to/data  

Convert a Nero .nrg file to .iso

Nero simply adds a 300Kb header to a normal iso image. This can be trimmed with dd.
# dd bs=1k if=imagefile.nrg of=imagefile.iso skip=300

Convert a bin/cue image to .iso

The little bchunk program http://freshmeat.net/projects/bchunk/ can do this. It is in the FreeBSD ports in sysutils/bchunk.
# bchunk imagefile.bin imagefile.cue imagefile.iso

Create a file based image

For example a partition of 1GB using the file /usr/vdisk.img. Here we use the vnode 0, but it could also be 1.

FreeBSD

# dd if=/dev/random of=/usr/vdisk.img bs=1K count=1M
# mdconfig -a -t vnode -f /usr/vdisk.img -u 0         
# bsdlabel -w /dev/md0
# newfs /dev/md0c
# mount /dev/md0c /mnt
# umount /mnt; mdconfig -d -u 0; rm /usr/vdisk.img    
The file based image can be automatically mounted during boot with an entry in /etc/rc.conf and /etc/fstab. Test your setup with # /etc/rc.d/mdconfig start (first delete the md0 device with # mdconfig -d -u 0).
Note however that this automatic setup will only work if the file image is NOT on the root partition. The reason is that the /etc/rc.d/mdconfig script is executed very early during boot and the root partition is still read-only. Images located outside the root partition will be mounted later with the script /etc/rc.d/mdconfig2.
/boot/loader.conf:
md_load="YES"
/etc/rc.conf:
# mdconfig_md0="-t vnode -f /usr/vdisk.img"          
/etc/fstab: (The 0 0 at the end is important, it tell fsck to ignore this device, as is does not exist yet)
/dev/md0                /usr/vdisk      ufs     rw              0       0
It is also possible to increase the size of the image afterward, say for example 300 MB larger.
# umount /mnt; mdconfig -d -u 0
# dd if=/dev/zero bs=1m count=300 >> /usr/vdisk.img
# mdconfig -a -t vnode -f /usr/vdisk.img -u 0
# growfs /dev/md0
# mount /dev/md0c /mnt                                

Linux

# dd if=/dev/zero of=/usr/vdisk.img bs=1024k count=1024
# mkfs.ext3 /usr/vdisk.img
# mount -o loop /usr/vdisk.img /mnt
# umount /mnt; rm /usr/vdisk.img                      

Linux with losetup

/dev/zero is much faster than urandom, but less secure for encryption.
# dd if=/dev/urandom of=/usr/vdisk.img bs=1024k count=1024
# losetup /dev/loop0 /usr/vdisk.img                   
# mkfs.ext3 /dev/loop0
# mount /dev/loop0 /mnt
# losetup -a                                          
# umount /mnt
# losetup -d /dev/loop0                               
# rm /usr/vdisk.img

Create a memory file system

A memory based file system is very fast for heavy IO application. How to create a 64 MB partition mounted on /memdisk:

FreeBSD

# mount_mfs -o rw -s 64M md /memdisk
# umount /memdisk; mdconfig -d -u 0                   
md     /memdisk     mfs     rw,-s64M    0   0         

Linux

# mount -t tmpfs -osize=64m tmpfs /memdisk

Disk performance

Read and write a 1 GB file on partition ad4s3c (/home)
# time dd if=/dev/ad4s3c of=/dev/null bs=1024k count=1000
# time dd if=/dev/zero bs=1024k count=1000 of=/home/1Gb.file
# hdparm -tT /dev/hda      

Network

Routing | Additional IP | Change MAC | Ports | Firewall | IP Forward | NAT | DNS | DHCP | Traffic | QoS | NIS | Netcat

Debugging (See also Traffic analysis)

Linux

# ethtool eth0              
# ethtool -s eth0 speed 100 duplex full 
# ethtool -s eth0 autoneg off 
# ethtool -p eth1           
# ip link show              
# ip link set eth0 up       
# ip addr show              
# ip neigh show             

Other OSes

# ifconfig fxp0             
# arp -a                    
# ping cb.vu                
# traceroute cb.vu          
# ifconfig fxp0 media 100baseTX mediaopt full-duplex 
# netstat -s                
Additional commands which are not always installed per default but easy to find:
# arping 192.168.16.254     
# tcptraceroute -f 5 cb.vu  

Routing

Print routing table

# route -n                  
# netstat -rn               
# route print               

Add and delete a route

FreeBSD

# route add 212.117.0.0/16 192.168.1.1
# route delete 212.117.0.0/16
# route add default 192.168.1.1
Add the route permanently in /etc/rc.conf
static_routes="myroute"
route_myroute="-net 212.117.0.0/16 192.168.1.1"

Linux

# route add -net 192.168.20.0 netmask 255.255.255.0 gw 192.168.16.254
# ip route add 192.168.20.0/24 via 192.168.16.254       
# route add -net 192.168.20.0 netmask 255.255.255.0 dev eth0
# route add default gw 192.168.51.254
# ip route add default via 192.168.51.254 dev eth0      
# route delete -net 192.168.20.0 netmask 255.255.255.0

Solaris

# route add -net 192.168.20.0 -netmask 255.255.255.0 192.168.16.254
# route add default 192.168.51.254 1                    
# route change default 192.168.50.254 1
Permanent entries are set in entry in /etc/defaultrouter.

Windows

# Route add 192.168.50.0 mask 255.255.255.0 192.168.51.253
# Route add 0.0.0.0 mask 0.0.0.0 192.168.51.254
Use add -p to make the route persistent.

Configure additional IP addresses

Linux

# ifconfig eth0 192.168.50.254 netmask 255.255.255.0       
# ifconfig eth0:0 192.168.51.254 netmask 255.255.255.0     
# ip addr add 192.168.50.254/24 dev eth0                   
# ip addr add 192.168.51.254/24 dev eth0 label eth0:1

FreeBSD

# ifconfig fxp0 inet 192.168.50.254/24                     
# ifconfig fxp0 alias 192.168.51.254 netmask 255.255.255.0 
# ifconfig fxp0 -alias 192.168.51.254                      
Permanent entries in /etc/rc.conf
ifconfig_fxp0="inet 192.168.50.254  netmask 255.255.255.0"
ifconfig_fxp0_alias0="192.168.51.254 netmask 255.255.255.0"

Solaris

Check the settings with ifconfig -a
# ifconfig hme0 plumb                                      
# ifconfig hme0 192.168.50.254 netmask 255.255.255.0 up    
# ifconfig hme0:1 192.168.51.254 netmask 255.255.255.0 up  

Change MAC address

Normally you have to bring the interface down before the change. Don't tell me why you want to change the MAC address...
# ifconfig eth0 down
# ifconfig eth0 hw ether 00:01:02:03:04:05      
# ifconfig fxp0 link 00:01:02:03:04:05          
# ifconfig hme0 ether 00:01:02:03:04:05         
# sudo ifconfig en0 ether 00:01:02:03:04:05     
# sudo ifconfig en0 lladdr 00:01:02:03:04:05    
Many tools exist for Windows. For example etherchange http://ntsecurity.nu/toolbox/etherchange. Or look for "Mac Makeup", "smac".

Ports in use

Listening open ports:
# netstat -an | grep LISTEN
# lsof -i                  
# socklist                 
# sockstat -4              
# netstat -anp --udp --tcp | grep LISTEN        
# netstat -tup             
# netstat -tupl            
# netstat -ano             

Firewall

Check if a firewall is running (typical configuration only):

Linux

# iptables -L -n -v                  
Open the iptables firewall
# iptables -P INPUT       ACCEPT     
# iptables -P FORWARD     ACCEPT
# iptables -P OUTPUT      ACCEPT
# iptables -Z                        
# iptables -F                        
# iptables -X                        

FreeBSD

# ipfw show                          
# ipfw list 65535 
# sysctl net.inet.ip.fw.enable=0     
# sysctl net.inet.ip.fw.enable=1     

IP Forward for routing

Linux

Check and then enable IP forward with:
# cat /proc/sys/net/ipv4/ip_forward  
# echo 1 > /proc/sys/net/ipv4/ip_forward
or edit /etc/sysctl.conf with:
net.ipv4.ip_forward = 1

FreeBSD

Check and enable with:
# sysctl net.inet.ip.forwarding      
# sysctl net.inet.ip.forwarding=1
# sysctl net.inet.ip.fastforwarding=1	
Permanent with entry in /etc/rc.conf:
gateway_enable="YES"                 

Solaris

# ndd -set /dev/ip ip_forwarding 1   

NAT Network Address Translation

Linux

# iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE	
# iptables -t nat -A PREROUTING -p tcp -d 78.31.70.238 --dport 20022 -j DNAT \
--to 192.168.16.44:22           
# iptables -t nat -A PREROUTING -p tcp -d 78.31.70.238 --dport 993:995 -j DNAT \
--to 192.168.16.254:993-995     
# ip route flush cache
# iptables -L -t nat            
Delete the port forward with -D instead of -A. The program netstat-nat http://tweegy.nl/projects/netstat-nat is very useful to track connections (it uses /proc/net/ip_conntrack or /proc/net/nf_conntrack).
# netstat-nat -n                

FreeBSD

# natd -s -m -u -dynamic -f /etc/natd.conf -n fxp0
Or edit /etc/rc.conf with:
firewall_enable="YES"           
firewall_type="open"            
natd_enable="YES"               
natd_interface="tun0"           
natd_flags="-s -m -u -dynamic -f /etc/natd.conf"
Port forward with:
# cat /etc/natd.conf 
same_ports yes
use_sockets yes
unregistered_only
# redirect_port tcp insideIP:2300-2399 3300-3399  
redirect_port udp 192.168.51.103:7777 7777

DNS

On Unix the DNS entries are valid for all interfaces and are stored in /etc/resolv.conf. The domain to which the host belongs is also stored in this file. A minimal configuration is:
nameserver 78.31.70.238
search sleepyowl.net intern.lab
domain sleepyowl.net
Check the system domain name with:
# hostname -d                        

Windows

On Windows the DNS are configured per interface. To display the configured DNS and to flush the DNS cache use:
# ipconfig /?                        
# ipconfig /all                      

Flush DNS

Flush the OS DNS cache, some application using their own cache (e.g. Firefox) and will be unaffected.
# /etc/init.d/nscd restart           
# lookupd -flushcache                
# dscacheutil -flushcache            
# ipconfig /flushdns                 

Forward queries

Dig is you friend to test the DNS settings. For example the public DNS server 213.133.105.2 ns.second-ns.de can be used for testing. See from which server the client receives the answer (simplified answer).
# dig sleepyowl.net
sleepyowl.net.          600     IN      A       78.31.70.238
;; SERVER: 192.168.51.254#53(192.168.51.254)
The router 192.168.51.254 answered and the response is the A entry. Any entry can be queried and the DNS server can be selected with @:
# dig MX google.com
# dig @127.0.0.1 NS sun.com          
# dig @204.97.212.10 NS MX heise.de  
# dig AXFR @ns1.xname.org cb.vu      
The program host is also powerful.
# host -t MX cb.vu                   
# host -t NS -T sun.com              
# host -a sleepyowl.net              

Reverse queries

Find the name belonging to an IP address (in-addr.arpa.). This can be done with dig, host and nslookup:
# dig -x 78.31.70.238
# host 78.31.70.238
# nslookup 78.31.70.238

/etc/hosts

Single hosts can be configured in the file /etc/hosts instead of running named locally to resolve the hostname queries. The format is simple, for example:
78.31.70.238   sleepyowl.net   sleepyowl
The priority between hosts and a dns query, that is the name resolution order, can be configured in /etc/nsswitch.conf AND /etc/host.conf. The file also exists on Windows, it is usually in:
C:\WINDOWS\SYSTEM32\DRIVERS\ETC

DHCP

Linux

Some distributions (SuSE) use dhcpcd as client. The default interface is eth0.
# dhcpcd -n eth0           
# dhcpcd -k eth0           
The lease with the full information is stored in:
/var/lib/dhcpcd/dhcpcd-eth0.info

FreeBSD

FreeBSD (and Debian) uses dhclient. To configure an interface (for example bge0) run:
# dhclient bge0
The lease with the full information is stored in:
/var/db/dhclient.leases.bge0
Use
/etc/dhclient.conf
to prepend options or force different options:
# cat /etc/dhclient.conf
interface "rl0" {
    prepend domain-name-servers 127.0.0.1;
    default domain-name "sleepyowl.net";
    supersede domain-name "sleepyowl.net";
}

Windows

The dhcp lease can be renewed with ipconfig:
# ipconfig /renew          
# ipconfig /renew LAN      
# ipconfig /release WLAN   
Yes it is a good idea to rename you adapter with simple names!

