# 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.
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.
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.
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
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:
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.
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):
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.
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)
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.
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*.
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.
# 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.
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.
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.
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.
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
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).
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:
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 @:
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:
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.
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.
# 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.
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:
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):
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
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
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):
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.
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
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.
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.
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.
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:
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.
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.
-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.
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 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:
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.
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.
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.
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:
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.
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:
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.
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:
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):
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.
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.
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:
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):
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
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:
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.
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.
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:
# 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
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.
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):
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:
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:
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:
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 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 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.
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.
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.
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.
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.
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.
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.
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).
The program cdparanoiahttp://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
# 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:
# 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;
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.
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
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:
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:
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):
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
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.
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.
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 ${i%%.cxx}.cpp; done
Programming
C basics
strcpy(newstr,str)
expr1 ? expr2 : expr3
x = (y > z) ? y : z;
int a[]={0,1,2};
int a[2][3]={{1,2,3},{4,5,6}};
int i = 12345;
char str[10];
sprintf(str, "%d", i);
C example
A minimal c program simple.c:
main() {
number=42;
printf("The answer is %i\n", number);
}
Compile with:
# gcc simple.c -o simple
# ./simple
The answer is 42
C++ basics
*pointer
&obj
obj.x
pobj->x
C++ example
As a slightly more realistic program in C++: a class in its own header (IPv4.h) and implementation (IPv4.cpp) and a program which uses the class functionality. The class converts an IP address in integer format to the known quad format.
Use ldd to check which libraries are used by the executable and where they are located. Also used to check if a shared library is missing or if the executable is static.
# ldd /sbin/ifconfig
# ar rcs staticlib.a *.o
# ar t staticlib.a
# ar x /usr/lib/libc.a version.o
# nm version.o
Simple Makefile
The minimal Makefile for the multi-source program is shown below. The lines with instructions must begin with a tab! The back slash "\" can be used to cut long lines.
Create the Google Play Account
Having a Google account, pay 25$, then you get your google developer account.
References:
http://developer.android.com/distribute/googleplay/start.html
https://p