Traffic analysis

Bmon http://people.suug.ch/~tgr/bmon/ is a small console bandwidth monitor and can display the flow on different interfaces.

Sniff with tcpdump

# tcpdump -nl -i bge0 not port ssh and src \(192.168.16.121 or 192.168.16.54\)
# tcpdump -n -i eth1 net 192.168.16.121           
# tcpdump -n -i eth1 net 192.168.16.0/24          
# tcpdump -l > dump && tail -f dump               
# tcpdump -i rl0 -w traffic.rl0                   
# tcpdump -i rl0 -s 0 -w traffic.rl0              
# tcpdump -r traffic.rl0                          
# tcpdump port 80                                 
# tcpdump host google.com
# tcpdump -i eth0 -X port \(110 or 143\)          
# tcpdump -n -i eth0 icmp                         
# tcpdump -i eth0 -s 0 -A port 80 | grep GET      
Additional important options:
  • -A     Print each packets in clear text (without header)
  • -X     Print packets in hex and ASCII
  • -l     Make stdout line buffered
  • -D     Print all interfaces available
On Windows use windump from www.winpcap.org. Use windump -D to list the interfaces.

Scan with nmap

Nmap http://insecure.org/nmap/ is a port scanner with OS detection, it is usually installed on most distributions and is also available for Windows. If you don't scan your servers, hackers do it for you...
# nmap cb.vu               
# nmap -sP 192.168.16.0/24 
# nmap -sS -sV -O cb.vu    
PORT      STATE  SERVICE             VERSION
22/tcp    open   ssh                 OpenSSH 3.8.1p1 FreeBSD-20060930 (protocol 2.0)
25/tcp    open   smtp                Sendmail smtpd 8.13.6/8.13.6
80/tcp    open   http                Apache httpd 2.0.59 ((FreeBSD) DAV/2 PHP/4.
[...]
Running: FreeBSD 5.X
Uptime 33.120 days (since Fri Aug 31 11:41:04 2007)
Other non standard but useful tools are hping (www.hping.org) an IP packet assembler/analyzer and fping (fping.sourceforge.net). fping can check multiple hosts in a round-robin fashion.

Traffic control (QoS)

Traffic control manages the queuing, policing, scheduling, and other traffic parameters for a network. The following examples are simple practical uses of the Linux and FreeBSD capabilities to better use the available bandwidth.

Limit upload

DSL or cable modems have a long queue to improve the upload throughput. However filling the queue with a fast device (e.g. ethernet) will dramatically decrease the interactivity. It is therefore useful to limit the device upload rate to match the physical capacity of the modem, this should greatly improve the interactivity. Set to about 90% of the modem maximal (cable) speed.

Linux

For a 512 Kbit upload modem.
# tc qdisc add dev eth0 root tbf rate 480kbit latency 50ms burst 1540
# tc -s qdisc ls dev eth0                         
# tc qdisc del dev eth0 root                      
# tc qdisc change dev eth0 root tbf rate 220kbit latency 50ms burst 1540

FreeBSD

FreeBSD uses the dummynet traffic shaper which is configured with ipfw. Pipes are used to set limits the bandwidth in units of [K|M]{bit/s|Byte/s}, 0 means unlimited bandwidth. Using the same pipe number will reconfigure it. For example limit the upload bandwidth to 500 Kbit.
# kldload dummynet                                
# ipfw pipe 1 config bw 500Kbit/s                 
# ipfw add pipe 1 ip from me to any               

Quality of service

Linux

Priority queuing with tc to optimize VoIP. See the full example on voip-info.org or www.howtoforge.com. Suppose VoIP uses udp on ports 10000:11024 and device eth0 (could also be ppp0 or so). The following commands define the QoS to three queues and force the VoIP traffic to queue 1 with QoS 0x1e (all bits set). The default traffic flows into queue 3 and QoS Minimize-Delay flows into queue 2.
# tc qdisc add dev eth0 root handle 1: prio priomap 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 0
# tc qdisc add dev eth0 parent 1:1 handle 10: sfq
# tc qdisc add dev eth0 parent 1:2 handle 20: sfq
# tc qdisc add dev eth0 parent 1:3 handle 30: sfq
# tc filter add dev eth0 protocol ip parent 1: prio 1 u32 \
  match ip dport 10000 0x3C00 flowid 1:1          
  match ip dst 123.23.0.1 flowid 1:1              
Status and remove with
# tc -s qdisc ls dev eth0                         
# tc qdisc del dev eth0 root                      

Calculate port range and mask

The tc filter defines the port range with port and mask which you have to calculate. Find the 2^N ending of the port range, deduce the range and convert to HEX. This is your mask. Example for 10000 -> 11024, the range is 1024.
# 2^13 (8192) < 10000 < 2^14 (16384)              
# echo "obase=16;(2^14)-1024" | bc                

FreeBSD

The max link bandwidth is 500Kbit/s and we define 3 queues with priority 100:10:1 for VoIP:ssh:all the rest.
# ipfw pipe 1 config bw 500Kbit/s 
# ipfw queue 1 config pipe 1 weight 100
# ipfw queue 2 config pipe 1 weight 10
# ipfw queue 3 config pipe 1 weight 1
# ipfw add 10 queue 1 proto udp dst-port 10000-11024
# ipfw add 11 queue 1 proto udp dst-ip 123.23.0.1 
# ipfw add 20 queue 2 dsp-port ssh
# ipfw add 30 queue 3 from me to any              
Status and remove with
# ipfw list                                       
# ipfw pipe list                                  
# ipfw flush                                      

NIS Debugging

Some commands which should work on a well configured NIS client:
# ypwhich                  
# domainname               
# ypcat group              
# cd /var/yp && make       
# rpcinfo -p servername    
Is ypbind running?
# ps auxww | grep ypbind
/usr/sbin/ypbind -s -m -S servername1,servername2	
/usr/sbin/ypbind           
# yppoll passwd.byname
Map passwd.byname has order number 1190635041. Mon Sep 24 13:57:21 2007
The master server is servername.domain.net.

Linux

# cat /etc/yp.conf
ypserver servername
domain domain.net broadcast

Netcat

Netcat http://netcat.sourceforge.net (nc) is better known as the "network Swiss Army Knife", it can manipulate, create or read/write TCP/IP connections. Here some useful examples, there are many more on the net, for example g-loaded.eu[...] http://www.g-loaded.eu/2006/11/06/netcat-a-couple-of-useful-examples and here http://www.terminally-incoherent.com/blog/2007/08/07/few-useful-netcat-tricks.
You might need to use the command netcat instead of nc. Also see the similar command socat.

File transfer

Copy a large folder over a raw tcp connection. The transfer is very quick (no protocol overhead) and you don't need to mess up with NFS or SMB or FTP or so, simply make the file available on the server, and get it from the client. Here 192.168.1.1 is the server IP address.
 tar -cf - -C VIDEO_TS . | nc -l -p 4444         
 nc 192.168.1.1 4444 | tar xpf - -C VIDEO_TS     
 cat largefile | nc -l 5678                      
 nc 192.168.1.1 5678 > largefile                 
 dd if=/dev/da0 | nc -l 4444                     
 nc 192.168.1.1 4444 | dd of=/dev/da0            
 nc 192.168.1.1 4444 | dd of=da0.img             

Other hacks

Specially here, you must know what you are doing.

Remote shell

Option -e only on the Windows version? Or use nc 1.10.
# nc -lp 4444 -e /bin/bash                        
# nc -lp 4444 -e cmd.exe                          

Emergency web server

Serve a single file on port 80 in a loop.
# while true; do nc -l -p 80 < unixtoolbox.xhtml; done

Chat

Alice and Bob can chat over a simple TCP socket. The text is transferred with the enter key.
 nc -lp 4444
 nc 192.168.1.1 4444

SSH SCP

Public key | Fingerprint | SCP | Tunneling

Public key authentication

Connect to a host without password using public key authentication. The idea is to append your public key to the authorized_keys2 file on the remote host. For this example let's connect host-client to host-server, the key is generated on the client. With cygwin you might have to create your home directoy and the .ssh directory with # mkdir -p /home/USER/.ssh
  • Use ssh-keygen to generate a key pair. ~/.ssh/id_dsa is the private key, ~/.ssh/id_dsa.pub is the public key.
  • Copy only the public key to the server and append it to the file ~/.ssh/authorized_keys2 on your home on the server.
# ssh-keygen -t dsa -N ''
# cat ~/.ssh/id_dsa.pub | ssh you@host-server "cat - >> ~/.ssh/authorized_keys2"

Using the Windows client from ssh.com

The non commercial version of the ssh.com client can be downloaded the main ftp site: ftp.ssh.com/pub/ssh/. Keys generated by the ssh.com client need to be converted for the OpenSSH server. This can be done with the ssh-keygen command.
  • Create a key pair with the ssh.com client: Settings - User Authentication - Generate New....
  • I use Key type DSA; key length 2048.
  • Copy the public key generated by the ssh.com client to the server into the ~/.ssh folder.
  • The keys are in C:\Documents and Settings\%USERNAME%\Application Data\SSH\UserKeys.
  • Use the ssh-keygen command on the server to convert the key:
    # cd ~/.ssh
    # ssh-keygen -i -f keyfilename.pub >> authorized_keys2
    
Notice: We used a DSA key, RSA is also possible. The key is not protected by a password.

Using putty for Windows

Putty http://www.chiark.greenend.org.uk/~sgtatham/putty/download.html is a simple and free ssh client for Windows.
  • Create a key pair with the puTTYgen program.
  • Save the public and private keys (for example into C:\Documents and Settings\%USERNAME%\.ssh).
  • Copy the public key to the server into the ~/.ssh folder:
    # scp .ssh/puttykey.pub [email protected]:.ssh/
  • Use the ssh-keygen command on the server to convert the key for OpenSSH:
    # cd ~/.ssh
    # ssh-keygen -i -f puttykey.pub >> authorized_keys2
    
  • Point the private key location in the putty settings: Connection - SSH - Auth

Check fingerprint

At the first login, ssh will ask if the unknown host with the fingerprint has to be stored in the known hosts. To avoid a man-in-the-middle attack the administrator of the server can send you the server fingerprint which is then compared on the first login. Use ssh-keygen -l to get the fingerprint (on the server):
# ssh-keygen -l -f /etc/ssh/ssh_host_rsa_key.pub      
2048 61:33:be:9b:ae:6c:36:31:fd:83:98:b7:99:2d:9f:cd /etc/ssh/ssh_host_rsa_key.pub
# ssh-keygen -l -f /etc/ssh/ssh_host_dsa_key.pub      
2048 14:4a:aa:d9:73:25:46:6d:0a:48:35:c7:f4:16:d4:ee /etc/ssh/ssh_host_dsa_key.pub
Now the client connecting to this server can verify that he is connecting to the right server:
# ssh linda
The authenticity of host 'linda (192.168.16.54)' can't be established.
DSA key fingerprint is 14:4a:aa:d9:73:25:46:6d:0a:48:35:c7:f4:16:d4:ee.
Are you sure you want to continue connecting (yes/no)? yes

Secure file transfer

Some simple commands:
# scp file.txt host-two:/tmp
# scp joe@host-two:/www/*.html /www/tmp
# scp -r joe@host-two:/www /www/tmp
In Konqueror or Midnight Commander it is possible to access a remote file system with the address fish://user@gate. However the implementation is very slow.
Furthermore it is possible to mount a remote folder with sshfs a file system client based on SCP. See fuse sshfs http://fuse.sourceforge.net/sshfs.html.
ssh_exchange_identification: Connection closed by remote host
With this error try the following on the server:
echo 'SSHD: ALL' >> /etc/hosts.allow
/etc/init.d/sshd restart

Tunneling

SSH tunneling allows to forward or reverse forward a port over the SSH connection, thus securing the traffic and accessing ports which would otherwise be blocked. This only works with TCP. The general nomenclature for forward and reverse is (see also ssh and NAT example):
# ssh -L localport:desthost:destport user@gate  
# ssh -R destport:desthost:localport user@gate  
    
# ssh -X user@gate   
This will connect to gate and forward the local port to the host desthost:destport. Note desthost is the destination host as seen by the gate, so if the connection is to the gate, then desthost is localhost. More than one port forward is possible.

Direct forward on the gate

Let say we want to access the CVS (port 2401) and http (port 80) which are running on the gate. This is the simplest example, desthost is thus localhost, and we use the port 8080 locally instead of 80 so we don't need to be root. Once the ssh session is open, both services are accessible on the local ports.
# ssh -L 2401:localhost:2401 -L 8080:localhost:80 user@gate

Netbios and remote desktop forward to a second server

Let say a Windows smb server is behind the gate and is not running ssh. We need access to the smb share and also remote desktop to the server.
# ssh -L 139:smbserver:139 -L 3388:smbserver:3389 user@gate
The smb share can now be accessed with \\127.0.0.1\, but only if the local share is disabled, because the local share is listening on port 139.
It is possible to keep the local share enabled, for this we need to create a new virtual device with a new IP address for the tunnel, the smb share will be connected over this address. Furthermore the local RDP is already listening on 3389, so we choose 3388. For this example let's use a virtual IP of 10.1.1.1.
  • With putty use Source port=10.1.1.1:139. It is possible to create multiple loop devices and tunnel. On Windows 2000, only putty worked for me. On Windows Vista also forward the port 445 in addition to the port 139. Also on Vista the patch KB942624 prevents the port 445 to be forwarded, so I had to uninstall this path in Vista.
  • With the ssh.com client, disable "Allow local connections only". Since ssh.com will bind to all addresses, only a single share can be connected.
Now create the loopback interface with IP 10.1.1.1:
  • # System->Control Panel->Add Hardware # Yes, Hardware is already connected # Add a new hardware device (at bottom).
  • # Install the hardware that I manually select # Network adapters # Microsoft , Microsoft Loopback Adapter.
  • Configure the IP address of the fake device to 10.1.1.1 mask 255.255.255.0, no gateway.
  • advanced->WINS, Enable LMHosts Lookup; Disable NetBIOS over TCP/IP.
  • # Enable Client for Microsoft Networks. # Disable File and Printer Sharing for Microsoft Networks.
I HAD to reboot for this to work. Now connect to the smb share with \\10.1.1.1 and remote desktop to 10.1.1.1:3388.

Debug

If it is not working:
  • Are the ports forwarded: netstat -an? Look at 0.0.0.0:139 or 10.1.1.1:139
  • Does telnet 10.1.1.1 139 connect?
  • You need the checkbox "Local ports accept connections from other hosts".
  • Is "File and Printer Sharing for Microsoft Networks" disabled on the loopback interface?

Connect two clients behind NAT

Suppose two clients are behind a NAT gateway and client cliadmin has to connect to client cliuser (the destination), both can login to the gate with ssh and are running Linux with sshd. You don't need root access anywhere as long as the ports on gate are above 1024. We use 2022 on gate. Also since the gate is used locally, the option GatewayPorts is not necessary.
On client cliuser (from destination to gate):
# ssh -R 2022:localhost:22 user@gate            
On client cliadmin (from host to gate):
# ssh -L 3022:localhost:2022 admin@gate         
Now the admin can connect directly to the client cliuser with:
# ssh -p 3022 admin@localhost                   

Connect to VNC behind NAT

Suppose a Windows client with VNC listening on port 5900 has to be accessed from behind NAT. On client cliwin to gate:
# ssh -R 15900:localhost:5900 user@gate
On client cliadmin (from host to gate):
# ssh -L 5900:localhost:15900 admin@gate
Now the admin can connect directly to the client VNC with:
# vncconnect -display :0 localhost

Dig a multi-hop ssh tunnel

Suppose you can not reach a server directly with ssh, but only via multiple intermediate hosts (for example because of routing issues). Sometimes it is still necessary to get a direct client - server connection, for example to copy files with scp, or forward other ports like smb or vnc. One way to do this is to chain tunnels together to forward a port to the server along the hops. This "carrier" port only reaches its final destination on the last connection to the server.
Suppose we want to forward the ssh port from a client to a server over two hops. Once the tunnel is build, it is possible to connect to the server directly from the client (and also add an other port forward).

Create tunnel in one shell

client -> host1 -> host2 -> server and dig tunnel 5678
# ssh -L5678:localhost:5678 host1        
# ssh -L5678:localhost:5678 host2        
# ssh -L5678:localhost:22 server         

Use tunnel with an other shell

client -> server using tunnel 5678
# ssh -p 5678 localhost                         
# scp -P 5678 myfile localhost:/tmp/            
# rsync -e 'ssh -p 5678' myfile localhost:/tmp/ 

Autoconnect and keep alive script

I use variations of the following script to keep a machine reacheable over a reverse ssh tunnel. The connection is automatically rebuilt if closed. You can add multiple -L or -R tunnels on one line.
#!/bin/sh
COMMAND="ssh -N -f -g -R 3022:localhost:22 [email protected]"
pgrep -f -x "$COMMAND" > /dev/null 2>&1 || $COMMAND
exit 0
1 * * * * colin /home/colin/port_forward.sh     

VPN with SSH

As of version 4.3, OpenSSH can use the tun/tap device to encrypt a tunnel. This is very similar to other TLS based VPN solutions like OpenVPN. One advantage with SSH is that there is no need to install and configure additional software. Additionally the tunnel uses the SSH authentication like pre shared keys. The drawback is that the encapsulation is done over TCP which might result in poor performance on a slow link. Also the tunnel is relying on a single (fragile) TCP connection. This technique is very useful for a quick IP based VPN setup. There is no limitation as with the single TCP port forward, all layer 3/4 protocols like ICMP, TCP/UDP, etc. are forwarded over the VPN. In any case, the following options are needed in the sshd_conf file:
PermitRootLogin yes
PermitTunnel yes

Single P2P connection

Here we are connecting two hosts, hclient and hserver with a peer to peer tunnel. The connection is started from hclient to hserver and is done as root. The tunnel end points are 10.0.1.1 (server) and 10.0.1.2 (client) and we create a device tun5 (this could also be an other number). The procedure is very simple:
  • Connect with SSH using the tunnel option -w
  • Configure the IP addresses of the tunnel. Once on the server and once on the client.

Connect to the server

Connection started on the client and commands are executed on the server.

Server is on Linux

 ssh -w5:5 root@hserver
 ifconfig tun5 10.0.1.1 netmask 255.255.255.252   

Server is on FreeBSD

 ssh -w5:5 root@hserver
 ifconfig tun5 10.0.1.1 10.0.1.2                  

Configure the client

Commands executed on the client:
 ifconfig tun5 10.0.1.2 netmask 255.255.255.252   
 ifconfig tun5 10.0.1.2 10.0.1.1                  
The two hosts are now connected and can transparently communicate with any layer 3/4 protocol using the tunnel IP addresses.

Connect two networks

In addition to the p2p setup above, it is more useful to connect two private networks with an SSH VPN using two gates. Suppose for the example, netA is 192.168.51.0/24 and netB 192.168.16.0/24. The procedure is similar as above, we only need to add the routing. NAT must be activated on the private interface only if the gates are not the same as the default gateway of their network.
192.168.51.0/24 (netA)|gateA <-> gateB|192.168.16.0/24 (netB)
  • Connect with SSH using the tunnel option -w.
  • Configure the IP addresses of the tunnel. Once on the server and once on the client.
  • Add the routing for the two networks.
  • If necessary, activate NAT on the private interface of the gate.
The setup is started from gateA in netA.

Connect from gateA to gateB

Connection is started from gateA and commands are executed on gateB.

gateB is on Linux

 ssh -w5:5 root@gateB
 ifconfig tun5 10.0.1.1 netmask 255.255.255.252 
 route add -net 192.168.51.0 netmask 255.255.255.0 dev tun5
 echo 1 > /proc/sys/net/ipv4/ip_forward        
 iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE

gateB is on FreeBSD

 ssh -w5:5 root@gateB                          
 ifconfig tun5 10.0.1.1 10.0.1.2               
 route add 192.168.51.0/24 10.0.1.2
 sysctl net.inet.ip.forwarding=1               
 natd -s -m -u -dynamic -n fxp0                
 sysctl net.inet.ip.fw.enable=1

Configure gateA

Commands executed on gateA:

gateA is on Linux

 ifconfig tun5 10.0.1.2 netmask 255.255.255.252
 route add -net 192.168.16.0 netmask 255.255.255.0 dev tun5
 echo 1 > /proc/sys/net/ipv4/ip_forward
 iptables -t nat -A POSTROUTING -o eth0 -j MASQUERADE

gateA is on FreeBSD

 ifconfig tun5 10.0.1.2 10.0.1.1
 route add 192.168.16.0/24 10.0.1.2
 sysctl net.inet.ip.forwarding=1
 natd -s -m -u -dynamic -n fxp0                
 sysctl net.inet.ip.fw.enable=1
The two private networks are now transparently connected via the SSH VPN. The IP forward and NAT settings are only necessary if the gates are not the default gateways. In this case the clients would not know where to forward the response, and nat must be activated.

RSYNC

Rsync can almost completely replace cp and scp, furthermore interrupted transfers are efficiently restarted. A trailing slash (and the absence thereof) has different meanings, the man page is good... Here some examples:
Copy the directories with full content:
# rsync -a /home/colin/ /backup/colin/                
# rsync -a /var/ /var_bak/
# rsync -aR --delete-during /home/user/ /backup/      
Same as before but over the network and with compression. Rsync uses SSH for the transport per default and will use the ssh key if they are set. Use ":" as with SCP. A typical remote copy:
# rsync -axSRzv /home/user/ user@server:/backup/user/ 
# rsync -a 'user@server:My\ Documents' My\ Documents  
Exclude any directory tmp within /home/user/ and keep the relative folders hierarchy, that is the remote directory will have the structure /backup/home/user/. This is typically used for backups.
# rsync -azR --exclude=tmp/ /home/user/ user@server:/backup/
Use port 20022 for the ssh connection:
# rsync -az -e 'ssh -p 20022' /home/colin/ user@server:/backup/colin/
Using the rsync daemon (used with "::") is much faster, but not encrypted over ssh. The location of /backup is defined by the configuration in /etc/rsyncd.conf. The variable RSYNC_PASSWORD can be set to avoid the need to enter the password manually.
# rsync -axSRz /home/ ruser@hostname::rmodule/backup/
# rsync -axSRz ruser@hostname::rmodule/backup/ /home/    
Some important options:
  • -a, --archive       archive mode; same as -rlptgoD (no -H)
  • -r, --recursive       recurse into directories
  • -R, --relative       use relative path names
  • -H, --hard-links       preserve hard links
  • -S, --sparse       handle sparse files efficiently
  • -x, --one-file-system       don't cross file system boundaries
  • --exclude=PATTERN       exclude files matching PATTERN
  • --delete-during       receiver deletes during xfer, not before
  • --delete-after       receiver deletes after transfer, not before

Rsync on Windows

Rsync is available for Windows through cygwin or as stand-alone packaged in cwrsync http://sourceforge.net/projects/sereds. This is very convenient for automated backups. Install one of them ( not both) and add the path to the Windows system variables: # Control Panel -> System -> tab Advanced, button Environment Variables. Edit the "Path" system variable and add the full path to the installed rsync, e.g. C:\Program Files\cwRsync\bin or C:\cygwin\bin. This way the commands rsync and ssh are available in a Windows command shell.

Public key authentication

Rsync is automatically tunneled over SSH and thus uses the SSH authentication on the server. Automatic backups have to avoid a user interaction, for this the SSH public key authentication can be used and the rsync command will run without a password.
All the following commands are executed within a Windows console. In a console (Start -> Run -> cmd) create and upload the key as described in SSH, change "user" and "server" as appropriate. If the file authorized_keys2 does not exist yet, simply copy id_dsa.pub to authorized_keys2 and upload it.
# ssh-keygen -t dsa -N ''                   
# rsync user@server:.ssh/authorized_keys2 . 
# cat id_dsa.pub >> authorized_keys2        
# rsync authorized_keys2 user@server:.ssh/  
# del authorized_keys2                      
Now test it with (in one line):
rsync -rv "/cygdrive/c/Documents and Settings/%USERNAME%/My Documents/" \
'user@server:My\ Documents/'

Automatic backup

Use a batch file to automate the backup and add the file in the scheduled tasks (Programs -> Accessories -> System Tools -> Scheduled Tasks). For example create the file backup.bat and replace user@server.
@ECHO OFF
REM rsync the directory My Documents
SETLOCAL
SET CWRSYNCHOME=C:\PROGRAM FILES\CWRSYNC
SET CYGWIN=nontsec
SET CWOLDPATH=%PATH%
REM uncomment the next line when using cygwin
SET PATH=%CWRSYNCHOME%\BIN;%PATH%
echo Press Control-C to abort
rsync -av "/cygdrive/c/Documents and Settings/%USERNAME%/My Documents/" \
'user@server:My\ Documents/'
pause

SUDO

Sudo is a standard way to give users some administrative rights without giving out the root password. Sudo is very useful in a multi user environment with a mix of server and workstations. Simply call the command with sudo:
# sudo /etc/init.d/dhcpd restart            
# sudo -u sysadmin whoami                   

Configuration

Sudo is configured in /etc/sudoers and must only be edited with visudo. The basic syntax is (the lists are comma separated):
user hosts = (runas) commands          
  • users one or more users or %group (like %wheel) to gain the rights
  • hosts list of hosts (or ALL)
  • runas list of users (or ALL) that the command rule can be run as. It is enclosed in ( )!
  • commands list of commands (or ALL) that will be run as root or as (runas)
Additionally those keywords can be defined as alias, they are called User_Alias, Host_Alias, Runas_Alias and Cmnd_Alias. This is useful for larger setups. Here a sudoers example:
# cat /etc/sudoers

Host_Alias   DMZ     = 212.118.81.40/28
Host_Alias   DESKTOP = work1, work2


User_Alias   ADMINS  = colin, luca, admin
User_Alias   DEVEL   = joe, jack, julia
Runas_Alias  DBA     = oracle,pgsql


Cmnd_Alias   SYSTEM  = /sbin/reboot,/usr/bin/kill,/sbin/halt,/sbin/shutdown,/etc/init.d/
Cmnd_Alias   PW      = /usr/bin/passwd [A-z]*, !/usr/bin/passwd root 
Cmnd_Alias   DEBUG   = /usr/sbin/tcpdump,/usr/bin/wireshark,/usr/bin/nmap

root,ADMINS  ALL     = (ALL) NOPASSWD: ALL    
DEVEL        DESKTOP = (ALL) NOPASSWD: ALL    
DEVEL        DMZ     = (ALL) NOPASSWD: DEBUG  


sysadmin     DMZ     = (ALL) NOPASSWD: SYSTEM,PW,DEBUG
sysadmin     ALL,!DMZ = (ALL) NOPASSWD: ALL   
%dba         ALL     = (DBA) ALL              


ALL          DESKTOP = NOPASSWD: /sbin/mount /cdrom,/sbin/umount /cdrom

Encrypt Files

OpenSSL

A single file

Encrypt and decrypt:
# openssl aes-128-cbc -salt -in file -out file.aes
# openssl aes-128-cbc -d -salt -in file.aes -out file
Note that the file can of course be a tar archive.

tar and encrypt a whole directory

# tar -cf - directory | openssl aes-128-cbc -salt -out directory.tar.aes      
# openssl aes-128-cbc -d -salt -in directory.tar.aes | tar -x -f -            

tar zip and encrypt a whole directory

# tar -zcf - directory | openssl aes-128-cbc -salt -out directory.tar.gz.aes  
# openssl aes-128-cbc -d -salt -in directory.tar.gz.aes | tar -xz -f -        
  • Use -k mysecretpassword after aes-128-cbc to avoid the interactive password request. However note that this is highly insecure.
  • Use aes-256-cbc instead of aes-128-cbc to get even stronger encryption. This uses also more CPU.

GPG

GnuPG is well known to encrypt and sign emails or any data. Furthermore gpg and also provides an advanced key management system. This section only covers files encryption, not email usage, signing or the Web-Of-Trust.
The simplest encryption is with a symmetric cipher. In this case the file is encrypted with a password and anyone who knows the password can decrypt it, thus the keys are not needed. Gpg adds an extention ".gpg" to the encrypted file names.
# gpg -c file                        
# gpg file.gpg                       

Using keys

For more details see GPG Quick Start http://www.madboa.com/geek/gpg-quickstart and GPG/PGP Basics http://aplawrence.com/Basics/gpg.html and the gnupg documentation http://gnupg.org/documentation among others.
The private and public keys are the heart of asymmetric cryptography. What is important to remember:
  • Your public key is used by others to encrypt files that only you as the receiver can decrypt (not even the one who encrypted the file can decrypt it). The public key is thus meant to be distributed.
  • Your private key is encrypted with your passphrase and is used to decrypt files which were encrypted with your public key. The private key must be kept secure. Also if the key or passphrase is lost, so are all the files encrypted with your public key.
  • The key files are called keyrings as they can contain more than one key.
First generate a key pair. The defaults are fine, however you will have to enter at least your full name and email and optionally a comment. The comment is useful to create more than one key with the same name and email. Also you should use a "passphrase", not a simple password.
# gpg --gen-key                      
The keys are stored in ~/.gnupg/ on Unix, on Windows they are typically stored in
C:/Documents and Settings/%USERNAME%/Application Data/gnupg/.
~/.gnupg/pubring.gpg                 
~/.gnupg/secring.gpg                 
Short reminder on most used options:
  • -e encrypt data
  • -d decrypt data
  • -r NAME encrypt for recipient NAME (or 'Full Name' or 'email@domain')
  • -a create ascii armored output of a key
  • -o use as output file
The examples use 'Your Name' and 'Alice' as the keys are referred to by the email or full name or partial name. For example I can use 'Colin' or '[email protected]' for my key [Colin Barschel (cb.vu) ].

Encrypt for personal use only

No need to export/import any key for this. You have both already.
# gpg -e -r 'Your Name' file                  
# gpg -o file -d file.gpg                     

Encrypt - Decrypt with keys

First you need to export your public key for someone else to use it. And you need to import the public say from Alice to encrypt a file for her. You can either handle the keys in simple ascii files or use a public key server.
For example Alice export her public key and you import it, you can then encrypt a file for her. That is only Alice will be able to decrypt it.
# gpg -a -o alicekey.asc --export 'Alice'     
# gpg --send-keys --keyserver subkeys.pgp.net KEYID   
# gpg --import alicekey.asc                   
# gpg --search-keys --keyserver subkeys.pgp.net 'Alice' 
Once the keys are imported it is very easy to encrypt or decrypt a file:
# gpg -e -r 'Alice' file                      
# gpg -d file.gpg -o file                     

Key administration

# gpg --list-keys                             
    
# gpg --gen-revoke 'Your Name'                
# gpg --list-secret-keys                      
# gpg --delete-keys NAME                      
# gpg --delete-secret-key NAME                
# gpg --fingerprint KEYID                     
# gpg --edit-key KEYID                        

Encrypt Partitions

Linux with LUKS | Linux dm-crypt only | FreeBSD GELI | FBSD pwd only

There are (many) other alternative methods to encrypt disks, I only show here the methods I know and use. Keep in mind that the security is only good as long the OS has not been tempered with. An intruder could easily record the password from the keyboard events. Furthermore the data is freely accessible when the partition is attached and will not prevent an intruder to have access to it in this state.

Linux

Those instructions use the Linux dm-crypt (device-mapper) facility available on the 2.6 kernel. In this example, lets encrypt the partition /dev/sdc1, it could be however any other partition or disk, or USB or a file based partition created with losetup. In this case we would use /dev/loop0. See file image partition. The device mapper uses labels to identify a partition. We use sdc1 in this example, but it could be any string.

dm-crypt with LUKS

LUKS with dm-crypt has better encryption and makes it possible to have multiple passphrase for the same partition or to change the password easily. To test if LUKS is available, simply type # cryptsetup --help, if nothing about LUKS shows up, use the instructions below Without LUKS. First create a partition if necessary: fdisk /dev/sdc.

Create encrypted partition

# dd if=/dev/urandom of=/dev/sdc1          
# cryptsetup -y luksFormat /dev/sdc1       
# cryptsetup luksOpen /dev/sdc1 sdc1
# mkfs.ext3 /dev/mapper/sdc1               
# mount -t ext3 /dev/mapper/sdc1 /mnt
# umount /mnt
# cryptsetup luksClose sdc1                

Attach

# cryptsetup luksOpen /dev/sdc1 sdc1
# mount -t ext3 /dev/mapper/sdc1 /mnt

Detach

# umount /mnt
# cryptsetup luksClose sdc1

dm-crypt without LUKS

# cryptsetup -y create sdc1 /dev/sdc1      
# dmsetup ls                               
# mkfs.ext3 /dev/mapper/sdc1               
# mount -t ext3 /dev/mapper/sdc1 /mnt
# umount /mnt/
# cryptsetup remove sdc1                   
Do exactly the same (without the mkfs part!) to re-attach the partition. If the password is not correct, the mount command will fail. In this case simply remove the map sdc1 ( cryptsetup remove sdc1) and create it again.

FreeBSD

The two popular FreeBSD disk encryption modules are gbde and geli. I now use geli because it is faster and also uses the crypto device for hardware acceleration. See The FreeBSD handbook Chapter 18.6 http://www.freebsd.org/handbook/disks-encrypting.html for all the details. The geli module must be loaded or compiled into the kernel:
options GEOM_ELI
device crypto                                       
# echo 'geom_eli_load="YES"' >> /boot/loader.conf   

Use password and key

I use those settings for a typical disk encryption, it uses a passphrase AND a key to encrypt the master key. That is you need both the password and the generated key /root/ad1.key to attach the partition. The master key is stored inside the partition and is not visible. See below for typical USB or file based image.

Create encrypted partition

# dd if=/dev/random of=/root/ad1.key bs=64 count=1  
# geli init -s 4096 -K /root/ad1.key /dev/ad1       
# geli attach -k /root/ad1.key /dev/ad1             
# dd if=/dev/random of=/dev/ad1.eli bs=1m           
# newfs /dev/ad1.eli                                
# mount /dev/ad1.eli /mnt

Attach

# geli attach -k /root/ad1.key /dev/ad1
# fsck -ny -t ffs /dev/ad1.eli                      
# mount /dev/ad1.eli /mnt

Detach

The detach procedure is done automatically on shutdown.
# umount /mnt
# geli detach /dev/ad1.eli

/etc/fstab

The encrypted partition can be configured to be mounted with /etc/fstab. The password will be prompted when booting. The following settings are required for this example:
# grep geli /etc/rc.conf
geli_devices="ad1"
geli_ad1_flags="-k /root/ad1.key"
# grep geli /etc/fstab
/dev/ad1.eli         /home/private              ufs             rw      0       0

Use password only

It is more convenient to encrypt a USB stick or file based image with a passphrase only and no key. In this case it is not necessary to carry the additional key file around. The procedure is very much the same as above, simply without the key file. Let's encrypt a file based image /cryptedfile of 1 GB.
# dd if=/dev/zero of=/cryptedfile bs=1M count=1000  
# mdconfig -at vnode -f /cryptedfile
# geli init /dev/md0                                
# geli attach /dev/md0
# newfs -U -m 0 /dev/md0.eli
# mount /dev/md0.eli /mnt
# umount /dev/md0.eli
# geli detach md0.eli
It is now possible to mount this image on an other system with the password only.
# mdconfig -at vnode -f /cryptedfile
# geli attach /dev/md0
# mount /dev/md0.eli /mnt

SSL Certificates

So called SSL/TLS certificates are cryptographic public key certificates and are composed of a public and a private key. The certificates are used to authenticate the endpoints and encrypt the data. They are used for example on a web server (https) or mail server (imaps).

Procedure

  • We need a certificate authority to sign our certificate. This step is usually provided by a vendor like Thawte, Verisign, etc., however we can also create our own.
  • Create a certificate signing request. This request is like an unsigned certificate (the public part) and already contains all necessary information. The certificate request is normally sent to the authority vendor for signing. This step also creates the private key on the local machine.
  • Sign the certificate with the certificate authority.
  • If necessary join the certificate and the key in a single file to be used by the application (web server, mail server etc.).

Configure OpenSSL

We use /usr/local/certs as directory for this example check or edit /etc/ssl/openssl.cnf accordingly to your settings so you know where the files will be created. Here are the relevant part of openssl.cnf:
[ CA_default ]
dir             = /usr/local/certs/CA       
certs           = $dir/certs                
crl_dir         = $dir/crl                  
database        = $dir/index.txt            
Make sure the directories exist or create them
# mkdir -p /usr/local/certs/CA
# cd /usr/local/certs/CA
# mkdir certs crl newcerts private
# echo "01" > serial                        
# touch index.txt
If you intend to get a signed certificate from a vendor, you only need a certificate signing request (CSR). This CSR will then be signed by the vendor for a limited time (e.g. 1 year).

Create a certificate authority

If you do not have a certificate authority from a vendor, you'll have to create your own. This step is not necessary if one intend to use a vendor to sign the request. To make a certificate authority (CA):
# openssl req -new -x509 -days 730 -config /etc/ssl/openssl.cnf \
-keyout CA/private/cakey.pem -out CA/cacert.pem

Create a certificate signing request

To make a new certificate (for mail server or web server for example), first create a request certificate with its private key. If your application do not support encrypted private key (for example UW-IMAP does not), then disable encryption with -nodes.
# openssl req -new -keyout newkey.pem -out newreq.pem \
-config /etc/ssl/openssl.cnf
# openssl req -nodes -new -keyout newkey.pem -out newreq.pem \
-config /etc/ssl/openssl.cnf                
Keep this created CSR ( newreq.pem) as it can be signed again at the next renewal, the signature onlt will limit the validity of the certificate. This process also created the private key newkey.pem.

Sign the certificate

The certificate request has to be signed by the CA to be valid, this step is usually done by the vendor. Note: replace "servername" with the name of your server in the next commands.
# cat newreq.pem newkey.pem > new.pem
# openssl ca -policy policy_anything -out servernamecert.pem \
-config /etc/ssl/openssl.cnf -infiles new.pem
# mv newkey.pem servernamekey.pem
Now servernamekey.pem is the private key and servernamecert.pem is the server certificate.

Create united certificate

The IMAP server wants to have both private key and server certificate in the same file. And in general, this is also easier to handle, but the file has to be kept securely!. Apache also can deal with it well. Create a file servername.pem containing both the certificate and key.
  • Open the private key (servernamekey.pem) with a text editor and copy the private key into the "servername.pem" file.
  • Do the same with the server certificate (servernamecert.pem).
The final servername.pem file should look like this:

-----BEGIN RSA PRIVATE KEY-----
MIICXQIBAAKBgQDutWy+o/XZ/[...]qK5LqQgT3c9dU6fcR+WuSs6aejdEDDqBRQ
-----END RSA PRIVATE KEY-----
-----BEGIN CERTIFICATE-----
MIIERzCCA7CgAwIBAgIBBDANB[...]iG9w0BAQQFADCBxTELMAkGA1UEBhMCREUx
-----END CERTIFICATE-----
What we have now in the directory /usr/local/certs/:
  • CA/private/cakey.pem (CA server private key)
  • CA/cacert.pem (CA server public key)
  • certs/servernamekey.pem (server private key)
  • certs/servernamecert.pem (server signed certificate)
  • certs/servername.pem (server certificate with private key)
Keep the private key secure!

View certificate information

To view the certificate information simply do:
# openssl x509 -text -in servernamecert.pem      
# openssl req -noout -text -in server.csr        
# openssl s_client -connect cb.vu:443            

CVS

Server setup | CVS test | SSH tunneling | CVS usage

Server setup

Initiate the CVS

Decide where the main repository will rest and create a root cvs. For example /usr/local/cvs (as root):
# mkdir -p /usr/local/cvs
# setenv CVSROOT /usr/local/cvs      
# cvs init                           
# cd /root
# cvs checkout CVSROOT               
# cd CVSROOT
edit config ( fine as it is)
# cvs commit config
cat >> writers                       
colin
^D                                   
# cvs add writers                    
# cvs edit checkoutlist
# cat >> checkoutlist
writers
^D                                   
# cvs commit                         
Add a readers file if you want to differentiate read and write permissions Note: Do not (ever) edit files directly into the main cvs, but rather checkout the file, modify it and check it in. We did this with the file writers to define the write access.
There are three popular ways to access the CVS at this point. The first two don't need any further configuration. See the examples on CVSROOT below for how to use them:
  • Direct local access to the file system. The user(s) need sufficient file permission to access the CS directly and there is no further authentication in addition to the OS login. However this is only useful if the repository is local.
  • Remote access with ssh with the ext protocol. Any use with an ssh shell account and read/write permissions on the CVS server can access the CVS directly with ext over ssh without any additional tunnel. There is no server process running on the CVS for this to work. The ssh login does the authentication.
  • Remote access with pserver (default port: 2401/tcp). This is the preferred use for larger user base as the users are authenticated by the CVS pserver with a dedicated password database, there is therefore no need for local users accounts. This setup is explained below.

Network setup with inetd

The CVS can be run locally only if a network access is not needed. For a remote access, the daemon inetd can start the pserver with the following line in /etc/inetd.conf (/etc/xinetd.d/cvs on SuSE):
cvspserver	stream  tcp  nowait  cvs  /usr/bin/cvs	cvs \
--allow-root=/usr/local/cvs pserver
It is a good idea to block the cvs port from the Internet with the firewall and use an ssh tunnel to access the repository remotely.

Separate authentication

It is possible to have cvs users which are not part of the OS (no local users). This is actually probably wanted too from the security point of view. Simply add a file named passwd (in the CVSROOT directory) containing the users login and password in the crypt format. This is can be done with the apache htpasswd tool.
Note: This passwd file is the only file which has to be edited directly in the CVSROOT directory. Also it won't be checked out. More info with htpasswd --help
# htpasswd -cb passwd user1 password1  
# htpasswd -b passwd user2 password2
Now add :cvs at the end of each line to tell the cvs server to change the user to cvs (or whatever your cvs server is running under). It looks like this:
# cat passwd
user1:xsFjhU22u8Fuo:cvs
user2:vnefJOsnnvToM:cvs

Test it

Test the login as normal user (for example here me)
# cvs -d :pserver:[email protected]:/usr/local/cvs login
Logging in to :pserver:[email protected]:2401/usr/local/cvs
CVS password:

CVSROOT variable

This is an environment variable used to specify the location of the repository we're doing operations on. For local use, it can be just set to the directory of the repository. For use over the network, the transport protocol must be specified. Set the CVSROOT variable with setenv CVSROOT string on a csh, tcsh shell, or with export CVSROOT=string on a sh, bash shell.
# setenv CVSROOT :pserver:@:/cvsdirectory

# setenv CVSROOT /usr/local/cvs                               
# setenv CVSROOT :local:/usr/local/cvs                        
# setenv CVSROOT :ext:user@cvsserver:/usr/local/cvs           
# setenv CVS_RSH ssh                                          
# setenv CVSROOT :pserver:[email protected]:/usr/local/cvs   
When the login succeeded one can import a new project into the repository: cd into your project root directory
cvs import   
cvs -d :pserver:[email protected]:/usr/local/cvs import MyProject MyCompany START
Where MyProject is the name of the new project in the repository (used later to checkout). Cvs will import the current directory content into the new project.

To checkout:
# cvs -d :pserver:[email protected]:/usr/local/cvs checkout MyProject

# setenv CVSROOT :pserver:[email protected]:/usr/local/cvs
# cvs checkout MyProject

SSH tunneling for CVS

We need 2 shells for this. On the first shell we connect to the cvs server with ssh and port-forward the cvs connection. On the second shell we use the cvs normally as if it where running locally.
on shell 1:
# ssh -L2401:localhost:2401 colin@cvs_server   
# ssh -L2401:cvs_server:2401 colin@gateway     
on shell 2:
# setenv CVSROOT :pserver:colin@localhost:/usr/local/cvs
# cvs login
Logging in to :pserver:colin@localhost:2401/usr/local/cvs
CVS password:
# cvs checkout MyProject/src

CVS commands and usage

Import

The import command is used to add a whole directory, it must be run from within the directory to be imported. Say the directory /devel/ contains all files and subdirectories to be imported. The directory name on the CVS (the module) will be called "myapp".
# cvs import [options] directory-name vendor-tag release-tag
# cd /devel                          
# cvs import myapp Company R1_0      
After a while a new directory "/devel/tools/" was added and it has to be imported too.
# cd /devel/tools
# cvs import myapp/tools Company R1_0

Checkout update add commit

# cvs co myapp/tools                 
# cvs co -r R1_1 myapp               
# cvs -q -d update -P                
# cvs update -A                      
# cvs add newfile                    
# cvs add -kb newfile                
# cvs commit file1 file2             
# cvs commit -m "message"            

Create a patch

It is best to create and apply a patch from the working development directory related to the project, or from within the source directory.
# cd /devel/project
# diff -Naur olddir newdir > patchfile 
# diff -Naur oldfile newfile > patchfile

Apply a patch

Sometimes it is necessary to strip a directory level from the patch, depending how it was created. In case of difficulties, simply look at the first lines of the patch and try -p0, -p1 or -p2.
# cd /devel/project
# patch --dry-run -p0 < patchfile    
# patch -p0 < patchfile
# patch -p1 < patchfile              

SVN

Server setup | SVN+SSH | SVN over http | SVN usage

Subversion (SVN) http://subversion.tigris.org/ is a version control system designed to be the successor of CVS (Concurrent Versions System). The concept is similar to CVS, but many shortcomings where improved. See also the SVN book http://svnbook.red-bean.com/en/1.4/.

Server setup

The initiation of the repository is fairly simple (here for example /home/svn/ must exist):
# svnadmin create --fs-type fsfs /home/svn/project1
Now the access to the repository is made possible with:
  • file:// Direct file system access with the svn client with. This requires local permissions on the file system.
  • svn:// or svn+ssh:// Remote access with the svnserve server (also over SSH). This requires local permissions on the file system (default port: 2690/tcp).
  • http:// Remote access with webdav using apache. No local users are necessary for this method.
Using the local file system, it is now possible to import and then check out an existing project. Unlike with CVS it is not necessary to cd into the project directory, simply give the full path:
# svn import /project1/ file:///home/svn/project1/trunk -m 'Initial import'
# svn checkout file:///home/svn/project1
The new directory "trunk" is only a convention, this is not required.

Remote access with ssh

No special setup is required to access the repository via ssh, simply replace file:// with svn+ssh/hostname. For example:
# svn checkout svn+ssh://hostname/home/svn/project1
As with the local file access, every user needs an ssh access to the server (with a local account) and also read/write access. This method might be suitable for a small group. All users could belong to a subversion group which owns the repository, for example:
# groupadd subversion
# groupmod -A user1 subversion
# chown -R root:subversion /home/svn
# chmod -R 770 /home/svn

Remote access with http (apache)

Remote access over http (https) is the only good solution for a larger user group. This method uses the apache authentication, not the local accounts. This is a typical but small apache configuration:
LoadModule dav_module         modules/mod_dav.so
LoadModule dav_svn_module     modules/mod_dav_svn.so
LoadModule authz_svn_module   modules/mod_authz_svn.so   

  DAV svn
  
  SVNParentPath /home/svn
  AuthType Basic
  AuthName "Subversion repository"
  AuthzSVNAccessFile /etc/apache2/svn.acl
  AuthUserFile /etc/apache2/svn-passwd
  Require valid-user

The apache server needs full access to the repository:
# chown -R www:www /home/svn
Create a user with htpasswd2:
# htpasswd -c /etc/svn-passwd user1 

Access control svn.acl example


[/]
* = r
[groups]
project1-developers = joe, jack, jane

[project1:]
@project1-developers = rw

SVN commands and usage

See also the Subversion Quick Reference Card http://www.cs.put.poznan.pl/csobaniec/Papers/svn-refcard.pdf. Tortoise SVN http://tortoisesvn.tigris.org is a nice Windows interface.

Import

A new project, that is a directory with some files, is imported into the repository with the import command. Import is also used to add a directory with its content to an existing project.
# svn help import                               

# svn import /project1/newdir http://host.url/svn/project1/trunk/src -m 'add newdir'

Typical SVN commands

# svn co http://host.url/svn/project1/trunk     

# svn mkdir http://host.url/svn/project1/tags/  
# svn copy -m "Tag rc1 rel." http://host.url/svn/project1/trunk \
                             http://host.url/svn/project1/tags/1.0rc1
# svn status [--verbose]                        
# svn add src/file.h src/file.cpp               
# svn commit -m 'Added new class file'          
# svn ls http://host.url/svn/project1/tags/     
# svn move foo.c bar.c                          
# svn delete some_old_file                      

Useful Commands

less | vi | mail | tar | dd | screen | find | Miscellaneous

less

The less command displays a text document on the console. It is present on most installation.
# less unixtoolbox.xhtml
Some important commands are (^N stands for [control]-[N]):
  • h H       good help on display
  • f ^F ^V SPACE       Forward one window (or N lines).
  • b ^B ESC-v       Backward one window (or N lines).
  • F       Forward forever; like "tail -f".
  • /pattern       Search forward for (N-th) matching line.
  • ?pattern       Search backward for (N-th) matching line.
  • n       Repeat previous search (for N-th occurrence).
  • N       Repeat previous search in reverse direction.
  • q       quit

vi

Vi is present on ANY Linux/Unix installation (not gentoo?) and it is therefore useful to know some basic commands. There are two modes: command mode and insertion mode. The commands mode is accessed with [ESC], the insertion mode with i. Use : help if you are lost.
The editors nano and pico are usually available too and are easier (IMHO) to use.

Quit

  • :w newfilename       save the file to newfilename
  • :wq or :x       save and quit
  • :q!       quit without saving

Search and move

  • /string       Search forward for string
  • ?string       Search back for string
  • n       Search for next instance of string
  • N       Search for previous instance of string
  • {       Move a paragraph back
  • }       Move a paragraph forward
  • 1G       Move to the first line of the file
  • nG       Move to the n th line of the file
  • G       Move to the last line of the file
  • :%s/OLD/NEW/g       Search and replace every occurrence

Delete copy paste text

  • dd (dw)       Cut current line (word)
  • D       Cut to the end of the line
  • x       Delete (cut) character
  • yy (yw)       Copy line (word) after cursor
  • P       Paste after cursor
  • u       Undo last modification
  • U       Undo all changes to current line

mail

The mail command is a basic application to read and send email, it is usually installed. To send an email simply type "mail user@domain". The first line is the subject, then the mail content. Terminate and send the email with a single dot (.) in a new line. Example:
# mail [email protected]
Subject: Your text is full of typos
"For a moment, nothing happened. Then, after a second or so, 
nothing continued to happen."
.
EOT
#
This is also working with a pipe:
# echo "This is the mail body" | mail [email protected]
This is also a simple way to test the mail server.

tar

The command tar (tape archive) creates and extracts archives of file and directories. The archive .tar is uncompressed, a compressed archive has the extension .tgz or .tar.gz (zip) or .tbz (bzip2). Do not use absolute path when creating an archive, you probably want to unpack it somewhere else. Some typical commands are:

Create

# cd /
# tar -cf home.tar home/        
# tar -czf home.tgz home/       
# tar -cjf home.tbz home/       
Only include one (or two) directories from a tree, but keep the relative structure. For example archive /usr/local/etc and /usr/local/www and the first directory in the archive should be local/.
# tar -C /usr -czf local.tgz local/etc local/www
# tar -C /usr -xzf local.tgz    
# cd /usr; tar -xzf local.tgz   

Extract

# tar -tzf home.tgz             
# tar -xf home.tar              
# tar -xzf home.tgz             
                                
# tar --strip-components 1 -zxvf gallery2.tgz -C gallery/ 
# tar -xjf home.tbz home/colin/file.txt    

More advanced

# tar c dir/ | gzip | ssh user@remote 'dd of=dir.tgz' 
# tar cvf - `find . -print` > backup.tar              
# tar -cf - -C /etc . | tar xpf - -C /backup/etc      
# tar -cf - -C /etc . | ssh user@remote tar xpf - -C /backup/etc      
# tar -czf home.tgz --exclude '*.o' --exclude 'tmp/' home/

dd

The program dd (disk dump or destroy disk or see the meaning of dd) is used to copy partitions and disks and for other copy tricks. Typical usage:
# dd if= of= bs= conv=
Important conv options:
  • notrunc       do not truncate the output file, all zeros will be written as zeros.
  • noerror       continue after read errors (e.g. bad blocks)
  • sync       pad every input block with Nulls to ibs-size
The default byte size is 512 (one block). The MBR, where the partition table is located, is on the first block, the first 63 blocks of a disk are empty. Larger byte sizes are faster to copy but require also more memory.

Backup and restore

# dd if=/dev/hda of=/dev/hdc bs=16065b                
# dd if=/dev/sda7 of=/home/root.img bs=4096 conv=notrunc,noerror 
# dd if=/home/root.img of=/dev/sda7 bs=4096 conv=notrunc,noerror 
# dd bs=1M if=/dev/ad4s3e | gzip -c > ad4s3e.gz                  
# gunzip -dc ad4s3e.gz | dd of=/dev/ad0s3e bs=1M                 
# dd bs=1M if=/dev/ad4s3e | gzip | ssh eedcoba@fry 'dd of=ad4s3e.gz' 
# gunzip -dc ad4s3e.gz | ssh eedcoba@host 'dd of=/dev/ad0s3e bs=1M'
# dd if=/dev/ad0 of=/dev/ad2 skip=1 seek=1 bs=4k conv=noerror    
    

Recover

The command dd will read every single block of the partition. In case of problems it is better to use the option conv=sync,noerror so dd will skip the bad block and write zeros at the destination. Accordingly it is important to set the block size equal or smaller than the disk block size. A 1k size seems safe, set it with bs=1k. If a disk has bad sectors and the data should be recovered from a partition, create an image file with dd, mount the image and copy the content to a new disk. With the option noerror, dd will skip the bad sectors and write zeros instead, thus only the data contained in the bad sectors will be lost.
# dd if=/dev/hda of=/dev/null bs=1m                   
# dd bs=1k if=/dev/hda1 conv=sync,noerror,notrunc | gzip | ssh \ 
root@fry 'dd of=hda1.gz bs=1k'
# dd bs=1k if=/dev/hda1 conv=sync,noerror,notrunc of=hda1.img    
# mount -o loop /hda1.img /mnt                        
# rsync -ax /mnt/ /newdisk/                           
# dd if=/dev/hda of=/dev/hda                          
  

Delete

# dd if=/dev/zero of=/dev/hdc                         
# dd if=/dev/urandom of=/dev/hdc                      
# kill -USR1 PID                                      
# kill -INFO PID                                      

MBR tricks

The MBR contains the boot loader and the partition table and is 512 bytes small. The first 446 are for the boot loader, the bytes 446 to 512 are for the partition table.
# dd if=/dev/sda of=/mbr_sda.bak bs=512 count=1       
# dd if=/dev/zero of=/dev/sda bs=512 count=1          
# dd if=/mbr_sda.bak of=/dev/sda bs=512 count=1       
# dd if=/mbr_sda.bak of=/dev/sda bs=446 count=1       
# dd if=/mbr_sda.bak of=/dev/sda bs=1 count=64 skip=446 seek=446 

screen

Screen (a must have) has two main functionalities:
  • Run multiple terminal session within a single terminal.
  • A started program is decoupled from the real terminal and can thus run in the background. The real terminal can be closed and reattached later.

Short start example

start screen with:
# screen
Within the screen session we can start a long lasting program (like top).
# top
Now detach with Ctrl-a Ctrl-d. Reattach the terminal with:
# screen -R -D
In detail this means: If a session is running, then reattach. If necessary detach and logout remotely first. If it was not running create it and notify the user. Or:
# screen -x
Attach to a running screen in a multi display mode. The console is thus shared among multiple users. Very useful for team work/debug!

Screen commands (within screen)

All screen commands start with Ctrl-a.
  • Ctrl-a ? help and summary of functions
  • Ctrl-a c create an new window (terminal)
  • Ctrl-a Ctrl-n and Ctrl-a Ctrl-p to switch to the next or previous window in the list, by number.
  • Ctrl-a Ctrl-N where N is a number from 0 to 9, to switch to the corresponding window.
  • Ctrl-a " to get a navigable list of running windows
  • Ctrl-a a to clear a missed Ctrl-a
  • Ctrl-a Ctrl-d to disconnect and leave the session running in the background
  • Ctrl-a x lock the screen terminal with a password
The screen session is terminated when the program within the running terminal is closed and you logout from the terminal.

Find

Some important options:
  • -x (on BSD) -xdev (on Linux)       Stay on the same file system (dev in fstab).
  • -exec cmd {} \;       Execute the command and replace {} with the full path
  • -iname       Like -name but is case insensitive
  • -ls       Display information about the file (like ls -la)
  • -size n       n is +-n (k M G T P)
  • -cmin n       File's status was last changed n minutes ago.
# find . -type f ! -perm -444        
# find . -type d ! -perm -111        
# find /home/user/ -cmin 10 -print   
# find . -name '*.[ch]' | xargs grep -E 'expr' 
# find / -name "*.core" | xargs rm   
# find / -name "*.core" -print -exec rm {} \;  
      
# find . \( -iname "*.png" -o -iname "*.jpg" \) -print -exec tar -rf images.tar {} \;
# find . -type f -name "*.txt" ! -name README.txt -print  
# find /var/ -size +10M -exec ls -lh {} \;     
# find /var/ -size +10M -ls           
# find . -size +10M -size -50M -print
# find /usr/ports/ -name work -type d -print -exec rm -rf {} \;  
      
# find / -type f -user root -perm -4000 -exec ls -l {} \; 
Be careful with xarg or exec as it might or might not honor quotings and can return wrong results when files or directories contain spaces. In doubt use "-print0 | xargs -0" instead of "| xargs". The option -print0 must be the last in the find command. See this nice mini tutorial for find http://www.hccfl.edu/pollock/Unix/FindCmd.htm.
# find . -type f | xargs ls -l       
# find . -type f -print0 | xargs -0 ls -l  
# find . -type f -exec ls -l '{}' \; 

Miscellaneous

# which command                      
# time command                       
# time cat                           
# set | grep $USER                   
# cal -3                             
# date [-u|--utc|--universal] [MMDDhhmm[[CC]YY][.ss]]
# date 10022155                      
# whatis grep                        
# whereis java                       
# setenv varname value               
# export varname="value"             
# pwd                                
# mkdir -p /path/to/dir              
# mkdir -p project/{bin,src,obj,doc/{html,man,pdf},debug/some/more/dirs}
# rmdir /path/to/dir                 
# rm -rf /path/to/dir                
# cp -la /dir1 /dir2                 
# cp -lpR /dir1 /dir2                
# cp unixtoolbox.xhtml{,.bak}        
# mv /dir1 /dir2                     
# ls -1                              
# history | tail -50                 
Check file hashes with openssl. This is a nice alternative to the commands md5sum or sha1sum (FreeBSD uses md5 and sha1) which are not always installed.
# openssl md5 file.tar.gz            
# openssl sha1 file.tar.gz           
# openssl rmd160 file.tar.gz         

Install Software

Usually the package manager uses the proxy variable for http/ftp requests. In .bashrc:
export http_proxy=http://proxy_server:3128
export ftp_proxy=http://proxy_server:3128

List installed packages

# rpm -qa                            
# dpkg -l                            
# pkg_info                           
# pkg_info -W smbd                   
# pkginfo                            

Add/remove software

Front ends: yast2/yast for SuSE, redhat-config-packages for Red Hat.
# rpm -i pkgname.rpm                 
# rpm -e pkgname                     

SuSE zypper (see doc and cheet sheet)http://en.opensuse.org/Zypper/Usage

# zypper refresh                     
# zypper install vim                 
# zypper remove vim                  
# zypper search vim                  
# zypper update vim                  

Debian

# apt-get update                     
# apt-get install emacs              
# dpkg --remove emacs                
# dpkg -S file                       

Gentoo

Gentoo uses emerge as the heart of its "Portage" package management system.
# emerge --sync                      
# emerge -u packagename              
# emerge -C packagename              
# revdep-rebuild                     

Solaris

The path is usually /cdrom/cdrom0.
# pkgadd -d /Solaris_9/Product SUNWgtar
# pkgadd -d SUNWgtar                 
# pkgrm SUNWgtar                     

FreeBSD

# pkg_add -r rsync                   
# pkg_delete /var/db/pkg/rsync-xx    
Set where the packages are fetched from with the PACKAGESITE variable. For example:
# export PACKAGESITE=ftp://ftp.freebsd.org/pub/FreeBSD/ports/i386/packages/Latest/ 
# or ftp://ftp.freebsd.org/pub/FreeBSD/ports/i386/packages-6-stable/Latest/

FreeBSD portshttp://www.freebsd.org/handbook/ports.html

The port tree /usr/ports/ is a collection of software ready to compile and install (see man ports). The ports are updated with the program portsnap.
# portsnap fetch extract             
# portsnap fetch update              
# cd /usr/ports/net/rsync/           
# make install distclean             
# make package                       
# pkgdb -F                           

Library path

Due to complex dependencies and runtime linking, programs are difficult to copy to an other system or distribution. However for small programs with little dependencies, the missing libraries can be copied over. The runtime libraries (and the missing one) are checked with ldd and managed with ldconfig.
# ldd /usr/bin/rsync                 
# ldconfig -n /path/to/libs/         
# ldconfig -m /path/to/libs/         
# LD_LIBRARY_PATH                    

Convert Media

Sometimes one simply need to convert a video, audio file or document to another format.

Text encoding

Text encoding can get totally wrong, specially when the language requires special characters like àäç. The command iconv can convert from one encoding to an other.
# iconv -f  -t  
# iconv -f ISO8859-1 -t UTF-8 -o file.input > file_utf8
# iconv -l                           
Without the -f option, iconv will use the local char-set, which is usually fine if the document displays well.

Unix - DOS newlines

Convert DOS (CR/LF) to Unix (LF) newlines and back within a Unix shell. See also dos2unix and unix2dos if you have them.
# sed 's/.$//' dosfile.txt > unixfile.txt                  
# awk '{sub(/\r$/,"");print}' dosfile.txt > unixfile.txt   
# awk '{sub(/$/,"\r");print}' unixfile.txt > dosfile.txt   
Convert Unix to DOS newlines within a Windows environment. Use sed or awk from mingw or cygwin.
# sed -n p unixfile.txt > dosfile.txt
# awk 1 unixfile.txt > dosfile.txt   

PDF to Jpeg and concatenate PDF files

Convert a PDF document with gs (GhostScript) to jpeg (or png) images for each page. Also much shorter with convert and mogrify (from ImageMagick or GraphicsMagick).
# gs -dBATCH -dNOPAUSE -sDEVICE=jpeg -r150 -dTextAlphaBits=4 -dGraphicsAlphaBits=4 \
 -dMaxStripSize=8192 -sOutputFile=unixtoolbox_%d.jpg unixtoolbox.pdf
# convert unixtoolbox.pdf unixtoolbox-%03d.png
# convert *.jpeg images.pdf          
# convert image000* -resample 120x120 -compress JPEG -quality 80 images.pdf
# mogrify -format png *.ppm          
Ghostscript can also concatenate multiple pdf files into a single one. This only works well if the PDF files are "well behaved".
# gs -q -sPAPERSIZE=a4 -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -sOutputFile=all.pdf \
file1.pdf file2.pdf ...              

Convert video

Compress the Canon digicam video with an mpeg4 codec and repair the crappy sound.
# mencoder -o videoout.avi -oac mp3lame -ovc lavc -srate 11025 \
-channels 1 -af-adv force=1 -lameopts preset=medium -lavcopts \
vcodec=msmpeg4v2:vbitrate=600 -mc 0 vidoein.AVI
See sox for sound processing.

Copy an audio cd

The program cdparanoia http://xiph.org/paranoia/ can save the audio tracks (FreeBSD port in audio/cdparanoia/), oggenc can encode in Ogg Vorbis format, lame converts to mp3.
# cdparanoia -B                      
# lame -b 256 in.wav out.mp3         
# for i in *.wav; do lame -b 256 $i `basename $i .wav`.mp3; done
# oggenc in.wav -b 256 out.ogg       

Printing

Print with lpr

# lpr unixtoolbox.ps                 
# export PRINTER=hp4600              
# lpr -Php4500 #2 unixtoolbox.ps     
# lpr -o Duplex=DuplexNoTumble ...   
# lpr -o PageSize=A4,Duplex=DuplexNoTumble ...
# lpq                                
# lpq -l -Php4500                    
# lprm -                             
# lprm -Php4500 3186                 
# lpc status                         
# lpc status hp4500                  
Some devices are not postscript and will print garbage when fed with a pdf file. This might be solved with:
# gs -dSAFER -dNOPAUSE -sDEVICE=deskjet -sOutputFile=\|lpr file.pdf
Print to a PDF file even if the application does not support it. Use gs on the print command instead of lpr.
# gs -q -sPAPERSIZE=a4 -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -sOutputFile=/path/file.pdf

Databases

PostgreSQL

Change root or a username password

# psql -d template1 -U pgsql
> alter user pgsql with password 'pgsql_password';  

Create user and database

The commands createuser, dropuser, createdb and dropdb are convenient shortcuts equivalent to the SQL commands. The new user is bob with database bobdb ; use as root with pgsql the database super user:
# createuser -U pgsql -P bob         
# createdb -U pgsql -O bob bobdb     
# dropdb bobdb                       
# dropuser bob                       
The general database authentication mechanism is configured in pg_hba.conf

Grant remote access

The file $PGSQL_DATA_D/postgresql.conf specifies the address to bind to. Typically listen_addresses = '*' for Postgres 8.x.
The file $PGSQL_DATA_D/pg_hba.conf defines the access control. Examples:
# TYPE  DATABASE    USER        IP-ADDRESS        IP-MASK          METHOD
host    bobdb       bob        212.117.81.42     255.255.255.255   password
host    all         all        0.0.0.0/0                           password

Backup and restore

The backups and restore are done with the user pgsql or postgres. Backup and restore a single database:
# pg_dump --clean dbname > dbname_sql.dump
# psql dbname < dbname_sql.dump
Backup and restore all databases (including users):
# pg_dumpall --clean > full.dump
# psql -f full.dump postgres
In this case the restore is started with the database postgres which is better when reloading an empty cluster.

MySQL

Change mysql root or username password

Method 1

# /etc/init.d/mysql stop

# killall mysqld
# mysqld --skip-grant-tables
# mysqladmin -u root password 'newpasswd'
# /etc/init.d/mysql start

Method 2

# mysql -u root mysql
 UPDATE USER SET PASSWORD=PASSWORD("newpassword") where user='root';
 FLUSH PRIVILEGES;                           
 quit

Create user and database (see MySQL dochttp://dev.mysql.com/doc/refman/5.1/en/adding-users.html)

# mysql -u root mysql
 CREATE USER 'bob'@'localhost' IDENTIFIED BY 'pwd'; 
 CREATE DATABASE bobdb;
 GRANT ALL ON *.* TO 'bob'@'%' IDENTIFIED BY 'pwd'; 
                                                   
 DROP DATABASE bobdb;                        
 DROP USER bob;                              
 DELETE FROM mysql.user WHERE user='bob and host='hostname'; 
 FLUSH PRIVILEGES;

Grant remote access

Remote access is typically permitted for a database, and not all databases. The file /etc/my.cnf contains the IP address to bind to. Typically comment the line bind-address = out.
# mysql -u root mysql
 GRANT ALL ON bobdb.* TO bob@'xxx.xxx.xxx.xxx' IDENTIFIED BY 'PASSWORD';
 REVOKE GRANT OPTION ON foo.* FROM bar@'xxx.xxx.xxx.xxx';
 FLUSH PRIVILEGES;                  

Backup and restore

Backup and restore a single database:
# mysqldump -u root -psecret --add-drop-database dbname > dbname_sql.dump
# mysql -u root -psecret -D dbname < dbname_sql.dump
Backup and restore all databases:
# mysqldump -u root -psecret --add-drop-database --all-databases > full.dump
# mysql -u root -psecret < full.dump
Here is "secret" the mysql root password, there is no space after -p. When the -p option is used alone (w/o password), the password is asked at the command prompt.

SQLite

SQLite http://www.sqlite.org is a small powerful self-contained, serverless, zero-configuration SQL database.

Dump and restore

It can be useful to dump and restore an SQLite database. For example you can edit the dump file to change a column attribute or type and then restore the database. This is easier than messing with SQL commands. Use the command sqlite3 for a 3.x database.
# sqlite database.db .dump > dump.sql              
# sqlite database.db < dump.sql                    

Convert 2.x to 3.x database

sqlite database_v2.db .dump | sqlite3 database_v3.db

Disk Quota

A disk quota allows to limit the amount of disk space and/or the number of files a user or (or member of group) can use. The quotas are allocated on a per-file system basis and are enforced by the kernel.

Linux setup

The quota tools package usually needs to be installed, it contains the command line tools.
Activate the user quota in the fstab and remount the partition. If the partition is busy, either all locked files must be closed, or the system must be rebooted. Add usrquota to the fstab mount options, for example:
/dev/sda2     /home    reiserfs     rw,acl,user_xattr,usrquota 1 1
# mount -o remount /home
# mount                              
Initialize the quota.user file with quotacheck.
# quotacheck -vum /home
# chmod 644 /home/aquota.user        
Activate the quota either with the provided script (e.g. /etc/init.d/quotad on SuSE) or with quotaon:
quotaon -vu /home
Check that the quota is active with:
quota -v

FreeBSD setup

The quota tools are part of the base system, however the kernel needs the option quota. If it is not there, add it and recompile the kernel.
options QUOTA
As with Linux, add the quota to the fstab options (userquota, not usrquota):
/dev/ad0s1d    /home    ufs     rw,noatime,userquota    2  2
# mount /home                        
Enable disk quotas in /etc/rc.conf and start the quota.
# grep quotas /etc/rc.conf
enable_quotas="YES"                  
check_quotas="YES"                   
# /etc/rc.d/quota start

Assign quota limits

The quotas are not limited per default (set to 0). The limits are set with edquota for single users. A quota can be also duplicated to many users. The file structure is different between the quota implementations, but the principle is the same: the values of blocks and inodes can be limited. Only change the values of soft and hard. If not specified, the blocks are 1k. The grace period is set with edquota -t. For example:
# edquota -u colin

Linux

Disk quotas for user colin (uid 1007):
  Filesystem         blocks       soft       hard     inodes     soft     hard
  /dev/sda8            108       1000       2000          1        0        0

FreeBSD

Quotas for user colin:
/home: kbytes in use: 504184, limits (soft = 700000, hard = 800000)
   inodes in use: 1792, limits (soft = 0, hard = 0)

For many users

The command edquota -p is used to duplicate a quota to other users. For example to duplicate a reference quota to all users:
# edquota -p refuser `awk -F: '$3 > 499 {print $1}' /etc/passwd`
# edquota -p refuser user1 user2     

Checks

Users can check their quota by simply typing quota (the file quota.user must be readable). Root can check all quotas.
# quota -u colin                     
# repquota /home                     

Shells

Most Linux distributions use the bash shell while the BSDs use tcsh, the bourne shell is only used for scripts. Filters are very useful and can be piped:
  • grep   Pattern matching
  • sed   Search and Replace strings or characters
  • cut   Print specific columns from a marker
  • sort   Sort alphabetically or numerically
  • uniq   Remove duplicate lines from a file
For example used all at once:
# ifconfig | sed 's/  / /g' | cut -d" " -f1 | uniq | grep -E "[a-z0-9]+" | sort -r
# ifconfig | sed '/.*inet addr:/!d;s///;s/ .*//'|sort -t. -k1,1n -k2,2n -k3,3n -k4,4n
The first character in the sed pattern is a tab. To write a tab on the console, use ctrl-v ctrl-tab.

bash

Redirects and pipes for bash and sh:
# cmd 1> file                         
# cmd 2> file                         
# cmd 1>> file                        
# cmd &> file                         
# cmd >file 2>&1                      
# cmd1 | cmd2                         
# cmd1 2>&1 | cmd2                    
Modify your configuration in ~/.bashrc (it can also be ~/.bash_profile). The following entries are useful, reload with ". .bashrc". With cygwin use ~/.bash_profile; with rxvt past with shift + left-click.
# in .bashrc
bind '"\e[A"':history-search-backward 
bind '"\e[B"':history-search-forward  
set -o emacs                          
set bell-style visible                
    
PS1="\[\033[1;30m\][\[\033[1;34m\]\u\[\033[1;30m\]"
PS1="$PS1@\[\033[0;33m\]\h\[\033[1;30m\]]\[\033[0;37m\]"
PS1="$PS1\w\[\033[1;30m\]>\[\033[0m\]"

alias  ls='ls -aF'                    
alias  ll='ls -aFls'                  
alias  la='ls -all'
alias ..='cd ..'
alias ...='cd ../..'
export HISTFILESIZE=5000              
export CLICOLOR=1                     
export LSCOLORS=ExGxFxdxCxDxDxBxBxExEx

tcsh

Redirects and pipes for tcsh and csh (simple > and >> are the same as sh):
# cmd >& file                         
# cmd >>& file                        
# cmd1 | cmd2                         
# cmd1 |& cmd2                        
The settings for csh/tcsh are set in ~/.cshrc, reload with "source .cshrc". Examples:
# in .cshrc
alias  ls      'ls -aF'
alias  ll      'ls -aFls'
alias  la      'ls -all'
alias  ..      'cd ..'
alias  ...     'cd ../..'
set   prompt    = "%B%n%b@%B%m%b%/> " 
set   history   =  5000
set   savehist  = ( 6000 merge )
set   autolist                        
set   visiblebell                     

bindkey -e     Select Emacs bindings  
bindkey -k up history-search-backward 
bindkey -k down history-search-forward
setenv CLICOLOR 1                     
setenv LSCOLORS ExGxFxdxCxDxDxBxBxExEx
The emacs mode enables to use the emacs keys shortcuts to modify the command prompt line. This is extremely useful (not only for emacs users). The most used commands are:
  • C-a       Move cursor to beginning of line
  • C-e       Move cursor to end of line
  • M-b       Move cursor back one word
  • M-f       Move cursor forward one word
  • M-d       Cut the next word
  • C-w       Cut the last word
  • C-u       Cut everything before the cursor
  • C-k       Cut everything after the cursor (rest of the line)
  • C-y       Paste the last thing to be cut (simply paste)
  • C-_       Undo
Note: C- = hold control, M- = hold meta (which is usually the alt or escape key).

Scripting

Basics | Script example | awk | sed | Regular Expressions | useful commands

The Bourne shell (/bin/sh) is present on all Unix installations and scripts written in this language are (quite) portable; man 1 sh is a good reference.

Basics

Variables and arguments

Assign with variable=value and get content with $variable
MESSAGE="Hello World"                        
PI=3.1415                                    
N=8
TWON=`expr $N * 2`                           
TWON=$(($N * 2))                             
TWOPI=`echo "$PI * 2" | bc -l`               
ZERO=`echo "c($PI/4)-sqrt(2)/2" | bc -l`
The command line arguments are
$0, $1, $2, ...                               
$#                                           
$*                                           

Special Variables

$$                                           
$?                                           
  command
   [ $ != 0 ]; 
    echo "command failed"
  
=`pwd`
=${}/file.txt
echo ${##*/}                           
echo ${%%.*}                           
=${:=string}                          

Constructs

  in `ls`

    echo $


=0
 [ $ -lt 5 ]; 
    echo $
    sleep 1
    =$(($ + 1))


myfunction() {
    find . -type f -name "*.$" -print       
}
myfunction "txt"

Generate a file

=/home/colin
cat > testhome.sh << _EOF

 [ -d "$" ] ; 
    echo $ exists

    echo $ does not exist

_EOF
sh testhome.sh

Bourne script example

As a small example, the script used to create a PDF booklet from this xhtml document:


 [ $  1 ];                         
  echo 1>&2 "Usage: $0 HtmlFile"
   1                                     


=$                                      
=${%.*}                             
=${#*.}                              

prince $ -o $.pdf                   
pdftops -paper A4 -noshrink $.pdf $.ps 
cat $.ps |psbook|psnup -Pa4 -2 |pstops -b "2:0,1U(21cm,29.7cm)" > $.book.ps

ps2pdf13 -sPAPERSIZE=a4 -sAutoRotatePages=None $.book.ps $.book.pdf
                                             
 0                                       

Some awk commands

Awk is useful for field stripping, like cut in a more powerful way. Search this document for other examples. See for example gnulamp.com and one-liners for awk for some nice examples.
awk '{ print $2, $1 }' file                  
awk '{printf("%5d : %s\n", NR,$0)}' file     
awk '{print FNR "\t" $0}' files              
awk NF test.txt                              
awk 'length > 80'                            

Some sed commands

Here is the one liner gold mine http://student.northpark.edu/pemente/sed/sed1line.txt. And a good introduction and tutorial to sed http://www.grymoire.com/Unix/Sed.html.
sed 's/string1/string2/g'                    
sed -i 's/wroong/wrong/g' *.txt              
sed 's/\(.*\)1/\12/g'                        
sed '/

/,/<\/p>/d' t.xhtml sed '/ *#/d; /^ *$/d' sed 's/[ \t]*$//' sed 's/^[ \t]*//;s/[ \t]*$//' sed 's/[^*]/[&]/' sed = file | sed 'N;s/\n/\t/' > file.num

Regular Expressions

Some basic regular expression useful for sed too. See Basic Regex Syntax http://www.regular-expressions.info/reference.html for a good primer.
[\^$.|?*+()                          
\                                    
*                                    
.                                    
.*                                   
^                                    
$                                    
.$                                   
^ $                                  
[^A-Z]                               

Some useful commands

The following commands are useful to include in a script or as one liners.
sort -t. -k1,1n -k2,2n -k3,3n -k4,4n         
echo 'Test' | tr '[:lower:]' '[:upper:]'     
echo foo.bar | cut -d . -f 1                 
PID=$(ps | grep script.sh | grep bin | awk '{print $1}')    
PID=$(ps axww | grep [p]ing | awk '{print $1}')             
IP=$(ifconfig $INTERFACE | sed '/.*inet addr:/!d;s///;s/ .*//')   
IP=$(ifconfig $INTERFACE | sed '/.*inet /!d;s///;s/ .*//')        
if [ `diff file1 file2 | wc -l` != 0 ]; then [...] fi       
cat /etc/master.passwd | grep -v root | grep -v \*: | awk -F":" \ 
'{ printf("%s:%s\n", $1, $2) }' > /usr/local/etc/apache2/passwd

testuser=$(cat /usr/local/etc/apache2/passwd | grep -v \    
root | grep -v \*: | awk -F":" '{ printf("%s\n", $1) }' | grep ^user$)
:(){ :|:& };:                                
tail +2 file > file2                         
I use this little trick to change the file extension for many files at once. For example from .cxx to .cpp. Test it first without the | sh at the end. You can also do this with the command rename if installed. Or with bash builtins.
# ls *.cxx | awk -F. '{print "mv "$0" "$1".cpp"}' | sh
# ls *.c | sed "s/.*/cp & &.$(date "+%Y%m%d")/" | sh 
# rename .cxx .cpp *.cxx                             
# for i in *.cxx; do mv $i ${i%%.cxx}.cpp; done      

Programming

C basics

strcpy(newstr,str)                        
expr1 ? expr2 : expr3                     
x = (y > z) ? y : z;                      
int a[]={0,1,2};                          
int a[2][3]={{1,2,3},{4,5,6}};            
int i = 12345;                            
char str[10];
sprintf(str, "%d", i);

C example

A minimal c program simple.c:
 
main() {
     number=42;
    printf("The answer is %i\n", number);  
}
Compile with:
# gcc simple.c -o simple
# ./simple
The answer is 42

C++ basics

*pointer                                  
&obj                                      
obj.x                                     
pobj->x                                   
                                          

C++ example

As a slightly more realistic program in C++: a class in its own header (IPv4.h) and implementation (IPv4.cpp) and a program which uses the class functionality. The class converts an IP address in integer format to the known quad format.

IPv4 class

IPv4.h:

 IPV4_H
 IPV4_H
 

 GenericUtils {                           IPv4 {                                      :
    IPv4(); ~IPv4();
    std::string IPint_to_IPquad(  ip);};
} 
 

IPv4.cpp:

 "IPv4.h"
 
 
  std;                                GenericUtils;

IPv4::IPv4() {}                                   IPv4::~IPv4() {}
string IPv4::IPint_to_IPquad(  ip) {      ostringstream ipstr;                              ipstr << ((ip &0xff000000) >> 24)                       << "." << ((ip &0x00ff0000) >> 16)
          << "." << ((ip &0x0000ff00) >> 8)
          << "." << ((ip &0x000000ff));
     ipstr.str();
}

The program simplecpp.cpp

 "IPv4.h"
 
 
  std;
 main ( argc, * argv[]) {
    string ipstr;                                       ipint = 1347861486;             
    GenericUtils::IPv4 iputils;                       ipstr = iputils.IPint_to_IPquad(ipint);           cout << ipint << " = " << ipstr << endl;      

     0;
}
Compile and execute with:
# g++ -c IPv4.cpp simplecpp.cpp                
# g++ IPv4.o simplecpp.o -o simplecpp.exe      
# ./simplecpp.exe 
1347861486 = 80.86.187.238
Use ldd to check which libraries are used by the executable and where they are located. Also used to check if a shared library is missing or if the executable is static.
# ldd /sbin/ifconfig                           
# ar rcs staticlib.a *.o                       
# ar t staticlib.a                             
# ar x /usr/lib/libc.a version.o               
# nm version.o                                 

Simple Makefile

The minimal Makefile for the multi-source program is shown below. The lines with instructions must begin with a tab! The back slash "\" can be used to cut long lines.
 = g++
 = -O
 = IPv4.o simplecpp.o

: ${OBJS}
	${CC} -o simplecpp ${CFLAGS} ${OBJS}
:
	rm -f ${TARGET} ${OBJS}

Online Help

Documentation

Linux Documentation en.tldp.org
Linux Man Pages www.linuxmanpages.com
Linux commands directory www.oreillynet.com/linux/cmd
Linux doc man howtos linux.die.net
FreeBSD Handbook www.freebsd.org/handbook
FreeBSD Man Pages www.freebsd.org/cgi/man.cgi
FreeBSD user wiki www.freebsdwiki.net
Solaris Man Pages docs.sun.com/app/docs/coll/40.10

Other Unix/Linux references

Rosetta Stone for Unix bhami.com/rosetta.html (a Unix command translator)
Unix guide cross reference unixguide.net/unixguide.shtml
Linux commands line list www.linuxcmd.org
Short Linux reference www.pixelbeat.org/cmdline.html
Little command line goodies www.shell-fu.org

转载于:https://www.cnblogs.com/sxfmol/archive/2011/02/19/1958704.html

你可能感兴趣的:(运维,操作系统,shell)